diff --git a/llvm/lib/Target/AMDGPU/AMDGPU.h b/llvm/lib/Target/AMDGPU/AMDGPU.h
index 41ab0eba8b125c..c4680cbedadf62 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPU.h
+++ b/llvm/lib/Target/AMDGPU/AMDGPU.h
@@ -299,6 +299,9 @@ extern char &SIMemoryLegalizerID;
 void initializeSIModeRegisterPass(PassRegistry&);
 extern char &SIModeRegisterID;
 
+void initializeAMDGPUReleaseVGPRsPass(PassRegistry &);
+extern char &AMDGPUReleaseVGPRsID;
+
 void initializeAMDGPUInsertDelayAluPass(PassRegistry &);
 extern char &AMDGPUInsertDelayAluID;
 
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUReleaseVGPRs.cpp b/llvm/lib/Target/AMDGPU/AMDGPUReleaseVGPRs.cpp
new file mode 100644
index 00000000000000..a86871a4a653f6
--- /dev/null
+++ b/llvm/lib/Target/AMDGPU/AMDGPUReleaseVGPRs.cpp
@@ -0,0 +1,140 @@
+//===- AMDGPUReleaseVGPRs.cpp - Automatically release vgprs on GFX11+ -----===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+/// \file
+/// Insert S_SENDMSG instructions to release vgprs on GFX11+.
+//
+//===----------------------------------------------------------------------===//
+
+#include "AMDGPU.h"
+#include "AMDGPUSubtarget.h"
+#include "GCNSubtarget.h"
+#include "MCTargetDesc/AMDGPUMCTargetDesc.h"
+#include "SIDefines.h"
+#include "llvm/ADT/STLExtras.h"
+#include "llvm/CodeGen/MachineBasicBlock.h"
+#include "llvm/CodeGen/MachineOperand.h"
+using namespace llvm;
+
+#define DEBUG_TYPE "release-vgprs"
+
+namespace {
+
+class AMDGPUReleaseVGPRs : public MachineFunctionPass {
+public:
+  static char ID;
+
+  const SIInstrInfo *SII;
+  const SIRegisterInfo *TRI;
+
+  AMDGPUReleaseVGPRs() : MachineFunctionPass(ID) {}
+
+  void getAnalysisUsage(AnalysisUsage &AU) const override {
+    AU.setPreservesAll();
+    MachineFunctionPass::getAnalysisUsage(AU);
+  }
+
+  // Used to cache the result of isLastInstructionVMEMStore for each block
+  using BlockVMEMStoreType = DenseMap<MachineBasicBlock *, bool>;
+  BlockVMEMStoreType BlockVMEMStore;
+
+  // Return true if the last instruction referencing a vgpr in this MBB
+  // is a VMEM store, otherwise return false.
+  // Visit previous basic blocks to find this last instruction if needed.
+  // Because this pass is late in the pipeline, it is expected that the
+  // last vgpr use will likely be one of vmem store, ds, exp.
+  // Loads and others vgpr operations would have been
+  // deleted by this point, except for complex control flow involving loops.
+  // This is why we are just testing the type of instructions rather
+  // than the operands.
+  bool isLastVGPRUseVMEMStore(MachineBasicBlock &MBB) {
+    // Use the cache to break infinite loop and save some time. Initialize to
+    // false in case we have a cycle.
+    BlockVMEMStoreType::iterator It;
+    bool Inserted;
+    std::tie(It, Inserted) = BlockVMEMStore.insert({&MBB, false});
+    bool &CacheEntry = It->second;
+    if (!Inserted)
+      return CacheEntry;
+
+    for (auto &MI : reverse(MBB.instrs())) {
+      // If it's a VMEM store, a vgpr will be used, return true.
+      if ((SIInstrInfo::isVMEM(MI) || SIInstrInfo::isFLAT(MI)) && MI.mayStore())
+        return CacheEntry = true;
+
+      // If it's referencing a VGPR but is not a VMEM store, return false.
+      if (SIInstrInfo::isDS(MI) || SIInstrInfo::isEXP(MI) ||
+          SIInstrInfo::isVMEM(MI) || SIInstrInfo::isFLAT(MI) ||
+          SIInstrInfo::isVALU(MI))
+        return CacheEntry = false;
+    }
+
+    // Recursive call into parent blocks. Look into predecessors if there is no
+    // vgpr used in this block.
+    return CacheEntry = llvm::any_of(MBB.predecessors(),
+                                     [this](MachineBasicBlock *Parent) {
+                                       return isLastVGPRUseVMEMStore(*Parent);
+                                     });
+  }
+
+  bool runOnMachineBasicBlock(MachineBasicBlock &MBB) {
+
+    bool Changed = false;
+
+    for (auto &MI : MBB.terminators()) {
+      // Look for S_ENDPGM instructions
+      if (MI.getOpcode() == AMDGPU::S_ENDPGM ||
+          MI.getOpcode() == AMDGPU::S_ENDPGM_SAVED) {
+        // If the last instruction using a VGPR in the block is a VMEM store,
+        // release VGPRs. The VGPRs release will be placed just before ending
+        // the program
+        if (isLastVGPRUseVMEMStore(MBB)) {
+          BuildMI(MBB, MI, DebugLoc(), SII->get(AMDGPU::S_SENDMSG))
+              .addImm(AMDGPU::SendMsg::ID_DEALLOC_VGPRS_GFX11Plus);
+          Changed = true;
+        }
+      }
+    }
+
+    return Changed;
+  }
+
+  bool runOnMachineFunction(MachineFunction &MF) override {
+    Function &F = MF.getFunction();
+    if (skipFunction(F) || !AMDGPU::isEntryFunctionCC(F.getCallingConv()))
+      return false;
+
+    // This pass only runs on GFX11+
+    const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
+    if (ST.getGeneration() < AMDGPUSubtarget::GFX11)
+      return false;
+
+    LLVM_DEBUG(dbgs() << "AMDGPUReleaseVGPRs running on " << MF.getName()
+                      << "\n");
+
+    SII = ST.getInstrInfo();
+    TRI = ST.getRegisterInfo();
+
+    bool Changed = false;
+    for (auto &MBB : MF) {
+      Changed |= runOnMachineBasicBlock(MBB);
+    }
+
+    BlockVMEMStore.clear();
+
+    return Changed;
+  }
+};
+
+} // namespace
+
+char AMDGPUReleaseVGPRs::ID = 0;
+
+char &llvm::AMDGPUReleaseVGPRsID = AMDGPUReleaseVGPRs::ID;
+
+INITIALIZE_PASS(AMDGPUReleaseVGPRs, DEBUG_TYPE, "Release VGPRs", false, false)
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
index 137e7048390b5b..1c6b9d35695ae7 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUTargetMachine.cpp
@@ -369,6 +369,7 @@ extern "C" LLVM_EXTERNAL_VISIBILITY void LLVMInitializeAMDGPUTarget() {
   initializeAMDGPURewriteOutArgumentsPass(*PR);
   initializeAMDGPUUnifyMetadataPass(*PR);
   initializeSIAnnotateControlFlowPass(*PR);
+  initializeAMDGPUReleaseVGPRsPass(*PR);
   initializeAMDGPUInsertDelayAluPass(*PR);
   initializeSIInsertHardClausesPass(*PR);
   initializeSIInsertWaitcntsPass(*PR);
@@ -1421,6 +1422,9 @@ void GCNPassConfig::addPreEmitPass() {
   // cases.
   addPass(&PostRAHazardRecognizerID);
 
+  if (getOptLevel() > CodeGenOpt::Less)
+    addPass(&AMDGPUReleaseVGPRsID);
+
   if (isPassEnabled(EnableInsertDelayAlu, CodeGenOpt::Less))
     addPass(&AMDGPUInsertDelayAluID);
 
diff --git a/llvm/lib/Target/AMDGPU/CMakeLists.txt b/llvm/lib/Target/AMDGPU/CMakeLists.txt
index 717bd5f5c3eab6..5cda7819746177 100644
--- a/llvm/lib/Target/AMDGPU/CMakeLists.txt
+++ b/llvm/lib/Target/AMDGPU/CMakeLists.txt
@@ -88,6 +88,7 @@ add_llvm_target(AMDGPUCodeGen
   AMDGPUPromoteKernelArguments.cpp
   AMDGPURegBankCombiner.cpp
   AMDGPURegisterBankInfo.cpp
+  AMDGPUReleaseVGPRs.cpp
   AMDGPUReplaceLDSUseWithPointer.cpp
   AMDGPUResourceUsageAnalysis.cpp
   AMDGPURewriteOutArguments.cpp
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.atomic.dim.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.atomic.dim.ll
index fbac801709ee06..d95f8eb5d7a452 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.atomic.dim.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.atomic.dim.ll
@@ -3,8 +3,8 @@
 ; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -o - %s | FileCheck -check-prefix=GFX8 %s
 ; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx900 -o - %s | FileCheck -check-prefix=GFX900 %s
 ; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx90a -o - %s | FileCheck -check-prefix=GFX90A %s
-; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -o - %s | FileCheck -check-prefix=GFX10 %s
-; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -o - %s | FileCheck -check-prefix=GFX10 %s
+; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -o - %s | FileCheck -check-prefixes=GFX10PLUS,GFX10 %s
+; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -o - %s | FileCheck -check-prefixes=GFX10PLUS,GFX11 %s
 
 define amdgpu_ps float @atomic_swap_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %s) {
 ; GFX6-LABEL: atomic_swap_i32_1d:
@@ -64,19 +64,19 @@ define amdgpu_ps float @atomic_swap_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_swap_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_swap v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_swap_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_swap v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.swap.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -141,19 +141,19 @@ define amdgpu_ps float @atomic_add_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -218,19 +218,19 @@ define amdgpu_ps float @atomic_sub_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_sub_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_sub v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_sub_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_sub v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.sub.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -295,19 +295,19 @@ define amdgpu_ps float @atomic_smin_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_smin_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_smin v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_smin_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_smin v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.smin.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -372,19 +372,19 @@ define amdgpu_ps float @atomic_umin_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_umin_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_umin v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_umin_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_umin v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.umin.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -449,19 +449,19 @@ define amdgpu_ps float @atomic_smax_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_smax_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_smax v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_smax_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_smax v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.smax.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -526,19 +526,19 @@ define amdgpu_ps float @atomic_umax_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_umax_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_umax v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_umax_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_umax v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.umax.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -603,19 +603,19 @@ define amdgpu_ps float @atomic_and_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_and_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_and v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_and_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_and v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.and.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -680,19 +680,19 @@ define amdgpu_ps float @atomic_or_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32 %
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_or_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_or v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_or_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_or v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.or.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -757,19 +757,19 @@ define amdgpu_ps float @atomic_xor_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_xor_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_xor v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_xor_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_xor v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.xor.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -834,19 +834,19 @@ define amdgpu_ps float @atomic_inc_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_inc_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_inc v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_inc_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_inc v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.inc.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -911,19 +911,19 @@ define amdgpu_ps float @atomic_dec_i32_1d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_dec_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_dec v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_dec_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_dec v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.dec.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -987,19 +987,19 @@ define amdgpu_ps float @atomic_cmpswap_i32_1d(<8 x i32> inreg %rsrc, i32 %cmp, i
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_cmpswap_i32_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_cmpswap v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_cmpswap_i32_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_cmpswap v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.cmpswap.1d.i32.i32(i32 %cmp, i32 %swap, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -1071,6 +1071,20 @@ define amdgpu_ps void @atomic_cmpswap_i32_1d_no_return(<8 x i32> inreg %rsrc, i3
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_atomic_cmpswap v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: atomic_cmpswap_i32_1d_no_return:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_atomic_cmpswap v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.cmpswap.1d.i32.i32(i32 %cmp, i32 %swap, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -1135,19 +1149,19 @@ define amdgpu_ps float @atomic_add_i32_2d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i32_2d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i32_2d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.2d.i32.i32(i32 %data, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -1214,19 +1228,19 @@ define amdgpu_ps float @atomic_add_i32_3d(<8 x i32> inreg %rsrc, i32 %data, i32
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i32_3d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v0, v[1:3], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_3D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i32_3d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v0, v[1:3], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_3D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.3d.i32.i32(i32 %data, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -1293,19 +1307,19 @@ define amdgpu_ps float @atomic_add_i32_cube(<8 x i32> inreg %rsrc, i32 %data, i3
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i32_cube:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v0, v[1:3], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_CUBE unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i32_cube:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v0, v[1:3], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_CUBE unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.cube.i32.i32(i32 %data, i32 %s, i32 %t, i32 %face, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -1371,19 +1385,19 @@ define amdgpu_ps float @atomic_add_i32_1darray(<8 x i32> inreg %rsrc, i32 %data,
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i32_1darray:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D_ARRAY unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i32_1darray:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v0, v[1:2], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D_ARRAY unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.1darray.i32.i32(i32 %data, i32 %s, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -1450,19 +1464,19 @@ define amdgpu_ps float @atomic_add_i32_2darray(<8 x i32> inreg %rsrc, i32 %data,
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i32_2darray:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v0, v[1:3], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D_ARRAY unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i32_2darray:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v0, v[1:3], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D_ARRAY unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.2darray.i32.i32(i32 %data, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -1529,19 +1543,19 @@ define amdgpu_ps float @atomic_add_i32_2dmsaa(<8 x i32> inreg %rsrc, i32 %data,
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i32_2dmsaa:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v0, v[1:3], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D_MSAA unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i32_2dmsaa:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v0, v[1:3], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D_MSAA unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.2dmsaa.i32.i32(i32 %data, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -1609,19 +1623,19 @@ define amdgpu_ps float @atomic_add_i32_2darraymsaa(<8 x i32> inreg %rsrc, i32 %d
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i32_2darraymsaa:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i32_2darraymsaa:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v0, v[1:4], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.2darraymsaa.i32.i32(i32 %data, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i32 %v to float
@@ -1686,19 +1700,19 @@ define amdgpu_ps float @atomic_add_i32_1d_slc(<8 x i32> inreg %rsrc, i32 %data,
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i32_1d_slc:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc slc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i32_1d_slc:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v0, v1, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm glc slc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i32 @llvm.amdgcn.image.atomic.add.1d.i32.i32(i32 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 2)
   %out = bitcast i32 %v to float
@@ -1762,19 +1776,19 @@ define amdgpu_ps <2 x float> @atomic_swap_i64_1d(<8 x i32> inreg %rsrc, i64 %dat
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_swap_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_swap v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_swap_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_swap v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.swap.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -1838,19 +1852,19 @@ define amdgpu_ps <2 x float> @atomic_add_i64_1d(<8 x i32> inreg %rsrc, i64 %data
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.add.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -1914,19 +1928,19 @@ define amdgpu_ps <2 x float> @atomic_sub_i64_1d(<8 x i32> inreg %rsrc, i64 %data
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_sub_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_sub v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_sub_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_sub v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.sub.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -1990,19 +2004,19 @@ define amdgpu_ps <2 x float> @atomic_smin_i64_1d(<8 x i32> inreg %rsrc, i64 %dat
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_smin_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_smin v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_smin_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_smin v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.smin.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -2066,19 +2080,19 @@ define amdgpu_ps <2 x float> @atomic_umin_i64_1d(<8 x i32> inreg %rsrc, i64 %dat
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_umin_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_umin v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_umin_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_umin v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.umin.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -2142,19 +2156,19 @@ define amdgpu_ps <2 x float> @atomic_smax_i64_1d(<8 x i32> inreg %rsrc, i64 %dat
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_smax_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_smax v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_smax_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_smax v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.smax.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -2218,19 +2232,19 @@ define amdgpu_ps <2 x float> @atomic_umax_i64_1d(<8 x i32> inreg %rsrc, i64 %dat
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_umax_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_umax v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_umax_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_umax v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.umax.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -2294,19 +2308,19 @@ define amdgpu_ps <2 x float> @atomic_and_i64_1d(<8 x i32> inreg %rsrc, i64 %data
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_and_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_and v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_and_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_and v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.and.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -2370,19 +2384,19 @@ define amdgpu_ps <2 x float> @atomic_or_i64_1d(<8 x i32> inreg %rsrc, i64 %data,
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_or_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_or v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_or_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_or v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.or.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -2446,19 +2460,19 @@ define amdgpu_ps <2 x float> @atomic_xor_i64_1d(<8 x i32> inreg %rsrc, i64 %data
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_xor_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_xor v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_xor_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_xor v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.xor.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -2522,19 +2536,19 @@ define amdgpu_ps <2 x float> @atomic_inc_i64_1d(<8 x i32> inreg %rsrc, i64 %data
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_inc_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_inc v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_inc_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_inc v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.inc.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -2598,19 +2612,19 @@ define amdgpu_ps <2 x float> @atomic_dec_i64_1d(<8 x i32> inreg %rsrc, i64 %data
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_dec_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_dec v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_dec_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_dec v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.dec.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -2674,19 +2688,19 @@ define amdgpu_ps <2 x float> @atomic_cmpswap_i64_1d(<8 x i32> inreg %rsrc, i64 %
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_cmpswap_i64_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_cmpswap v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_cmpswap_i64_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_cmpswap v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.cmpswap.1d.i64.i32(i64 %cmp, i64 %swap, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -2758,6 +2772,20 @@ define amdgpu_ps void @atomic_cmpswap_i64_1d_no_return(<8 x i32> inreg %rsrc, i6
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_atomic_cmpswap v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: atomic_cmpswap_i64_1d_no_return:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_atomic_cmpswap v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.cmpswap.1d.i64.i32(i64 %cmp, i64 %swap, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2820,19 +2848,19 @@ define amdgpu_ps <2 x float> @atomic_add_i64_2d(<8 x i32> inreg %rsrc, i64 %data
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i64_2d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i64_2d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.add.2d.i64.i32(i64 %data, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -2896,19 +2924,19 @@ define amdgpu_ps <2 x float> @atomic_add_i64_3d(<8 x i32> inreg %rsrc, i64 %data
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i64_3d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v[0:1], v[2:4], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_3D unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i64_3d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v[0:1], v[2:4], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_3D unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.add.3d.i64.i32(i64 %data, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -2972,19 +3000,19 @@ define amdgpu_ps <2 x float> @atomic_add_i64_cube(<8 x i32> inreg %rsrc, i64 %da
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i64_cube:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v[0:1], v[2:4], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_CUBE unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i64_cube:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v[0:1], v[2:4], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_CUBE unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.add.cube.i64.i32(i64 %data, i32 %s, i32 %t, i32 %face, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -3048,19 +3076,19 @@ define amdgpu_ps <2 x float> @atomic_add_i64_1darray(<8 x i32> inreg %rsrc, i64
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i64_1darray:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D_ARRAY unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i64_1darray:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v[0:1], v[2:3], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D_ARRAY unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.add.1darray.i64.i32(i64 %data, i32 %s, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -3124,19 +3152,19 @@ define amdgpu_ps <2 x float> @atomic_add_i64_2darray(<8 x i32> inreg %rsrc, i64
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i64_2darray:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v[0:1], v[2:4], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D_ARRAY unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i64_2darray:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v[0:1], v[2:4], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D_ARRAY unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.add.2darray.i64.i32(i64 %data, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -3200,19 +3228,19 @@ define amdgpu_ps <2 x float> @atomic_add_i64_2dmsaa(<8 x i32> inreg %rsrc, i64 %
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i64_2dmsaa:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v[0:1], v[2:4], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i64_2dmsaa:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v[0:1], v[2:4], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.add.2dmsaa.i64.i32(i64 %data, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -3276,19 +3304,19 @@ define amdgpu_ps <2 x float> @atomic_add_i64_2darraymsaa(<8 x i32> inreg %rsrc,
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i64_2darraymsaa:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v[0:1], v[2:5], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm glc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i64_2darraymsaa:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v[0:1], v[2:5], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm glc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.add.2darraymsaa.i64.i32(i64 %data, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
   %out = bitcast i64 %v to <2 x float>
@@ -3352,19 +3380,19 @@ define amdgpu_ps <2 x float> @atomic_add_i64_1d_slc(<8 x i32> inreg %rsrc, i64 %
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: atomic_add_i64_1d_slc:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    s_mov_b32 s0, s2
-; GFX10-NEXT:    s_mov_b32 s1, s3
-; GFX10-NEXT:    s_mov_b32 s2, s4
-; GFX10-NEXT:    s_mov_b32 s3, s5
-; GFX10-NEXT:    s_mov_b32 s4, s6
-; GFX10-NEXT:    s_mov_b32 s5, s7
-; GFX10-NEXT:    s_mov_b32 s6, s8
-; GFX10-NEXT:    s_mov_b32 s7, s9
-; GFX10-NEXT:    image_atomic_add v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc slc
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: atomic_add_i64_1d_slc:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    s_mov_b32 s0, s2
+; GFX10PLUS-NEXT:    s_mov_b32 s1, s3
+; GFX10PLUS-NEXT:    s_mov_b32 s2, s4
+; GFX10PLUS-NEXT:    s_mov_b32 s3, s5
+; GFX10PLUS-NEXT:    s_mov_b32 s4, s6
+; GFX10PLUS-NEXT:    s_mov_b32 s5, s7
+; GFX10PLUS-NEXT:    s_mov_b32 s6, s8
+; GFX10PLUS-NEXT:    s_mov_b32 s7, s9
+; GFX10PLUS-NEXT:    image_atomic_add v[0:1], v2, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm glc slc
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %v = call i64 @llvm.amdgcn.image.atomic.add.1d.i64.i32(i64 %data, i32 %s, <8 x i32> %rsrc, i32 0, i32 2)
   %out = bitcast i64 %v to <2 x float>
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.store.2d.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.store.2d.ll
index 0fa38536eed0c0..26d7b0c4fd065b 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.store.2d.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.image.store.2d.ll
@@ -2,7 +2,7 @@
 ; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=tahiti -o - %s | FileCheck -check-prefix=GFX6 %s
 ; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=fiji -o - %s | FileCheck -check-prefix=GFX8 %s
 ; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1010 -o - %s | FileCheck -check-prefix=GFX10 %s
-; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -o - %s | FileCheck -check-prefix=GFX10 %s
+; RUN: llc -global-isel -mtriple=amdgcn-mesa-mesa3d -mcpu=gfx1100 -o - %s | FileCheck -check-prefix=GFX11 %s
 
 define amdgpu_ps void @image_store_f32(<8 x i32> inreg %rsrc, i32 %s, i32 %t, float %data) {
 ; GFX6-LABEL: image_store_f32:
@@ -43,6 +43,20 @@ define amdgpu_ps void @image_store_f32(<8 x i32> inreg %rsrc, i32 %s, i32 %t, fl
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_store v2, v[0:1], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: image_store_f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_store v2, v[0:1], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   call void @llvm.amdgcn.image.store.2d.f32.i32(float %data, i32 1, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
@@ -86,6 +100,20 @@ define amdgpu_ps void @image_store_v2f32(<8 x i32> inreg %rsrc, i32 %s, i32 %t,
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_store v[2:3], v[0:1], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: image_store_v2f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_store v[2:3], v[0:1], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   call void @llvm.amdgcn.image.store.2d.v2f32.i32(<2 x float> %in, i32 3, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
@@ -129,6 +157,20 @@ define amdgpu_ps void @image_store_v3f32(<8 x i32> inreg %rsrc, i32 %s, i32 %t,
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_store v[2:4], v[0:1], s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_2D unorm
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: image_store_v3f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_store v[2:4], v[0:1], s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   call void @llvm.amdgcn.image.store.2d.v3f32.i32(<3 x float> %in, i32 7, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
@@ -172,6 +214,20 @@ define amdgpu_ps void @image_store_v4f32(<8 x i32> inreg %rsrc, i32 %s, i32 %t,
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: image_store_v4f32:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   call void @llvm.amdgcn.image.store.2d.v4f32.i32(<4 x float> %in, i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
@@ -215,6 +271,20 @@ define amdgpu_ps void @image_store_v4f32_dmask_0001(<8 x i32> inreg %rsrc, i32 %
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: image_store_v4f32_dmask_0001:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   call void @llvm.amdgcn.image.store.2d.v4f32.i32(<4 x float> %in, i32 1, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
@@ -258,6 +328,20 @@ define amdgpu_ps void @image_store_v4f32_dmask_0010(<8 x i32> inreg %rsrc, i32 %
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x2 dim:SQ_RSRC_IMG_2D unorm
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: image_store_v4f32_dmask_0010:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x2 dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   call void @llvm.amdgcn.image.store.2d.v4f32.i32(<4 x float> %in, i32 2, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
@@ -301,6 +385,20 @@ define amdgpu_ps void @image_store_v4f32_dmask_0100(<8 x i32> inreg %rsrc, i32 %
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x4 dim:SQ_RSRC_IMG_2D unorm
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: image_store_v4f32_dmask_0100:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x4 dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   call void @llvm.amdgcn.image.store.2d.v4f32.i32(<4 x float> %in, i32 4, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
@@ -344,6 +442,20 @@ define amdgpu_ps void @image_store_v4f32_dmask_1000(<8 x i32> inreg %rsrc, i32 %
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x8 dim:SQ_RSRC_IMG_2D unorm
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: image_store_v4f32_dmask_1000:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x8 dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   call void @llvm.amdgcn.image.store.2d.v4f32.i32(<4 x float> %in, i32 8, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
@@ -387,6 +499,20 @@ define amdgpu_ps void @image_store_v4f32_dmask_0011(<8 x i32> inreg %rsrc, i32 %
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: image_store_v4f32_dmask_0011:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   call void @llvm.amdgcn.image.store.2d.v4f32.i32(<4 x float> %in, i32 3, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
@@ -430,6 +556,20 @@ define amdgpu_ps void @image_store_v4f32_dmask_0110(<8 x i32> inreg %rsrc, i32 %
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x6 dim:SQ_RSRC_IMG_2D unorm
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: image_store_v4f32_dmask_0110:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x6 dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   call void @llvm.amdgcn.image.store.2d.v4f32.i32(<4 x float> %in, i32 6, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
@@ -479,6 +619,22 @@ define amdgpu_ps void @image_store_f32_dmask_1111(<8 x i32> inreg %rsrc, i32 inr
 ; GFX10-NEXT:    s_mov_b32 s7, s9
 ; GFX10-NEXT:    image_store v0, v[1:2], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: image_store_f32_dmask_1111:
+; GFX11:       ; %bb.0:
+; GFX11-NEXT:    v_mov_b32_e32 v1, s10
+; GFX11-NEXT:    v_mov_b32_e32 v2, s11
+; GFX11-NEXT:    s_mov_b32 s0, s2
+; GFX11-NEXT:    s_mov_b32 s1, s3
+; GFX11-NEXT:    s_mov_b32 s2, s4
+; GFX11-NEXT:    s_mov_b32 s3, s5
+; GFX11-NEXT:    s_mov_b32 s4, s6
+; GFX11-NEXT:    s_mov_b32 s5, s7
+; GFX11-NEXT:    s_mov_b32 s6, s8
+; GFX11-NEXT:    s_mov_b32 s7, s9
+; GFX11-NEXT:    image_store v0, v[1:2], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
   tail call void @llvm.amdgcn.image.store.2d.f32.i32(float %in, i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
 }
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.intersect_ray.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.intersect_ray.ll
index e43dbe868a70f5..da7b9d2f8b1981 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.intersect_ray.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.intersect_ray.ll
@@ -743,6 +743,7 @@ define amdgpu_kernel void @image_bvh_intersect_ray_nsa_reassign(i32* %p_node_ptr
 ; GFX11-NEXT:    image_bvh_intersect_ray v[0:3], [v9, v10, v[0:2], v[3:5], v[6:8]], s[0:3]
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    flat_store_b128 v[0:1], v[0:3]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep_node_ptr = getelementptr inbounds i32, i32* %p_node_ptr, i32 %lid
@@ -891,6 +892,7 @@ define amdgpu_kernel void @image_bvh_intersect_ray_a16_nsa_reassign(i32* %p_node
 ; GFX11-NEXT:    image_bvh_intersect_ray v[0:3], [v6, v7, v[0:2], v[3:5]], s[0:3] a16
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    flat_store_b128 v[0:1], v[0:3]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep_node_ptr = getelementptr inbounds i32, i32* %p_node_ptr, i32 %lid
@@ -1009,6 +1011,7 @@ define amdgpu_kernel void @image_bvh64_intersect_ray_nsa_reassign(float* %p_ray,
 ; GFX11-NEXT:    image_bvh64_intersect_ray v[0:3], [v[9:10], v11, v[0:2], v[3:5], v[6:8]], s[0:3]
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    flat_store_b128 v[0:1], v[0:3]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep_ray = getelementptr inbounds float, float* %p_ray, i32 %lid
@@ -1149,6 +1152,7 @@ define amdgpu_kernel void @image_bvh64_intersect_ray_a16_nsa_reassign(float* %p_
 ; GFX11-NEXT:    image_bvh64_intersect_ray v[0:3], [v[6:7], v8, v[0:2], v[3:5]], s[0:3] a16
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    flat_store_b128 v[0:1], v[0:3]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
   %gep_ray = getelementptr inbounds float, float* %p_ray, i32 %lid
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.wmma_32.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.wmma_32.ll
index 9778a953ba0ce7..6b8c6922c15f5a 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.wmma_32.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.wmma_32.ll
@@ -17,6 +17,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16(<8 x float> %A, <8 x float> %B
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[24:25], v[16:19], off
 ; W32-NEXT:    global_store_b128 v[24:25], v[20:23], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<8 x float> %A, <8 x float> %B, <8 x float> %C)
@@ -33,6 +34,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16(<8 x i32> %A, <8 x i32> %B, <
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[24:25], v[16:19], off
 ; W32-NEXT:    global_store_b128 v[24:25], v[20:23], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <8 x float> %C)
@@ -49,6 +51,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_lo(<8 x float> %A, <8 x float>
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[24:25], v[16:19], off
 ; W32-NEXT:    global_store_b128 v[24:25], v[20:23], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x float> @llvm.amdgcn.wmma.f16.16x16x16.f16(<8 x float> %A, <8 x float> %B, <8 x float> %C, i1 0)
@@ -63,6 +66,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_hi(<8 x float> %A, <8 x float>
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[24:25], v[16:19], off
 ; W32-NEXT:    global_store_b128 v[24:25], v[20:23], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x float> @llvm.amdgcn.wmma.f16.16x16x16.f16(<8 x float> %A, <8 x float> %B, <8 x float> %C, i1 1)
@@ -79,6 +83,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_lo(<8 x i32> %A, <8 x i32> %
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[24:25], v[16:19], off
 ; W32-NEXT:    global_store_b128 v[24:25], v[20:23], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <8 x i32> %C, i1 0)
@@ -93,6 +98,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_hi(<8 x i32> %A, <8 x i32> %
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[24:25], v[16:19], off
 ; W32-NEXT:    global_store_b128 v[24:25], v[20:23], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <8 x i32> %C, i1 1)
@@ -109,6 +115,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned(<4 x i32> %A
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
@@ -123,6 +130,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed(<4 x i32> %A,
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
@@ -137,6 +145,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned(<4 x i32> %A,
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
@@ -151,6 +160,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed(<4 x i32> %A, <4
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
@@ -165,6 +175,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned_clamp(<4 x i
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
@@ -179,6 +190,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed_clamp(<4 x i32
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
@@ -193,6 +205,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned_clamp(<4 x i32
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
@@ -207,6 +220,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed_clamp(<4 x i32>
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
@@ -223,6 +237,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned(<2 x i32> %A
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
@@ -237,6 +252,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed(<2 x i32> %A,
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
@@ -251,6 +267,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned(<2 x i32> %A,
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
@@ -265,6 +282,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed(<2 x i32> %A, <2
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
@@ -280,6 +298,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned_clamp(<2 x i
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
@@ -294,6 +313,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed_clamp(<2 x i32
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
@@ -308,6 +328,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned_clamp(<2 x i32
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
@@ -322,6 +343,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed_clamp(<2 x i32>
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.wmma_64.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.wmma_64.ll
index b99c22c48b9f00..864e9bb57df99a 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.wmma_64.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/llvm.amdgcn.wmma_64.ll
@@ -15,6 +15,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16(<8 x float> %A, <8 x float> %B
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_f32_16x16x16_f16 v[16:19], v[0:7], v[8:15], v[16:19]
 ; W64-NEXT:    global_store_b128 v[20:21], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<8 x float> %A, <8 x float> %B, <4 x float> %C)
@@ -29,6 +30,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16(<8 x i32> %A, <8 x i32> %B, <
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_f32_16x16x16_bf16 v[16:19], v[0:7], v[8:15], v[16:19]
 ; W64-NEXT:    global_store_b128 v[20:21], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <4 x float> %C)
@@ -43,6 +45,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_lo(<8 x float> %A, <8 x float>
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_f16_16x16x16_f16 v[16:19], v[0:7], v[8:15], v[16:19]
 ; W64-NEXT:    global_store_b128 v[20:21], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x float> @llvm.amdgcn.wmma.f16.16x16x16.f16(<8 x float> %A, <8 x float> %B, <4 x float> %C, i1 0)
@@ -55,6 +58,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_hi(<8 x float> %A, <8 x float>
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_f16_16x16x16_f16 v[16:19], v[0:7], v[8:15], v[16:19] op_sel:[0,0,1]
 ; W64-NEXT:    global_store_b128 v[20:21], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x float> @llvm.amdgcn.wmma.f16.16x16x16.f16(<8 x float> %A, <8 x float> %B, <4 x float> %C, i1 1)
@@ -69,6 +73,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_lo(<8 x i32> %A, <8 x i32> %
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_bf16_16x16x16_bf16 v[16:19], v[0:7], v[8:15], v[16:19]
 ; W64-NEXT:    global_store_b128 v[20:21], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <4 x i32> %C, i1 0)
@@ -81,6 +86,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_hi(<8 x i32> %A, <8 x i32> %
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_bf16_16x16x16_bf16 v[16:19], v[0:7], v[8:15], v[16:19] op_sel:[0,0,1]
 ; W64-NEXT:    global_store_b128 v[20:21], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <4 x i32> %C, i1 1)
@@ -95,6 +101,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned(<4 x i32> %A
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11]
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
@@ -108,6 +115,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed(<4 x i32> %A,
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] neg_lo:[0,1,0]
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
@@ -120,6 +128,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned(<4 x i32> %A,
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] neg_lo:[1,0,0]
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
@@ -132,6 +141,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed(<4 x i32> %A, <4
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] neg_lo:[1,1,0]
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
@@ -144,6 +154,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned_clamp(<4 x i
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] clamp
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
@@ -156,6 +167,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed_clamp(<4 x i32
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] neg_lo:[0,1,0] clamp
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
@@ -168,6 +180,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned_clamp(<4 x i32
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] neg_lo:[1,0,0] clamp
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
@@ -180,6 +193,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed_clamp(<4 x i32>
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] neg_lo:[1,1,0] clamp
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
@@ -194,6 +208,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned(<2 x i32> %A
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7]
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
@@ -206,6 +221,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed(<2 x i32> %A,
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] neg_lo:[0,1,0]
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
@@ -218,6 +234,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned(<2 x i32> %A,
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] neg_lo:[1,0,0]
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
@@ -230,6 +247,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed(<2 x i32> %A, <2
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] neg_lo:[1,1,0]
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
@@ -242,6 +260,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned_clamp(<2 x i
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] clamp
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
@@ -254,6 +273,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed_clamp(<2 x i32
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] neg_lo:[0,1,0] clamp
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
@@ -266,6 +286,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned_clamp(<2 x i32
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] neg_lo:[1,0,0] clamp
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
@@ -278,6 +299,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed_clamp(<2 x i32>
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] neg_lo:[1,1,0] clamp
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
diff --git a/llvm/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll b/llvm/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll
index 1945d4f6baf4bb..5c140183008031 100644
--- a/llvm/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll
+++ b/llvm/test/CodeGen/AMDGPU/atomic_optimizations_local_pointer.ll
@@ -197,6 +197,7 @@ define amdgpu_kernel void @add_i32_constant(i32 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: add_i32_constant:
@@ -229,6 +230,7 @@ define amdgpu_kernel void @add_i32_constant(i32 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw add i32 addrspace(3)* @local_var32, i32 5 acq_rel
@@ -435,6 +437,7 @@ define amdgpu_kernel void @add_i32_uniform(i32 addrspace(1)* %out, i32 %additive
 ; GFX1164-NEXT:    v_mad_u64_u32 v[1:2], null, s6, v0, s[0:1]
 ; GFX1164-NEXT:    s_mov_b32 s6, -1
 ; GFX1164-NEXT:    buffer_store_b32 v1, off, s[4:7], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: add_i32_uniform:
@@ -470,6 +473,7 @@ define amdgpu_kernel void @add_i32_uniform(i32 addrspace(1)* %out, i32 %additive
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mad_u64_u32 v[1:2], null, s0, v0, s[2:3]
 ; GFX1132-NEXT:    buffer_store_b32 v1, off, s[4:7], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw add i32 addrspace(3)* @local_var32, i32 %additive acq_rel
@@ -772,6 +776,7 @@ define amdgpu_kernel void @add_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: add_i32_varying:
@@ -828,6 +833,7 @@ define amdgpu_kernel void @add_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
@@ -1262,6 +1268,7 @@ define amdgpu_kernel void @add_i64_constant(i64 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: add_i64_constant:
@@ -1295,6 +1302,7 @@ define amdgpu_kernel void @add_i64_constant(i64 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw add i64 addrspace(3)* @local_var64, i64 5 acq_rel
@@ -1540,6 +1548,7 @@ define amdgpu_kernel void @add_i64_uniform(i64 addrspace(1)* %out, i64 %additive
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, v3
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: add_i64_uniform:
@@ -1580,6 +1589,7 @@ define amdgpu_kernel void @add_i64_uniform(i64 addrspace(1)* %out, i64 %additive
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, v3
 ; GFX1132-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw add i64 addrspace(3)* @local_var64, i64 %additive acq_rel
@@ -1654,6 +1664,7 @@ define amdgpu_kernel void @add_i64_varying(i64 addrspace(1)* %out) {
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
@@ -1852,6 +1863,7 @@ define amdgpu_kernel void @sub_i32_constant(i32 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: sub_i32_constant:
@@ -1885,6 +1897,7 @@ define amdgpu_kernel void @sub_i32_constant(i32 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw sub i32 addrspace(3)* @local_var32, i32 5 acq_rel
@@ -2094,6 +2107,7 @@ define amdgpu_kernel void @sub_i32_uniform(i32 addrspace(1)* %out, i32 %subitive
 ; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1164-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[4:7], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: sub_i32_uniform:
@@ -2130,6 +2144,7 @@ define amdgpu_kernel void @sub_i32_uniform(i32 addrspace(1)* %out, i32 %subitive
 ; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1132-NEXT:    v_sub_nc_u32_e32 v0, s0, v0
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[4:7], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw sub i32 addrspace(3)* @local_var32, i32 %subitive acq_rel
@@ -2432,6 +2447,7 @@ define amdgpu_kernel void @sub_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: sub_i32_varying:
@@ -2488,6 +2504,7 @@ define amdgpu_kernel void @sub_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
@@ -2933,6 +2950,7 @@ define amdgpu_kernel void @sub_i64_constant(i64 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: sub_i64_constant:
@@ -2969,6 +2987,7 @@ define amdgpu_kernel void @sub_i64_constant(i64 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw sub i64 addrspace(3)* @local_var64, i64 5 acq_rel
@@ -3225,6 +3244,7 @@ define amdgpu_kernel void @sub_i64_uniform(i64 addrspace(1)* %out, i64 %subitive
 ; GFX1164-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX1164-NEXT:    v_sub_co_ci_u32_e32 v1, vcc, s4, v1, vcc
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: sub_i64_uniform:
@@ -3267,6 +3287,7 @@ define amdgpu_kernel void @sub_i64_uniform(i64 addrspace(1)* %out, i64 %subitive
 ; GFX1132-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX1132-NEXT:    v_sub_co_ci_u32_e32 v1, vcc_lo, s4, v1, vcc_lo
 ; GFX1132-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw sub i64 addrspace(3)* @local_var64, i64 %subitive acq_rel
@@ -3341,6 +3362,7 @@ define amdgpu_kernel void @sub_i64_varying(i64 addrspace(1)* %out) {
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    buffer_gl0_inv
 ; GFX11-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
@@ -3645,6 +3667,7 @@ define amdgpu_kernel void @and_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: and_i32_varying:
@@ -3701,6 +3724,7 @@ define amdgpu_kernel void @and_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
@@ -4004,6 +4028,7 @@ define amdgpu_kernel void @or_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: or_i32_varying:
@@ -4060,6 +4085,7 @@ define amdgpu_kernel void @or_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
@@ -4363,6 +4389,7 @@ define amdgpu_kernel void @xor_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: xor_i32_varying:
@@ -4419,6 +4446,7 @@ define amdgpu_kernel void @xor_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
@@ -4722,6 +4750,7 @@ define amdgpu_kernel void @max_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: max_i32_varying:
@@ -4778,6 +4807,7 @@ define amdgpu_kernel void @max_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
@@ -4993,6 +5023,7 @@ define amdgpu_kernel void @max_i64_constant(i64 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: max_i64_constant:
@@ -5027,6 +5058,7 @@ define amdgpu_kernel void @max_i64_constant(i64 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw max i64 addrspace(3)* @local_var64, i64 5 acq_rel
@@ -5329,6 +5361,7 @@ define amdgpu_kernel void @min_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: min_i32_varying:
@@ -5385,6 +5418,7 @@ define amdgpu_kernel void @min_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
@@ -5600,6 +5634,7 @@ define amdgpu_kernel void @min_i64_constant(i64 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: min_i64_constant:
@@ -5634,6 +5669,7 @@ define amdgpu_kernel void @min_i64_constant(i64 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw min i64 addrspace(3)* @local_var64, i64 5 acq_rel
@@ -5936,6 +5972,7 @@ define amdgpu_kernel void @umax_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: umax_i32_varying:
@@ -5992,6 +6029,7 @@ define amdgpu_kernel void @umax_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
@@ -6204,6 +6242,7 @@ define amdgpu_kernel void @umax_i64_constant(i64 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s2, -1
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: umax_i64_constant:
@@ -6238,6 +6277,7 @@ define amdgpu_kernel void @umax_i64_constant(i64 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s2, -1
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw umax i64 addrspace(3)* @local_var64, i64 5 acq_rel
@@ -6540,6 +6580,7 @@ define amdgpu_kernel void @umin_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: umin_i32_varying:
@@ -6596,6 +6637,7 @@ define amdgpu_kernel void @umin_i32_varying(i32 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b32 v0, off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %lane = call i32 @llvm.amdgcn.workitem.id.x()
@@ -6808,6 +6850,7 @@ define amdgpu_kernel void @umin_i64_constant(i64 addrspace(1)* %out) {
 ; GFX1164-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1164-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1164-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1164-NEXT:    s_endpgm
 ;
 ; GFX1132-LABEL: umin_i64_constant:
@@ -6842,6 +6885,7 @@ define amdgpu_kernel void @umin_i64_constant(i64 addrspace(1)* %out) {
 ; GFX1132-NEXT:    s_mov_b32 s3, 0x31016000
 ; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX1132-NEXT:    buffer_store_b64 v[0:1], off, s[0:3], 0
+; GFX1132-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX1132-NEXT:    s_endpgm
 entry:
   %old = atomicrmw umin i64 addrspace(3)* @local_var64, i64 5 acq_rel
diff --git a/llvm/test/CodeGen/AMDGPU/clamp.ll b/llvm/test/CodeGen/AMDGPU/clamp.ll
index 92b4fc24fba9e6..76534a91a8abfc 100644
--- a/llvm/test/CodeGen/AMDGPU/clamp.ll
+++ b/llvm/test/CodeGen/AMDGPU/clamp.ll
@@ -58,6 +58,7 @@ define amdgpu_kernel void @v_clamp_f32(float addrspace(1)* %out, float addrspace
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f32_e64 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -124,6 +125,7 @@ define amdgpu_kernel void @v_clamp_neg_f32(float addrspace(1)* %out, float addrs
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f32_e64 v1, -v1, -v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -191,6 +193,7 @@ define amdgpu_kernel void @v_clamp_negabs_f32(float addrspace(1)* %out, float ad
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f32_e64 v1, -|v1|, -|v1| clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -269,6 +272,7 @@ define amdgpu_kernel void @v_clamp_negzero_f32(float addrspace(1)* %out, float a
 ; GFX11-NEXT:    v_max_f32_e32 v1, 0x80000000, v1
 ; GFX11-NEXT:    v_min_f32_e32 v1, 1.0, v1
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -347,6 +351,7 @@ define amdgpu_kernel void @v_clamp_negzero_maybe_snan_f32(float addrspace(1)* %o
 ; GFX11-NEXT:    v_max_f32_e32 v1, 0x80000000, v1
 ; GFX11-NEXT:    v_min_f32_e32 v1, 1.0, v1
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -431,6 +436,7 @@ define amdgpu_kernel void @v_clamp_multi_use_max_f32(float addrspace(1)* %out, f
 ; GFX11-NEXT:    global_store_b32 v0, v2, s[0:1]
 ; GFX11-NEXT:    global_store_b32 v[0:1], v1, off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -499,6 +505,7 @@ define amdgpu_kernel void @v_clamp_f16(half addrspace(1)* %out, half addrspace(1
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f16_e64 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b16 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr half, half addrspace(1)* %aptr, i32 %tid
@@ -566,6 +573,7 @@ define amdgpu_kernel void @v_clamp_neg_f16(half addrspace(1)* %out, half addrspa
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f16_e64 v1, -v1, -v1 clamp
 ; GFX11-NEXT:    global_store_b16 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr half, half addrspace(1)* %aptr, i32 %tid
@@ -634,6 +642,7 @@ define amdgpu_kernel void @v_clamp_negabs_f16(half addrspace(1)* %out, half addr
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f16_e64 v1, -|v1|, -|v1| clamp
 ; GFX11-NEXT:    global_store_b16 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr half, half addrspace(1)* %aptr, i32 %tid
@@ -703,6 +712,7 @@ define amdgpu_kernel void @v_clamp_f64(double addrspace(1)* %out, double addrspa
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f64 v[0:1], v[0:1], v[0:1] clamp
 ; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr double, double addrspace(1)* %aptr, i32 %tid
@@ -769,6 +779,7 @@ define amdgpu_kernel void @v_clamp_neg_f64(double addrspace(1)* %out, double add
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f64 v[0:1], -v[0:1], -v[0:1] clamp
 ; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr double, double addrspace(1)* %aptr, i32 %tid
@@ -836,6 +847,7 @@ define amdgpu_kernel void @v_clamp_negabs_f64(double addrspace(1)* %out, double
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f64 v[0:1], -|v[0:1]|, -|v[0:1]| clamp
 ; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr double, double addrspace(1)* %aptr, i32 %tid
@@ -908,6 +920,7 @@ define amdgpu_kernel void @v_clamp_med3_aby_negzero_f32(float addrspace(1)* %out
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_med3_f32 v1, 0x80000000, 1.0, v1
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -972,6 +985,7 @@ define amdgpu_kernel void @v_clamp_med3_aby_f32(float addrspace(1)* %out, float
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f32_e64 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -1036,6 +1050,7 @@ define amdgpu_kernel void @v_clamp_med3_bay_f32(float addrspace(1)* %out, float
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f32_e64 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -1100,6 +1115,7 @@ define amdgpu_kernel void @v_clamp_med3_yab_f32(float addrspace(1)* %out, float
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f32_e64 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -1164,6 +1180,7 @@ define amdgpu_kernel void @v_clamp_med3_yba_f32(float addrspace(1)* %out, float
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f32_e64 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -1228,6 +1245,7 @@ define amdgpu_kernel void @v_clamp_med3_ayb_f32(float addrspace(1)* %out, float
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f32_e64 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -1292,6 +1310,7 @@ define amdgpu_kernel void @v_clamp_med3_bya_f32(float addrspace(1)* %out, float
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f32_e64 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -1343,6 +1362,7 @@ define amdgpu_kernel void @v_clamp_constants_to_one_f32(float addrspace(1)* %out
 ; GFX11-NEXT:    v_mov_b32_e32 v1, 1.0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %out.gep = getelementptr float, float addrspace(1)* %out, i32 %tid
@@ -1391,6 +1411,7 @@ define amdgpu_kernel void @v_clamp_constants_to_zero_f32(float addrspace(1)* %ou
 ; GFX11-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %out.gep = getelementptr float, float addrspace(1)* %out, i32 %tid
@@ -1440,6 +1461,7 @@ define amdgpu_kernel void @v_clamp_constant_preserve_f32(float addrspace(1)* %ou
 ; GFX11-NEXT:    v_mov_b32_e32 v1, 0.5
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %out.gep = getelementptr float, float addrspace(1)* %out, i32 %tid
@@ -1489,6 +1511,7 @@ define amdgpu_kernel void @v_clamp_constant_preserve_denorm_f32(float addrspace(
 ; GFX11-NEXT:    v_mov_b32_e32 v1, 0x7fffff
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %out.gep = getelementptr float, float addrspace(1)* %out, i32 %tid
@@ -1537,6 +1560,7 @@ define amdgpu_kernel void @v_clamp_constant_qnan_f32(float addrspace(1)* %out) #
 ; GFX11-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %out.gep = getelementptr float, float addrspace(1)* %out, i32 %tid
@@ -1585,6 +1609,7 @@ define amdgpu_kernel void @v_clamp_constant_snan_f32(float addrspace(1)* %out) #
 ; GFX11-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %out.gep = getelementptr float, float addrspace(1)* %out, i32 %tid
@@ -1656,6 +1681,7 @@ define amdgpu_kernel void @v_clamp_f32_no_dx10_clamp(float addrspace(1)* %out, f
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_med3_f32 v1, v1, 0, 1.0
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -1723,6 +1749,7 @@ define amdgpu_kernel void @v_clamp_f32_snan_dx10clamp(float addrspace(1)* %out,
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_add_f32_e64 v1, v1, 0.5 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -1795,6 +1822,7 @@ define amdgpu_kernel void @v_clamp_f32_snan_no_dx10clamp(float addrspace(1)* %ou
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_med3_f32 v1, v1, 0, 1.0
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -1866,6 +1894,7 @@ define amdgpu_kernel void @v_clamp_f32_snan_no_dx10clamp_nnan_src(float addrspac
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_med3_f32 v1, v1, 0, 1.0
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -1933,6 +1962,7 @@ define amdgpu_kernel void @v_clamp_med3_aby_f32_no_dx10_clamp(float addrspace(1)
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f32_e64 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -1997,6 +2027,7 @@ define amdgpu_kernel void @v_clamp_med3_bay_f32_no_dx10_clamp(float addrspace(1)
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_max_f32_e64 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -2061,6 +2092,7 @@ define amdgpu_kernel void @v_clamp_med3_yab_f32_no_dx10_clamp(float addrspace(1)
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_med3_f32 v1, v1, 0, 1.0
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -2125,6 +2157,7 @@ define amdgpu_kernel void @v_clamp_med3_yba_f32_no_dx10_clamp(float addrspace(1)
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_med3_f32 v1, v1, 1.0, 0
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -2189,6 +2222,7 @@ define amdgpu_kernel void @v_clamp_med3_ayb_f32_no_dx10_clamp(float addrspace(1)
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_med3_f32 v1, 0, v1, 1.0
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -2253,6 +2287,7 @@ define amdgpu_kernel void @v_clamp_med3_bya_f32_no_dx10_clamp(float addrspace(1)
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_med3_f32 v1, 1.0, v1, 0
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 %tid
@@ -2304,6 +2339,7 @@ define amdgpu_kernel void @v_clamp_constant_qnan_f32_no_dx10_clamp(float addrspa
 ; GFX11-NEXT:    v_mov_b32_e32 v1, 0x7fc00000
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %out.gep = getelementptr float, float addrspace(1)* %out, i32 %tid
@@ -2353,6 +2389,7 @@ define amdgpu_kernel void @v_clamp_constant_snan_f32_no_dx10_clamp(float addrspa
 ; GFX11-NEXT:    v_mov_b32_e32 v1, 0x7f800001
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %out.gep = getelementptr float, float addrspace(1)* %out, i32 %tid
@@ -2423,6 +2460,7 @@ define amdgpu_kernel void @v_clamp_v2f16(<2 x half> addrspace(1)* %out, <2 x hal
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr <2 x half>, <2 x half> addrspace(1)* %aptr, i32 %tid
@@ -2508,6 +2546,7 @@ define amdgpu_kernel void @v_clamp_v2f16_undef_elt(<2 x half> addrspace(1)* %out
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr <2 x half>, <2 x half> addrspace(1)* %aptr, i32 %tid
@@ -2592,6 +2631,7 @@ define amdgpu_kernel void @v_clamp_v2f16_not_zero(<2 x half> addrspace(1)* %out,
 ; GFX11-NEXT:    v_pk_max_f16 v1, v1, 2.0
 ; GFX11-NEXT:    v_pk_min_f16 v1, v1, 1.0 op_sel_hi:[1,0]
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr <2 x half>, <2 x half> addrspace(1)* %aptr, i32 %tid
@@ -2675,6 +2715,7 @@ define amdgpu_kernel void @v_clamp_v2f16_not_one(<2 x half> addrspace(1)* %out,
 ; GFX11-NEXT:    v_pk_max_f16 v1, v1, 0
 ; GFX11-NEXT:    v_pk_min_f16 v1, v1, 1.0 op_sel:[0,1] op_sel_hi:[1,0]
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr <2 x half>, <2 x half> addrspace(1)* %aptr, i32 %tid
@@ -2750,6 +2791,7 @@ define amdgpu_kernel void @v_clamp_neg_v2f16(<2 x half> addrspace(1)* %out, <2 x
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 neg_lo:[1,1] neg_hi:[1,1] clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr <2 x half>, <2 x half> addrspace(1)* %aptr, i32 %tid
@@ -2829,6 +2871,7 @@ define amdgpu_kernel void @v_clamp_negabs_v2f16(<2 x half> addrspace(1)* %out, <
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 neg_lo:[1,1] neg_hi:[1,1] clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr <2 x half>, <2 x half> addrspace(1)* %aptr, i32 %tid
@@ -2907,6 +2950,7 @@ define amdgpu_kernel void @v_clamp_neglo_v2f16(<2 x half> addrspace(1)* %out, <2
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 neg_lo:[1,1] clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr <2 x half>, <2 x half> addrspace(1)* %aptr, i32 %tid
@@ -2984,6 +3028,7 @@ define amdgpu_kernel void @v_clamp_neghi_v2f16(<2 x half> addrspace(1)* %out, <2
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 neg_hi:[1,1] clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr <2 x half>, <2 x half> addrspace(1)* %aptr, i32 %tid
@@ -3061,6 +3106,7 @@ define amdgpu_kernel void @v_clamp_v2f16_shuffle(<2 x half> addrspace(1)* %out,
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 op_sel:[1,1] op_sel_hi:[0,0] clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr <2 x half>, <2 x half> addrspace(1)* %aptr, i32 %tid
@@ -3146,6 +3192,7 @@ define amdgpu_kernel void @v_clamp_v2f16_undef_limit_elts0(<2 x half> addrspace(
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr <2 x half>, <2 x half> addrspace(1)* %aptr, i32 %tid
@@ -3231,6 +3278,7 @@ define amdgpu_kernel void @v_clamp_v2f16_undef_limit_elts1(<2 x half> addrspace(
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    v_pk_max_f16 v1, v1, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep0 = getelementptr <2 x half>, <2 x half> addrspace(1)* %aptr, i32 %tid
@@ -3310,6 +3358,7 @@ define amdgpu_kernel void @v_clamp_diff_source_f32(float addrspace(1)* %out, flo
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_max_f32_e64 v0, v0, v1 clamp
 ; GFX11-NEXT:    global_store_b32 v2, v0, s[0:1] offset:12
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 {
   %gep0 = getelementptr float, float addrspace(1)* %aptr, i32 0
diff --git a/llvm/test/CodeGen/AMDGPU/cluster_stores.ll b/llvm/test/CodeGen/AMDGPU/cluster_stores.ll
index 70262d66a19514..544d77f61218c4 100644
--- a/llvm/test/CodeGen/AMDGPU/cluster_stores.ll
+++ b/llvm/test/CodeGen/AMDGPU/cluster_stores.ll
@@ -123,6 +123,7 @@ define amdgpu_kernel void @cluster_load_cluster_store(i32* noalias %lb, i32* noa
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v4 offset:16
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(3)
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v5 offset:24
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 bb:
   %la0 = getelementptr inbounds i32, i32* %lb, i32 0
@@ -264,6 +265,7 @@ define amdgpu_kernel void @cluster_load_valu_cluster_store(i32* noalias %lb, i32
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v4 offset:16
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(3)
 ; GFX11-NEXT:    flat_store_b32 v[0:1], v5 offset:24
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 bb:
   %la0 = getelementptr inbounds i32, i32* %lb, i32 0
@@ -344,6 +346,7 @@ define amdgpu_ps void @cluster_image_load(<8 x i32> inreg %src, <8 x i32> inreg
 ; GFX11-NEXT:    v_add_f32_e32 v3, v3, v7
 ; GFX11-NEXT:    v_add_f32_e32 v2, v2, v6
 ; GFX11-NEXT:    image_store v[2:5], v[0:1], s[8:15] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 entry:
   %x1 = add i32 %x, 1
@@ -400,6 +403,7 @@ define amdgpu_ps void @no_cluster_image_load(<8 x i32> inreg %src1, <8 x i32> in
 ; GFX11-NEXT:    v_add_f32_e32 v3, v3, v7
 ; GFX11-NEXT:    v_add_f32_e32 v2, v2, v6
 ; GFX11-NEXT:    image_store v[2:5], v[0:1], s[16:23] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 entry:
   %val1 = call <4 x float> @llvm.amdgcn.image.load.mip.2d.v4f32.i32(i32 15, i32 %x, i32 %y, i32 0, <8 x i32> %src1, i32 0, i32 0)
@@ -497,6 +501,7 @@ define amdgpu_ps void @cluster_image_sample(<8 x i32> inreg %src, <4 x i32> inre
 ; GFX11-NEXT:    v_add_f32_e32 v3, v3, v7
 ; GFX11-NEXT:    v_add_f32_e32 v2, v2, v6
 ; GFX11-NEXT:    image_store v[2:5], v[0:1], s[12:19] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 entry:
   %s = sitofp i32 %x to float
diff --git a/llvm/test/CodeGen/AMDGPU/flat-scratch-svs.ll b/llvm/test/CodeGen/AMDGPU/flat-scratch-svs.ll
index 3a058774c15a0e..103bd82157527a 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-scratch-svs.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-scratch-svs.ll
@@ -58,6 +58,7 @@ define amdgpu_kernel void @soff1_voff1(i32 %soff) {
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-SDAG-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: soff1_voff1:
@@ -76,6 +77,7 @@ define amdgpu_kernel void @soff1_voff1(i32 %soff) {
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-GISEL-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 bb:
   %soff1 = mul i32 %soff, 1
@@ -145,6 +147,7 @@ define amdgpu_kernel void @soff1_voff2(i32 %soff) {
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-SDAG-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: soff1_voff2:
@@ -164,6 +167,7 @@ define amdgpu_kernel void @soff1_voff2(i32 %soff) {
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-GISEL-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 bb:
   %soff1 = mul i32 %soff, 1
@@ -233,6 +237,7 @@ define amdgpu_kernel void @soff1_voff4(i32 %soff) {
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-SDAG-NEXT:    scratch_store_b8 v0, v3, s0 offset:4 dlc
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: soff1_voff4:
@@ -252,6 +257,7 @@ define amdgpu_kernel void @soff1_voff4(i32 %soff) {
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-GISEL-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 bb:
   %soff1 = mul i32 %soff, 1
@@ -322,6 +328,7 @@ define amdgpu_kernel void @soff2_voff1(i32 %soff) {
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-SDAG-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: soff2_voff1:
@@ -341,6 +348,7 @@ define amdgpu_kernel void @soff2_voff1(i32 %soff) {
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-GISEL-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 bb:
   %soff2 = mul i32 %soff, 2
@@ -414,6 +422,7 @@ define amdgpu_kernel void @soff2_voff2(i32 %soff) {
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-SDAG-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: soff2_voff2:
@@ -434,6 +443,7 @@ define amdgpu_kernel void @soff2_voff2(i32 %soff) {
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-GISEL-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 bb:
   %soff2 = mul i32 %soff, 2
@@ -507,6 +517,7 @@ define amdgpu_kernel void @soff2_voff4(i32 %soff) {
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-SDAG-NEXT:    scratch_store_b8 v0, v3, s0 offset:4 dlc
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: soff2_voff4:
@@ -527,6 +538,7 @@ define amdgpu_kernel void @soff2_voff4(i32 %soff) {
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-GISEL-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 bb:
   %soff2 = mul i32 %soff, 2
@@ -598,6 +610,7 @@ define amdgpu_kernel void @soff4_voff1(i32 %soff) {
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-SDAG-NEXT:    scratch_store_b8 v0, v4, s0 offset:4 dlc
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: soff4_voff1:
@@ -617,6 +630,7 @@ define amdgpu_kernel void @soff4_voff1(i32 %soff) {
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-GISEL-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 bb:
   %soff4 = mul i32 %soff, 4
@@ -691,6 +705,7 @@ define amdgpu_kernel void @soff4_voff2(i32 %soff) {
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-SDAG-NEXT:    scratch_store_b8 v0, v4, s0 offset:4 dlc
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: soff4_voff2:
@@ -711,6 +726,7 @@ define amdgpu_kernel void @soff4_voff2(i32 %soff) {
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-GISEL-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 bb:
   %soff4 = mul i32 %soff, 4
@@ -784,6 +800,7 @@ define amdgpu_kernel void @soff4_voff4(i32 %soff) {
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-SDAG-NEXT:    scratch_store_b8 v0, v3, s0 offset:4 dlc
 ; GFX11-SDAG-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: soff4_voff4:
@@ -804,6 +821,7 @@ define amdgpu_kernel void @soff4_voff4(i32 %soff) {
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-GISEL-NEXT:    scratch_store_b8 v0, v3, off offset:4 dlc
 ; GFX11-GISEL-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
 bb:
   %soff4 = mul i32 %soff, 4
diff --git a/llvm/test/CodeGen/AMDGPU/flat-scratch.ll b/llvm/test/CodeGen/AMDGPU/flat-scratch.ll
index a0641558c193c3..015731a15183b1 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-scratch.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-scratch.ll
@@ -67,6 +67,7 @@ define amdgpu_kernel void @zero_init_kernel() {
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], off offset:48
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], off offset:32
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], off offset:16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
 ; GFX9-PAL-LABEL: zero_init_kernel:
@@ -183,6 +184,7 @@ define amdgpu_kernel void @zero_init_kernel() {
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:48
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:32
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:16
+; GFX11-PAL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-PAL-NEXT:    s_endpgm
   %alloca = alloca [32 x i16], align 2, addrspace(5)
   %cast = bitcast [32 x i16] addrspace(5)* %alloca to i8 addrspace(5)*
@@ -1042,6 +1044,7 @@ define amdgpu_kernel void @zero_init_small_offset_kernel() {
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], off offset:288
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], off offset:304
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], off offset:320
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
 ; GFX9-PAL-LABEL: zero_init_small_offset_kernel:
@@ -1170,6 +1173,7 @@ define amdgpu_kernel void @zero_init_small_offset_kernel() {
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:288
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:304
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], off offset:320
+; GFX11-PAL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-PAL-NEXT:    s_endpgm
   %padding = alloca [64 x i32], align 4, addrspace(5)
   %alloca = alloca [32 x i16], align 2, addrspace(5)
@@ -2102,6 +2106,7 @@ define amdgpu_kernel void @zero_init_large_offset_kernel() {
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], vcc_lo offset:32
 ; GFX11-NEXT:    s_movk_i32 vcc_lo, 0x4010
 ; GFX11-NEXT:    scratch_store_b128 off, v[0:3], vcc_lo offset:48
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
 ; GFX9-PAL-LABEL: zero_init_large_offset_kernel:
@@ -2240,6 +2245,7 @@ define amdgpu_kernel void @zero_init_large_offset_kernel() {
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], vcc_lo offset:32
 ; GFX11-PAL-NEXT:    s_movk_i32 vcc_lo, 0x4010
 ; GFX11-PAL-NEXT:    scratch_store_b128 off, v[0:3], vcc_lo offset:48
+; GFX11-PAL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-PAL-NEXT:    s_endpgm
   %padding = alloca [4096 x i32], align 4, addrspace(5)
   %alloca = alloca [32 x i16], align 2, addrspace(5)
diff --git a/llvm/test/CodeGen/AMDGPU/image-load-d16-tfe.ll b/llvm/test/CodeGen/AMDGPU/image-load-d16-tfe.ll
index 45c446ccf48f0e..60cff45c3e72c6 100644
--- a/llvm/test/CodeGen/AMDGPU/image-load-d16-tfe.ll
+++ b/llvm/test/CodeGen/AMDGPU/image-load-d16-tfe.ll
@@ -63,6 +63,7 @@ define amdgpu_ps void @load_1d_f16_tfe_dmask0(<8 x i32> inreg %rsrc, i32 %s) {
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    global_store_b32 v[0:1], v2, off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
 ; GFX8-UNPACKED-LABEL: load_1d_f16_tfe_dmask0:
@@ -151,6 +152,7 @@ define amdgpu_ps void @load_1d_f16_tfe_dmask1(<8 x i32> inreg %rsrc, i32 %s) {
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    global_store_b32 v[0:1], v2, off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
 ; GFX8-UNPACKED-LABEL: load_1d_f16_tfe_dmask1:
@@ -239,6 +241,7 @@ define amdgpu_ps void @load_1d_v2f16_tfe_dmask0(<8 x i32> inreg %rsrc, i32 %s) {
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    global_store_b32 v[0:1], v2, off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
 ; GFX8-UNPACKED-LABEL: load_1d_v2f16_tfe_dmask0:
@@ -327,6 +330,7 @@ define amdgpu_ps void @load_1d_v2f16_tfe_dmask1(<8 x i32> inreg %rsrc, i32 %s) {
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    global_store_b32 v[0:1], v2, off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
 ; GFX8-UNPACKED-LABEL: load_1d_v2f16_tfe_dmask1:
@@ -415,6 +419,7 @@ define amdgpu_ps void @load_1d_v2f16_tfe_dmask3(<8 x i32> inreg %rsrc, i32 %s) {
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    global_store_b32 v[0:1], v2, off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
 ; GFX8-UNPACKED-LABEL: load_1d_v2f16_tfe_dmask3:
@@ -515,6 +520,7 @@ define amdgpu_ps void @load_1d_v3f16_tfe_dmask7(<8 x i32> inreg %rsrc, i32 %s) {
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    global_store_b32 v[0:1], v3, off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
 ; GFX8-UNPACKED-LABEL: load_1d_v3f16_tfe_dmask7:
@@ -612,6 +618,7 @@ define amdgpu_ps void @load_1d_v4f16_tfe_dmask15(<8 x i32> inreg %rsrc, i32 %s)
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-NEXT:    global_store_b32 v[0:1], v3, off dlc
 ; GFX11-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 ;
 ; GFX8-UNPACKED-LABEL: load_1d_v4f16_tfe_dmask15:
diff --git a/llvm/test/CodeGen/AMDGPU/llc-pipeline.ll b/llvm/test/CodeGen/AMDGPU/llc-pipeline.ll
index 01135bdf806088..800e906f1dee5c 100644
--- a/llvm/test/CodeGen/AMDGPU/llc-pipeline.ll
+++ b/llvm/test/CodeGen/AMDGPU/llc-pipeline.ll
@@ -968,6 +968,7 @@
 ; GCN-O2-NEXT:        SI Final Branch Preparation
 ; GCN-O2-NEXT:        SI peephole optimizations
 ; GCN-O2-NEXT:        Post RA hazard recognizer
+; GCN-O2-NEXT:        Release VGPRs
 ; GCN-O2-NEXT:        AMDGPU Insert Delay ALU
 ; GCN-O2-NEXT:        Branch relaxation pass
 ; GCN-O2-NEXT:        Register Usage Information Collector Pass
@@ -1271,6 +1272,7 @@
 ; GCN-O3-NEXT:        SI Final Branch Preparation
 ; GCN-O3-NEXT:        SI peephole optimizations
 ; GCN-O3-NEXT:        Post RA hazard recognizer
+; GCN-O3-NEXT:        Release VGPRs
 ; GCN-O3-NEXT:        AMDGPU Insert Delay ALU
 ; GCN-O3-NEXT:        Branch relaxation pass
 ; GCN-O3-NEXT:        Register Usage Information Collector Pass
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.add.gs.reg.rtn.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.add.gs.reg.rtn.ll
index 67bfe6cc0d750e..76751028868eb5 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.add.gs.reg.rtn.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.add.gs.reg.rtn.ll
@@ -31,6 +31,7 @@ define amdgpu_gs void @test_add_32_use(i32 %arg, i32 addrspace(1)* %out) {
 ; CHECK-NEXT:    buffer_gl0_inv
 ; CHECK-NEXT:    buffer_gl1_inv
 ; CHECK-NEXT:    global_store_b32 v[1:2], v3, off
+; CHECK-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; CHECK-NEXT:    s_endpgm
   %res = call i32 @llvm.amdgcn.ds.add.gs.reg.rtn.i32(i32 %arg, i32 16)
   store i32 %res, i32 addrspace(1)* %out, align 4
@@ -63,6 +64,7 @@ define amdgpu_gs void @test_add_64_use(i32 %arg, i64 addrspace(1)* %out) {
 ; CHECK-NEXT:    buffer_gl0_inv
 ; CHECK-NEXT:    buffer_gl1_inv
 ; CHECK-NEXT:    global_store_b64 v[1:2], v[3:4], off
+; CHECK-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; CHECK-NEXT:    s_endpgm
   %res = call i64 @llvm.amdgcn.ds.add.gs.reg.rtn.i64(i32 %arg, i32 32)
   store i64 %res, i64 addrspace(1)* %out, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.sub.gs.reg.rtn.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.sub.gs.reg.rtn.ll
index e5fcc7f09f8dd1..8d99b08a86d6a0 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.sub.gs.reg.rtn.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.ds.sub.gs.reg.rtn.ll
@@ -31,6 +31,7 @@ define amdgpu_gs void @test_sub_32_use(i32 %arg, i32 addrspace(1)* %out) {
 ; CHECK-NEXT:    buffer_gl0_inv
 ; CHECK-NEXT:    buffer_gl1_inv
 ; CHECK-NEXT:    global_store_b32 v[1:2], v3, off
+; CHECK-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; CHECK-NEXT:    s_endpgm
   %res = call i32 @llvm.amdgcn.ds.sub.gs.reg.rtn.i32(i32 %arg, i32 16)
   store i32 %res, i32 addrspace(1)* %out, align 4
@@ -63,6 +64,7 @@ define amdgpu_gs void @test_sub_64_use(i32 %arg, i64 addrspace(1)* %out) {
 ; CHECK-NEXT:    buffer_gl0_inv
 ; CHECK-NEXT:    buffer_gl1_inv
 ; CHECK-NEXT:    global_store_b64 v[1:2], v[3:4], off
+; CHECK-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; CHECK-NEXT:    s_endpgm
   %res = call i64 @llvm.amdgcn.ds.sub.gs.reg.rtn.i64(i32 %arg, i32 32)
   store i64 %res, i64 addrspace(1)* %out, align 4
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.bf16.bf16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.bf16.bf16.ll
index 65925e0112abf9..e48fc945d049fe 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.bf16.bf16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.bf16.bf16.ll
@@ -16,6 +16,7 @@ define amdgpu_kernel void @test_llvm_amdgcn_fdot2_f16_f16(
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    v_dot2_bf16_bf16 v1, s2, s3, v1
 ; GFX11-NEXT:    global_store_b16 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
     i16 addrspace(1)* %r,
     <2 x i16> addrspace(1)* %a,
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.f16.f16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.f16.f16.ll
index fe3de5a99bdf6c..15387bf7b6aaa2 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.f16.f16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.f16.f16.ll
@@ -16,6 +16,7 @@ define amdgpu_kernel void @test_llvm_amdgcn_fdot2_f16_f16(
 ; GFX11-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-NEXT:    v_dot2_f16_f16 v1, s2, s3, v1
 ; GFX11-NEXT:    global_store_b16 v0, v1, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
     half addrspace(1)* %r,
     <2 x half> addrspace(1)* %a,
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.f32.bf16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.f32.bf16.ll
index 843deff83ab9aa..ebe207f5de192d 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.f32.bf16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.fdot2.f32.bf16.ll
@@ -18,6 +18,7 @@ define amdgpu_kernel void @test_llvm_amdgcn_fdot2_f32_bf16_clamp(
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_dot2_f32_bf16 v0, s2, s3, v0 clamp
 ; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
     float addrspace(1)* %r,
     <2 x i16> addrspace(1)* %a,
@@ -47,6 +48,7 @@ define amdgpu_kernel void @test_llvm_amdgcn_fdot2_f32_bf16_no_clamp(
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_dot2_f32_bf16 v0, s2, s3, v0
 ; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
     float addrspace(1)* %r,
     <2 x i16> addrspace(1)* %a,
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.a16.dim.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.a16.dim.ll
index 8ad50aab16de43..96509e725951d2 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.a16.dim.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.a16.dim.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10 %s
+; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10PLUS,GFX10 %s
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10PLUS,GFX11 %s
 
 define amdgpu_ps <4 x float> @load_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
 ; GFX9-LABEL: load_1d:
@@ -10,11 +10,11 @@ define amdgpu_ps <4 x float> @load_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords)
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -28,11 +28,11 @@ define amdgpu_ps <4 x float> @load_2d(<8 x i32> inreg %rsrc, <2 x i16> %coords)
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_2d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_2d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   %t = extractelement <2 x i16> %coords, i32 1
@@ -47,11 +47,11 @@ define amdgpu_ps <4 x float> @load_3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_l
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_3d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_3d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -67,11 +67,11 @@ define amdgpu_ps <4 x float> @load_cube(<8 x i32> inreg %rsrc, <2 x i16> %coords
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_cube:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_cube:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -87,11 +87,11 @@ define amdgpu_ps <4 x float> @load_1darray(<8 x i32> inreg %rsrc, <2 x i16> %coo
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_1darray:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_1darray:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   %slice = extractelement <2 x i16> %coords, i32 1
@@ -106,11 +106,11 @@ define amdgpu_ps <4 x float> @load_2darray(<8 x i32> inreg %rsrc, <2 x i16> %coo
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_2darray:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_2darray:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -126,11 +126,11 @@ define amdgpu_ps <4 x float> @load_2dmsaa(<8 x i32> inreg %rsrc, <2 x i16> %coor
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_2dmsaa:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_2dmsaa:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -146,11 +146,11 @@ define amdgpu_ps <4 x float> @load_2darraymsaa(<8 x i32> inreg %rsrc, <2 x i16>
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_2darraymsaa:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_2darraymsaa:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -167,11 +167,11 @@ define amdgpu_ps <4 x float> @load_mip_1d(<8 x i32> inreg %rsrc, <2 x i16> %coor
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_mip_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load_mip v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_mip_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load_mip v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   %mip = extractelement <2 x i16> %coords, i32 1
@@ -186,11 +186,11 @@ define amdgpu_ps <4 x float> @load_mip_2d(<8 x i32> inreg %rsrc, <2 x i16> %coor
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_mip_2d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_mip_2d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -206,11 +206,11 @@ define amdgpu_ps <4 x float> @load_mip_3d(<8 x i32> inreg %rsrc, <2 x i16> %coor
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_mip_3d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_mip_3d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -227,11 +227,11 @@ define amdgpu_ps <4 x float> @load_mip_cube(<8 x i32> inreg %rsrc, <2 x i16> %co
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_mip_cube:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_mip_cube:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -248,11 +248,11 @@ define amdgpu_ps <4 x float> @load_mip_1darray(<8 x i32> inreg %rsrc, <2 x i16>
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_mip_1darray:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_mip_1darray:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %slice = extractelement <2 x i16> %coords_lo, i32 1
@@ -268,11 +268,11 @@ define amdgpu_ps <4 x float> @load_mip_2darray(<8 x i32> inreg %rsrc, <2 x i16>
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_mip_2darray:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_mip_2darray:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load_mip v[0:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -292,6 +292,12 @@ define amdgpu_ps void @store_1d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -308,6 +314,12 @@ define amdgpu_ps void @store_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_2d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   %t = extractelement <2 x i16> %coords, i32 1
@@ -325,6 +337,12 @@ define amdgpu_ps void @store_3d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_3d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -343,6 +361,12 @@ define amdgpu_ps void @store_cube(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_cube:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -361,6 +385,12 @@ define amdgpu_ps void @store_1darray(<8 x i32> inreg %rsrc, <4 x float> %vdata,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1darray:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   %slice = extractelement <2 x i16> %coords, i32 1
@@ -378,6 +408,12 @@ define amdgpu_ps void @store_2darray(<8 x i32> inreg %rsrc, <4 x float> %vdata,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_2darray:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -396,6 +432,12 @@ define amdgpu_ps void @store_2dmsaa(<8 x i32> inreg %rsrc, <4 x float> %vdata, <
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_2dmsaa:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -414,6 +456,12 @@ define amdgpu_ps void @store_2darraymsaa(<8 x i32> inreg %rsrc, <4 x float> %vda
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_2darraymsaa:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -433,6 +481,12 @@ define amdgpu_ps void @store_mip_1d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store_mip v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_mip_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store_mip v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   %mip = extractelement <2 x i16> %coords, i32 1
@@ -450,6 +504,12 @@ define amdgpu_ps void @store_mip_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_mip_2d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -468,6 +528,12 @@ define amdgpu_ps void @store_mip_3d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_mip_3d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -487,6 +553,12 @@ define amdgpu_ps void @store_mip_cube(<8 x i32> inreg %rsrc, <4 x float> %vdata,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_mip_cube:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -506,6 +578,12 @@ define amdgpu_ps void @store_mip_1darray(<8 x i32> inreg %rsrc, <4 x float> %vda
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_mip_1darray:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %slice = extractelement <2 x i16> %coords_lo, i32 1
@@ -524,6 +602,12 @@ define amdgpu_ps void @store_mip_2darray(<8 x i32> inreg %rsrc, <4 x float> %vda
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_mip_2darray:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
   %t = extractelement <2 x i16> %coords_lo, i32 1
@@ -540,11 +624,11 @@ define amdgpu_ps <4 x float> @getresinfo_1d(<8 x i32> inreg %rsrc, <2 x i16> %co
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: getresinfo_1d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: getresinfo_1d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %mip = extractelement <2 x i16> %coords, i32 0
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.1d.v4f32.i16(i32 15, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
@@ -558,11 +642,11 @@ define amdgpu_ps <4 x float> @getresinfo_2d(<8 x i32> inreg %rsrc, <2 x i16> %co
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: getresinfo_2d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: getresinfo_2d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %mip = extractelement <2 x i16> %coords, i32 0
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.2d.v4f32.i16(i32 15, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
@@ -576,11 +660,11 @@ define amdgpu_ps <4 x float> @getresinfo_3d(<8 x i32> inreg %rsrc, <2 x i16> %co
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: getresinfo_3d:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: getresinfo_3d:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %mip = extractelement <2 x i16> %coords, i32 0
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.3d.v4f32.i16(i32 15, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
@@ -594,11 +678,11 @@ define amdgpu_ps <4 x float> @getresinfo_cube(<8 x i32> inreg %rsrc, <2 x i16> %
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: getresinfo_cube:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: getresinfo_cube:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %mip = extractelement <2 x i16> %coords, i32 0
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.cube.v4f32.i16(i32 15, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
@@ -612,11 +696,11 @@ define amdgpu_ps <4 x float> @getresinfo_1darray(<8 x i32> inreg %rsrc, <2 x i16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: getresinfo_1darray:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: getresinfo_1darray:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %mip = extractelement <2 x i16> %coords, i32 0
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.1darray.v4f32.i16(i32 15, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
@@ -630,11 +714,11 @@ define amdgpu_ps <4 x float> @getresinfo_2darray(<8 x i32> inreg %rsrc, <2 x i16
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: getresinfo_2darray:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: getresinfo_2darray:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %mip = extractelement <2 x i16> %coords, i32 0
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.2darray.v4f32.i16(i32 15, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
@@ -648,11 +732,11 @@ define amdgpu_ps <4 x float> @getresinfo_2dmsaa(<8 x i32> inreg %rsrc, <2 x i16>
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: getresinfo_2dmsaa:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: getresinfo_2dmsaa:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %mip = extractelement <2 x i16> %coords, i32 0
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.2dmsaa.v4f32.i16(i32 15, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
@@ -666,11 +750,11 @@ define amdgpu_ps <4 x float> @getresinfo_2darraymsaa(<8 x i32> inreg %rsrc, <2 x
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: getresinfo_2darraymsaa:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: getresinfo_2darraymsaa:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_get_resinfo v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %mip = extractelement <2 x i16> %coords, i32 0
   %v = call <4 x float> @llvm.amdgcn.image.getresinfo.2darraymsaa.v4f32.i16(i32 15, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
@@ -684,11 +768,11 @@ define amdgpu_ps float @load_1d_V1(<8 x i32> inreg %rsrc, <2 x i16> %coords) {
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_1d_V1:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v0, v0, s[0:7] dmask:0x8 dim:SQ_RSRC_IMG_1D unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_1d_V1:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v0, v0, s[0:7] dmask:0x8 dim:SQ_RSRC_IMG_1D unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   %v = call float @llvm.amdgcn.image.load.1d.f32.i16(i32 8, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -702,11 +786,11 @@ define amdgpu_ps <2 x float> @load_1d_V2(<8 x i32> inreg %rsrc, <2 x i16> %coord
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_1d_V2:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v[0:1], v0, s[0:7] dmask:0x9 dim:SQ_RSRC_IMG_1D unorm a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_1d_V2:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v[0:1], v0, s[0:7] dmask:0x9 dim:SQ_RSRC_IMG_1D unorm a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   %v = call <2 x float> @llvm.amdgcn.image.load.1d.v2f32.i16(i32 9, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -723,6 +807,12 @@ define amdgpu_ps void @store_1d_V1(<8 x i32> inreg %rsrc, float %vdata, <2 x i16
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v0, v1, s[0:7] dmask:0x2 dim:SQ_RSRC_IMG_1D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1d_V1:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v0, v1, s[0:7] dmask:0x2 dim:SQ_RSRC_IMG_1D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   call void @llvm.amdgcn.image.store.1d.f32.i16(float %vdata, i32 2, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -739,6 +829,12 @@ define amdgpu_ps void @store_1d_V2(<8 x i32> inreg %rsrc, <2 x float> %vdata, <2
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[0:1], v2, s[0:7] dmask:0xc dim:SQ_RSRC_IMG_1D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1d_V2:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:1], v2, s[0:7] dmask:0xc dim:SQ_RSRC_IMG_1D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   call void @llvm.amdgcn.image.store.1d.v2f32.i16(<2 x float> %vdata, i32 12, i16 %s, <8 x i32> %rsrc, i32 0, i32 0)
@@ -752,11 +848,11 @@ define amdgpu_ps <4 x float> @load_1d_glc(<8 x i32> inreg %rsrc, <2 x i16> %coor
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_1d_glc:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_1d_glc:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 1)
@@ -770,11 +866,11 @@ define amdgpu_ps <4 x float> @load_1d_slc(<8 x i32> inreg %rsrc, <2 x i16> %coor
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_1d_slc:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm slc a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_1d_slc:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm slc a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 2)
@@ -788,11 +884,11 @@ define amdgpu_ps <4 x float> @load_1d_glc_slc(<8 x i32> inreg %rsrc, <2 x i16> %
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: load_1d_glc_slc:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc slc a16
-; GFX10-NEXT:    s_waitcnt vmcnt(0)
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: load_1d_glc_slc:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    image_load v[0:3], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc slc a16
+; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   %v = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i16(i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 3)
@@ -809,6 +905,12 @@ define amdgpu_ps void @store_1d_glc(<8 x i32> inreg %rsrc, <4 x float> %vdata, <
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1d_glc:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 1)
@@ -825,6 +927,12 @@ define amdgpu_ps void @store_1d_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, <
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm slc a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1d_slc:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm slc a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 2)
@@ -841,6 +949,12 @@ define amdgpu_ps void @store_1d_glc_slc(<8 x i32> inreg %rsrc, <4 x float> %vdat
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc slc a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1d_glc_slc:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc slc a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
   call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %vdata, i32 15, i16 %s, <8 x i32> %rsrc, i32 0, i32 3)
@@ -852,9 +966,9 @@ define amdgpu_ps <4 x float> @getresinfo_dmask0(<8 x i32> inreg %rsrc, <4 x floa
 ; GFX9:       ; %bb.0: ; %main_body
 ; GFX9-NEXT:    ; return to shader part epilog
 ;
-; GFX10-LABEL: getresinfo_dmask0:
-; GFX10:       ; %bb.0: ; %main_body
-; GFX10-NEXT:    ; return to shader part epilog
+; GFX10PLUS-LABEL: getresinfo_dmask0:
+; GFX10PLUS:       ; %bb.0: ; %main_body
+; GFX10PLUS-NEXT:    ; return to shader part epilog
 main_body:
   %mip = extractelement <2 x i16> %coords, i32 0
   %r = call <4 x float> @llvm.amdgcn.image.getresinfo.1d.v4f32.i16(i32 0, i16 %mip, <8 x i32> %rsrc, i32 0, i32 0)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.a16.encode.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.a16.encode.ll
index 1085d430321c7b..a4884731f269f1 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.a16.encode.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.a16.encode.ll
@@ -380,6 +380,7 @@ define amdgpu_ps void @store_1d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x
 ; GFX11-LABEL: store_1d:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16 ; encoding: [0x80,0x0f,0x19,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
@@ -401,6 +402,7 @@ define amdgpu_ps void @store_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x
 ; GFX11-LABEL: store_2d:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16 ; encoding: [0x84,0x0f,0x19,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
@@ -423,6 +425,7 @@ define amdgpu_ps void @store_3d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2 x
 ; GFX11-LABEL: store_3d:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16 ; encoding: [0x88,0x0f,0x19,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
@@ -446,6 +449,7 @@ define amdgpu_ps void @store_cube(<8 x i32> inreg %rsrc, <4 x float> %vdata, <2
 ; GFX11-LABEL: store_cube:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm a16 ; encoding: [0x8c,0x0f,0x19,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
@@ -469,6 +473,7 @@ define amdgpu_ps void @store_1darray(<8 x i32> inreg %rsrc, <4 x float> %vdata,
 ; GFX11-LABEL: store_1darray:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm a16 ; encoding: [0x90,0x0f,0x19,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
@@ -491,6 +496,7 @@ define amdgpu_ps void @store_2darray(<8 x i32> inreg %rsrc, <4 x float> %vdata,
 ; GFX11-LABEL: store_2darray:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm a16 ; encoding: [0x94,0x0f,0x19,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
@@ -514,6 +520,7 @@ define amdgpu_ps void @store_2dmsaa(<8 x i32> inreg %rsrc, <4 x float> %vdata, <
 ; GFX11-LABEL: store_2dmsaa:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm a16 ; encoding: [0x98,0x0f,0x19,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
@@ -537,6 +544,7 @@ define amdgpu_ps void @store_2darraymsaa(<8 x i32> inreg %rsrc, <4 x float> %vda
 ; GFX11-LABEL: store_2darraymsaa:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm a16 ; encoding: [0x9c,0x0f,0x19,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
@@ -561,6 +569,7 @@ define amdgpu_ps void @store_mip_1d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <
 ; GFX11-LABEL: store_mip_1d:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store_mip v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16 ; encoding: [0x80,0x0f,0x1d,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
@@ -583,6 +592,7 @@ define amdgpu_ps void @store_mip_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <
 ; GFX11-LABEL: store_mip_2d:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16 ; encoding: [0x84,0x0f,0x1d,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
@@ -606,6 +616,7 @@ define amdgpu_ps void @store_mip_3d(<8 x i32> inreg %rsrc, <4 x float> %vdata, <
 ; GFX11-LABEL: store_mip_3d:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16 ; encoding: [0x88,0x0f,0x1d,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
@@ -630,6 +641,7 @@ define amdgpu_ps void @store_mip_cube(<8 x i32> inreg %rsrc, <4 x float> %vdata,
 ; GFX11-LABEL: store_mip_cube:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm a16 ; encoding: [0x8c,0x0f,0x1d,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
@@ -654,6 +666,7 @@ define amdgpu_ps void @store_mip_1darray(<8 x i32> inreg %rsrc, <4 x float> %vda
 ; GFX11-LABEL: store_mip_1darray:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm a16 ; encoding: [0x90,0x0f,0x1d,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
@@ -677,6 +690,7 @@ define amdgpu_ps void @store_mip_2darray(<8 x i32> inreg %rsrc, <4 x float> %vda
 ; GFX11-LABEL: store_mip_2darray:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm a16 ; encoding: [0x94,0x0f,0x1d,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords_lo, i32 0
@@ -941,6 +955,7 @@ define amdgpu_ps void @store_1d_V1(<8 x i32> inreg %rsrc, float %vdata, <2 x i16
 ; GFX11-LABEL: store_1d_V1:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v0, v1, s[0:7] dmask:0x2 dim:SQ_RSRC_IMG_1D unorm a16 ; encoding: [0x80,0x02,0x19,0xf0,0x01,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
@@ -962,6 +977,7 @@ define amdgpu_ps void @store_1d_V2(<8 x i32> inreg %rsrc, <2 x float> %vdata, <2
 ; GFX11-LABEL: store_1d_V2:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v[0:1], v2, s[0:7] dmask:0xc dim:SQ_RSRC_IMG_1D unorm a16 ; encoding: [0x80,0x0c,0x19,0xf0,0x02,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
@@ -1055,6 +1071,7 @@ define amdgpu_ps void @store_1d_glc(<8 x i32> inreg %rsrc, <4 x float> %vdata, <
 ; GFX11-LABEL: store_1d_glc:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc a16 ; encoding: [0x80,0x4f,0x19,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
@@ -1076,6 +1093,7 @@ define amdgpu_ps void @store_1d_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, <
 ; GFX11-LABEL: store_1d_slc:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm slc a16 ; encoding: [0x80,0x1f,0x19,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
@@ -1097,6 +1115,7 @@ define amdgpu_ps void @store_1d_glc_slc(<8 x i32> inreg %rsrc, <4 x float> %vdat
 ; GFX11-LABEL: store_1d_glc_slc:
 ; GFX11:       ; %bb.0: ; %main_body
 ; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc slc a16 ; encoding: [0x80,0x5f,0x19,0xf0,0x04,0x00,0x00,0x00]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS) ; encoding: [0x03,0x00,0xb6,0xbf]
 ; GFX11-NEXT:    s_endpgm ; encoding: [0x00,0x00,0xb0,0xbf]
 main_body:
   %s = extractelement <2 x i16> %coords, i32 0
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.dim.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.dim.ll
index 1548e32a265d4c..614e5d769c7e30 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.dim.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.dim.ll
@@ -2437,10 +2437,16 @@ define amdgpu_ps void @store_1d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %
 ; NOPRT-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf unorm
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_1d:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_1d:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2467,10 +2473,16 @@ define amdgpu_ps void @store_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %
 ; NOPRT-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf unorm
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_2d:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_2d:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_2d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.2d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2497,10 +2509,16 @@ define amdgpu_ps void @store_3d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32 %
 ; NOPRT-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf unorm
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_3d:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_3d:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_3d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.3d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %r, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2527,10 +2545,16 @@ define amdgpu_ps void @store_cube(<8 x i32> inreg %rsrc, <4 x float> %vdata, i32
 ; NOPRT-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf unorm da
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_cube:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_cube:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_cube:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.cube.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2557,10 +2581,16 @@ define amdgpu_ps void @store_1darray(<8 x i32> inreg %rsrc, <4 x float> %vdata,
 ; NOPRT-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf unorm da
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_1darray:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_1darray:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1darray:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.1darray.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2587,10 +2617,16 @@ define amdgpu_ps void @store_2darray(<8 x i32> inreg %rsrc, <4 x float> %vdata,
 ; NOPRT-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf unorm da
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_2darray:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_2darray:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_2darray:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.2darray.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2617,10 +2653,16 @@ define amdgpu_ps void @store_2dmsaa(<8 x i32> inreg %rsrc, <4 x float> %vdata, i
 ; NOPRT-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf unorm
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_2dmsaa:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_2dmsaa:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_2dmsaa:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.2dmsaa.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2647,10 +2689,16 @@ define amdgpu_ps void @store_2darraymsaa(<8 x i32> inreg %rsrc, <4 x float> %vda
 ; NOPRT-NEXT:    image_store v[0:3], v[4:7], s[0:7] dmask:0xf unorm da
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_2darraymsaa:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_2darraymsaa:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_2darraymsaa:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_MSAA_ARRAY unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.2darraymsaa.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, i32 %fragid, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2677,10 +2725,16 @@ define amdgpu_ps void @store_mip_1d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i
 ; NOPRT-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf unorm
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_mip_1d:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_mip_1d:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_mip_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store_mip v[0:3], v[4:5], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.mip.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2707,10 +2761,16 @@ define amdgpu_ps void @store_mip_2d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i
 ; NOPRT-NEXT:    image_store_mip v[0:3], v[4:6], s[0:7] dmask:0xf unorm
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_mip_2d:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store_mip v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_mip_2d:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store_mip v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_mip_2d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store_mip v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.mip.2d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2737,10 +2797,16 @@ define amdgpu_ps void @store_mip_3d(<8 x i32> inreg %rsrc, <4 x float> %vdata, i
 ; NOPRT-NEXT:    image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_mip_3d:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_mip_3d:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_mip_3d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.mip.3d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %r, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2767,10 +2833,16 @@ define amdgpu_ps void @store_mip_cube(<8 x i32> inreg %rsrc, <4 x float> %vdata,
 ; NOPRT-NEXT:    image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm da
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_mip_cube:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_mip_cube:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_mip_cube:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_CUBE unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.mip.cube.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2797,10 +2869,16 @@ define amdgpu_ps void @store_mip_1darray(<8 x i32> inreg %rsrc, <4 x float> %vda
 ; NOPRT-NEXT:    image_store_mip v[0:3], v[4:6], s[0:7] dmask:0xf unorm da
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_mip_1darray:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store_mip v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_mip_1darray:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store_mip v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_mip_1darray:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store_mip v[0:3], v[4:6], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D_ARRAY unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.mip.1darray.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -2827,10 +2905,16 @@ define amdgpu_ps void @store_mip_2darray(<8 x i32> inreg %rsrc, <4 x float> %vda
 ; NOPRT-NEXT:    image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf unorm da
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_mip_2darray:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_mip_2darray:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_mip_2darray:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store_mip v[0:3], v[4:7], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D_ARRAY unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.mip.2darray.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, i32 %t, i32 %slice, i32 %mip, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -3207,10 +3291,16 @@ define amdgpu_ps void @store_1d_V1(<8 x i32> inreg %rsrc, float %vdata, i32 %s)
 ; NOPRT-NEXT:    image_store v0, v1, s[0:7] dmask:0x2 unorm
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_1d_V1:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v0, v1, s[0:7] dmask:0x2 dim:SQ_RSRC_IMG_1D unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_1d_V1:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v0, v1, s[0:7] dmask:0x2 dim:SQ_RSRC_IMG_1D unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1d_V1:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v0, v1, s[0:7] dmask:0x2 dim:SQ_RSRC_IMG_1D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.1d.f32.i32(float %vdata, i32 2, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -3237,10 +3327,16 @@ define amdgpu_ps void @store_1d_V2(<8 x i32> inreg %rsrc, <2 x float> %vdata, i3
 ; NOPRT-NEXT:    image_store v[0:1], v2, s[0:7] dmask:0xc unorm
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_1d_V2:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:1], v2, s[0:7] dmask:0xc dim:SQ_RSRC_IMG_1D unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_1d_V2:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:1], v2, s[0:7] dmask:0xc dim:SQ_RSRC_IMG_1D unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1d_V2:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:1], v2, s[0:7] dmask:0xc dim:SQ_RSRC_IMG_1D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.1d.v2f32.i32(<2 x float> %vdata, i32 12, i32 %s, <8 x i32> %rsrc, i32 0, i32 0)
   ret void
@@ -3372,10 +3468,16 @@ define amdgpu_ps void @store_1d_glc(<8 x i32> inreg %rsrc, <4 x float> %vdata, i
 ; NOPRT-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf unorm glc
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_1d_glc:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_1d_glc:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1d_glc:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 1)
   ret void
@@ -3402,10 +3504,16 @@ define amdgpu_ps void @store_1d_slc(<8 x i32> inreg %rsrc, <4 x float> %vdata, i
 ; NOPRT-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf unorm slc
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_1d_slc:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm slc
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_1d_slc:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm slc
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1d_slc:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm slc
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 2)
   ret void
@@ -3432,10 +3540,16 @@ define amdgpu_ps void @store_1d_glc_slc(<8 x i32> inreg %rsrc, <4 x float> %vdat
 ; NOPRT-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf unorm glc slc
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: store_1d_glc_slc:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc slc
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: store_1d_glc_slc:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc slc
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_1d_glc_slc:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm glc slc
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %vdata, i32 15, i32 %s, <8 x i32> %rsrc, i32 0, i32 3)
   ret void
@@ -3690,13 +3804,22 @@ define amdgpu_ps void @image_store_wait(<8 x i32> inreg %arg, <8 x i32> inreg %a
 ; NOPRT-NEXT:    image_store v[0:3], v4, s[16:23] dmask:0xf unorm
 ; NOPRT-NEXT:    s_endpgm
 ;
-; GFX10PLUS-LABEL: image_store_wait:
-; GFX10PLUS:       ; %bb.0: ; %main_body
-; GFX10PLUS-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
-; GFX10PLUS-NEXT:    image_load v[0:3], v4, s[8:15] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
-; GFX10PLUS-NEXT:    s_waitcnt vmcnt(0)
-; GFX10PLUS-NEXT:    image_store v[0:3], v4, s[16:23] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
-; GFX10PLUS-NEXT:    s_endpgm
+; GFX10-LABEL: image_store_wait:
+; GFX10:       ; %bb.0: ; %main_body
+; GFX10-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
+; GFX10-NEXT:    image_load v[0:3], v4, s[8:15] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
+; GFX10-NEXT:    s_waitcnt vmcnt(0)
+; GFX10-NEXT:    image_store v[0:3], v4, s[16:23] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
+; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: image_store_wait:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[0:3], v4, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
+; GFX11-NEXT:    image_load v[0:3], v4, s[8:15] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
+; GFX11-NEXT:    s_waitcnt vmcnt(0)
+; GFX11-NEXT:    image_store v[0:3], v4, s[16:23] dmask:0xf dim:SQ_RSRC_IMG_1D unorm
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   call void @llvm.amdgcn.image.store.1d.v4f32.i32(<4 x float> %arg3, i32 15, i32 %arg4, <8 x i32> %arg, i32 0, i32 0)
   %data = call <4 x float> @llvm.amdgcn.image.load.1d.v4f32.i32(i32 15, i32 %arg4, <8 x i32> %arg1, i32 0, i32 0)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.store.a16.d16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.store.a16.d16.ll
index 637ae0b3c26a6d..5a9ec1d4598ac6 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.store.a16.d16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.store.a16.d16.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX9 %s
 ; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10 %s
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11 %s
 
 define amdgpu_ps void @store_f16_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords, <2 x i32> %val) {
 ; GFX9-LABEL: store_f16_1d:
@@ -13,6 +13,12 @@ define amdgpu_ps void @store_f16_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords, <2
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm a16 d16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_f16_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm a16 d16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   %bitcast = bitcast <2 x i32> %val to <4 x half>
@@ -30,6 +36,12 @@ define amdgpu_ps void @store_v2f16_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm a16 d16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v2f16_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm a16 d16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   %bitcast = bitcast <2 x i32> %val to <4 x half>
@@ -47,6 +59,12 @@ define amdgpu_ps void @store_v3f16_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_1D unorm a16 d16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v3f16_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_1D unorm a16 d16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   %bitcast = bitcast <2 x i32> %val to <4 x half>
@@ -64,6 +82,12 @@ define amdgpu_ps void @store_v4f16_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16 d16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v4f16_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16 d16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   %bitcast = bitcast <2 x i32> %val to <4 x half>
@@ -81,6 +105,12 @@ define amdgpu_ps void @store_f16_2d(<8 x i32> inreg %rsrc, <2 x i16> %coords, <2
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm a16 d16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_f16_2d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm a16 d16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   %y = extractelement <2 x i16> %coords, i32 1
@@ -99,6 +129,12 @@ define amdgpu_ps void @store_v2f16_2d(<8 x i32> inreg %rsrc, <2 x i16> %coords,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm a16 d16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v2f16_2d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm a16 d16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   %y = extractelement <2 x i16> %coords, i32 1
@@ -117,6 +153,12 @@ define amdgpu_ps void @store_v3f16_2d(<8 x i32> inreg %rsrc, <2 x i16> %coords,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_2D unorm a16 d16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v3f16_2d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_2D unorm a16 d16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   %y = extractelement <2 x i16> %coords, i32 1
@@ -135,6 +177,12 @@ define amdgpu_ps void @store_v4f16_2d(<8 x i32> inreg %rsrc, <2 x i16> %coords,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16 d16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v4f16_2d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:2], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16 d16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   %y = extractelement <2 x i16> %coords, i32 1
@@ -153,6 +201,12 @@ define amdgpu_ps void @store_f16_3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[2:3], v[0:1], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_3D unorm a16 d16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_f16_3d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[2:3], v[0:1], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_3D unorm a16 d16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords_lo, i32 0
   %y = extractelement <2 x i16> %coords_lo, i32 1
@@ -172,6 +226,12 @@ define amdgpu_ps void @store_v2f16_3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_l
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[2:3], v[0:1], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_3D unorm a16 d16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v2f16_3d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[2:3], v[0:1], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_3D unorm a16 d16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords_lo, i32 0
   %y = extractelement <2 x i16> %coords_lo, i32 1
@@ -191,6 +251,12 @@ define amdgpu_ps void @store_v3f16_3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_l
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[2:3], v[0:1], s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_3D unorm a16 d16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v3f16_3d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[2:3], v[0:1], s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_3D unorm a16 d16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords_lo, i32 0
   %y = extractelement <2 x i16> %coords_lo, i32 1
@@ -210,6 +276,12 @@ define amdgpu_ps void @store_v4f16_3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_l
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[2:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16 d16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v4f16_3d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[2:3], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16 d16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords_lo, i32 0
   %y = extractelement <2 x i16> %coords_lo, i32 1
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.store.a16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.store.a16.ll
index 086c3ef96c8885..24540039563196 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.store.a16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.store.a16.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -march=amdgcn -mcpu=gfx900 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX9 %s
 ; RUN: llc -march=amdgcn -mcpu=gfx1010 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX10 %s
+; RUN: llc -march=amdgcn -mcpu=gfx1100 -verify-machineinstrs < %s | FileCheck -check-prefixes=GFX11 %s
 
 define amdgpu_ps void @store_f32_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords, <4 x float> %val) {
 ; GFX9-LABEL: store_f32_1d:
@@ -13,6 +13,12 @@ define amdgpu_ps void @store_f32_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords, <4
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_f32_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_1D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %val, i32 1, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
@@ -29,6 +35,12 @@ define amdgpu_ps void @store_v2f32_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v2f32_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_1D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %val, i32 3, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
@@ -45,6 +57,12 @@ define amdgpu_ps void @store_v3f32_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_1D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v3f32_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_1D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %val, i32 7, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
@@ -61,6 +79,12 @@ define amdgpu_ps void @store_v4f32_1d(<8 x i32> inreg %rsrc, <2 x i16> %coords,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v4f32_1d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_1D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   call void @llvm.amdgcn.image.store.1d.v4f32.i16(<4 x float> %val, i32 15, i16 %x, <8 x i32> %rsrc, i32 0, i32 0)
@@ -77,6 +101,12 @@ define amdgpu_ps void @store_f32_2d(<8 x i32> inreg %rsrc, <2 x i16> %coords, <4
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_f32_2d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_2D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   %y = extractelement <2 x i16> %coords, i32 1
@@ -94,6 +124,12 @@ define amdgpu_ps void @store_v2f32_2d(<8 x i32> inreg %rsrc, <2 x i16> %coords,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v2f32_2d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_2D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   %y = extractelement <2 x i16> %coords, i32 1
@@ -111,6 +147,12 @@ define amdgpu_ps void @store_v3f32_2d(<8 x i32> inreg %rsrc, <2 x i16> %coords,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_2D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v3f32_2d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_2D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   %y = extractelement <2 x i16> %coords, i32 1
@@ -128,6 +170,12 @@ define amdgpu_ps void @store_v4f32_2d(<8 x i32> inreg %rsrc, <2 x i16> %coords,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v4f32_2d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[1:4], v0, s[0:7] dmask:0xf dim:SQ_RSRC_IMG_2D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords, i32 0
   %y = extractelement <2 x i16> %coords, i32 1
@@ -145,6 +193,12 @@ define amdgpu_ps void @store_f32_3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_lo,
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_3D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_f32_3d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x1 dim:SQ_RSRC_IMG_3D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords_lo, i32 0
   %y = extractelement <2 x i16> %coords_lo, i32 1
@@ -163,6 +217,12 @@ define amdgpu_ps void @store_v2f32_3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_l
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_3D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v2f32_3d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x3 dim:SQ_RSRC_IMG_3D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords_lo, i32 0
   %y = extractelement <2 x i16> %coords_lo, i32 1
@@ -181,6 +241,12 @@ define amdgpu_ps void @store_v3f32_3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_l
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_3D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v3f32_3d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0x7 dim:SQ_RSRC_IMG_3D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords_lo, i32 0
   %y = extractelement <2 x i16> %coords_lo, i32 1
@@ -199,6 +265,12 @@ define amdgpu_ps void @store_v4f32_3d(<8 x i32> inreg %rsrc, <2 x i16> %coords_l
 ; GFX10:       ; %bb.0: ; %main_body
 ; GFX10-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
 ; GFX10-NEXT:    s_endpgm
+;
+; GFX11-LABEL: store_v4f32_3d:
+; GFX11:       ; %bb.0: ; %main_body
+; GFX11-NEXT:    image_store v[2:5], v[0:1], s[0:7] dmask:0xf dim:SQ_RSRC_IMG_3D unorm a16
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 main_body:
   %x = extractelement <2 x i16> %coords_lo, i32 0
   %y = extractelement <2 x i16> %coords_lo, i32 1
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll
index a3786758d21562..542708dee1da69 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.intersect_ray.ll
@@ -247,6 +247,7 @@ define amdgpu_kernel void @image_bvh_intersect_ray_nsa_reassign(i32* %p_node_ptr
 ; GFX11-NEXT:    image_bvh_intersect_ray v[0:3], [v9, v10, v[6:8], v[3:5], v[0:2]], s[0:3]
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    flat_store_b128 v[0:1], v[0:3]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 main_body:
   %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
@@ -340,6 +341,7 @@ define amdgpu_kernel void @image_bvh_intersect_ray_a16_nsa_reassign(i32* %p_node
 ; GFX11-NEXT:    image_bvh_intersect_ray v[0:3], [v6, v7, v[3:5], v[0:2]], s[0:3] a16
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    flat_store_b128 v[0:1], v[0:3]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 main_body:
   %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
@@ -439,6 +441,7 @@ define amdgpu_kernel void @image_bvh64_intersect_ray_nsa_reassign(float* %p_ray,
 ; GFX11-NEXT:    image_bvh64_intersect_ray v[0:3], [v[9:10], v11, v[6:8], v[3:5], v[0:2]], s[0:3]
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    flat_store_b128 v[0:1], v[0:3]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 main_body:
   %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
@@ -527,6 +530,7 @@ define amdgpu_kernel void @image_bvh64_intersect_ray_a16_nsa_reassign(float* %p_
 ; GFX11-NEXT:    image_bvh64_intersect_ray v[0:3], [v[6:7], v8, v[3:5], v[0:2]], s[0:3] a16
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    flat_store_b128 v[0:1], v[0:3]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 main_body:
   %lid = tail call i32 @llvm.amdgcn.workitem.id.x()
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll
index fca7fdae148f5f..6fe3d6e937b56f 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.permlane64.ll
@@ -17,6 +17,7 @@ define amdgpu_kernel void @test_s(i32 addrspace(1)* %out, i32 %src0) {
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX11-NEXT:    v_permlane64_b32 v0, v0
 ; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %v = call i32 @llvm.amdgcn.permlane64(i32 %src0)
   store i32 %v, i32 addrspace(1)* %out
@@ -33,6 +34,7 @@ define amdgpu_kernel void @test_i(i32 addrspace(1)* %out) {
 ; GFX11-NEXT:    v_permlane64_b32 v0, v0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %v = call i32 @llvm.amdgcn.permlane64(i32 99)
   store i32 %v, i32 addrspace(1)* %out
@@ -47,6 +49,7 @@ define amdgpu_kernel void @test_v(i32 addrspace(1)* %out, i32 %src0) #1 {
 ; GFX11-SDAG-NEXT:    v_permlane64_b32 v0, v0
 ; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-SDAG-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: test_v:
@@ -56,6 +59,7 @@ define amdgpu_kernel void @test_v(i32 addrspace(1)* %out, i32 %src0) #1 {
 ; GFX11-GISEL-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
   %tidx = call i32 @llvm.amdgcn.workitem.id.x()
   %v = call i32 @llvm.amdgcn.permlane64(i32 %tidx)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sendmsg.rtn.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sendmsg.rtn.ll
index 2d3742699ab35d..a77a18034e4046 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sendmsg.rtn.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.sendmsg.rtn.ll
@@ -11,6 +11,7 @@ define amdgpu_kernel void @test_get_doorbell(i32 addrspace(1)* %out) {
 ; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: test_get_doorbell:
@@ -21,6 +22,7 @@ define amdgpu_kernel void @test_get_doorbell(i32 addrspace(1)* %out) {
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
   %ret = call i32 @llvm.amdgcn.s.sendmsg.rtn.i32(i32 128)
   store i32 %ret, i32 addrspace(1)* %out
@@ -36,6 +38,7 @@ define amdgpu_kernel void @test_get_ddid(i32 addrspace(1)* %out) {
 ; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: test_get_ddid:
@@ -46,6 +49,7 @@ define amdgpu_kernel void @test_get_ddid(i32 addrspace(1)* %out) {
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
   %ret = call i32 @llvm.amdgcn.s.sendmsg.rtn.i32(i32 129)
   store i32 %ret, i32 addrspace(1)* %out
@@ -62,6 +66,7 @@ define amdgpu_kernel void @test_get_tma(i64 addrspace(1)* %out) {
 ; GFX11-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %ret = call i64 @llvm.amdgcn.s.sendmsg.rtn.i64(i32 130)
   store i64 %ret, i64 addrspace(1)* %out
@@ -78,6 +83,7 @@ define amdgpu_kernel void @test_get_realtime(i64 addrspace(1)* %out) {
 ; GFX11-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %ret = call i64 @llvm.amdgcn.s.sendmsg.rtn.i64(i32 131)
   store i64 %ret, i64 addrspace(1)* %out
@@ -93,6 +99,7 @@ define amdgpu_kernel void @test_savewave(i32 addrspace(1)* %out) {
 ; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: test_savewave:
@@ -103,6 +110,7 @@ define amdgpu_kernel void @test_savewave(i32 addrspace(1)* %out) {
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
   %ret = call i32 @llvm.amdgcn.s.sendmsg.rtn.i32(i32 132)
   store i32 %ret, i32 addrspace(1)* %out
@@ -119,6 +127,7 @@ define amdgpu_kernel void @test_get_tba(i64 addrspace(1)* %out) {
 ; GFX11-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %ret = call i64 @llvm.amdgcn.s.sendmsg.rtn.i64(i32 133)
   store i64 %ret, i64 addrspace(1)* %out
@@ -134,6 +143,7 @@ define amdgpu_kernel void @test_get_0_i32(i32 addrspace(1)* %out) {
 ; GFX11-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-SDAG-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-SDAG-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-SDAG-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-SDAG-NEXT:    s_endpgm
 ;
 ; GFX11-GISEL-LABEL: test_get_0_i32:
@@ -144,6 +154,7 @@ define amdgpu_kernel void @test_get_0_i32(i32 addrspace(1)* %out) {
 ; GFX11-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-GISEL-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-GISEL-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-GISEL-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-GISEL-NEXT:    s_endpgm
   %ret = call i32 @llvm.amdgcn.s.sendmsg.rtn.i32(i32 0)
   store i32 %ret, i32 addrspace(1)* %out
@@ -160,6 +171,7 @@ define amdgpu_kernel void @test_get_99999_i64(i64 addrspace(1)* %out) {
 ; GFX11-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %ret = call i64 @llvm.amdgcn.s.sendmsg.rtn.i64(i32 99999)
   store i64 %ret, i64 addrspace(1)* %out
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma_32.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma_32.ll
index 7f52d3f2690750..22614565cd311c 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma_32.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma_32.ll
@@ -17,6 +17,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16(<8 x float> %A, <8 x float> %B
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[24:25], v[20:23], off offset:16
 ; W32-NEXT:    global_store_b128 v[24:25], v[16:19], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<8 x float> %A, <8 x float> %B, <8 x float> %C)
@@ -33,6 +34,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16(<8 x i32> %A, <8 x i32> %B, <
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[24:25], v[20:23], off offset:16
 ; W32-NEXT:    global_store_b128 v[24:25], v[16:19], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <8 x float> %C)
@@ -49,6 +51,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_lo(<8 x float> %A, <8 x float>
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[24:25], v[20:23], off offset:16
 ; W32-NEXT:    global_store_b128 v[24:25], v[16:19], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x float> @llvm.amdgcn.wmma.f16.16x16x16.f16(<8 x float> %A, <8 x float> %B, <8 x float> %C, i1 0)
@@ -63,6 +66,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_hi(<8 x float> %A, <8 x float>
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[24:25], v[20:23], off offset:16
 ; W32-NEXT:    global_store_b128 v[24:25], v[16:19], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x float> @llvm.amdgcn.wmma.f16.16x16x16.f16(<8 x float> %A, <8 x float> %B, <8 x float> %C, i1 1)
@@ -79,6 +83,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_lo(<8 x i32> %A, <8 x i32> %
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[24:25], v[20:23], off offset:16
 ; W32-NEXT:    global_store_b128 v[24:25], v[16:19], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <8 x i32> %C, i1 0)
@@ -93,6 +98,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_hi(<8 x i32> %A, <8 x i32> %
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[24:25], v[20:23], off offset:16
 ; W32-NEXT:    global_store_b128 v[24:25], v[16:19], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <8 x i32> %C, i1 1)
@@ -109,6 +115,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned(<4 x i32> %A
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
@@ -123,6 +130,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed(<4 x i32> %A,
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
@@ -137,6 +145,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned(<4 x i32> %A,
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
@@ -151,6 +160,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed(<4 x i32> %A, <4
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
@@ -165,6 +175,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned_clamp(<4 x i
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
@@ -179,6 +190,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed_clamp(<4 x i32
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
@@ -193,6 +205,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned_clamp(<4 x i32
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
@@ -207,6 +220,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed_clamp(<4 x i32>
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[16:17], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[16:17], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
@@ -223,6 +237,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned(<2 x i32> %A
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
@@ -237,6 +252,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed(<2 x i32> %A,
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
@@ -251,6 +267,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned(<2 x i32> %A,
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
@@ -265,6 +282,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed(<2 x i32> %A, <2
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
@@ -280,6 +298,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned_clamp(<2 x i
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
@@ -294,6 +313,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed_clamp(<2 x i32
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
@@ -308,6 +328,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned_clamp(<2 x i32
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
@@ -322,6 +343,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed_clamp(<2 x i32>
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[12:13], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[12:13], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma_64.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma_64.ll
index 1b1e5d064ac2b9..a2240f24306a8b 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma_64.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.wmma_64.ll
@@ -15,6 +15,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16(<8 x float> %A, <8 x float> %B
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_f32_16x16x16_f16 v[16:19], v[0:7], v[8:15], v[16:19]
 ; W64-NEXT:    global_store_b128 v[20:21], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<8 x float> %A, <8 x float> %B, <4 x float> %C)
@@ -29,6 +30,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16(<8 x i32> %A, <8 x i32> %B, <
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_f32_16x16x16_bf16 v[16:19], v[0:7], v[8:15], v[16:19]
 ; W64-NEXT:    global_store_b128 v[20:21], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <4 x float> %C)
@@ -43,6 +45,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_lo(<8 x float> %A, <8 x float>
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_f16_16x16x16_f16 v[16:19], v[0:7], v[8:15], v[16:19]
 ; W64-NEXT:    global_store_b128 v[20:21], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x float> @llvm.amdgcn.wmma.f16.16x16x16.f16(<8 x float> %A, <8 x float> %B, <4 x float> %C, i1 0)
@@ -55,6 +58,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_hi(<8 x float> %A, <8 x float>
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_f16_16x16x16_f16 v[16:19], v[0:7], v[8:15], v[16:19] op_sel:[0,0,1]
 ; W64-NEXT:    global_store_b128 v[20:21], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x float> @llvm.amdgcn.wmma.f16.16x16x16.f16(<8 x float> %A, <8 x float> %B, <4 x float> %C, i1 1)
@@ -69,6 +73,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_lo(<8 x i32> %A, <8 x i32> %
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_bf16_16x16x16_bf16 v[16:19], v[0:7], v[8:15], v[16:19]
 ; W64-NEXT:    global_store_b128 v[20:21], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <4 x i32> %C, i1 0)
@@ -81,6 +86,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_hi(<8 x i32> %A, <8 x i32> %
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_bf16_16x16x16_bf16 v[16:19], v[0:7], v[8:15], v[16:19] op_sel:[0,0,1]
 ; W64-NEXT:    global_store_b128 v[20:21], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <4 x i32> %C, i1 1)
@@ -95,6 +101,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned(<4 x i32> %A
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11]
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
@@ -108,6 +115,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed(<4 x i32> %A,
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] neg_lo:[0,1,0]
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
@@ -120,6 +128,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned(<4 x i32> %A,
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] neg_lo:[1,0,0]
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
@@ -132,6 +141,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed(<4 x i32> %A, <4
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] neg_lo:[1,1,0]
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
@@ -144,6 +154,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned_clamp(<4 x i
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] clamp
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
@@ -156,6 +167,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed_clamp(<4 x i32
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] neg_lo:[0,1,0] clamp
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
@@ -168,6 +180,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned_clamp(<4 x i32
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] neg_lo:[1,0,0] clamp
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
@@ -180,6 +193,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed_clamp(<4 x i32>
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[0:3], v[4:7], v[8:11] neg_lo:[1,1,0] clamp
 ; W64-NEXT:    global_store_b128 v[12:13], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
@@ -194,6 +208,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned(<2 x i32> %A
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7]
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
@@ -206,6 +221,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed(<2 x i32> %A,
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] neg_lo:[0,1,0]
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
@@ -218,6 +234,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned(<2 x i32> %A,
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] neg_lo:[1,0,0]
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
@@ -230,6 +247,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed(<2 x i32> %A, <2
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] neg_lo:[1,1,0]
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
@@ -242,6 +260,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned_clamp(<2 x i
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] clamp
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
@@ -254,6 +273,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed_clamp(<2 x i32
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] neg_lo:[0,1,0] clamp
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
@@ -266,6 +286,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned_clamp(<2 x i32
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] neg_lo:[1,0,0] clamp
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
@@ -278,6 +299,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed_clamp(<2 x i32>
 ; W64:       ; %bb.0: ; %bb
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[0:1], v[2:3], v[4:7] neg_lo:[1,1,0] clamp
 ; W64-NEXT:    global_store_b128 v[8:9], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.mulo.ll b/llvm/test/CodeGen/AMDGPU/llvm.mulo.ll
index 378af816dbad02..525f991bc5c731 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.mulo.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.mulo.ll
@@ -390,6 +390,7 @@ define amdgpu_kernel void @umulo_i64_s(i64 %x, i64 %y) {
 ; GFX11-NEXT:    v_cndmask_b32_e64 v1, s1, 0, s2
 ; GFX11-NEXT:    v_cndmask_b32_e64 v0, s0, 0, s2
 ; GFX11-NEXT:    global_store_b64 v[0:1], v[0:1], off
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 bb:
   %umulo = tail call { i64, i1 } @llvm.umul.with.overflow.i64(i64 %x, i64 %y)
@@ -582,6 +583,7 @@ define amdgpu_kernel void @smulo_i64_s(i64 %x, i64 %y) {
 ; GFX11-NEXT:    v_cndmask_b32_e64 v1, s1, 0, vcc_lo
 ; GFX11-NEXT:    v_cndmask_b32_e64 v0, s0, 0, vcc_lo
 ; GFX11-NEXT:    global_store_b64 v[0:1], v[0:1], off
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
 bb:
   %umulo = tail call { i64, i1 } @llvm.smul.with.overflow.i64(i64 %x, i64 %y)
diff --git a/llvm/test/CodeGen/AMDGPU/mad_64_32.ll b/llvm/test/CodeGen/AMDGPU/mad_64_32.ll
index fb370ab7b61d2d..68ef2763a5ad45 100644
--- a/llvm/test/CodeGen/AMDGPU/mad_64_32.ll
+++ b/llvm/test/CodeGen/AMDGPU/mad_64_32.ll
@@ -706,6 +706,7 @@ define amdgpu_kernel void @mad_i64_i32_uniform(i64 addrspace(1)* %out, i32 %arg0
 ; GFX11-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-NEXT:    global_store_b64 v2, v[0:1], s[0:1]
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-NEXT:    s_endpgm
   %ext0 = zext i32 %arg0 to i64
   %ext1 = zext i32 %arg1 to i64
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll
index 84bedfe3e10fdc..580ff9cdfb9045 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll
@@ -126,6 +126,7 @@ define amdgpu_kernel void @flat_agent_unordered_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_unordered_load:
@@ -139,6 +140,7 @@ define amdgpu_kernel void @flat_agent_unordered_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -263,6 +265,7 @@ define amdgpu_kernel void @flat_agent_monotonic_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_monotonic_load:
@@ -276,6 +279,7 @@ define amdgpu_kernel void @flat_agent_monotonic_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -411,6 +415,7 @@ define amdgpu_kernel void @flat_agent_acquire_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_acquire_load:
@@ -426,6 +431,7 @@ define amdgpu_kernel void @flat_agent_acquire_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -573,6 +579,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_seq_cst_load:
@@ -590,6 +597,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -695,6 +703,7 @@ define amdgpu_kernel void @flat_agent_unordered_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_unordered_store:
@@ -707,6 +716,7 @@ define amdgpu_kernel void @flat_agent_unordered_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -811,6 +821,7 @@ define amdgpu_kernel void @flat_agent_monotonic_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_monotonic_store:
@@ -823,6 +834,7 @@ define amdgpu_kernel void @flat_agent_monotonic_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -941,6 +953,7 @@ define amdgpu_kernel void @flat_agent_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_release_store:
@@ -955,6 +968,7 @@ define amdgpu_kernel void @flat_agent_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -1073,6 +1087,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_seq_cst_store:
@@ -1087,6 +1102,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -1191,6 +1207,7 @@ define amdgpu_kernel void @flat_agent_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_monotonic_atomicrmw:
@@ -1203,6 +1220,7 @@ define amdgpu_kernel void @flat_agent_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1464,6 +1482,7 @@ define amdgpu_kernel void @flat_agent_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_release_atomicrmw:
@@ -1478,6 +1497,7 @@ define amdgpu_kernel void @flat_agent_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1929,6 +1949,7 @@ define amdgpu_kernel void @flat_agent_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_acquire_ret_atomicrmw:
@@ -1945,6 +1966,7 @@ define amdgpu_kernel void @flat_agent_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -2093,6 +2115,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_acq_rel_ret_atomicrmw:
@@ -2111,6 +2134,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -2259,6 +2283,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_seq_cst_ret_atomicrmw:
@@ -2277,6 +2302,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -2395,6 +2421,7 @@ define amdgpu_kernel void @flat_agent_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_monotonic_monotonic_cmpxchg:
@@ -2408,6 +2435,7 @@ define amdgpu_kernel void @flat_agent_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2698,6 +2726,7 @@ define amdgpu_kernel void @flat_agent_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_release_monotonic_cmpxchg:
@@ -2713,6 +2742,7 @@ define amdgpu_kernel void @flat_agent_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4913,6 +4943,7 @@ define amdgpu_kernel void @flat_agent_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_monotonic_monotonic_ret_cmpxchg:
@@ -4928,6 +4959,7 @@ define amdgpu_kernel void @flat_agent_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5085,6 +5117,7 @@ define amdgpu_kernel void @flat_agent_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_acquire_monotonic_ret_cmpxchg:
@@ -5102,6 +5135,7 @@ define amdgpu_kernel void @flat_agent_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5262,6 +5296,7 @@ define amdgpu_kernel void @flat_agent_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_release_monotonic_ret_cmpxchg:
@@ -5279,6 +5314,7 @@ define amdgpu_kernel void @flat_agent_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5450,6 +5486,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_acq_rel_monotonic_ret_cmpxchg:
@@ -5469,6 +5506,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5640,6 +5678,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_seq_cst_monotonic_ret_cmpxchg:
@@ -5659,6 +5698,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5816,6 +5856,7 @@ define amdgpu_kernel void @flat_agent_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_monotonic_acquire_ret_cmpxchg:
@@ -5833,6 +5874,7 @@ define amdgpu_kernel void @flat_agent_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5990,6 +6032,7 @@ define amdgpu_kernel void @flat_agent_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_acquire_acquire_ret_cmpxchg:
@@ -6007,6 +6050,7 @@ define amdgpu_kernel void @flat_agent_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6178,6 +6222,7 @@ define amdgpu_kernel void @flat_agent_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_release_acquire_ret_cmpxchg:
@@ -6197,6 +6242,7 @@ define amdgpu_kernel void @flat_agent_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6368,6 +6414,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_acq_rel_acquire_ret_cmpxchg:
@@ -6387,6 +6434,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6558,6 +6606,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_seq_cst_acquire_ret_cmpxchg:
@@ -6577,6 +6626,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6748,6 +6798,7 @@ define amdgpu_kernel void @flat_agent_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_monotonic_seq_cst_ret_cmpxchg:
@@ -6767,6 +6818,7 @@ define amdgpu_kernel void @flat_agent_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6938,6 +6990,7 @@ define amdgpu_kernel void @flat_agent_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_acquire_seq_cst_ret_cmpxchg:
@@ -6957,6 +7010,7 @@ define amdgpu_kernel void @flat_agent_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -7128,6 +7182,7 @@ define amdgpu_kernel void @flat_agent_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_release_seq_cst_ret_cmpxchg:
@@ -7147,6 +7202,7 @@ define amdgpu_kernel void @flat_agent_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -7318,6 +7374,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_acq_rel_seq_cst_ret_cmpxchg:
@@ -7337,6 +7394,7 @@ define amdgpu_kernel void @flat_agent_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -7508,6 +7566,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_seq_cst_seq_cst_ret_cmpxchg:
@@ -7527,6 +7586,7 @@ define amdgpu_kernel void @flat_agent_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -7653,6 +7713,7 @@ define amdgpu_kernel void @flat_agent_one_as_unordered_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_unordered_load:
@@ -7666,6 +7727,7 @@ define amdgpu_kernel void @flat_agent_one_as_unordered_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -7790,6 +7852,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_monotonic_load:
@@ -7803,6 +7866,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -7945,6 +8009,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_acquire_load:
@@ -7961,6 +8026,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -8115,6 +8181,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_seq_cst_load:
@@ -8133,6 +8200,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -8238,6 +8306,7 @@ define amdgpu_kernel void @flat_agent_one_as_unordered_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_unordered_store:
@@ -8250,6 +8319,7 @@ define amdgpu_kernel void @flat_agent_one_as_unordered_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -8354,6 +8424,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_monotonic_store:
@@ -8366,6 +8437,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -8484,6 +8556,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_release_store:
@@ -8498,6 +8571,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -8616,6 +8690,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_seq_cst_store:
@@ -8630,6 +8705,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -8734,6 +8810,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_monotonic_atomicrmw:
@@ -8746,6 +8823,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -9003,6 +9081,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_release_atomicrmw:
@@ -9017,6 +9096,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -9466,6 +9546,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_acquire_ret_atomicrmw:
@@ -9483,6 +9564,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -9637,6 +9719,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_acq_rel_ret_atomicrmw:
@@ -9656,6 +9739,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -9810,6 +9894,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_seq_cst_ret_atomicrmw:
@@ -9829,6 +9914,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -9947,6 +10033,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_monotonic_monotonic_cmpxchg:
@@ -9960,6 +10047,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10246,6 +10334,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_release_monotonic_cmpxchg:
@@ -10261,6 +10350,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12413,6 +12503,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_monotonic_monotonic_ret_cmpxchg:
@@ -12428,6 +12519,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12592,6 +12684,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_acquire_monotonic_ret_cmpxchg:
@@ -12610,6 +12703,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12770,6 +12864,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_release_monotonic_ret_cmpxchg:
@@ -12787,6 +12882,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12965,6 +13061,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_acq_rel_monotonic_ret_cmpxchg:
@@ -12985,6 +13082,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -13163,6 +13261,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_seq_cst_monotonic_ret_cmpxchg:
@@ -13183,6 +13282,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -13347,6 +13447,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_monotonic_acquire_ret_cmpxchg:
@@ -13365,6 +13466,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -13529,6 +13631,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_acquire_acquire_ret_cmpxchg:
@@ -13547,6 +13650,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -13725,6 +13829,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_release_acquire_ret_cmpxchg:
@@ -13745,6 +13850,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -13923,6 +14029,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_acq_rel_acquire_ret_cmpxchg:
@@ -13943,6 +14050,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -14121,6 +14229,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_seq_cst_acquire_ret_cmpxchg:
@@ -14141,6 +14250,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -14319,6 +14429,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_monotonic_seq_cst_ret_cmpxchg:
@@ -14339,6 +14450,7 @@ define amdgpu_kernel void @flat_agent_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -14517,6 +14629,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_acquire_seq_cst_ret_cmpxchg:
@@ -14537,6 +14650,7 @@ define amdgpu_kernel void @flat_agent_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -14715,6 +14829,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_release_seq_cst_ret_cmpxchg:
@@ -14735,6 +14850,7 @@ define amdgpu_kernel void @flat_agent_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -14913,6 +15029,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_acq_rel_seq_cst_ret_cmpxchg:
@@ -14933,6 +15050,7 @@ define amdgpu_kernel void @flat_agent_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -15111,6 +15229,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_agent_one_as_seq_cst_seq_cst_ret_cmpxchg:
@@ -15131,6 +15250,7 @@ define amdgpu_kernel void @flat_agent_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-nontemporal.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-nontemporal.ll
index ed140d8c2e9332..1818b7ea7c667f 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-nontemporal.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-nontemporal.ll
@@ -126,6 +126,7 @@ define amdgpu_kernel void @flat_nontemporal_load_0(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_nontemporal_load_0:
@@ -139,6 +140,7 @@ define amdgpu_kernel void @flat_nontemporal_load_0(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -277,6 +279,7 @@ define amdgpu_kernel void @flat_nontemporal_load_1(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_nontemporal_load_1:
@@ -292,6 +295,7 @@ define amdgpu_kernel void @flat_nontemporal_load_1(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -418,6 +422,7 @@ define amdgpu_kernel void @flat_nontemporal_store_0(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2 glc slc dlc
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_nontemporal_store_0:
@@ -431,6 +436,7 @@ define amdgpu_kernel void @flat_nontemporal_store_0(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2 glc slc dlc
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -569,6 +575,7 @@ define amdgpu_kernel void @flat_nontemporal_store_1(
 ; GFX11-WGP-NEXT:    v_add_co_ci_u32_e64 v1, null, s3, 0, s0
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2 glc slc dlc
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_nontemporal_store_1:
@@ -584,6 +591,7 @@ define amdgpu_kernel void @flat_nontemporal_store_1(
 ; GFX11-CU-NEXT:    v_add_co_ci_u32_e64 v1, null, s3, 0, s0
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2 glc slc dlc
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-singlethread.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-singlethread.ll
index 08b230d818f6e1..b76de03786cfe2 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-singlethread.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-singlethread.ll
@@ -126,6 +126,7 @@ define amdgpu_kernel void @flat_singlethread_unordered_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_unordered_load:
@@ -139,6 +140,7 @@ define amdgpu_kernel void @flat_singlethread_unordered_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -263,6 +265,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_monotonic_load:
@@ -276,6 +279,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -400,6 +404,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acquire_load:
@@ -413,6 +418,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -537,6 +543,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_seq_cst_load:
@@ -550,6 +557,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -655,6 +663,7 @@ define amdgpu_kernel void @flat_singlethread_unordered_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_unordered_store:
@@ -667,6 +676,7 @@ define amdgpu_kernel void @flat_singlethread_unordered_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -771,6 +781,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_monotonic_store:
@@ -783,6 +794,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -887,6 +899,7 @@ define amdgpu_kernel void @flat_singlethread_release_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_release_store:
@@ -899,6 +912,7 @@ define amdgpu_kernel void @flat_singlethread_release_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -1003,6 +1017,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_seq_cst_store:
@@ -1015,6 +1030,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -1119,6 +1135,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_monotonic_atomicrmw:
@@ -1131,6 +1148,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1235,6 +1253,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acquire_atomicrmw:
@@ -1247,6 +1266,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1351,6 +1371,7 @@ define amdgpu_kernel void @flat_singlethread_release_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_release_atomicrmw:
@@ -1363,6 +1384,7 @@ define amdgpu_kernel void @flat_singlethread_release_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1467,6 +1489,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acq_rel_atomicrmw:
@@ -1479,6 +1502,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1583,6 +1607,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_seq_cst_atomicrmw:
@@ -1595,6 +1620,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1717,6 +1743,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acquire_ret_atomicrmw:
@@ -1731,6 +1758,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1854,6 +1882,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acq_rel_ret_atomicrmw:
@@ -1868,6 +1897,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1991,6 +2021,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_seq_cst_ret_atomicrmw:
@@ -2005,6 +2036,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -2123,6 +2155,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_monotonic_monotonic_cmpxchg:
@@ -2136,6 +2169,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2254,6 +2288,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acquire_monotonic_cmpxchg:
@@ -2267,6 +2302,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2385,6 +2421,7 @@ define amdgpu_kernel void @flat_singlethread_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_release_monotonic_cmpxchg:
@@ -2398,6 +2435,7 @@ define amdgpu_kernel void @flat_singlethread_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2516,6 +2554,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acq_rel_monotonic_cmpxchg:
@@ -2529,6 +2568,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2647,6 +2687,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_seq_cst_monotonic_cmpxchg:
@@ -2660,6 +2701,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2778,6 +2820,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_monotonic_acquire_cmpxchg:
@@ -2791,6 +2834,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2909,6 +2953,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acquire_acquire_cmpxchg:
@@ -2922,6 +2967,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3040,6 +3086,7 @@ define amdgpu_kernel void @flat_singlethread_release_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_release_acquire_cmpxchg:
@@ -3053,6 +3100,7 @@ define amdgpu_kernel void @flat_singlethread_release_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3171,6 +3219,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acq_rel_acquire_cmpxchg:
@@ -3184,6 +3233,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3302,6 +3352,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_seq_cst_acquire_cmpxchg:
@@ -3315,6 +3366,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3433,6 +3485,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_monotonic_seq_cst_cmpxchg:
@@ -3446,6 +3499,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3564,6 +3618,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acquire_seq_cst_cmpxchg:
@@ -3577,6 +3632,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3695,6 +3751,7 @@ define amdgpu_kernel void @flat_singlethread_release_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_release_seq_cst_cmpxchg:
@@ -3708,6 +3765,7 @@ define amdgpu_kernel void @flat_singlethread_release_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3826,6 +3884,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acq_rel_seq_cst_cmpxchg:
@@ -3839,6 +3898,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3957,6 +4017,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_seq_cst_seq_cst_cmpxchg:
@@ -3970,6 +4031,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4114,6 +4176,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_monotonic_monotonic_ret_cmpxchg:
@@ -4129,6 +4192,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4275,6 +4339,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acquire_monotonic_ret_cmpxchg:
@@ -4290,6 +4355,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4436,6 +4502,7 @@ define amdgpu_kernel void @flat_singlethread_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_release_monotonic_ret_cmpxchg:
@@ -4451,6 +4518,7 @@ define amdgpu_kernel void @flat_singlethread_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4597,6 +4665,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acq_rel_monotonic_ret_cmpxchg:
@@ -4612,6 +4681,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4758,6 +4828,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_seq_cst_monotonic_ret_cmpxchg:
@@ -4773,6 +4844,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4919,6 +4991,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_monotonic_acquire_ret_cmpxchg:
@@ -4934,6 +5007,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5080,6 +5154,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acquire_acquire_ret_cmpxchg:
@@ -5095,6 +5170,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5241,6 +5317,7 @@ define amdgpu_kernel void @flat_singlethread_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_release_acquire_ret_cmpxchg:
@@ -5256,6 +5333,7 @@ define amdgpu_kernel void @flat_singlethread_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5402,6 +5480,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acq_rel_acquire_ret_cmpxchg:
@@ -5417,6 +5496,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5563,6 +5643,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_seq_cst_acquire_ret_cmpxchg:
@@ -5578,6 +5659,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5724,6 +5806,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_monotonic_seq_cst_ret_cmpxchg:
@@ -5739,6 +5822,7 @@ define amdgpu_kernel void @flat_singlethread_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5885,6 +5969,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acquire_seq_cst_ret_cmpxchg:
@@ -5900,6 +5985,7 @@ define amdgpu_kernel void @flat_singlethread_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6046,6 +6132,7 @@ define amdgpu_kernel void @flat_singlethread_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_release_seq_cst_ret_cmpxchg:
@@ -6061,6 +6148,7 @@ define amdgpu_kernel void @flat_singlethread_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6207,6 +6295,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_acq_rel_seq_cst_ret_cmpxchg:
@@ -6222,6 +6311,7 @@ define amdgpu_kernel void @flat_singlethread_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6368,6 +6458,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_seq_cst_seq_cst_ret_cmpxchg:
@@ -6383,6 +6474,7 @@ define amdgpu_kernel void @flat_singlethread_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6509,6 +6601,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_unordered_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_unordered_load:
@@ -6522,6 +6615,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_unordered_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -6646,6 +6740,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_monotonic_load:
@@ -6659,6 +6754,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -6783,6 +6879,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acquire_load:
@@ -6796,6 +6893,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -6920,6 +7018,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_seq_cst_load:
@@ -6933,6 +7032,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -7038,6 +7138,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_unordered_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_unordered_store:
@@ -7050,6 +7151,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_unordered_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -7154,6 +7256,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_monotonic_store:
@@ -7166,6 +7269,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -7270,6 +7374,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_release_store:
@@ -7282,6 +7387,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -7386,6 +7492,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_seq_cst_store:
@@ -7398,6 +7505,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -7502,6 +7610,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_monotonic_atomicrmw:
@@ -7514,6 +7623,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -7618,6 +7728,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acquire_atomicrmw:
@@ -7630,6 +7741,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -7734,6 +7846,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_release_atomicrmw:
@@ -7746,6 +7859,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -7850,6 +7964,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acq_rel_atomicrmw:
@@ -7862,6 +7977,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -7966,6 +8082,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_seq_cst_atomicrmw:
@@ -7978,6 +8095,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8100,6 +8218,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acquire_ret_atomicrmw:
@@ -8114,6 +8233,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8237,6 +8357,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acq_rel_ret_atomicrmw:
@@ -8251,6 +8372,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8374,6 +8496,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_seq_cst_ret_atomicrmw:
@@ -8388,6 +8511,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8506,6 +8630,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_monotonic_monotonic_cmpxchg:
@@ -8519,6 +8644,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -8637,6 +8763,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acquire_monotonic_cmpxchg:
@@ -8650,6 +8777,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -8768,6 +8896,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_release_monotonic_cmpxchg:
@@ -8781,6 +8910,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -8899,6 +9029,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acq_rel_monotonic_cmpxchg:
@@ -8912,6 +9043,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9030,6 +9162,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_seq_cst_monotonic_cmpxchg:
@@ -9043,6 +9176,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9161,6 +9295,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_monotonic_acquire_cmpxchg:
@@ -9174,6 +9309,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9292,6 +9428,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acquire_acquire_cmpxchg:
@@ -9305,6 +9442,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9423,6 +9561,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_release_acquire_cmpxchg:
@@ -9436,6 +9575,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9554,6 +9694,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acq_rel_acquire_cmpxchg:
@@ -9567,6 +9708,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9685,6 +9827,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_seq_cst_acquire_cmpxchg:
@@ -9698,6 +9841,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9816,6 +9960,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_monotonic_seq_cst_cmpxchg:
@@ -9829,6 +9974,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9947,6 +10093,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acquire_seq_cst_cmpxchg:
@@ -9960,6 +10107,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10078,6 +10226,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_release_seq_cst_cmpxchg:
@@ -10091,6 +10240,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10209,6 +10359,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acq_rel_seq_cst_cmpxchg:
@@ -10222,6 +10373,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10340,6 +10492,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_seq_cst_seq_cst_cmpxchg:
@@ -10353,6 +10506,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10497,6 +10651,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_monotonic_ret_cmpx
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_monotonic_monotonic_ret_cmpxchg:
@@ -10512,6 +10667,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_monotonic_ret_cmpx
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10658,6 +10814,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_monotonic_ret_cmpxch
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acquire_monotonic_ret_cmpxchg:
@@ -10673,6 +10830,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_monotonic_ret_cmpxch
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10819,6 +10977,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_monotonic_ret_cmpxch
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_release_monotonic_ret_cmpxchg:
@@ -10834,6 +10993,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_monotonic_ret_cmpxch
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10980,6 +11140,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_monotonic_ret_cmpxch
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acq_rel_monotonic_ret_cmpxchg:
@@ -10995,6 +11156,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_monotonic_ret_cmpxch
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11141,6 +11303,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_monotonic_ret_cmpxch
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_seq_cst_monotonic_ret_cmpxchg:
@@ -11156,6 +11319,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_monotonic_ret_cmpxch
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11302,6 +11466,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_acquire_ret_cmpxch
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_monotonic_acquire_ret_cmpxchg:
@@ -11317,6 +11482,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_acquire_ret_cmpxch
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11463,6 +11629,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acquire_acquire_ret_cmpxchg:
@@ -11478,6 +11645,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11624,6 +11792,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_release_acquire_ret_cmpxchg:
@@ -11639,6 +11808,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11785,6 +11955,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acq_rel_acquire_ret_cmpxchg:
@@ -11800,6 +11971,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11946,6 +12118,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_seq_cst_acquire_ret_cmpxchg:
@@ -11961,6 +12134,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12107,6 +12281,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_seq_cst_ret_cmpxch
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_monotonic_seq_cst_ret_cmpxchg:
@@ -12122,6 +12297,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_monotonic_seq_cst_ret_cmpxch
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12268,6 +12444,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acquire_seq_cst_ret_cmpxchg:
@@ -12283,6 +12460,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12429,6 +12607,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_release_seq_cst_ret_cmpxchg:
@@ -12444,6 +12623,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12590,6 +12770,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_acq_rel_seq_cst_ret_cmpxchg:
@@ -12605,6 +12786,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12751,6 +12933,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_singlethread_one_as_seq_cst_seq_cst_ret_cmpxchg:
@@ -12766,6 +12949,7 @@ define amdgpu_kernel void @flat_singlethread_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-system.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-system.ll
index a99380d1131fbb..9fd2a333c5213c 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-system.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-system.ll
@@ -126,6 +126,7 @@ define amdgpu_kernel void @flat_system_unordered_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_unordered_load:
@@ -139,6 +140,7 @@ define amdgpu_kernel void @flat_system_unordered_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -263,6 +265,7 @@ define amdgpu_kernel void @flat_system_monotonic_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_monotonic_load:
@@ -276,6 +279,7 @@ define amdgpu_kernel void @flat_system_monotonic_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -413,6 +417,7 @@ define amdgpu_kernel void @flat_system_acquire_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_acquire_load:
@@ -428,6 +433,7 @@ define amdgpu_kernel void @flat_system_acquire_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -577,6 +583,7 @@ define amdgpu_kernel void @flat_system_seq_cst_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_seq_cst_load:
@@ -594,6 +601,7 @@ define amdgpu_kernel void @flat_system_seq_cst_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -699,6 +707,7 @@ define amdgpu_kernel void @flat_system_unordered_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_unordered_store:
@@ -711,6 +720,7 @@ define amdgpu_kernel void @flat_system_unordered_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -815,6 +825,7 @@ define amdgpu_kernel void @flat_system_monotonic_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_monotonic_store:
@@ -827,6 +838,7 @@ define amdgpu_kernel void @flat_system_monotonic_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -947,6 +959,7 @@ define amdgpu_kernel void @flat_system_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_release_store:
@@ -961,6 +974,7 @@ define amdgpu_kernel void @flat_system_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -1081,6 +1095,7 @@ define amdgpu_kernel void @flat_system_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_seq_cst_store:
@@ -1095,6 +1110,7 @@ define amdgpu_kernel void @flat_system_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -1199,6 +1215,7 @@ define amdgpu_kernel void @flat_system_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_monotonic_atomicrmw:
@@ -1211,6 +1228,7 @@ define amdgpu_kernel void @flat_system_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1476,6 +1494,7 @@ define amdgpu_kernel void @flat_system_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_release_atomicrmw:
@@ -1490,6 +1509,7 @@ define amdgpu_kernel void @flat_system_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1951,6 +1971,7 @@ define amdgpu_kernel void @flat_system_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_acquire_ret_atomicrmw:
@@ -1967,6 +1988,7 @@ define amdgpu_kernel void @flat_system_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -2119,6 +2141,7 @@ define amdgpu_kernel void @flat_system_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_acq_rel_ret_atomicrmw:
@@ -2137,6 +2160,7 @@ define amdgpu_kernel void @flat_system_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -2289,6 +2313,7 @@ define amdgpu_kernel void @flat_system_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_seq_cst_ret_atomicrmw:
@@ -2307,6 +2332,7 @@ define amdgpu_kernel void @flat_system_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -2425,6 +2451,7 @@ define amdgpu_kernel void @flat_system_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_monotonic_monotonic_cmpxchg:
@@ -2438,6 +2465,7 @@ define amdgpu_kernel void @flat_system_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2732,6 +2760,7 @@ define amdgpu_kernel void @flat_system_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_release_monotonic_cmpxchg:
@@ -2747,6 +2776,7 @@ define amdgpu_kernel void @flat_system_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4991,6 +5021,7 @@ define amdgpu_kernel void @flat_system_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_monotonic_monotonic_ret_cmpxchg:
@@ -5006,6 +5037,7 @@ define amdgpu_kernel void @flat_system_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5165,6 +5197,7 @@ define amdgpu_kernel void @flat_system_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_acquire_monotonic_ret_cmpxchg:
@@ -5182,6 +5215,7 @@ define amdgpu_kernel void @flat_system_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5344,6 +5378,7 @@ define amdgpu_kernel void @flat_system_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_release_monotonic_ret_cmpxchg:
@@ -5361,6 +5396,7 @@ define amdgpu_kernel void @flat_system_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5536,6 +5572,7 @@ define amdgpu_kernel void @flat_system_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_acq_rel_monotonic_ret_cmpxchg:
@@ -5555,6 +5592,7 @@ define amdgpu_kernel void @flat_system_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5730,6 +5768,7 @@ define amdgpu_kernel void @flat_system_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_seq_cst_monotonic_ret_cmpxchg:
@@ -5749,6 +5788,7 @@ define amdgpu_kernel void @flat_system_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5908,6 +5948,7 @@ define amdgpu_kernel void @flat_system_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_monotonic_acquire_ret_cmpxchg:
@@ -5925,6 +5966,7 @@ define amdgpu_kernel void @flat_system_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6084,6 +6126,7 @@ define amdgpu_kernel void @flat_system_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_acquire_acquire_ret_cmpxchg:
@@ -6101,6 +6144,7 @@ define amdgpu_kernel void @flat_system_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6276,6 +6320,7 @@ define amdgpu_kernel void @flat_system_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_release_acquire_ret_cmpxchg:
@@ -6295,6 +6340,7 @@ define amdgpu_kernel void @flat_system_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6470,6 +6516,7 @@ define amdgpu_kernel void @flat_system_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_acq_rel_acquire_ret_cmpxchg:
@@ -6489,6 +6536,7 @@ define amdgpu_kernel void @flat_system_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6664,6 +6712,7 @@ define amdgpu_kernel void @flat_system_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_seq_cst_acquire_ret_cmpxchg:
@@ -6683,6 +6732,7 @@ define amdgpu_kernel void @flat_system_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6858,6 +6908,7 @@ define amdgpu_kernel void @flat_system_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_monotonic_seq_cst_ret_cmpxchg:
@@ -6877,6 +6928,7 @@ define amdgpu_kernel void @flat_system_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -7052,6 +7104,7 @@ define amdgpu_kernel void @flat_system_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_acquire_seq_cst_ret_cmpxchg:
@@ -7071,6 +7124,7 @@ define amdgpu_kernel void @flat_system_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -7246,6 +7300,7 @@ define amdgpu_kernel void @flat_system_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_release_seq_cst_ret_cmpxchg:
@@ -7265,6 +7320,7 @@ define amdgpu_kernel void @flat_system_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -7440,6 +7496,7 @@ define amdgpu_kernel void @flat_system_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_acq_rel_seq_cst_ret_cmpxchg:
@@ -7459,6 +7516,7 @@ define amdgpu_kernel void @flat_system_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -7634,6 +7692,7 @@ define amdgpu_kernel void @flat_system_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_seq_cst_seq_cst_ret_cmpxchg:
@@ -7653,6 +7712,7 @@ define amdgpu_kernel void @flat_system_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -7779,6 +7839,7 @@ define amdgpu_kernel void @flat_system_one_as_unordered_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_unordered_load:
@@ -7792,6 +7853,7 @@ define amdgpu_kernel void @flat_system_one_as_unordered_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -7916,6 +7978,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_monotonic_load:
@@ -7929,6 +7992,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -8073,6 +8137,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_acquire_load:
@@ -8089,6 +8154,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -8245,6 +8311,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_seq_cst_load:
@@ -8263,6 +8330,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -8368,6 +8436,7 @@ define amdgpu_kernel void @flat_system_one_as_unordered_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_unordered_store:
@@ -8380,6 +8449,7 @@ define amdgpu_kernel void @flat_system_one_as_unordered_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -8484,6 +8554,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_monotonic_store:
@@ -8496,6 +8567,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -8616,6 +8688,7 @@ define amdgpu_kernel void @flat_system_one_as_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_release_store:
@@ -8630,6 +8703,7 @@ define amdgpu_kernel void @flat_system_one_as_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -8750,6 +8824,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_seq_cst_store:
@@ -8764,6 +8839,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -8868,6 +8944,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_monotonic_atomicrmw:
@@ -8880,6 +8957,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -9141,6 +9219,7 @@ define amdgpu_kernel void @flat_system_one_as_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_release_atomicrmw:
@@ -9155,6 +9234,7 @@ define amdgpu_kernel void @flat_system_one_as_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -9614,6 +9694,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_acquire_ret_atomicrmw:
@@ -9631,6 +9712,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -9789,6 +9871,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_acq_rel_ret_atomicrmw:
@@ -9808,6 +9891,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -9966,6 +10050,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_seq_cst_ret_atomicrmw:
@@ -9985,6 +10070,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -10103,6 +10189,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_monotonic_monotonic_cmpxchg:
@@ -10116,6 +10203,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10406,6 +10494,7 @@ define amdgpu_kernel void @flat_system_one_as_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_release_monotonic_cmpxchg:
@@ -10421,6 +10510,7 @@ define amdgpu_kernel void @flat_system_one_as_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12617,6 +12707,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_monotonic_monotonic_ret_cmpxchg:
@@ -12632,6 +12723,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12798,6 +12890,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_acquire_monotonic_ret_cmpxchg:
@@ -12816,6 +12909,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12978,6 +13072,7 @@ define amdgpu_kernel void @flat_system_one_as_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_release_monotonic_ret_cmpxchg:
@@ -12995,6 +13090,7 @@ define amdgpu_kernel void @flat_system_one_as_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -13177,6 +13273,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_acq_rel_monotonic_ret_cmpxchg:
@@ -13197,6 +13294,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -13379,6 +13477,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_seq_cst_monotonic_ret_cmpxchg:
@@ -13399,6 +13498,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -13565,6 +13665,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_monotonic_acquire_ret_cmpxchg:
@@ -13583,6 +13684,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -13749,6 +13851,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_acquire_acquire_ret_cmpxchg:
@@ -13767,6 +13870,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -13949,6 +14053,7 @@ define amdgpu_kernel void @flat_system_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_release_acquire_ret_cmpxchg:
@@ -13969,6 +14074,7 @@ define amdgpu_kernel void @flat_system_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -14151,6 +14257,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_acq_rel_acquire_ret_cmpxchg:
@@ -14171,6 +14278,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -14353,6 +14461,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_seq_cst_acquire_ret_cmpxchg:
@@ -14373,6 +14482,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -14555,6 +14665,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_monotonic_seq_cst_ret_cmpxchg:
@@ -14575,6 +14686,7 @@ define amdgpu_kernel void @flat_system_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -14757,6 +14869,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_acquire_seq_cst_ret_cmpxchg:
@@ -14777,6 +14890,7 @@ define amdgpu_kernel void @flat_system_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -14959,6 +15073,7 @@ define amdgpu_kernel void @flat_system_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_release_seq_cst_ret_cmpxchg:
@@ -14979,6 +15094,7 @@ define amdgpu_kernel void @flat_system_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -15161,6 +15277,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_acq_rel_seq_cst_ret_cmpxchg:
@@ -15181,6 +15298,7 @@ define amdgpu_kernel void @flat_system_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -15363,6 +15481,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_system_one_as_seq_cst_seq_cst_ret_cmpxchg:
@@ -15383,6 +15502,7 @@ define amdgpu_kernel void @flat_system_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-volatile.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-volatile.ll
index e1729e0bfd1c2b..2fd78b987aa11d 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-volatile.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-volatile.ll
@@ -75,6 +75,7 @@ define amdgpu_kernel void @flat_nontemporal_load_0(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_nontemporal_load_0:
@@ -89,6 +90,7 @@ define amdgpu_kernel void @flat_nontemporal_load_0(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -174,6 +176,7 @@ define amdgpu_kernel void @flat_nontemporal_load_1(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_nontemporal_load_1:
@@ -190,6 +193,7 @@ define amdgpu_kernel void @flat_nontemporal_load_1(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -269,6 +273,7 @@ define amdgpu_kernel void @flat_nontemporal_store_0(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2 dlc
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_nontemporal_store_0:
@@ -283,6 +288,7 @@ define amdgpu_kernel void @flat_nontemporal_store_0(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2 dlc
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -368,6 +374,7 @@ define amdgpu_kernel void @flat_nontemporal_store_1(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2 dlc
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_nontemporal_store_1:
@@ -384,6 +391,7 @@ define amdgpu_kernel void @flat_nontemporal_store_1(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2 dlc
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -463,6 +471,7 @@ define amdgpu_kernel void @flat_volatile_workgroup_acquire_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_volatile_workgroup_acquire_load:
@@ -477,6 +486,7 @@ define amdgpu_kernel void @flat_volatile_workgroup_acquire_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -549,6 +559,7 @@ define amdgpu_kernel void @flat_volatile_workgroup_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_volatile_workgroup_release_store:
@@ -562,6 +573,7 @@ define amdgpu_kernel void @flat_volatile_workgroup_release_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
    i32 %in, i32* %out) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-wavefront.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-wavefront.ll
index 7b1b5a1899c032..3267731e60a62a 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-wavefront.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-wavefront.ll
@@ -126,6 +126,7 @@ define amdgpu_kernel void @flat_wavefront_unordered_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_unordered_load:
@@ -139,6 +140,7 @@ define amdgpu_kernel void @flat_wavefront_unordered_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -263,6 +265,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_monotonic_load:
@@ -276,6 +279,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -400,6 +404,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acquire_load:
@@ -413,6 +418,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -537,6 +543,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_seq_cst_load:
@@ -550,6 +557,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -655,6 +663,7 @@ define amdgpu_kernel void @flat_wavefront_unordered_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_unordered_store:
@@ -667,6 +676,7 @@ define amdgpu_kernel void @flat_wavefront_unordered_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -771,6 +781,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_monotonic_store:
@@ -783,6 +794,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -887,6 +899,7 @@ define amdgpu_kernel void @flat_wavefront_release_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_release_store:
@@ -899,6 +912,7 @@ define amdgpu_kernel void @flat_wavefront_release_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -1003,6 +1017,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_seq_cst_store:
@@ -1015,6 +1030,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -1119,6 +1135,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_monotonic_atomicrmw:
@@ -1131,6 +1148,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1235,6 +1253,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acquire_atomicrmw:
@@ -1247,6 +1266,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1351,6 +1371,7 @@ define amdgpu_kernel void @flat_wavefront_release_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_release_atomicrmw:
@@ -1363,6 +1384,7 @@ define amdgpu_kernel void @flat_wavefront_release_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1467,6 +1489,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acq_rel_atomicrmw:
@@ -1479,6 +1502,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1583,6 +1607,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_seq_cst_atomicrmw:
@@ -1595,6 +1620,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1717,6 +1743,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acquire_ret_atomicrmw:
@@ -1731,6 +1758,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1854,6 +1882,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acq_rel_ret_atomicrmw:
@@ -1868,6 +1897,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1991,6 +2021,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_seq_cst_ret_atomicrmw:
@@ -2005,6 +2036,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -2123,6 +2155,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_monotonic_monotonic_cmpxchg:
@@ -2136,6 +2169,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2254,6 +2288,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acquire_monotonic_cmpxchg:
@@ -2267,6 +2302,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2385,6 +2421,7 @@ define amdgpu_kernel void @flat_wavefront_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_release_monotonic_cmpxchg:
@@ -2398,6 +2435,7 @@ define amdgpu_kernel void @flat_wavefront_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2516,6 +2554,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acq_rel_monotonic_cmpxchg:
@@ -2529,6 +2568,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2647,6 +2687,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_seq_cst_monotonic_cmpxchg:
@@ -2660,6 +2701,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2778,6 +2820,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_monotonic_acquire_cmpxchg:
@@ -2791,6 +2834,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2909,6 +2953,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acquire_acquire_cmpxchg:
@@ -2922,6 +2967,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3040,6 +3086,7 @@ define amdgpu_kernel void @flat_wavefront_release_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_release_acquire_cmpxchg:
@@ -3053,6 +3100,7 @@ define amdgpu_kernel void @flat_wavefront_release_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3171,6 +3219,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acq_rel_acquire_cmpxchg:
@@ -3184,6 +3233,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3302,6 +3352,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_seq_cst_acquire_cmpxchg:
@@ -3315,6 +3366,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3433,6 +3485,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_monotonic_seq_cst_cmpxchg:
@@ -3446,6 +3499,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3564,6 +3618,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acquire_seq_cst_cmpxchg:
@@ -3577,6 +3632,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3695,6 +3751,7 @@ define amdgpu_kernel void @flat_wavefront_release_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_release_seq_cst_cmpxchg:
@@ -3708,6 +3765,7 @@ define amdgpu_kernel void @flat_wavefront_release_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3826,6 +3884,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acq_rel_seq_cst_cmpxchg:
@@ -3839,6 +3898,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3957,6 +4017,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_seq_cst_seq_cst_cmpxchg:
@@ -3970,6 +4031,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4114,6 +4176,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_monotonic_monotonic_ret_cmpxchg:
@@ -4129,6 +4192,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4275,6 +4339,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acquire_monotonic_ret_cmpxchg:
@@ -4290,6 +4355,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4436,6 +4502,7 @@ define amdgpu_kernel void @flat_wavefront_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_release_monotonic_ret_cmpxchg:
@@ -4451,6 +4518,7 @@ define amdgpu_kernel void @flat_wavefront_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4597,6 +4665,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acq_rel_monotonic_ret_cmpxchg:
@@ -4612,6 +4681,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4758,6 +4828,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_seq_cst_monotonic_ret_cmpxchg:
@@ -4773,6 +4844,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4919,6 +4991,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_monotonic_acquire_ret_cmpxchg:
@@ -4934,6 +5007,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5080,6 +5154,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acquire_acquire_ret_cmpxchg:
@@ -5095,6 +5170,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5241,6 +5317,7 @@ define amdgpu_kernel void @flat_wavefront_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_release_acquire_ret_cmpxchg:
@@ -5256,6 +5333,7 @@ define amdgpu_kernel void @flat_wavefront_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5402,6 +5480,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acq_rel_acquire_ret_cmpxchg:
@@ -5417,6 +5496,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5563,6 +5643,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_seq_cst_acquire_ret_cmpxchg:
@@ -5578,6 +5659,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5724,6 +5806,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_monotonic_seq_cst_ret_cmpxchg:
@@ -5739,6 +5822,7 @@ define amdgpu_kernel void @flat_wavefront_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5885,6 +5969,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acquire_seq_cst_ret_cmpxchg:
@@ -5900,6 +5985,7 @@ define amdgpu_kernel void @flat_wavefront_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6046,6 +6132,7 @@ define amdgpu_kernel void @flat_wavefront_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_release_seq_cst_ret_cmpxchg:
@@ -6061,6 +6148,7 @@ define amdgpu_kernel void @flat_wavefront_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6207,6 +6295,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_acq_rel_seq_cst_ret_cmpxchg:
@@ -6222,6 +6311,7 @@ define amdgpu_kernel void @flat_wavefront_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6368,6 +6458,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_seq_cst_seq_cst_ret_cmpxchg:
@@ -6383,6 +6474,7 @@ define amdgpu_kernel void @flat_wavefront_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6509,6 +6601,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_unordered_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_unordered_load:
@@ -6522,6 +6615,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_unordered_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -6646,6 +6740,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_monotonic_load:
@@ -6659,6 +6754,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -6783,6 +6879,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acquire_load:
@@ -6796,6 +6893,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -6920,6 +7018,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_seq_cst_load:
@@ -6933,6 +7032,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -7038,6 +7138,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_unordered_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_unordered_store:
@@ -7050,6 +7151,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_unordered_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -7154,6 +7256,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_monotonic_store:
@@ -7166,6 +7269,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -7270,6 +7374,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_release_store:
@@ -7282,6 +7387,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -7386,6 +7492,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_seq_cst_store:
@@ -7398,6 +7505,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -7502,6 +7610,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_monotonic_atomicrmw:
@@ -7514,6 +7623,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -7618,6 +7728,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acquire_atomicrmw:
@@ -7630,6 +7741,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -7734,6 +7846,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_release_atomicrmw:
@@ -7746,6 +7859,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -7850,6 +7964,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acq_rel_atomicrmw:
@@ -7862,6 +7977,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -7966,6 +8082,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_seq_cst_atomicrmw:
@@ -7978,6 +8095,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8100,6 +8218,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acquire_ret_atomicrmw:
@@ -8114,6 +8233,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8237,6 +8357,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acq_rel_ret_atomicrmw:
@@ -8251,6 +8372,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8374,6 +8496,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_seq_cst_ret_atomicrmw:
@@ -8388,6 +8511,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8506,6 +8630,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_monotonic_monotonic_cmpxchg:
@@ -8519,6 +8644,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -8637,6 +8763,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acquire_monotonic_cmpxchg:
@@ -8650,6 +8777,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -8768,6 +8896,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_release_monotonic_cmpxchg:
@@ -8781,6 +8910,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -8899,6 +9029,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acq_rel_monotonic_cmpxchg:
@@ -8912,6 +9043,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9030,6 +9162,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_seq_cst_monotonic_cmpxchg:
@@ -9043,6 +9176,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9161,6 +9295,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_monotonic_acquire_cmpxchg:
@@ -9174,6 +9309,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9292,6 +9428,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acquire_acquire_cmpxchg:
@@ -9305,6 +9442,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9423,6 +9561,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_release_acquire_cmpxchg:
@@ -9436,6 +9575,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9554,6 +9694,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acq_rel_acquire_cmpxchg:
@@ -9567,6 +9708,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9685,6 +9827,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_seq_cst_acquire_cmpxchg:
@@ -9698,6 +9841,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9816,6 +9960,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_monotonic_seq_cst_cmpxchg:
@@ -9829,6 +9974,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9947,6 +10093,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acquire_seq_cst_cmpxchg:
@@ -9960,6 +10107,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10078,6 +10226,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_release_seq_cst_cmpxchg:
@@ -10091,6 +10240,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10209,6 +10359,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acq_rel_seq_cst_cmpxchg:
@@ -10222,6 +10373,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10340,6 +10492,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_seq_cst_seq_cst_cmpxchg:
@@ -10353,6 +10506,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10497,6 +10651,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_monotonic_ret_cmpxchg
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_monotonic_monotonic_ret_cmpxchg:
@@ -10512,6 +10667,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_monotonic_ret_cmpxchg
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10658,6 +10814,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acquire_monotonic_ret_cmpxchg:
@@ -10673,6 +10830,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10819,6 +10977,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg:
@@ -10834,6 +10993,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10980,6 +11140,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg:
@@ -10995,6 +11156,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11141,6 +11303,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_monotonic_acquire_ret_cmpxchg:
@@ -11156,6 +11319,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11302,6 +11466,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acquire_acquire_ret_cmpxchg:
@@ -11317,6 +11482,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11463,6 +11629,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_release_acquire_ret_cmpxchg:
@@ -11478,6 +11645,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11624,6 +11792,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acq_rel_acquire_ret_cmpxchg:
@@ -11639,6 +11808,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11785,6 +11955,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_seq_cst_acquire_ret_cmpxchg:
@@ -11800,6 +11971,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11946,6 +12118,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg:
@@ -11961,6 +12134,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
    i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12107,6 +12281,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acquire_seq_cst_ret_cmpxchg:
@@ -12122,6 +12297,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
    i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12268,6 +12444,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_release_seq_cst_ret_cmpxchg:
@@ -12283,6 +12460,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
    i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12429,6 +12607,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_relc_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_acq_relc_seq_cst_ret_cmpxchg:
@@ -12444,6 +12623,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_acq_relc_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
    i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12590,6 +12770,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg:
@@ -12605,6 +12786,7 @@ define amdgpu_kernel void @flat_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-workgroup.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-workgroup.ll
index efc7a8294b04ef..59e57d6a18f574 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-workgroup.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-workgroup.ll
@@ -126,6 +126,7 @@ define amdgpu_kernel void @flat_workgroup_unordered_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_unordered_load:
@@ -139,6 +140,7 @@ define amdgpu_kernel void @flat_workgroup_unordered_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -263,6 +265,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_monotonic_load:
@@ -276,6 +279,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -409,6 +413,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_acquire_load:
@@ -423,6 +428,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -567,6 +573,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_seq_cst_load:
@@ -582,6 +589,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -687,6 +695,7 @@ define amdgpu_kernel void @flat_workgroup_unordered_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_unordered_store:
@@ -699,6 +708,7 @@ define amdgpu_kernel void @flat_workgroup_unordered_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -803,6 +813,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_monotonic_store:
@@ -815,6 +826,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -930,6 +942,7 @@ define amdgpu_kernel void @flat_workgroup_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_release_store:
@@ -943,6 +956,7 @@ define amdgpu_kernel void @flat_workgroup_release_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -1058,6 +1072,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_seq_cst_store:
@@ -1071,6 +1086,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -1175,6 +1191,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_monotonic_atomicrmw:
@@ -1187,6 +1204,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1319,6 +1337,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1434,6 +1453,7 @@ define amdgpu_kernel void @flat_workgroup_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_release_atomicrmw:
@@ -1447,6 +1467,7 @@ define amdgpu_kernel void @flat_workgroup_release_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1591,6 +1612,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1735,6 +1757,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -1861,6 +1884,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_acquire_ret_atomicrmw:
@@ -1875,6 +1899,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -2013,6 +2038,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_acq_rel_ret_atomicrmw:
@@ -2028,6 +2054,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -2166,6 +2193,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_seq_cst_ret_atomicrmw:
@@ -2181,6 +2209,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -2299,6 +2328,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_monotonic_monotonic_cmpxchg:
@@ -2312,6 +2342,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2459,6 +2490,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2588,6 +2620,7 @@ define amdgpu_kernel void @flat_workgroup_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_release_monotonic_cmpxchg:
@@ -2602,6 +2635,7 @@ define amdgpu_kernel void @flat_workgroup_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2761,6 +2795,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -2920,6 +2955,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3067,6 +3103,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3214,6 +3251,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3373,6 +3411,7 @@ define amdgpu_kernel void @flat_workgroup_release_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3532,6 +3571,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3691,6 +3731,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3850,6 +3891,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -3994,6 +4036,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_monotonic_monotonic_ret_cmpxchg:
@@ -4009,6 +4052,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4162,6 +4206,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_acquire_monotonic_ret_cmpxchg:
@@ -4177,6 +4222,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4334,6 +4380,7 @@ define amdgpu_kernel void @flat_workgroup_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_release_monotonic_ret_cmpxchg:
@@ -4350,6 +4397,7 @@ define amdgpu_kernel void @flat_workgroup_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4514,6 +4562,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_acq_rel_monotonic_ret_cmpxchg:
@@ -4530,6 +4579,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4694,6 +4744,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_seq_cst_monotonic_ret_cmpxchg:
@@ -4710,6 +4761,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -4863,6 +4915,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_monotonic_acquire_ret_cmpxchg:
@@ -4878,6 +4931,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5031,6 +5085,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_acquire_acquire_ret_cmpxchg:
@@ -5046,6 +5101,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5210,6 +5266,7 @@ define amdgpu_kernel void @flat_workgroup_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_release_acquire_ret_cmpxchg:
@@ -5226,6 +5283,7 @@ define amdgpu_kernel void @flat_workgroup_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5390,6 +5448,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_acq_rel_acquire_ret_cmpxchg:
@@ -5406,6 +5465,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5570,6 +5630,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_seq_cst_acquire_ret_cmpxchg:
@@ -5586,6 +5647,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5750,6 +5812,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_monotonic_seq_cst_ret_cmpxchg:
@@ -5766,6 +5829,7 @@ define amdgpu_kernel void @flat_workgroup_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -5930,6 +5994,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_acquire_seq_cst_ret_cmpxchg:
@@ -5946,6 +6011,7 @@ define amdgpu_kernel void @flat_workgroup_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6110,6 +6176,7 @@ define amdgpu_kernel void @flat_workgroup_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_release_seq_cst_ret_cmpxchg:
@@ -6126,6 +6193,7 @@ define amdgpu_kernel void @flat_workgroup_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6290,6 +6358,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_acq_rel_seq_cst_ret_cmpxchg:
@@ -6306,6 +6375,7 @@ define amdgpu_kernel void @flat_workgroup_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6470,6 +6540,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_seq_cst_seq_cst_ret_cmpxchg:
@@ -6486,6 +6557,7 @@ define amdgpu_kernel void @flat_workgroup_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -6612,6 +6684,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_unordered_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_unordered_load:
@@ -6625,6 +6698,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_unordered_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -6749,6 +6823,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_monotonic_load:
@@ -6762,6 +6837,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -6892,6 +6968,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_acquire_load:
@@ -6905,6 +6982,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -7041,6 +7119,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_load(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_seq_cst_load:
@@ -7054,6 +7133,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_load(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %in, i32* %out) {
 entry:
@@ -7159,6 +7239,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_unordered_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_unordered_store:
@@ -7171,6 +7252,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_unordered_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -7275,6 +7357,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_store(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_monotonic_store:
@@ -7287,6 +7370,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -7397,6 +7481,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_release_store:
@@ -7409,6 +7494,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -7519,6 +7605,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_seq_cst_store:
@@ -7531,6 +7618,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32* %out) {
 entry:
@@ -7635,6 +7723,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_monotonic_atomicrmw:
@@ -7647,6 +7736,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -7771,6 +7861,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -7881,6 +7972,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_release_atomicrmw:
@@ -7893,6 +7985,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8023,6 +8116,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8153,6 +8247,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8281,6 +8376,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_acquire_ret_atomicrmw:
@@ -8295,6 +8391,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8430,6 +8527,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_acq_rel_ret_atomicrmw:
@@ -8444,6 +8542,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8579,6 +8678,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_seq_cst_ret_atomicrmw:
@@ -8593,6 +8693,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    flat_atomic_swap_b32 v2, v[0:1], v2 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in) {
 entry:
@@ -8711,6 +8812,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_monotonic_monotonic_cmpxchg:
@@ -8724,6 +8826,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -8863,6 +8966,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -8987,6 +9091,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_release_monotonic_cmpxchg:
@@ -9000,6 +9105,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9145,6 +9251,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9290,6 +9397,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9429,6 +9537,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9568,6 +9677,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9713,6 +9823,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -9858,6 +9969,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10003,6 +10115,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10148,6 +10261,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10293,6 +10407,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10438,6 +10553,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10583,6 +10699,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10728,6 +10845,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v[0:1], v[2:3] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -10872,6 +10990,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonicmonotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_monotonicmonotonic_ret_cmpxchg:
@@ -10887,6 +11006,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonicmonotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11039,6 +11159,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_acquire_monotonic_ret_cmpxchg:
@@ -11054,6 +11175,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11206,6 +11328,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_release_monotonic_ret_cmpxchg:
@@ -11221,6 +11344,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11379,6 +11503,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_acq_rel_monotonic_ret_cmpxchg:
@@ -11394,6 +11519,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11552,6 +11678,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_seq_cst_monotonic_ret_cmpxchg:
@@ -11567,6 +11694,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11719,6 +11847,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_monotonic_acquire_ret_cmpxchg:
@@ -11734,6 +11863,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -11886,6 +12016,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_acquire_acquire_ret_cmpxchg:
@@ -11901,6 +12032,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12059,6 +12191,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_release_acquire_ret_cmpxchg:
@@ -12074,6 +12207,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12232,6 +12366,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_acq_rel_acquire_ret_cmpxchg:
@@ -12247,6 +12382,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12405,6 +12541,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_seq_cst_acquire_ret_cmpxchg:
@@ -12420,6 +12557,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12578,6 +12716,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_monotonic_seq_cst_ret_cmpxchg:
@@ -12593,6 +12732,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12751,6 +12891,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_acquire_seq_cst_ret_cmpxchg:
@@ -12766,6 +12907,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -12924,6 +13066,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_release_seq_cst_ret_cmpxchg:
@@ -12939,6 +13082,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -13097,6 +13241,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_acq_rel_seq_cst_ret_cmpxchg:
@@ -13112,6 +13257,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
@@ -13270,6 +13416,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: flat_workgroup_one_as_seq_cst_seq_cst_ret_cmpxchg:
@@ -13285,6 +13432,7 @@ define amdgpu_kernel void @flat_workgroup_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    flat_atomic_cmpswap_b32 v2, v[0:1], v[2:3] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    flat_store_b32 v[0:1], v2
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32* %out, i32 %in, i32 %old) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-agent.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-agent.ll
index 44af934e17fc8e..93f7df0f0ec703 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-agent.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-agent.ll
@@ -123,6 +123,7 @@ define amdgpu_kernel void @global_agent_unordered_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_unordered_load:
@@ -133,6 +134,7 @@ define amdgpu_kernel void @global_agent_unordered_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -253,6 +255,7 @@ define amdgpu_kernel void @global_agent_monotonic_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_monotonic_load:
@@ -263,6 +266,7 @@ define amdgpu_kernel void @global_agent_monotonic_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -395,6 +399,7 @@ define amdgpu_kernel void @global_agent_acquire_load(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_acquire_load:
@@ -407,6 +412,7 @@ define amdgpu_kernel void @global_agent_acquire_load(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -545,6 +551,7 @@ define amdgpu_kernel void @global_agent_seq_cst_load(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_seq_cst_load:
@@ -558,6 +565,7 @@ define amdgpu_kernel void @global_agent_seq_cst_load(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -671,6 +679,7 @@ define amdgpu_kernel void @global_agent_unordered_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_unordered_store:
@@ -682,6 +691,7 @@ define amdgpu_kernel void @global_agent_unordered_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -794,6 +804,7 @@ define amdgpu_kernel void @global_agent_monotonic_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_monotonic_store:
@@ -805,6 +816,7 @@ define amdgpu_kernel void @global_agent_monotonic_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -932,6 +944,7 @@ define amdgpu_kernel void @global_agent_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_release_store:
@@ -945,6 +958,7 @@ define amdgpu_kernel void @global_agent_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -1072,6 +1086,7 @@ define amdgpu_kernel void @global_agent_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_seq_cst_store:
@@ -1085,6 +1100,7 @@ define amdgpu_kernel void @global_agent_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -1197,6 +1213,7 @@ define amdgpu_kernel void @global_agent_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_monotonic_atomicrmw:
@@ -1208,6 +1225,7 @@ define amdgpu_kernel void @global_agent_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1483,6 +1501,7 @@ define amdgpu_kernel void @global_agent_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_release_atomicrmw:
@@ -1496,6 +1515,7 @@ define amdgpu_kernel void @global_agent_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1970,6 +1990,7 @@ define amdgpu_kernel void @global_agent_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_acquire_ret_atomicrmw:
@@ -1985,6 +2006,7 @@ define amdgpu_kernel void @global_agent_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2145,6 +2167,7 @@ define amdgpu_kernel void @global_agent_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_acq_rel_ret_atomicrmw:
@@ -2162,6 +2185,7 @@ define amdgpu_kernel void @global_agent_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2322,6 +2346,7 @@ define amdgpu_kernel void @global_agent_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_seq_cst_ret_atomicrmw:
@@ -2339,6 +2364,7 @@ define amdgpu_kernel void @global_agent_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2460,6 +2486,7 @@ define amdgpu_kernel void @global_agent_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_monotonic_monotonic_cmpxchg:
@@ -2472,6 +2499,7 @@ define amdgpu_kernel void @global_agent_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2766,6 +2794,7 @@ define amdgpu_kernel void @global_agent_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_release_monotonic_cmpxchg:
@@ -2780,6 +2809,7 @@ define amdgpu_kernel void @global_agent_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4989,6 +5019,7 @@ define amdgpu_kernel void @global_agent_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_monotonic_monotonic_ret_cmpxchg:
@@ -5003,6 +5034,7 @@ define amdgpu_kernel void @global_agent_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5160,6 +5192,7 @@ define amdgpu_kernel void @global_agent_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_acquire_monotonic_ret_cmpxchg:
@@ -5176,6 +5209,7 @@ define amdgpu_kernel void @global_agent_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5336,6 +5370,7 @@ define amdgpu_kernel void @global_agent_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_release_monotonic_ret_cmpxchg:
@@ -5352,6 +5387,7 @@ define amdgpu_kernel void @global_agent_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5524,6 +5560,7 @@ define amdgpu_kernel void @global_agent_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_acq_rel_monotonic_ret_cmpxchg:
@@ -5542,6 +5579,7 @@ define amdgpu_kernel void @global_agent_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5714,6 +5752,7 @@ define amdgpu_kernel void @global_agent_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_seq_cst_monotonic_ret_cmpxchg:
@@ -5732,6 +5771,7 @@ define amdgpu_kernel void @global_agent_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5889,6 +5929,7 @@ define amdgpu_kernel void @global_agent_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_monotonic_acquire_ret_cmpxchg:
@@ -5905,6 +5946,7 @@ define amdgpu_kernel void @global_agent_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6062,6 +6104,7 @@ define amdgpu_kernel void @global_agent_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_acquire_acquire_ret_cmpxchg:
@@ -6078,6 +6121,7 @@ define amdgpu_kernel void @global_agent_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6250,6 +6294,7 @@ define amdgpu_kernel void @global_agent_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_release_acquire_ret_cmpxchg:
@@ -6268,6 +6313,7 @@ define amdgpu_kernel void @global_agent_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6440,6 +6486,7 @@ define amdgpu_kernel void @global_agent_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_acq_rel_acquire_ret_cmpxchg:
@@ -6458,6 +6505,7 @@ define amdgpu_kernel void @global_agent_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6630,6 +6678,7 @@ define amdgpu_kernel void @global_agent_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_seq_cst_acquire_ret_cmpxchg:
@@ -6648,6 +6697,7 @@ define amdgpu_kernel void @global_agent_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6820,6 +6870,7 @@ define amdgpu_kernel void @global_agent_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_monotonic_seq_cst_ret_cmpxchg:
@@ -6838,6 +6889,7 @@ define amdgpu_kernel void @global_agent_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -7010,6 +7062,7 @@ define amdgpu_kernel void @global_agent_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_acquire_seq_cst_ret_cmpxchg:
@@ -7028,6 +7081,7 @@ define amdgpu_kernel void @global_agent_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -7200,6 +7254,7 @@ define amdgpu_kernel void @global_agent_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_release_seq_cst_ret_cmpxchg:
@@ -7218,6 +7273,7 @@ define amdgpu_kernel void @global_agent_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -7390,6 +7446,7 @@ define amdgpu_kernel void @global_agent_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_acq_rel_seq_cst_ret_cmpxchg:
@@ -7408,6 +7465,7 @@ define amdgpu_kernel void @global_agent_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -7580,6 +7638,7 @@ define amdgpu_kernel void @global_agent_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_seq_cst_seq_cst_ret_cmpxchg:
@@ -7598,6 +7657,7 @@ define amdgpu_kernel void @global_agent_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -7720,6 +7780,7 @@ define amdgpu_kernel void @global_agent_one_as_unordered_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_unordered_load:
@@ -7730,6 +7791,7 @@ define amdgpu_kernel void @global_agent_one_as_unordered_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -7850,6 +7912,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_monotonic_load:
@@ -7860,6 +7923,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -7992,6 +8056,7 @@ define amdgpu_kernel void @global_agent_one_as_acquire_load(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_acquire_load:
@@ -8004,6 +8069,7 @@ define amdgpu_kernel void @global_agent_one_as_acquire_load(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -8142,6 +8208,7 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_load(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_seq_cst_load:
@@ -8155,6 +8222,7 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_load(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -8268,6 +8336,7 @@ define amdgpu_kernel void @global_agent_one_as_unordered_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_unordered_store:
@@ -8279,6 +8348,7 @@ define amdgpu_kernel void @global_agent_one_as_unordered_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -8391,6 +8461,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_monotonic_store:
@@ -8402,6 +8473,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -8529,6 +8601,7 @@ define amdgpu_kernel void @global_agent_one_as_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_release_store:
@@ -8542,6 +8615,7 @@ define amdgpu_kernel void @global_agent_one_as_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -8669,6 +8743,7 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_seq_cst_store:
@@ -8682,6 +8757,7 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -8794,6 +8870,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_monotonic_atomicrmw:
@@ -8805,6 +8882,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -9080,6 +9158,7 @@ define amdgpu_kernel void @global_agent_one_as_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_release_atomicrmw:
@@ -9093,6 +9172,7 @@ define amdgpu_kernel void @global_agent_one_as_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -9567,6 +9647,7 @@ define amdgpu_kernel void @global_agent_one_as_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_acquire_ret_atomicrmw:
@@ -9582,6 +9663,7 @@ define amdgpu_kernel void @global_agent_one_as_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -9742,6 +9824,7 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_acq_rel_ret_atomicrmw:
@@ -9759,6 +9842,7 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -9919,6 +10003,7 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_seq_cst_ret_atomicrmw:
@@ -9936,6 +10021,7 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -10057,6 +10143,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_monotonic_monotonic_cmpxchg:
@@ -10069,6 +10156,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10363,6 +10451,7 @@ define amdgpu_kernel void @global_agent_one_as_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_release_monotonic_cmpxchg:
@@ -10377,6 +10466,7 @@ define amdgpu_kernel void @global_agent_one_as_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12586,6 +12676,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_monotonic_monotonic_ret_cmpxchg:
@@ -12600,6 +12691,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12757,6 +12849,7 @@ define amdgpu_kernel void @global_agent_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_acquire_monotonic_ret_cmpxchg:
@@ -12773,6 +12866,7 @@ define amdgpu_kernel void @global_agent_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12945,6 +13039,7 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_acq_rel_monotonic_ret_cmpxchg:
@@ -12963,6 +13058,7 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13135,6 +13231,7 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_seq_cst_monotonic_ret_cmpxchg:
@@ -13153,6 +13250,7 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13310,6 +13408,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_monotonic_acquire_ret_cmpxchg:
@@ -13326,6 +13425,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13483,6 +13583,7 @@ define amdgpu_kernel void @global_agent_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_acquire_acquire_ret_cmpxchg:
@@ -13499,6 +13600,7 @@ define amdgpu_kernel void @global_agent_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13671,6 +13773,7 @@ define amdgpu_kernel void @global_agent_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_release_acquire_ret_cmpxchg:
@@ -13689,6 +13792,7 @@ define amdgpu_kernel void @global_agent_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13861,6 +13965,7 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_acq_rel_acquire_ret_cmpxchg:
@@ -13879,6 +13984,7 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -14051,6 +14157,7 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_seq_cst_acquire_ret_cmpxchg:
@@ -14069,6 +14176,7 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -14241,6 +14349,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_monotonic_seq_cst_ret_cmpxchg:
@@ -14259,6 +14368,7 @@ define amdgpu_kernel void @global_agent_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -14431,6 +14541,7 @@ define amdgpu_kernel void @global_agent_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_acquire_seq_cst_ret_cmpxchg:
@@ -14449,6 +14560,7 @@ define amdgpu_kernel void @global_agent_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -14621,6 +14733,7 @@ define amdgpu_kernel void @global_agent_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_release_seq_cst_ret_cmpxchg:
@@ -14639,6 +14752,7 @@ define amdgpu_kernel void @global_agent_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -14811,6 +14925,7 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_acq_rel_seq_cst_ret_cmpxchg:
@@ -14829,6 +14944,7 @@ define amdgpu_kernel void @global_agent_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -15001,6 +15117,7 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_agent_one_as_seq_cst_seq_cst_ret_cmpxchg:
@@ -15019,6 +15136,7 @@ define amdgpu_kernel void @global_agent_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-nontemporal.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-nontemporal.ll
index c48e5e60cd4c6c..31f2c90ec42739 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-nontemporal.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-nontemporal.ll
@@ -127,6 +127,7 @@ define amdgpu_kernel void @global_nontemporal_load_0(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_nontemporal_load_0:
@@ -138,6 +139,7 @@ define amdgpu_kernel void @global_nontemporal_load_0(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -271,6 +273,7 @@ define amdgpu_kernel void @global_nontemporal_load_1(
 ; GFX11-WGP-NEXT:    global_load_b32 v0, v0, s[0:1] slc dlc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_nontemporal_load_1:
@@ -282,6 +285,7 @@ define amdgpu_kernel void @global_nontemporal_load_1(
 ; GFX11-CU-NEXT:    global_load_b32 v0, v0, s[0:1] slc dlc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -408,6 +412,7 @@ define amdgpu_kernel void @global_nontemporal_store_0(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3] glc slc dlc
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_nontemporal_store_0:
@@ -419,6 +424,7 @@ define amdgpu_kernel void @global_nontemporal_store_0(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3] glc slc dlc
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -547,6 +553,7 @@ define amdgpu_kernel void @global_nontemporal_store_1(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3] glc slc dlc
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_nontemporal_store_1:
@@ -558,6 +565,7 @@ define amdgpu_kernel void @global_nontemporal_store_1(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3] glc slc dlc
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-singlethread.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-singlethread.ll
index b89c91a3cf28ba..dbea4adca0f193 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-singlethread.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-singlethread.ll
@@ -123,6 +123,7 @@ define amdgpu_kernel void @global_singlethread_unordered_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_unordered_load:
@@ -133,6 +134,7 @@ define amdgpu_kernel void @global_singlethread_unordered_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -253,6 +255,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_monotonic_load:
@@ -263,6 +266,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -383,6 +387,7 @@ define amdgpu_kernel void @global_singlethread_acquire_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acquire_load:
@@ -393,6 +398,7 @@ define amdgpu_kernel void @global_singlethread_acquire_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -513,6 +519,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_seq_cst_load:
@@ -523,6 +530,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -636,6 +644,7 @@ define amdgpu_kernel void @global_singlethread_unordered_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_unordered_store:
@@ -647,6 +656,7 @@ define amdgpu_kernel void @global_singlethread_unordered_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -759,6 +769,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_monotonic_store:
@@ -770,6 +781,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -882,6 +894,7 @@ define amdgpu_kernel void @global_singlethread_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_release_store:
@@ -893,6 +906,7 @@ define amdgpu_kernel void @global_singlethread_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -1005,6 +1019,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_seq_cst_store:
@@ -1016,6 +1031,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -1128,6 +1144,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_monotonic_atomicrmw:
@@ -1139,6 +1156,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1251,6 +1269,7 @@ define amdgpu_kernel void @global_singlethread_acquire_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acquire_atomicrmw:
@@ -1262,6 +1281,7 @@ define amdgpu_kernel void @global_singlethread_acquire_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1374,6 +1394,7 @@ define amdgpu_kernel void @global_singlethread_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_release_atomicrmw:
@@ -1385,6 +1406,7 @@ define amdgpu_kernel void @global_singlethread_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1497,6 +1519,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acq_rel_atomicrmw:
@@ -1508,6 +1531,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1620,6 +1644,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_seq_cst_atomicrmw:
@@ -1631,6 +1656,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1763,6 +1789,7 @@ define amdgpu_kernel void @global_singlethread_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acquire_ret_atomicrmw:
@@ -1776,6 +1803,7 @@ define amdgpu_kernel void @global_singlethread_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1909,6 +1937,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acq_rel_ret_atomicrmw:
@@ -1922,6 +1951,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2055,6 +2085,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_seq_cst_ret_atomicrmw:
@@ -2068,6 +2099,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2189,6 +2221,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_monotonic_monotonic_cmpxchg:
@@ -2201,6 +2234,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2322,6 +2356,7 @@ define amdgpu_kernel void @global_singlethread_acquire_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acquire_monotonic_cmpxchg:
@@ -2334,6 +2369,7 @@ define amdgpu_kernel void @global_singlethread_acquire_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2455,6 +2491,7 @@ define amdgpu_kernel void @global_singlethread_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_release_monotonic_cmpxchg:
@@ -2467,6 +2504,7 @@ define amdgpu_kernel void @global_singlethread_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2588,6 +2626,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acq_rel_monotonic_cmpxchg:
@@ -2600,6 +2639,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2721,6 +2761,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_seq_cst_monotonic_cmpxchg:
@@ -2733,6 +2774,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2854,6 +2896,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_monotonic_acquire_cmpxchg:
@@ -2866,6 +2909,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2987,6 +3031,7 @@ define amdgpu_kernel void @global_singlethread_acquire_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acquire_acquire_cmpxchg:
@@ -2999,6 +3044,7 @@ define amdgpu_kernel void @global_singlethread_acquire_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3120,6 +3166,7 @@ define amdgpu_kernel void @global_singlethread_release_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_release_acquire_cmpxchg:
@@ -3132,6 +3179,7 @@ define amdgpu_kernel void @global_singlethread_release_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3253,6 +3301,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acq_rel_acquire_cmpxchg:
@@ -3265,6 +3314,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3386,6 +3436,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_seq_cst_acquire_cmpxchg:
@@ -3398,6 +3449,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3519,6 +3571,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_monotonic_seq_cst_cmpxchg:
@@ -3531,6 +3584,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3652,6 +3706,7 @@ define amdgpu_kernel void @global_singlethread_acquire_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acquire_seq_cst_cmpxchg:
@@ -3664,6 +3719,7 @@ define amdgpu_kernel void @global_singlethread_acquire_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3785,6 +3841,7 @@ define amdgpu_kernel void @global_singlethread_release_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_release_seq_cst_cmpxchg:
@@ -3797,6 +3854,7 @@ define amdgpu_kernel void @global_singlethread_release_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3918,6 +3976,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acq_rel_seq_cst_cmpxchg:
@@ -3930,6 +3989,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4051,6 +4111,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_seq_cst_seq_cst_cmpxchg:
@@ -4063,6 +4124,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4206,6 +4268,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_monotonic_monotonic_ret_cmpxchg:
@@ -4220,6 +4283,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4365,6 +4429,7 @@ define amdgpu_kernel void @global_singlethread_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acquire_monotonic_ret_cmpxchg:
@@ -4379,6 +4444,7 @@ define amdgpu_kernel void @global_singlethread_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4524,6 +4590,7 @@ define amdgpu_kernel void @global_singlethread_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_release_monotonic_ret_cmpxchg:
@@ -4538,6 +4605,7 @@ define amdgpu_kernel void @global_singlethread_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4683,6 +4751,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acq_rel_monotonic_ret_cmpxchg:
@@ -4697,6 +4766,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4842,6 +4912,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_seq_cst_monotonic_ret_cmpxchg:
@@ -4856,6 +4927,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5001,6 +5073,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_monotonic_acquire_ret_cmpxchg:
@@ -5015,6 +5088,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5160,6 +5234,7 @@ define amdgpu_kernel void @global_singlethread_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acquire_acquire_ret_cmpxchg:
@@ -5174,6 +5249,7 @@ define amdgpu_kernel void @global_singlethread_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5319,6 +5395,7 @@ define amdgpu_kernel void @global_singlethread_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_release_acquire_ret_cmpxchg:
@@ -5333,6 +5410,7 @@ define amdgpu_kernel void @global_singlethread_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5478,6 +5556,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acq_rel_acquire_ret_cmpxchg:
@@ -5492,6 +5571,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5637,6 +5717,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_seq_cst_acquire_ret_cmpxchg:
@@ -5651,6 +5732,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5796,6 +5878,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_monotonic_seq_cst_ret_cmpxchg:
@@ -5810,6 +5893,7 @@ define amdgpu_kernel void @global_singlethread_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5955,6 +6039,7 @@ define amdgpu_kernel void @global_singlethread_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acquire_seq_cst_ret_cmpxchg:
@@ -5969,6 +6054,7 @@ define amdgpu_kernel void @global_singlethread_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6114,6 +6200,7 @@ define amdgpu_kernel void @global_singlethread_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_release_seq_cst_ret_cmpxchg:
@@ -6128,6 +6215,7 @@ define amdgpu_kernel void @global_singlethread_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6273,6 +6361,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_acq_rel_seq_cst_ret_cmpxchg:
@@ -6287,6 +6376,7 @@ define amdgpu_kernel void @global_singlethread_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6432,6 +6522,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_seq_cst_seq_cst_ret_cmpxchg:
@@ -6446,6 +6537,7 @@ define amdgpu_kernel void @global_singlethread_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6568,6 +6660,7 @@ define amdgpu_kernel void @global_singlethread_one_as_unordered_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_unordered_load:
@@ -6578,6 +6671,7 @@ define amdgpu_kernel void @global_singlethread_one_as_unordered_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -6698,6 +6792,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_monotonic_load:
@@ -6708,6 +6803,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -6828,6 +6924,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acquire_load:
@@ -6838,6 +6935,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -6958,6 +7056,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_seq_cst_load:
@@ -6968,6 +7067,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -7081,6 +7181,7 @@ define amdgpu_kernel void @global_singlethread_one_as_unordered_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_unordered_store:
@@ -7092,6 +7193,7 @@ define amdgpu_kernel void @global_singlethread_one_as_unordered_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7204,6 +7306,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_monotonic_store:
@@ -7215,6 +7318,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7327,6 +7431,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_release_store:
@@ -7338,6 +7443,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7450,6 +7556,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_seq_cst_store:
@@ -7461,6 +7568,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7573,6 +7681,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_monotonic_atomicrmw:
@@ -7584,6 +7693,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -7696,6 +7806,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acquire_atomicrmw:
@@ -7707,6 +7818,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -7819,6 +7931,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_release_atomicrmw:
@@ -7830,6 +7943,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -7942,6 +8056,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acq_rel_atomicrmw:
@@ -7953,6 +8068,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8065,6 +8181,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_seq_cst_atomicrmw:
@@ -8076,6 +8193,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8208,6 +8326,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acquire_ret_atomicrmw:
@@ -8221,6 +8340,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8354,6 +8474,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acq_rel_ret_atomicrmw:
@@ -8367,6 +8488,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8500,6 +8622,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_seq_cst_ret_atomicrmw:
@@ -8513,6 +8636,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8634,6 +8758,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_monotonic_cmpxch
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_monotonic_monotonic_cmpxchg:
@@ -8646,6 +8771,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_monotonic_cmpxch
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -8767,6 +8893,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acquire_monotonic_cmpxchg:
@@ -8779,6 +8906,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -8900,6 +9028,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_release_monotonic_cmpxchg:
@@ -8912,6 +9041,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9033,6 +9163,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acq_rel_monotonic_cmpxchg:
@@ -9045,6 +9176,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9166,6 +9298,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_seq_cst_monotonic_cmpxchg:
@@ -9178,6 +9311,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9299,6 +9433,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_monotonic_acquire_cmpxchg:
@@ -9311,6 +9446,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9432,6 +9568,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acquire_acquire_cmpxchg:
@@ -9444,6 +9581,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9565,6 +9703,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_release_acquire_cmpxchg:
@@ -9577,6 +9716,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9698,6 +9838,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acq_rel_acquire_cmpxchg:
@@ -9710,6 +9851,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9831,6 +9973,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_seq_cst_acquire_cmpxchg:
@@ -9843,6 +9986,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9964,6 +10108,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_monotonic_seq_cst_cmpxchg:
@@ -9976,6 +10121,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10097,6 +10243,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acquire_seq_cst_cmpxchg:
@@ -10109,6 +10256,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10230,6 +10378,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_release_seq_cst_cmpxchg:
@@ -10242,6 +10391,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10363,6 +10513,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acq_rel_seq_cst_cmpxchg:
@@ -10375,6 +10526,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10496,6 +10648,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_seq_cst_seq_cst_cmpxchg:
@@ -10508,6 +10661,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10651,6 +10805,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_monotonic_ret_cm
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_monotonic_monotonic_ret_cmpxchg:
@@ -10665,6 +10820,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_monotonic_ret_cm
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10810,6 +10966,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_monotonic_ret_cmpx
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acquire_monotonic_ret_cmpxchg:
@@ -10824,6 +10981,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_monotonic_ret_cmpx
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10969,6 +11127,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_monotonic_ret_cmpx
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_release_monotonic_ret_cmpxchg:
@@ -10983,6 +11142,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_monotonic_ret_cmpx
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11128,6 +11288,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_monotonic_ret_cmpx
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acq_rel_monotonic_ret_cmpxchg:
@@ -11142,6 +11303,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_monotonic_ret_cmpx
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11287,6 +11449,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_monotonic_ret_cmpx
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_seq_cst_monotonic_ret_cmpxchg:
@@ -11301,6 +11464,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_monotonic_ret_cmpx
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11446,6 +11610,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_acquire_ret_cmpx
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_monotonic_acquire_ret_cmpxchg:
@@ -11460,6 +11625,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_acquire_ret_cmpx
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11605,6 +11771,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_acquire_ret_cmpxch
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acquire_acquire_ret_cmpxchg:
@@ -11619,6 +11786,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_acquire_ret_cmpxch
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11764,6 +11932,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_acquire_ret_cmpxch
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_release_acquire_ret_cmpxchg:
@@ -11778,6 +11947,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_acquire_ret_cmpxch
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11923,6 +12093,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_acquire_ret_cmpxch
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acq_rel_acquire_ret_cmpxchg:
@@ -11937,6 +12108,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_acquire_ret_cmpxch
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12082,6 +12254,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_acquire_ret_cmpxch
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_seq_cst_acquire_ret_cmpxchg:
@@ -12096,6 +12269,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_acquire_ret_cmpxch
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12241,6 +12415,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_seq_cst_ret_cmpx
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_monotonic_seq_cst_ret_cmpxchg:
@@ -12255,6 +12430,7 @@ define amdgpu_kernel void @global_singlethread_one_as_monotonic_seq_cst_ret_cmpx
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12400,6 +12576,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_seq_cst_ret_cmpxch
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acquire_seq_cst_ret_cmpxchg:
@@ -12414,6 +12591,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acquire_seq_cst_ret_cmpxch
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12559,6 +12737,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_seq_cst_ret_cmpxch
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_release_seq_cst_ret_cmpxchg:
@@ -12573,6 +12752,7 @@ define amdgpu_kernel void @global_singlethread_one_as_release_seq_cst_ret_cmpxch
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12718,6 +12898,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_seq_cst_ret_cmpxch
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_acq_rel_seq_cst_ret_cmpxchg:
@@ -12732,6 +12913,7 @@ define amdgpu_kernel void @global_singlethread_one_as_acq_rel_seq_cst_ret_cmpxch
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12877,6 +13059,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_seq_cst_ret_cmpxch
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_singlethread_one_as_seq_cst_seq_cst_ret_cmpxchg:
@@ -12891,6 +13074,7 @@ define amdgpu_kernel void @global_singlethread_one_as_seq_cst_seq_cst_ret_cmpxch
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-system.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-system.ll
index 274ffbd1a6de51..d29df184e43bdf 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-system.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-system.ll
@@ -123,6 +123,7 @@ define amdgpu_kernel void @global_system_unordered_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_unordered_load:
@@ -133,6 +134,7 @@ define amdgpu_kernel void @global_system_unordered_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -253,6 +255,7 @@ define amdgpu_kernel void @global_system_monotonic_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_monotonic_load:
@@ -263,6 +266,7 @@ define amdgpu_kernel void @global_system_monotonic_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -397,6 +401,7 @@ define amdgpu_kernel void @global_system_acquire_load(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_acquire_load:
@@ -409,6 +414,7 @@ define amdgpu_kernel void @global_system_acquire_load(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -549,6 +555,7 @@ define amdgpu_kernel void @global_system_seq_cst_load(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_seq_cst_load:
@@ -562,6 +569,7 @@ define amdgpu_kernel void @global_system_seq_cst_load(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -675,6 +683,7 @@ define amdgpu_kernel void @global_system_unordered_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_unordered_store:
@@ -686,6 +695,7 @@ define amdgpu_kernel void @global_system_unordered_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -798,6 +808,7 @@ define amdgpu_kernel void @global_system_monotonic_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_monotonic_store:
@@ -809,6 +820,7 @@ define amdgpu_kernel void @global_system_monotonic_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -938,6 +950,7 @@ define amdgpu_kernel void @global_system_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_release_store:
@@ -951,6 +964,7 @@ define amdgpu_kernel void @global_system_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -1080,6 +1094,7 @@ define amdgpu_kernel void @global_system_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_seq_cst_store:
@@ -1093,6 +1108,7 @@ define amdgpu_kernel void @global_system_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -1205,6 +1221,7 @@ define amdgpu_kernel void @global_system_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_monotonic_atomicrmw:
@@ -1216,6 +1233,7 @@ define amdgpu_kernel void @global_system_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1495,6 +1513,7 @@ define amdgpu_kernel void @global_system_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_release_atomicrmw:
@@ -1508,6 +1527,7 @@ define amdgpu_kernel void @global_system_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1992,6 +2012,7 @@ define amdgpu_kernel void @global_system_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_acquire_ret_atomicrmw:
@@ -2007,6 +2028,7 @@ define amdgpu_kernel void @global_system_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2171,6 +2193,7 @@ define amdgpu_kernel void @global_system_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_acq_rel_ret_atomicrmw:
@@ -2188,6 +2211,7 @@ define amdgpu_kernel void @global_system_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2352,6 +2376,7 @@ define amdgpu_kernel void @global_system_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_seq_cst_ret_atomicrmw:
@@ -2369,6 +2394,7 @@ define amdgpu_kernel void @global_system_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2490,6 +2516,7 @@ define amdgpu_kernel void @global_system_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_monotonic_monotonic_cmpxchg:
@@ -2502,6 +2529,7 @@ define amdgpu_kernel void @global_system_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2800,6 +2828,7 @@ define amdgpu_kernel void @global_system_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_release_monotonic_cmpxchg:
@@ -2814,6 +2843,7 @@ define amdgpu_kernel void @global_system_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4351,6 +4381,7 @@ define amdgpu_kernel void @global_system_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_monotonic_monotonic_ret_cmpxchg:
@@ -4365,6 +4396,7 @@ define amdgpu_kernel void @global_system_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4524,6 +4556,7 @@ define amdgpu_kernel void @global_system_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_acquire_monotonic_ret_cmpxchg:
@@ -4540,6 +4573,7 @@ define amdgpu_kernel void @global_system_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4716,6 +4750,7 @@ define amdgpu_kernel void @global_system_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_acq_rel_monotonic_ret_cmpxchg:
@@ -4734,6 +4769,7 @@ define amdgpu_kernel void @global_system_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4910,6 +4946,7 @@ define amdgpu_kernel void @global_system_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_seq_cst_monotonic_ret_cmpxchg:
@@ -4928,6 +4965,7 @@ define amdgpu_kernel void @global_system_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5087,6 +5125,7 @@ define amdgpu_kernel void @global_system_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_monotonic_acquire_ret_cmpxchg:
@@ -5103,6 +5142,7 @@ define amdgpu_kernel void @global_system_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5262,6 +5302,7 @@ define amdgpu_kernel void @global_system_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_acquire_acquire_ret_cmpxchg:
@@ -5278,6 +5319,7 @@ define amdgpu_kernel void @global_system_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5454,6 +5496,7 @@ define amdgpu_kernel void @global_system_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_release_acquire_ret_cmpxchg:
@@ -5472,6 +5515,7 @@ define amdgpu_kernel void @global_system_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5648,6 +5692,7 @@ define amdgpu_kernel void @global_system_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_acq_rel_acquire_ret_cmpxchg:
@@ -5666,6 +5711,7 @@ define amdgpu_kernel void @global_system_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5842,6 +5888,7 @@ define amdgpu_kernel void @global_system_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_seq_cst_acquire_ret_cmpxchg:
@@ -5860,6 +5907,7 @@ define amdgpu_kernel void @global_system_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6036,6 +6084,7 @@ define amdgpu_kernel void @global_system_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_monotonic_seq_cst_ret_cmpxchg:
@@ -6054,6 +6103,7 @@ define amdgpu_kernel void @global_system_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6230,6 +6280,7 @@ define amdgpu_kernel void @global_system_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_acquire_seq_cst_ret_cmpxchg:
@@ -6248,6 +6299,7 @@ define amdgpu_kernel void @global_system_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6424,6 +6476,7 @@ define amdgpu_kernel void @global_system_relese_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_relese_seq_cst_ret_cmpxchg:
@@ -6442,6 +6495,7 @@ define amdgpu_kernel void @global_system_relese_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6618,6 +6672,7 @@ define amdgpu_kernel void @global_system_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_acq_rel_seq_cst_ret_cmpxchg:
@@ -6636,6 +6691,7 @@ define amdgpu_kernel void @global_system_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6812,6 +6868,7 @@ define amdgpu_kernel void @global_system_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_seq_cst_seq_cst_ret_cmpxchg:
@@ -6830,6 +6887,7 @@ define amdgpu_kernel void @global_system_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6952,6 +7010,7 @@ define amdgpu_kernel void @global_system_one_as_unordered_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_unordered_load:
@@ -6962,6 +7021,7 @@ define amdgpu_kernel void @global_system_one_as_unordered_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -7082,6 +7142,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_monotonic_load:
@@ -7092,6 +7153,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -7226,6 +7288,7 @@ define amdgpu_kernel void @global_system_one_as_acquire_load(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_acquire_load:
@@ -7238,6 +7301,7 @@ define amdgpu_kernel void @global_system_one_as_acquire_load(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -7378,6 +7442,7 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_load(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_seq_cst_load:
@@ -7391,6 +7456,7 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_load(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -7504,6 +7570,7 @@ define amdgpu_kernel void @global_system_one_as_unordered_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_unordered_store:
@@ -7515,6 +7582,7 @@ define amdgpu_kernel void @global_system_one_as_unordered_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7627,6 +7695,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_monotonic_store:
@@ -7638,6 +7707,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7767,6 +7837,7 @@ define amdgpu_kernel void @global_system_one_as_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_release_store:
@@ -7780,6 +7851,7 @@ define amdgpu_kernel void @global_system_one_as_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7909,6 +7981,7 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_seq_cst_store:
@@ -7922,6 +7995,7 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -8034,6 +8108,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_monotonic_atomicrmw:
@@ -8045,6 +8120,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8324,6 +8400,7 @@ define amdgpu_kernel void @global_system_one_as_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_release_atomicrmw:
@@ -8337,6 +8414,7 @@ define amdgpu_kernel void @global_system_one_as_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8821,6 +8899,7 @@ define amdgpu_kernel void @global_system_one_as_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_acquire_ret_atomicrmw:
@@ -8836,6 +8915,7 @@ define amdgpu_kernel void @global_system_one_as_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -9000,6 +9080,7 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_acq_rel_ret_atomicrmw:
@@ -9017,6 +9098,7 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -9181,6 +9263,7 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_seq_cst_ret_atomicrmw:
@@ -9198,6 +9281,7 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -9319,6 +9403,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_monotonic_monotonic_cmpxchg:
@@ -9331,6 +9416,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9629,6 +9715,7 @@ define amdgpu_kernel void @global_system_one_as_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_release_monotonic_cmpxchg:
@@ -9643,6 +9730,7 @@ define amdgpu_kernel void @global_system_one_as_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11896,6 +11984,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_monotonic_monotonic_ret_cmpxchg:
@@ -11910,6 +11999,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12069,6 +12159,7 @@ define amdgpu_kernel void @global_system_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_acquire_monotonic_ret_cmpxchg:
@@ -12085,6 +12176,7 @@ define amdgpu_kernel void @global_system_one_as_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12247,6 +12339,7 @@ define amdgpu_kernel void @global_system_one_as_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_release_monotonic_ret_cmpxchg:
@@ -12263,6 +12356,7 @@ define amdgpu_kernel void @global_system_one_as_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12439,6 +12533,7 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_acq_rel_monotonic_ret_cmpxchg:
@@ -12457,6 +12552,7 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12633,6 +12729,7 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_seq_cst_monotonic_ret_cmpxchg:
@@ -12651,6 +12748,7 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12810,6 +12908,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_monotonic_acquire_ret_cmpxchg:
@@ -12826,6 +12925,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12985,6 +13085,7 @@ define amdgpu_kernel void @global_system_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_acquire_acquire_ret_cmpxchg:
@@ -13001,6 +13102,7 @@ define amdgpu_kernel void @global_system_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13177,6 +13279,7 @@ define amdgpu_kernel void @global_system_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_release_acquire_ret_cmpxchg:
@@ -13195,6 +13298,7 @@ define amdgpu_kernel void @global_system_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13371,6 +13475,7 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_acq_rel_acquire_ret_cmpxchg:
@@ -13389,6 +13494,7 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13565,6 +13671,7 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_seq_cst_acquire_ret_cmpxchg:
@@ -13583,6 +13690,7 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13759,6 +13867,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_monotonic_seq_cst_ret_cmpxchg:
@@ -13777,6 +13886,7 @@ define amdgpu_kernel void @global_system_one_as_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13953,6 +14063,7 @@ define amdgpu_kernel void @global_system_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_acquire_seq_cst_ret_cmpxchg:
@@ -13971,6 +14082,7 @@ define amdgpu_kernel void @global_system_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -14147,6 +14259,7 @@ define amdgpu_kernel void @global_system_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_release_seq_cst_ret_cmpxchg:
@@ -14165,6 +14278,7 @@ define amdgpu_kernel void @global_system_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -14341,6 +14455,7 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_acq_rel_seq_cst_ret_cmpxchg:
@@ -14359,6 +14474,7 @@ define amdgpu_kernel void @global_system_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -14535,6 +14651,7 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    buffer_gl1_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_system_one_as_seq_cst_seq_cst_ret_cmpxchg:
@@ -14553,6 +14670,7 @@ define amdgpu_kernel void @global_system_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    buffer_gl0_inv
 ; GFX11-CU-NEXT:    buffer_gl1_inv
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-volatile.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-volatile.ll
index fa87daac9880cc..bad7c6642084c3 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-volatile.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-volatile.ll
@@ -79,6 +79,7 @@ define amdgpu_kernel void @global_volatile_load_0(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] glc dlc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_volatile_load_0:
@@ -89,6 +90,7 @@ define amdgpu_kernel void @global_volatile_load_0(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1] glc dlc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -178,6 +180,7 @@ define amdgpu_kernel void @global_volatile_load_1(
 ; GFX11-WGP-NEXT:    global_load_b32 v0, v0, s[0:1] glc dlc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_volatile_load_1:
@@ -189,6 +192,7 @@ define amdgpu_kernel void @global_volatile_load_1(
 ; GFX11-CU-NEXT:    global_load_b32 v0, v0, s[0:1] glc dlc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v1, v0, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -277,6 +281,7 @@ define amdgpu_kernel void @global_volatile_store_0(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3] dlc
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_volatile_store_0:
@@ -289,6 +294,7 @@ define amdgpu_kernel void @global_volatile_store_0(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3] dlc
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -379,6 +385,7 @@ define amdgpu_kernel void @global_volatile_store_1(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3] dlc
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_volatile_store_1:
@@ -391,6 +398,7 @@ define amdgpu_kernel void @global_volatile_store_1(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s0
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3] dlc
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -475,6 +483,7 @@ define amdgpu_kernel void @global_volatile_workgroup_acquire_load(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_volatile_workgroup_acquire_load:
@@ -485,6 +494,7 @@ define amdgpu_kernel void @global_volatile_workgroup_acquire_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -566,6 +576,7 @@ define amdgpu_kernel void @global_volatile_workgroup_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_volatile_workgroup_release_store:
@@ -578,6 +589,7 @@ define amdgpu_kernel void @global_volatile_workgroup_release_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
    i32 %in, i32 addrspace(1)* %out) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-wavefront.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-wavefront.ll
index 037ee7c38fd601..ef63e4c7d3919b 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-wavefront.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-wavefront.ll
@@ -123,6 +123,7 @@ define amdgpu_kernel void @global_wavefront_unordered_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_unordered_load:
@@ -133,6 +134,7 @@ define amdgpu_kernel void @global_wavefront_unordered_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -253,6 +255,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_monotonic_load:
@@ -263,6 +266,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -383,6 +387,7 @@ define amdgpu_kernel void @global_wavefront_acquire_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acquire_load:
@@ -393,6 +398,7 @@ define amdgpu_kernel void @global_wavefront_acquire_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -513,6 +519,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_seq_cst_load:
@@ -523,6 +530,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -636,6 +644,7 @@ define amdgpu_kernel void @global_wavefront_unordered_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_unordered_store:
@@ -647,6 +656,7 @@ define amdgpu_kernel void @global_wavefront_unordered_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -759,6 +769,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_monotonic_store:
@@ -770,6 +781,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -882,6 +894,7 @@ define amdgpu_kernel void @global_wavefront_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_release_store:
@@ -893,6 +906,7 @@ define amdgpu_kernel void @global_wavefront_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -1005,6 +1019,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_seq_cst_store:
@@ -1016,6 +1031,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -1128,6 +1144,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_monotonic_atomicrmw:
@@ -1139,6 +1156,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1251,6 +1269,7 @@ define amdgpu_kernel void @global_wavefront_acquire_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acquire_atomicrmw:
@@ -1262,6 +1281,7 @@ define amdgpu_kernel void @global_wavefront_acquire_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1374,6 +1394,7 @@ define amdgpu_kernel void @global_wavefront_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_release_atomicrmw:
@@ -1385,6 +1406,7 @@ define amdgpu_kernel void @global_wavefront_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1497,6 +1519,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acq_rel_atomicrmw:
@@ -1508,6 +1531,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1620,6 +1644,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_seq_cst_atomicrmw:
@@ -1631,6 +1656,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1763,6 +1789,7 @@ define amdgpu_kernel void @global_wavefront_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acquire_ret_atomicrmw:
@@ -1776,6 +1803,7 @@ define amdgpu_kernel void @global_wavefront_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1909,6 +1937,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acq_rel_ret_atomicrmw:
@@ -1922,6 +1951,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2055,6 +2085,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_seq_cst_ret_atomicrmw:
@@ -2068,6 +2099,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2189,6 +2221,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_monotonic_monotonic_cmpxchg:
@@ -2201,6 +2234,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2322,6 +2356,7 @@ define amdgpu_kernel void @global_wavefront_acquire_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acquire_monotonic_cmpxchg:
@@ -2334,6 +2369,7 @@ define amdgpu_kernel void @global_wavefront_acquire_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2455,6 +2491,7 @@ define amdgpu_kernel void @global_wavefront_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_release_monotonic_cmpxchg:
@@ -2467,6 +2504,7 @@ define amdgpu_kernel void @global_wavefront_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2588,6 +2626,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acq_rel_monotonic_cmpxchg:
@@ -2600,6 +2639,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2721,6 +2761,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_seq_cst_monotonic_cmpxchg:
@@ -2733,6 +2774,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2854,6 +2896,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_monotonic_acquire_cmpxchg:
@@ -2866,6 +2909,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2987,6 +3031,7 @@ define amdgpu_kernel void @global_wavefront_acquire_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acquire_acquire_cmpxchg:
@@ -2999,6 +3044,7 @@ define amdgpu_kernel void @global_wavefront_acquire_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3120,6 +3166,7 @@ define amdgpu_kernel void @global_wavefront_release_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_release_acquire_cmpxchg:
@@ -3132,6 +3179,7 @@ define amdgpu_kernel void @global_wavefront_release_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3253,6 +3301,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acq_rel_acquire_cmpxchg:
@@ -3265,6 +3314,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3386,6 +3436,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_seq_cst_acquire_cmpxchg:
@@ -3398,6 +3449,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3519,6 +3571,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_monotonic_seq_cst_cmpxchg:
@@ -3531,6 +3584,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3652,6 +3706,7 @@ define amdgpu_kernel void @global_wavefront_acquire_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acquire_seq_cst_cmpxchg:
@@ -3664,6 +3719,7 @@ define amdgpu_kernel void @global_wavefront_acquire_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3785,6 +3841,7 @@ define amdgpu_kernel void @global_wavefront_release_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_release_seq_cst_cmpxchg:
@@ -3797,6 +3854,7 @@ define amdgpu_kernel void @global_wavefront_release_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3918,6 +3976,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acq_rel_seq_cst_cmpxchg:
@@ -3930,6 +3989,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4051,6 +4111,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_seq_cst_seq_cst_cmpxchg:
@@ -4063,6 +4124,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4206,6 +4268,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_monotonic_monotonic_ret_cmpxchg:
@@ -4220,6 +4283,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
    i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4365,6 +4429,7 @@ define amdgpu_kernel void @global_wavefront_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acquire_monotonic_ret_cmpxchg:
@@ -4379,6 +4444,7 @@ define amdgpu_kernel void @global_wavefront_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4524,6 +4590,7 @@ define amdgpu_kernel void @global_wavefront_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_release_monotonic_ret_cmpxchg:
@@ -4538,6 +4605,7 @@ define amdgpu_kernel void @global_wavefront_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
    i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4683,6 +4751,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acq_rel_monotonic_ret_cmpxchg:
@@ -4697,6 +4766,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4842,6 +4912,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_seq_cst_monotonic_ret_cmpxchg:
@@ -4856,6 +4927,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5001,6 +5073,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_monotonic_acquire_ret_cmpxchg:
@@ -5015,6 +5088,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5160,6 +5234,7 @@ define amdgpu_kernel void @global_wavefront_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acquire_acquire_ret_cmpxchg:
@@ -5174,6 +5249,7 @@ define amdgpu_kernel void @global_wavefront_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5319,6 +5395,7 @@ define amdgpu_kernel void @global_wavefront_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_release_acquire_ret_cmpxchg:
@@ -5333,6 +5410,7 @@ define amdgpu_kernel void @global_wavefront_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5478,6 +5556,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acq_rel_acquire_ret_cmpxchg:
@@ -5492,6 +5571,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5637,6 +5717,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_seq_cst_acquire_ret_cmpxchg:
@@ -5651,6 +5732,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5796,6 +5878,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_monotonic_seq_cst_ret_cmpxchg:
@@ -5810,6 +5893,7 @@ define amdgpu_kernel void @global_wavefront_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5955,6 +6039,7 @@ define amdgpu_kernel void @global_wavefront_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acquire_seq_cst_ret_cmpxchg:
@@ -5969,6 +6054,7 @@ define amdgpu_kernel void @global_wavefront_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6114,6 +6200,7 @@ define amdgpu_kernel void @global_wavefront_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_release_seq_cst_ret_cmpxchg:
@@ -6128,6 +6215,7 @@ define amdgpu_kernel void @global_wavefront_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6273,6 +6361,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_acq_rel_seq_cst_ret_cmpxchg:
@@ -6287,6 +6376,7 @@ define amdgpu_kernel void @global_wavefront_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6432,6 +6522,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_seq_cst_seq_cst_ret_cmpxchg:
@@ -6446,6 +6537,7 @@ define amdgpu_kernel void @global_wavefront_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6568,6 +6660,7 @@ define amdgpu_kernel void @global_wavefront_one_as_unordered_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_unordered_load:
@@ -6578,6 +6671,7 @@ define amdgpu_kernel void @global_wavefront_one_as_unordered_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -6698,6 +6792,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_monotonic_load:
@@ -6708,6 +6803,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -6828,6 +6924,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acquire_load:
@@ -6838,6 +6935,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -6958,6 +7056,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_seq_cst_load:
@@ -6968,6 +7067,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -7081,6 +7181,7 @@ define amdgpu_kernel void @global_wavefront_one_as_unordered_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_unordered_store:
@@ -7092,6 +7193,7 @@ define amdgpu_kernel void @global_wavefront_one_as_unordered_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7204,6 +7306,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_monotonic_store:
@@ -7215,6 +7318,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7327,6 +7431,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_release_store:
@@ -7338,6 +7443,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7450,6 +7556,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_seq_cst_store:
@@ -7461,6 +7568,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7573,6 +7681,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_monotonic_atomicrmw:
@@ -7584,6 +7693,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -7696,6 +7806,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acquire_atomicrmw:
@@ -7707,6 +7818,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -7819,6 +7931,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_release_atomicrmw:
@@ -7830,6 +7943,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -7942,6 +8056,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acq_rel_atomicrmw:
@@ -7953,6 +8068,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8065,6 +8181,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_seq_cst_atomicrmw:
@@ -8076,6 +8193,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8208,6 +8326,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acquire_ret_atomicrmw:
@@ -8221,6 +8340,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8354,6 +8474,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acq_rel_ret_atomicrmw:
@@ -8367,6 +8488,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8500,6 +8622,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_seq_cst_ret_atomicrmw:
@@ -8513,6 +8636,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8634,6 +8758,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_monotonic_monotonic_cmpxchg:
@@ -8646,6 +8771,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -8767,6 +8893,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acquire_monotonic_cmpxchg:
@@ -8779,6 +8906,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -8900,6 +9028,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_release_monotonic_cmpxchg:
@@ -8912,6 +9041,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9033,6 +9163,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acq_rel_monotonic_cmpxchg:
@@ -9045,6 +9176,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9166,6 +9298,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_seq_cst_monotonic_cmpxchg:
@@ -9178,6 +9311,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9299,6 +9433,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_monotonic_acquire_cmpxchg:
@@ -9311,6 +9446,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9432,6 +9568,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acquire_acquire_cmpxchg:
@@ -9444,6 +9581,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9565,6 +9703,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_release_acquire_cmpxchg:
@@ -9577,6 +9716,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9698,6 +9838,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acq_rel_acquire_cmpxchg:
@@ -9710,6 +9851,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9831,6 +9973,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_acquire_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_seq_cst_acquire_cmpxchg:
@@ -9843,6 +9986,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9964,6 +10108,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_monotonic_seq_cst_cmpxchg:
@@ -9976,6 +10121,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10097,6 +10243,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acquire_seq_cst_cmpxchg:
@@ -10109,6 +10256,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10230,6 +10378,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_release_seq_cst_cmpxchg:
@@ -10242,6 +10391,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10363,6 +10513,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acq_rel_seq_cst_cmpxchg:
@@ -10375,6 +10526,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10496,6 +10648,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_seq_cst_seq_cst_cmpxchg:
@@ -10508,6 +10661,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10651,6 +10805,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_monotonic_ret_cmpxc
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_monotonic_monotonic_ret_cmpxchg:
@@ -10665,6 +10820,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_monotonic_ret_cmpxc
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10810,6 +10966,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_monotonic_ret_cmpxchg
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acquire_monotonic_ret_cmpxchg:
@@ -10824,6 +10981,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_monotonic_ret_cmpxchg
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10969,6 +11127,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_monotonic_ret_cmpxchg
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_release_monotonic_ret_cmpxchg:
@@ -10983,6 +11142,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_monotonic_ret_cmpxchg
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11128,6 +11288,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg:
@@ -11142,6 +11303,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_monotonic_ret_cmpxchg
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11287,6 +11449,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg:
@@ -11301,6 +11464,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_monotonic_ret_cmpxchg
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11446,6 +11610,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_acquire_ret_cmpxchg
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_monotonic_acquire_ret_cmpxchg:
@@ -11460,6 +11625,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_acquire_ret_cmpxchg
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11605,6 +11771,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acquire_acquire_ret_cmpxchg:
@@ -11619,6 +11786,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11764,6 +11932,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_release_acquire_ret_cmpxchg:
@@ -11778,6 +11947,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11923,6 +12093,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acq_rel_acquire_ret_cmpxchg:
@@ -11937,6 +12108,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12082,6 +12254,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_seq_cst_acquire_ret_cmpxchg:
@@ -12096,6 +12269,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12241,6 +12415,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg:
@@ -12255,6 +12430,7 @@ define amdgpu_kernel void @global_wavefront_one_as_monotonic_seq_cst_ret_cmpxchg
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12400,6 +12576,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acquire_seq_cst_ret_cmpxchg:
@@ -12414,6 +12591,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12559,6 +12737,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_release_seq_cst_ret_cmpxchg:
@@ -12573,6 +12752,7 @@ define amdgpu_kernel void @global_wavefront_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12718,6 +12898,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_acq_rel_seq_cst_ret_cmpxchg:
@@ -12732,6 +12913,7 @@ define amdgpu_kernel void @global_wavefront_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12877,6 +13059,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg:
@@ -12891,6 +13074,7 @@ define amdgpu_kernel void @global_wavefront_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-workgroup.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-workgroup.ll
index a849e583cb6178..252e50d58c72d2 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-workgroup.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-global-workgroup.ll
@@ -123,6 +123,7 @@ define amdgpu_kernel void @global_workgroup_unordered_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_unordered_load:
@@ -133,6 +134,7 @@ define amdgpu_kernel void @global_workgroup_unordered_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -253,6 +255,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_monotonic_load:
@@ -263,6 +266,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -387,6 +391,7 @@ define amdgpu_kernel void @global_workgroup_acquire_load(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_acquire_load:
@@ -397,6 +402,7 @@ define amdgpu_kernel void @global_workgroup_acquire_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -526,6 +532,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_load(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_seq_cst_load:
@@ -536,6 +543,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -649,6 +657,7 @@ define amdgpu_kernel void @global_workgroup_unordered_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_unordered_store:
@@ -660,6 +669,7 @@ define amdgpu_kernel void @global_workgroup_unordered_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -772,6 +782,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_monotonic_store:
@@ -783,6 +794,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -907,6 +919,7 @@ define amdgpu_kernel void @global_workgroup_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_release_store:
@@ -919,6 +932,7 @@ define amdgpu_kernel void @global_workgroup_release_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -1043,6 +1057,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_seq_cst_store:
@@ -1055,6 +1070,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_store(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -1167,6 +1183,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_monotonic_atomicrmw:
@@ -1178,6 +1195,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1309,6 +1327,7 @@ define amdgpu_kernel void @global_workgroup_acquire_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1433,6 +1452,7 @@ define amdgpu_kernel void @global_workgroup_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_release_atomicrmw:
@@ -1445,6 +1465,7 @@ define amdgpu_kernel void @global_workgroup_release_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1589,6 +1610,7 @@ define amdgpu_kernel void @global_workgroup_acq_rel_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1733,6 +1755,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_atomicrmw(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -1869,6 +1892,7 @@ define amdgpu_kernel void @global_workgroup_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_acquire_ret_atomicrmw:
@@ -1882,6 +1906,7 @@ define amdgpu_kernel void @global_workgroup_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2031,6 +2056,7 @@ define amdgpu_kernel void @global_workgroup_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_acq_rel_ret_atomicrmw:
@@ -2045,6 +2071,7 @@ define amdgpu_kernel void @global_workgroup_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2194,6 +2221,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_seq_cst_ret_atomicrmw:
@@ -2208,6 +2236,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -2329,6 +2358,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_monotonic_monotonic_cmpxchg:
@@ -2341,6 +2371,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2482,6 +2513,7 @@ define amdgpu_kernel void @global_workgroup_acquire_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2615,6 +2647,7 @@ define amdgpu_kernel void @global_workgroup_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_release_monotonic_cmpxchg:
@@ -2628,6 +2661,7 @@ define amdgpu_kernel void @global_workgroup_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2782,6 +2816,7 @@ define amdgpu_kernel void @global_workgroup_acq_rel_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -2936,6 +2971,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3077,6 +3113,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3218,6 +3255,7 @@ define amdgpu_kernel void @global_workgroup_acquire_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3372,6 +3410,7 @@ define amdgpu_kernel void @global_workgroup_release_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3526,6 +3565,7 @@ define amdgpu_kernel void @global_workgroup_acq_rel_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3680,6 +3720,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3834,6 +3875,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -3988,6 +4030,7 @@ define amdgpu_kernel void @global_workgroup_acquire_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4142,6 +4185,7 @@ define amdgpu_kernel void @global_workgroup_release_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4296,6 +4340,7 @@ define amdgpu_kernel void @global_workgroup_acq_rel_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4450,6 +4495,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4593,6 +4639,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_monotonic_monotonic_ret_cmpxchg:
@@ -4607,6 +4654,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4756,6 +4804,7 @@ define amdgpu_kernel void @global_workgroup_acquire_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_acquire_monotonic_ret_cmpxchg:
@@ -4770,6 +4819,7 @@ define amdgpu_kernel void @global_workgroup_acquire_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -4927,6 +4977,7 @@ define amdgpu_kernel void @global_workgroup_release_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_release_monotonic_ret_cmpxchg:
@@ -4942,6 +4993,7 @@ define amdgpu_kernel void @global_workgroup_release_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5103,6 +5155,7 @@ define amdgpu_kernel void @global_workgroup_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_acq_rel_monotonic_ret_cmpxchg:
@@ -5118,6 +5171,7 @@ define amdgpu_kernel void @global_workgroup_acq_rel_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5279,6 +5333,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_seq_cst_monotonic_ret_cmpxchg:
@@ -5294,6 +5349,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_monotonic_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5443,6 +5499,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_monotonic_acquire_ret_cmpxchg:
@@ -5457,6 +5514,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5606,6 +5664,7 @@ define amdgpu_kernel void @global_workgroup_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_acquire_acquire_ret_cmpxchg:
@@ -5620,6 +5679,7 @@ define amdgpu_kernel void @global_workgroup_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5781,6 +5841,7 @@ define amdgpu_kernel void @global_workgroup_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_release_acquire_ret_cmpxchg:
@@ -5796,6 +5857,7 @@ define amdgpu_kernel void @global_workgroup_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -5957,6 +6019,7 @@ define amdgpu_kernel void @global_workgroup_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_acq_rel_acquire_ret_cmpxchg:
@@ -5972,6 +6035,7 @@ define amdgpu_kernel void @global_workgroup_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6133,6 +6197,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_seq_cst_acquire_ret_cmpxchg:
@@ -6148,6 +6213,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6309,6 +6375,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_monotonic_seq_cst_ret_cmpxchg:
@@ -6324,6 +6391,7 @@ define amdgpu_kernel void @global_workgroup_monotonic_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6485,6 +6553,7 @@ define amdgpu_kernel void @global_workgroup_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_acquire_seq_cst_ret_cmpxchg:
@@ -6500,6 +6569,7 @@ define amdgpu_kernel void @global_workgroup_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6661,6 +6731,7 @@ define amdgpu_kernel void @global_workgroup_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_release_seq_cst_ret_cmpxchg:
@@ -6676,6 +6747,7 @@ define amdgpu_kernel void @global_workgroup_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -6837,6 +6909,7 @@ define amdgpu_kernel void @global_workgroup_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_acq_rel_seq_cst_ret_cmpxchg:
@@ -6852,6 +6925,7 @@ define amdgpu_kernel void @global_workgroup_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -7013,6 +7087,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_seq_cst_seq_cst_ret_cmpxchg:
@@ -7028,6 +7103,7 @@ define amdgpu_kernel void @global_workgroup_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -7150,6 +7226,7 @@ define amdgpu_kernel void @global_workgroup_one_as_unordered_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_unordered_load:
@@ -7160,6 +7237,7 @@ define amdgpu_kernel void @global_workgroup_one_as_unordered_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -7280,6 +7358,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_load(
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_monotonic_load:
@@ -7290,6 +7369,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -7414,6 +7494,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_load(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_acquire_load:
@@ -7424,6 +7505,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -7550,6 +7632,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_load(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_seq_cst_load:
@@ -7560,6 +7643,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_load(
 ; GFX11-CU-NEXT:    global_load_b32 v1, v0, s[0:1]
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[2:3]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -7673,6 +7757,7 @@ define amdgpu_kernel void @global_workgroup_one_as_unordered_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_unordered_store:
@@ -7684,6 +7769,7 @@ define amdgpu_kernel void @global_workgroup_one_as_unordered_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7796,6 +7882,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_store(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_monotonic_store:
@@ -7807,6 +7894,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -7925,6 +8013,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_release_store:
@@ -7936,6 +8025,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -8054,6 +8144,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_store(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_seq_cst_store:
@@ -8065,6 +8156,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_store(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 %in, i32 addrspace(1)* %out) {
 entry:
@@ -8177,6 +8269,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_monotonic_atomicrmw:
@@ -8188,6 +8281,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8319,6 +8413,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8437,6 +8532,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_release_atomicrmw:
@@ -8448,6 +8544,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8585,6 +8682,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8722,6 +8820,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_atomicrmw(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s2
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -8858,6 +8957,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_acquire_ret_atomicrmw:
@@ -8871,6 +8971,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -9014,6 +9115,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_acq_rel_ret_atomicrmw:
@@ -9027,6 +9129,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -9170,6 +9273,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_seq_cst_ret_atomicrmw:
@@ -9183,6 +9287,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_ret_atomicrmw(
 ; GFX11-CU-NEXT:    global_atomic_swap_b32 v1, v0, v1, s[0:1] glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v0, v1, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in) {
 entry:
@@ -9304,6 +9409,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_monotonic_monotonic_cmpxchg:
@@ -9316,6 +9422,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9457,6 +9564,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9584,6 +9692,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_monotonic_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_release_monotonic_cmpxchg:
@@ -9596,6 +9705,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9743,6 +9853,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -9890,6 +10001,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_monotonic_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10031,6 +10143,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10172,6 +10285,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10319,6 +10433,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10466,6 +10581,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10613,6 +10729,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_acquire_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10760,6 +10877,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -10907,6 +11025,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11054,6 +11173,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11201,6 +11321,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11348,6 +11469,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_seq_cst_cmpxchg(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v2, v[0:1], s[0:1] offset:16
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11491,6 +11613,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_monotonic_ret_cmpxc
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_monotonic_monotonic_ret_cmpxchg:
@@ -11505,6 +11628,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_monotonic_ret_cmpxc
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11654,6 +11778,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_monotonic_ret_cmpxchg
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_acquire_monotonic_ret_cmpxchg:
@@ -11668,6 +11793,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_monotonic_ret_cmpxchg
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11819,6 +11945,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_monotonic_ret_cmpxchg
 ; GFX11-WGP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_release_monotonic_ret_cmpxchg:
@@ -11833,6 +11960,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_monotonic_ret_cmpxchg
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -11988,6 +12116,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_monotonic_ret_cmpxchg
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_acq_rel_monotonic_ret_cmpxchg:
@@ -12002,6 +12131,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_monotonic_ret_cmpxchg
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12157,6 +12287,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_monotonic_ret_cmpxchg
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_seq_cst_monotonic_ret_cmpxchg:
@@ -12171,6 +12302,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_monotonic_ret_cmpxchg
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12320,6 +12452,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_acquire_ret_cmpxchg
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_monotonic_acquire_ret_cmpxchg:
@@ -12334,6 +12467,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_acquire_ret_cmpxchg
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12483,6 +12617,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_acquire_acquire_ret_cmpxchg:
@@ -12497,6 +12632,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12652,6 +12788,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_release_acquire_ret_cmpxchg:
@@ -12666,6 +12803,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12821,6 +12959,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_acq_rel_acquire_ret_cmpxchg:
@@ -12835,6 +12974,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -12990,6 +13130,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_seq_cst_acquire_ret_cmpxchg:
@@ -13004,6 +13145,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_acquire_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13159,6 +13301,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_seq_cst_ret_cmpxchg
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_monotonic_seq_cst_ret_cmpxchg:
@@ -13173,6 +13316,7 @@ define amdgpu_kernel void @global_workgroup_one_as_monotonic_seq_cst_ret_cmpxchg
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13328,6 +13472,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_acquire_seq_cst_ret_cmpxchg:
@@ -13342,6 +13487,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acquire_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13497,6 +13643,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_release_seq_cst_ret_cmpxchg:
@@ -13511,6 +13658,7 @@ define amdgpu_kernel void @global_workgroup_one_as_release_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13666,6 +13814,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_acq_rel_seq_cst_ret_cmpxchg:
@@ -13680,6 +13829,7 @@ define amdgpu_kernel void @global_workgroup_one_as_acq_rel_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
@@ -13835,6 +13985,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    buffer_gl0_inv
 ; GFX11-WGP-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: global_workgroup_one_as_seq_cst_seq_cst_ret_cmpxchg:
@@ -13849,6 +14000,7 @@ define amdgpu_kernel void @global_workgroup_one_as_seq_cst_seq_cst_ret_cmpxchg(
 ; GFX11-CU-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] offset:16 glc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v2, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %out, i32 %in, i32 %old) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-nontemporal.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-nontemporal.ll
index 80d66c916f2d85..8ce2a0e535b8a1 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-nontemporal.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-nontemporal.ll
@@ -139,6 +139,7 @@ define amdgpu_kernel void @local_nontemporal_load_0(
 ; GFX11-WGP-NEXT:    ds_load_b32 v0, v0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: local_nontemporal_load_0:
@@ -152,6 +153,7 @@ define amdgpu_kernel void @local_nontemporal_load_0(
 ; GFX11-CU-NEXT:    ds_load_b32 v0, v0
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(3)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -291,6 +293,7 @@ define amdgpu_kernel void @local_nontemporal_load_1(
 ; GFX11-WGP-NEXT:    ds_load_b32 v0, v0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: local_nontemporal_load_1:
@@ -304,6 +307,7 @@ define amdgpu_kernel void @local_nontemporal_load_1(
 ; GFX11-CU-NEXT:    ds_load_b32 v0, v0
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(3)* %in, i32 addrspace(1)* %out) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-volatile.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-volatile.ll
index 98e0e013be5b8d..fbfeedbedb2d57 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-volatile.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-local-volatile.ll
@@ -87,6 +87,7 @@ define amdgpu_kernel void @local_volatile_load_0(
 ; GFX11-WGP-NEXT:    ds_load_b32 v0, v0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: local_volatile_load_0:
@@ -100,6 +101,7 @@ define amdgpu_kernel void @local_volatile_load_0(
 ; GFX11-CU-NEXT:    ds_load_b32 v0, v0
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(3)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -191,6 +193,7 @@ define amdgpu_kernel void @local_volatile_load_1(
 ; GFX11-WGP-NEXT:    ds_load_b32 v0, v0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: local_volatile_load_1:
@@ -204,6 +207,7 @@ define amdgpu_kernel void @local_volatile_load_1(
 ; GFX11-CU-NEXT:    ds_load_b32 v0, v0
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(3)* %in, i32 addrspace(1)* %out) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-private-nontemporal.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-private-nontemporal.ll
index 79d935bf1334a4..c15b70150c0dbb 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-private-nontemporal.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-private-nontemporal.ll
@@ -162,6 +162,7 @@ define amdgpu_kernel void @private_nontemporal_load_0(
 ; GFX11-WGP-NEXT:    scratch_load_b32 v0, off, s2 slc dlc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: private_nontemporal_load_0:
@@ -174,6 +175,7 @@ define amdgpu_kernel void @private_nontemporal_load_0(
 ; GFX11-CU-NEXT:    scratch_load_b32 v0, off, s2 slc dlc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(5)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -338,6 +340,7 @@ define amdgpu_kernel void @private_nontemporal_load_1(
 ; GFX11-WGP-NEXT:    scratch_load_b32 v0, v0, s2 slc dlc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: private_nontemporal_load_1:
@@ -350,6 +353,7 @@ define amdgpu_kernel void @private_nontemporal_load_1(
 ; GFX11-CU-NEXT:    scratch_load_b32 v0, v0, s2 slc dlc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(5)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -509,6 +513,7 @@ define amdgpu_kernel void @private_nontemporal_store_0(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s1
 ; GFX11-WGP-NEXT:    scratch_store_b32 off, v0, s0 glc slc dlc
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: private_nontemporal_store_0:
@@ -521,6 +526,7 @@ define amdgpu_kernel void @private_nontemporal_store_0(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s1
 ; GFX11-CU-NEXT:    scratch_store_b32 off, v0, s0 glc slc dlc
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(5)* %out) {
 entry:
@@ -684,6 +690,7 @@ define amdgpu_kernel void @private_nontemporal_store_1(
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX11-WGP-NEXT:    scratch_store_b32 v0, v1, s0 glc slc dlc
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: private_nontemporal_store_1:
@@ -697,6 +704,7 @@ define amdgpu_kernel void @private_nontemporal_store_1(
 ; GFX11-CU-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX11-CU-NEXT:    scratch_store_b32 v0, v1, s0 glc slc dlc
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(5)* %out) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/memory-legalizer-private-volatile.ll b/llvm/test/CodeGen/AMDGPU/memory-legalizer-private-volatile.ll
index e0dfca637564cc..194a248307cd02 100644
--- a/llvm/test/CodeGen/AMDGPU/memory-legalizer-private-volatile.ll
+++ b/llvm/test/CodeGen/AMDGPU/memory-legalizer-private-volatile.ll
@@ -106,6 +106,7 @@ define amdgpu_kernel void @private_volatile_load_0(
 ; GFX11-WGP-NEXT:    scratch_load_b32 v0, off, s2 glc dlc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: private_volatile_load_0:
@@ -118,6 +119,7 @@ define amdgpu_kernel void @private_volatile_load_0(
 ; GFX11-CU-NEXT:    scratch_load_b32 v0, off, s2 glc dlc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(5)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -228,6 +230,7 @@ define amdgpu_kernel void @private_volatile_load_1(
 ; GFX11-WGP-NEXT:    scratch_load_b32 v0, v0, s2 glc dlc
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-WGP-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: private_volatile_load_1:
@@ -240,6 +243,7 @@ define amdgpu_kernel void @private_volatile_load_1(
 ; GFX11-CU-NEXT:    scratch_load_b32 v0, v0, s2 glc dlc
 ; GFX11-CU-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-CU-NEXT:    global_store_b32 v1, v0, s[0:1]
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(5)* %in, i32 addrspace(1)* %out) {
 entry:
@@ -353,6 +357,7 @@ define amdgpu_kernel void @private_volatile_store_0(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, s1
 ; GFX11-WGP-NEXT:    scratch_store_b32 off, v0, s0 dlc
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: private_volatile_store_0:
@@ -366,6 +371,7 @@ define amdgpu_kernel void @private_volatile_store_0(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v0, s1
 ; GFX11-CU-NEXT:    scratch_store_b32 off, v0, s0 dlc
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(5)* %out) {
 entry:
@@ -481,6 +487,7 @@ define amdgpu_kernel void @private_volatile_store_1(
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX11-WGP-NEXT:    scratch_store_b32 v0, v1, s0 dlc
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-WGP-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-WGP-NEXT:    s_endpgm
 ;
 ; GFX11-CU-LABEL: private_volatile_store_1:
@@ -495,6 +502,7 @@ define amdgpu_kernel void @private_volatile_store_1(
 ; GFX11-CU-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX11-CU-NEXT:    scratch_store_b32 v0, v1, s0 dlc
 ; GFX11-CU-NEXT:    s_waitcnt_vscnt null, 0x0
+; GFX11-CU-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; GFX11-CU-NEXT:    s_endpgm
     i32 addrspace(1)* %in, i32 addrspace(5)* %out) {
 entry:
diff --git a/llvm/test/CodeGen/AMDGPU/release-vgprs.mir b/llvm/test/CodeGen/AMDGPU/release-vgprs.mir
new file mode 100644
index 00000000000000..258377f5668d10
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/release-vgprs.mir
@@ -0,0 +1,411 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
+
+# RUN: llc -march=amdgcn -mcpu=gfx1100 -run-pass=release-vgprs -verify-machineinstrs -o - %s | FileCheck %s
+
+--- |
+  define amdgpu_ps void @tbuffer_store1() { ret void }
+  define amdgpu_ps void @tbuffer_store2() { ret void }
+  define amdgpu_ps void @flat_store() { ret void }
+  define amdgpu_ps void @global_store() { ret void }
+  define amdgpu_ps void @buffer_store_format() { ret void }
+  define amdgpu_ps void @ds_write_b32() { ret void }
+  define amdgpu_ps void @global_store_dword() { ret void }
+  define amdgpu_ps void @multiple_basic_blocks1() { ret void }
+  define amdgpu_ps void @multiple_basic_blocks2() { ret void }
+  define amdgpu_ps void @multiple_basic_blocks3() { ret void }
+  define amdgpu_ps void @recursive_loop() { ret void }
+  define amdgpu_ps void @recursive_loop_vmem() { ret void }
+  define amdgpu_ps void @image_store() { ret void }
+  define amdgpu_ps void @scratch_store() { ret void }
+  define amdgpu_ps void @buffer_atomic() { ret void }
+  define amdgpu_ps void @flat_atomic() { ret void }
+  define amdgpu_ps void @global_atomic() { ret void }
+  define amdgpu_ps void @image_atomic() { ret void }
+...
+
+---
+name:            tbuffer_store1
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: tbuffer_store1
+    ; CHECK: TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, killed renamable $sgpr4, 42, 117, 0, 0, 0, implicit $exec
+    ; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0
+    ; CHECK-NEXT: S_ENDPGM 0
+    TBUFFER_STORE_FORMAT_XYZW_OFFSET_exact killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, killed renamable $sgpr4, 42, 117, 0, 0, 0, implicit $exec
+    S_ENDPGM 0
+...
+
+---
+name:            tbuffer_store2
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: tbuffer_store2
+    ; CHECK: TBUFFER_STORE_FORMAT_XYZW_OFFEN_exact killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, killed renamable $vgpr4, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 115, 0, 0, 0, implicit $exec :: (dereferenceable store (s128) into custom "BufferResource", align 1, addrspace 4)
+    ; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0
+    ; CHECK-NEXT: S_ENDPGM 0
+    TBUFFER_STORE_FORMAT_XYZW_OFFEN_exact killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, killed renamable $vgpr4, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 115, 0, 0, 0, implicit $exec :: (dereferenceable store (s128) into custom "BufferResource", align 1, addrspace 4)
+    S_ENDPGM 0
+...
+
+---
+name:            flat_store
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: flat_store
+    ; CHECK: FLAT_STORE_DWORDX4 $vgpr49_vgpr50, $vgpr26_vgpr27_vgpr28_vgpr29, 0, 0, implicit $exec, implicit $flat_scr
+    ; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0
+    ; CHECK-NEXT: S_ENDPGM 0
+    FLAT_STORE_DWORDX4 $vgpr49_vgpr50, $vgpr26_vgpr27_vgpr28_vgpr29, 0, 0, implicit $exec, implicit $flat_scr
+    S_ENDPGM 0
+...
+
+---
+name:            global_store
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: global_store
+    ; CHECK: GLOBAL_STORE_DWORD undef renamable $vgpr0_vgpr1, killed renamable $vgpr1, 0, 4, implicit $exec
+    ; CHECK-NEXT: S_WAITCNT_VSCNT undef $sgpr_null, 0
+    ; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0
+    ; CHECK-NEXT: S_ENDPGM 0
+    GLOBAL_STORE_DWORD undef renamable $vgpr0_vgpr1, killed renamable $vgpr1, 0, 4, implicit $exec
+    S_WAITCNT_VSCNT undef $sgpr_null, 0
+    S_ENDPGM 0
+...
+
+---
+name:            buffer_store_format
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: buffer_store_format
+    ; CHECK: BUFFER_STORE_FORMAT_D16_X_OFFEN_exact killed renamable $vgpr0, killed renamable $vgpr1, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, killed renamable $sgpr4, 0, 0, 0, 0, implicit $exec
+    ; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0
+    ; CHECK-NEXT: S_ENDPGM 0
+    BUFFER_STORE_FORMAT_D16_X_OFFEN_exact killed renamable $vgpr0, killed renamable $vgpr1, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, killed renamable $sgpr4, 0, 0, 0, 0, implicit $exec
+    S_ENDPGM 0
+...
+
+---
+name:            ds_write_b32
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: ds_write_b32
+    ; CHECK: renamable $vgpr0 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr1 = IMPLICIT_DEF
+    ; CHECK-NEXT: DS_WRITE_B32_gfx9 killed renamable $vgpr0, killed renamable $vgpr1, 12, 0, implicit $exec
+    ; CHECK-NEXT: S_ENDPGM 0
+    renamable $vgpr0 = IMPLICIT_DEF
+    renamable $vgpr1 = IMPLICIT_DEF
+    DS_WRITE_B32_gfx9 killed renamable $vgpr0, killed renamable $vgpr1, 12, 0, implicit $exec
+    S_ENDPGM 0
+
+...
+---
+name:            global_store_dword
+body:             |
+  bb.0:
+    liveins: $vgpr0, $sgpr0_sgpr1
+
+    ; CHECK-LABEL: name: global_store_dword
+    ; CHECK: liveins: $vgpr0, $sgpr0_sgpr1
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: renamable $vgpr0 = V_MAD_I32_I24_e64 killed $vgpr1, killed $vgpr0, killed $sgpr2, 0, implicit $exec
+    ; CHECK-NEXT: GLOBAL_STORE_DWORD_SADDR killed renamable $vgpr2, killed renamable $vgpr0, killed renamable $sgpr0_sgpr1, 0, 0, implicit $exec
+    ; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0
+    ; CHECK-NEXT: S_ENDPGM 0
+    renamable $vgpr0 = V_MAD_I32_I24_e64 killed $vgpr1, killed $vgpr0, killed $sgpr2, 0, implicit $exec
+    GLOBAL_STORE_DWORD_SADDR killed renamable $vgpr2, killed renamable $vgpr0, killed renamable $sgpr0_sgpr1, 0, 0, implicit $exec
+    S_ENDPGM 0
+...
+
+---
+name:            multiple_basic_blocks1
+body:             |
+  ; CHECK-LABEL: name: multiple_basic_blocks1
+  ; CHECK: bb.0:
+  ; CHECK-NEXT:   successors: %bb.1(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   renamable $vgpr0 = BUFFER_LOAD_FORMAT_X_IDXEN killed renamable $vgpr0, renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, 0, 0, implicit $exec
+  ; CHECK-NEXT:   S_BRANCH %bb.1
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.1:
+  ; CHECK-NEXT:   successors: %bb.1(0x40000000), %bb.2(0x40000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   $vgpr1 = V_ADD_U32_e32 $vgpr0, $vgpr2, implicit $exec
+  ; CHECK-NEXT:   S_CMP_LG_U32 killed renamable $sgpr3, renamable $sgpr4, implicit-def $scc
+  ; CHECK-NEXT:   S_CBRANCH_SCC1 %bb.1, implicit killed $scc
+  ; CHECK-NEXT:   S_BRANCH %bb.2
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.2:
+  ; CHECK-NEXT:   S_ENDPGM 0
+  bb.0:
+    successors: %bb.1
+
+    renamable $vgpr0 = BUFFER_LOAD_FORMAT_X_IDXEN killed renamable $vgpr0, renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, 0, 0, implicit $exec
+    S_BRANCH %bb.1
+
+  bb.1:
+    successors: %bb.1, %bb.2
+
+    $vgpr1 = V_ADD_U32_e32 renamable $vgpr0, renamable $vgpr2, implicit $exec
+    S_CMP_LG_U32 killed renamable $sgpr3, renamable $sgpr4, implicit-def $scc
+    S_CBRANCH_SCC1 %bb.1, implicit killed $scc
+    S_BRANCH %bb.2
+
+  bb.2:
+    S_ENDPGM 0
+
+...
+
+
+# One block has a VMEM store as the last instruction, we should release the VGPRS
+...
+---
+name:            multiple_basic_blocks2
+body:             |
+  ; CHECK-LABEL: name: multiple_basic_blocks2
+  ; CHECK: bb.0:
+  ; CHECK-NEXT:   successors: %bb.2(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   TBUFFER_STORE_FORMAT_X_OFFSET_exact killed renamable $vgpr0, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 125, 0, 0, 0, implicit $exec
+  ; CHECK-NEXT:   $vgpr1 = V_ADD_U32_e32 $vgpr0, $vgpr2, implicit $exec
+  ; CHECK-NEXT:   S_BRANCH %bb.2
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.1:
+  ; CHECK-NEXT:   successors: %bb.2(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   $vgpr1 = V_ADD_U32_e32 $vgpr0, $vgpr2, implicit $exec
+  ; CHECK-NEXT:   TBUFFER_STORE_FORMAT_X_OFFSET_exact killed renamable $vgpr0, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 125, 0, 0, 0, implicit $exec
+  ; CHECK-NEXT:   S_BRANCH %bb.2
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.2:
+  ; CHECK-NEXT:   S_SENDMSG 3, implicit $exec, implicit $m0
+  ; CHECK-NEXT:   S_ENDPGM 0
+  bb.0:
+    successors: %bb.2
+
+    TBUFFER_STORE_FORMAT_X_OFFSET_exact killed renamable $vgpr0, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 125, 0, 0, 0, implicit $exec
+    $vgpr1 = V_ADD_U32_e32 renamable $vgpr0, renamable $vgpr2, implicit $exec
+    S_BRANCH %bb.2
+
+  bb.1:
+    successors: %bb.2
+
+    $vgpr1 = V_ADD_U32_e32 renamable $vgpr0, renamable $vgpr2, implicit $exec
+    TBUFFER_STORE_FORMAT_X_OFFSET_exact killed renamable $vgpr0, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 125, 0, 0, 0, implicit $exec
+    S_BRANCH %bb.2
+
+  bb.2:
+    S_ENDPGM 0
+...
+
+
+# One parent block has a VMEM store, release VGPRs
+---
+name:            multiple_basic_blocks3
+body:             |
+  ; CHECK-LABEL: name: multiple_basic_blocks3
+  ; CHECK: bb.0:
+  ; CHECK-NEXT:   successors: %bb.2(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   $vgpr1 = V_ADD_U32_e32 $vgpr0, $vgpr2, implicit $exec
+  ; CHECK-NEXT:   TBUFFER_STORE_FORMAT_X_OFFSET_exact killed renamable $vgpr0, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 125, 0, 0, 0, implicit $exec
+  ; CHECK-NEXT:   S_BRANCH %bb.2
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.1:
+  ; CHECK-NEXT:   successors: %bb.2(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   $vgpr1 = V_ADD_U32_e32 $vgpr0, $vgpr2, implicit $exec
+  ; CHECK-NEXT:   S_BRANCH %bb.2
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.2:
+  ; CHECK-NEXT:   successors: %bb.4(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   S_BRANCH %bb.4
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.3:
+  ; CHECK-NEXT:   successors: %bb.4(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   $vgpr1 = V_ADD_U32_e32 $vgpr0, $vgpr2, implicit $exec
+  ; CHECK-NEXT:   S_BRANCH %bb.4
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.4:
+  ; CHECK-NEXT:   S_SENDMSG 3, implicit $exec, implicit $m0
+  ; CHECK-NEXT:   S_ENDPGM 0
+  bb.0:
+    successors: %bb.2
+
+    $vgpr1 = V_ADD_U32_e32 renamable $vgpr0, renamable $vgpr2, implicit $exec
+    TBUFFER_STORE_FORMAT_X_OFFSET_exact killed renamable $vgpr0, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 125, 0, 0, 0, implicit $exec
+    S_BRANCH %bb.2
+
+  bb.1:
+    successors: %bb.2
+
+    $vgpr1 = V_ADD_U32_e32 renamable $vgpr0, renamable $vgpr2, implicit $exec
+    S_BRANCH %bb.2
+
+  bb.2:
+    successors: %bb.4
+
+    S_BRANCH %bb.4
+
+  bb.3:
+    successors: %bb.4
+
+    $vgpr1 = V_ADD_U32_e32 renamable $vgpr0, renamable $vgpr2, implicit $exec
+    S_BRANCH %bb.4
+
+  bb.4:
+    S_ENDPGM 0
+...
+
+---
+name:            recursive_loop
+body:             |
+  ; CHECK-LABEL: name: recursive_loop
+  ; CHECK: bb.0:
+  ; CHECK-NEXT:   successors: %bb.1(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   renamable $vgpr0 = BUFFER_LOAD_FORMAT_X_IDXEN killed renamable $vgpr0, renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, 0, 0, implicit $exec
+  ; CHECK-NEXT:   S_BRANCH %bb.1
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.1:
+  ; CHECK-NEXT:   successors: %bb.1(0x40000000), %bb.2(0x40000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   S_CMP_LG_U32 killed renamable $sgpr3, renamable $sgpr4, implicit-def $scc
+  ; CHECK-NEXT:   S_CBRANCH_SCC1 %bb.1, implicit killed $scc
+  ; CHECK-NEXT:   S_BRANCH %bb.2
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.2:
+  ; CHECK-NEXT:   S_ENDPGM 0
+  bb.0:
+    successors: %bb.1
+
+    renamable $vgpr0 = BUFFER_LOAD_FORMAT_X_IDXEN killed renamable $vgpr0, renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, 0, 0, implicit $exec
+    S_BRANCH %bb.1
+
+  bb.1:
+    successors: %bb.1, %bb.2
+
+    S_CMP_LG_U32 killed renamable $sgpr3, renamable $sgpr4, implicit-def $scc
+    S_CBRANCH_SCC1 %bb.1, implicit killed $scc
+    S_BRANCH %bb.2
+
+  bb.2:
+    S_ENDPGM 0
+...
+
+---
+name:            recursive_loop_vmem
+body:             |
+  ; CHECK-LABEL: name: recursive_loop_vmem
+  ; CHECK: bb.0:
+  ; CHECK-NEXT:   successors: %bb.1(0x80000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   renamable $vgpr0 = BUFFER_LOAD_FORMAT_X_IDXEN killed renamable $vgpr0, renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, 0, 0, implicit $exec
+  ; CHECK-NEXT:   S_BRANCH %bb.1
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.1:
+  ; CHECK-NEXT:   successors: %bb.1(0x40000000), %bb.2(0x40000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   TBUFFER_STORE_FORMAT_XYZW_OFFEN_exact killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, killed renamable $vgpr4, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 115, 0, 0, 0, implicit $exec
+  ; CHECK-NEXT:   S_CMP_LG_U32 killed renamable $sgpr3, renamable $sgpr4, implicit-def $scc
+  ; CHECK-NEXT:   S_CBRANCH_SCC1 %bb.1, implicit killed $scc
+  ; CHECK-NEXT:   S_BRANCH %bb.2
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.2:
+  ; CHECK-NEXT:   S_SENDMSG 3, implicit $exec, implicit $m0
+  ; CHECK-NEXT:   S_ENDPGM 0
+  bb.0:
+    successors: %bb.1
+
+    renamable $vgpr0 = BUFFER_LOAD_FORMAT_X_IDXEN killed renamable $vgpr0, renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, 0, 0, implicit $exec
+    S_BRANCH %bb.1
+
+  bb.1:
+    successors: %bb.1, %bb.2
+
+    TBUFFER_STORE_FORMAT_XYZW_OFFEN_exact killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, killed renamable $vgpr4, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 115, 0, 0, 0, implicit $exec
+    S_CMP_LG_U32 killed renamable $sgpr3, renamable $sgpr4, implicit-def $scc
+    S_CBRANCH_SCC1 %bb.1, implicit killed $scc
+    S_BRANCH %bb.2
+
+  bb.2:
+    S_ENDPGM 0
+...
+
+---
+name:            image_store
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: image_store
+    ; CHECK: IMAGE_STORE_V2_V1_gfx11 killed renamable $vgpr0_vgpr1, killed renamable $vgpr2, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3_sgpr4_sgpr5_sgpr6_sgpr7, 12, 0, 1, 0, 0, -1, 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s32>) into custom "ImageResource")
+    ; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0
+    ; CHECK-NEXT: S_ENDPGM 0
+  IMAGE_STORE_V2_V1_gfx11 killed renamable $vgpr0_vgpr1, killed renamable $vgpr2, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3_sgpr4_sgpr5_sgpr6_sgpr7, 12, 0, 1, 0, 0, -1, 0, 0, 0, implicit $exec :: (dereferenceable store (<2 x s32>) into custom "ImageResource")
+  S_ENDPGM 0
+...
+
+---
+name:            scratch_store
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: scratch_store
+    ; CHECK: renamable $sgpr0 = S_AND_B32 killed renamable $sgpr0, -16, implicit-def dead $scc
+    ; CHECK-NEXT: SCRATCH_STORE_DWORD_SADDR killed renamable $vgpr0, killed renamable $sgpr0, 0, 0, implicit $exec, implicit $flat_scr
+    ; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0
+    ; CHECK-NEXT: S_ENDPGM 0
+    renamable $sgpr0 = S_AND_B32 killed renamable $sgpr0, -16, implicit-def dead $scc
+    SCRATCH_STORE_DWORD_SADDR killed renamable $vgpr0, killed renamable $sgpr0, 0, 0, implicit $exec, implicit $flat_scr
+    S_ENDPGM 0
+...
+
+---
+name:            buffer_atomic
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: buffer_atomic
+    ; CHECK: BUFFER_ATOMIC_ADD_F32_OFFEN killed renamable $vgpr0, killed renamable $vgpr2, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, implicit $exec :: (volatile dereferenceable load store (s64), align 1, addrspace 4)
+    ; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0
+    ; CHECK-NEXT: S_ENDPGM 0
+    BUFFER_ATOMIC_ADD_F32_OFFEN killed renamable $vgpr0, killed renamable $vgpr2, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, 0, 0, 0, implicit $exec :: (volatile dereferenceable load store (s64), align 1, addrspace 4)
+    S_ENDPGM 0
+...
+
+---
+name:            flat_atomic
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: flat_atomic
+    ; CHECK: renamable $vgpr0_vgpr1 = FLAT_ATOMIC_DEC_X2_RTN killed renamable $vgpr0_vgpr1, killed renamable $vgpr2_vgpr3, 40, 1, implicit $exec, implicit $flat_scr
+    ; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0
+    ; CHECK-NEXT: S_ENDPGM 0
+    renamable $vgpr0_vgpr1 = FLAT_ATOMIC_DEC_X2_RTN killed renamable $vgpr0_vgpr1, killed renamable $vgpr2_vgpr3, 40, 1, implicit $exec, implicit $flat_scr
+    S_ENDPGM 0
+...
+
+
+---
+name:            global_atomic
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: global_atomic
+    ; CHECK: renamable $vgpr0_vgpr1 = GLOBAL_ATOMIC_INC_X2_SADDR_RTN killed renamable $vgpr0, killed renamable $vgpr1_vgpr2, killed renamable $sgpr0_sgpr1, 40, 1, implicit $exec
+    ; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0
+    ; CHECK-NEXT: S_ENDPGM 0
+    renamable $vgpr0_vgpr1 = GLOBAL_ATOMIC_INC_X2_SADDR_RTN killed renamable $vgpr0, killed renamable $vgpr1_vgpr2, killed renamable $sgpr0_sgpr1, 40, 1, implicit $exec
+    S_ENDPGM 0
+...
+
+---
+name:            image_atomic
+body:             |
+  bb.0:
+    ; CHECK-LABEL: name: image_atomic
+    ; CHECK: renamable $vgpr0_vgpr1_vgpr2_vgpr3 = IMAGE_ATOMIC_CMPSWAP_V2_V1_gfx11 killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, killed renamable $vgpr4, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3_sgpr4_sgpr5_sgpr6_sgpr7, 15, 0, 1, 1, 0, 0, 0, 0, implicit $exec :: (volatile dereferenceable load store (s64) on custom "ImageResource")
+    ; CHECK-NEXT: S_SENDMSG 3, implicit $exec, implicit $m0
+    ; CHECK-NEXT: S_ENDPGM 0
+    renamable $vgpr0_vgpr1_vgpr2_vgpr3 = IMAGE_ATOMIC_CMPSWAP_V2_V1_gfx11 killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, killed renamable $vgpr4, killed renamable $sgpr0_sgpr1_sgpr2_sgpr3_sgpr4_sgpr5_sgpr6_sgpr7, 15, 0, 1, 1, 0, 0, 0, 0, implicit $exec :: (volatile dereferenceable load store (s64) on custom "ImageResource")
+    S_ENDPGM 0
+...
diff --git a/llvm/test/CodeGen/AMDGPU/wmma_multiple_32.ll b/llvm/test/CodeGen/AMDGPU/wmma_multiple_32.ll
index 6f4a69ee54d189..44af8f4074fc5d 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma_multiple_32.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma_multiple_32.ll
@@ -31,6 +31,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16(<8 x float> %A, <8 x float> %B
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[26:27], v[20:23], off offset:16
 ; W32-NEXT:    global_store_b128 v[26:27], v[16:19], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<8 x float> %A, <8 x float> %B, <8 x float> %C)
@@ -53,6 +54,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16(<8 x i32> %A, <8 x i32> %B, <
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[26:27], v[20:23], off offset:16
 ; W32-NEXT:    global_store_b128 v[26:27], v[16:19], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <8 x float> %C)
@@ -75,6 +77,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_lo(<8 x float> %A, <8 x float>
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[26:27], v[20:23], off offset:16
 ; W32-NEXT:    global_store_b128 v[26:27], v[16:19], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x float> @llvm.amdgcn.wmma.f16.16x16x16.f16(<8 x float> %A, <8 x float> %B, <8 x float> %C, i1 0)
@@ -95,6 +98,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_hi(<8 x float> %A, <8 x float>
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[26:27], v[20:23], off offset:16
 ; W32-NEXT:    global_store_b128 v[26:27], v[16:19], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x float> @llvm.amdgcn.wmma.f16.16x16x16.f16(<8 x float> %A, <8 x float> %B, <8 x float> %C, i1 1)
@@ -117,6 +121,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_lo(<8 x i32> %A, <8 x i32> %
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[26:27], v[20:23], off offset:16
 ; W32-NEXT:    global_store_b128 v[26:27], v[16:19], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <8 x i32> %C, i1 0)
@@ -137,6 +142,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_hi(<8 x i32> %A, <8 x i32> %
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[26:27], v[20:23], off offset:16
 ; W32-NEXT:    global_store_b128 v[26:27], v[16:19], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <8 x i32> %C, i1 1)
@@ -159,6 +165,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned(<4 x i32> %A
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[18:19], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[18:19], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
@@ -179,6 +186,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed(<4 x i32> %A,
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[18:19], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[18:19], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
@@ -199,6 +207,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned(<4 x i32> %A,
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[18:19], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[18:19], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 0)
@@ -219,6 +228,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed(<4 x i32> %A, <4
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[18:19], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[18:19], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 0)
@@ -239,6 +249,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned_clamp(<4 x i
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[18:19], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[18:19], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
@@ -259,6 +270,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed_clamp(<4 x i32
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[18:19], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[18:19], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
@@ -279,6 +291,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned_clamp(<4 x i32
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[18:19], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[18:19], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <8 x i32> %C, i1 1)
@@ -299,6 +312,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed_clamp(<4 x i32>
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[18:19], v[12:15], off offset:16
 ; W32-NEXT:    global_store_b128 v[18:19], v[8:11], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <8 x i32> %C, i1 1)
@@ -321,6 +335,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned(<2 x i32> %A
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[14:15], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[14:15], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
@@ -341,6 +356,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed(<2 x i32> %A,
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[14:15], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[14:15], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
@@ -361,6 +377,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned(<2 x i32> %A,
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[14:15], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[14:15], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 0)
@@ -381,6 +398,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed(<2 x i32> %A, <2
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[14:15], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[14:15], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 0)
@@ -402,6 +420,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned_clamp(<2 x i
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[14:15], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[14:15], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
@@ -422,6 +441,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed_clamp(<2 x i32
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[14:15], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[14:15], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
@@ -442,6 +462,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned_clamp(<2 x i32
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[14:15], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[14:15], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <8 x i32> %C, i1 1)
@@ -462,6 +483,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed_clamp(<2 x i32>
 ; W32-NEXT:    s_clause 0x1
 ; W32-NEXT:    global_store_b128 v[14:15], v[8:11], off offset:16
 ; W32-NEXT:    global_store_b128 v[14:15], v[4:7], off
+; W32-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W32-NEXT:    s_endpgm
 bb:
   %res = call <8 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <8 x i32> %C, i1 1)
diff --git a/llvm/test/CodeGen/AMDGPU/wmma_multiple_64.ll b/llvm/test/CodeGen/AMDGPU/wmma_multiple_64.ll
index 259f7bb7aa031e..e02cbcba9f2b1e 100644
--- a/llvm/test/CodeGen/AMDGPU/wmma_multiple_64.ll
+++ b/llvm/test/CodeGen/AMDGPU/wmma_multiple_64.ll
@@ -27,6 +27,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_f16(<8 x float> %A, <8 x float> %B
 ; W64-NEXT:    v_wmma_f32_16x16x16_f16 v[16:19], v[8:15], v[8:15], v[16:19]
 ; W64-NEXT:    global_store_b128 v[20:21], v[24:27], off
 ; W64-NEXT:    global_store_b128 v[22:23], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.f16(<8 x float> %A, <8 x float> %B, <4 x float> %C)
@@ -45,6 +46,7 @@ define amdgpu_ps void @test_wmma_f32_16x16x16_bf16(<8 x i32> %A, <8 x i32> %B, <
 ; W64-NEXT:    v_wmma_f32_16x16x16_bf16 v[16:19], v[8:15], v[8:15], v[16:19]
 ; W64-NEXT:    global_store_b128 v[20:21], v[24:27], off
 ; W64-NEXT:    global_store_b128 v[22:23], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x float> @llvm.amdgcn.wmma.f32.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <4 x float> %C)
@@ -63,6 +65,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_lo(<8 x float> %A, <8 x float>
 ; W64-NEXT:    v_wmma_f16_16x16x16_f16 v[16:19], v[8:15], v[8:15], v[16:19]
 ; W64-NEXT:    global_store_b128 v[20:21], v[24:27], off
 ; W64-NEXT:    global_store_b128 v[22:23], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x float> @llvm.amdgcn.wmma.f16.16x16x16.f16(<8 x float> %A, <8 x float> %B, <4 x float> %C, i1 0)
@@ -79,6 +82,7 @@ define amdgpu_ps void @test_wmma_f16_16x16x16_f16_hi(<8 x float> %A, <8 x float>
 ; W64-NEXT:    v_wmma_f16_16x16x16_f16 v[16:19], v[8:15], v[8:15], v[16:19] op_sel:[0,0,1]
 ; W64-NEXT:    global_store_b128 v[20:21], v[24:27], off
 ; W64-NEXT:    global_store_b128 v[22:23], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x float> @llvm.amdgcn.wmma.f16.16x16x16.f16(<8 x float> %A, <8 x float> %B, <4 x float> %C, i1 1)
@@ -97,6 +101,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_lo(<8 x i32> %A, <8 x i32> %
 ; W64-NEXT:    v_wmma_bf16_16x16x16_bf16 v[16:19], v[8:15], v[8:15], v[16:19]
 ; W64-NEXT:    global_store_b128 v[20:21], v[24:27], off
 ; W64-NEXT:    global_store_b128 v[22:23], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <4 x i32> %C, i1 0)
@@ -113,6 +118,7 @@ define amdgpu_ps void @test_wmma_bf16_16x16x16_bf16_hi(<8 x i32> %A, <8 x i32> %
 ; W64-NEXT:    v_wmma_bf16_16x16x16_bf16 v[16:19], v[8:15], v[8:15], v[16:19] op_sel:[0,0,1]
 ; W64-NEXT:    global_store_b128 v[20:21], v[24:27], off
 ; W64-NEXT:    global_store_b128 v[22:23], v[16:19], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.bf16.16x16x16.bf16(<8 x i32> %A, <8 x i32> %B, <4 x i32> %C, i1 1)
@@ -131,6 +137,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned(<4 x i32> %A
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[4:7], v[4:7], v[8:11]
 ; W64-NEXT:    global_store_b128 v[12:13], v[16:19], off
 ; W64-NEXT:    global_store_b128 v[14:15], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
@@ -148,6 +155,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed(<4 x i32> %A,
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[4:7], v[4:7], v[8:11] neg_lo:[0,1,0]
 ; W64-NEXT:    global_store_b128 v[12:13], v[16:19], off
 ; W64-NEXT:    global_store_b128 v[14:15], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
@@ -164,6 +172,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned(<4 x i32> %A,
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[4:7], v[4:7], v[8:11] neg_lo:[1,0,0]
 ; W64-NEXT:    global_store_b128 v[12:13], v[16:19], off
 ; W64-NEXT:    global_store_b128 v[14:15], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 0)
@@ -180,6 +189,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed(<4 x i32> %A, <4
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[4:7], v[4:7], v[8:11] neg_lo:[1,1,0]
 ; W64-NEXT:    global_store_b128 v[12:13], v[16:19], off
 ; W64-NEXT:    global_store_b128 v[14:15], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 0)
@@ -196,6 +206,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_unsigned_clamp(<4 x i
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[4:7], v[4:7], v[8:11] clamp
 ; W64-NEXT:    global_store_b128 v[12:13], v[16:19], off
 ; W64-NEXT:    global_store_b128 v[14:15], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
@@ -212,6 +223,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_unsigned_signed_clamp(<4 x i32
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[4:7], v[4:7], v[8:11] neg_lo:[0,1,0] clamp
 ; W64-NEXT:    global_store_b128 v[12:13], v[16:19], off
 ; W64-NEXT:    global_store_b128 v[14:15], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 0, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
@@ -228,6 +240,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_unsigned_clamp(<4 x i32
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[4:7], v[4:7], v[8:11] neg_lo:[1,0,0] clamp
 ; W64-NEXT:    global_store_b128 v[12:13], v[16:19], off
 ; W64-NEXT:    global_store_b128 v[14:15], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 0, <4 x i32> %B, <4 x i32> %C, i1 1)
@@ -244,6 +257,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui8_signed_signed_clamp(<4 x i32>
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu8 v[8:11], v[4:7], v[4:7], v[8:11] neg_lo:[1,1,0] clamp
 ; W64-NEXT:    global_store_b128 v[12:13], v[16:19], off
 ; W64-NEXT:    global_store_b128 v[14:15], v[8:11], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu8(i1 1, <4 x i32> %A, i1 1, <4 x i32> %B, <4 x i32> %C, i1 1)
@@ -262,6 +276,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned(<2 x i32> %A
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[2:3], v[2:3], v[4:7]
 ; W64-NEXT:    global_store_b128 v[8:9], v[12:15], off
 ; W64-NEXT:    global_store_b128 v[10:11], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
@@ -278,6 +293,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed(<2 x i32> %A,
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[2:3], v[2:3], v[4:7] neg_lo:[0,1,0]
 ; W64-NEXT:    global_store_b128 v[8:9], v[12:15], off
 ; W64-NEXT:    global_store_b128 v[10:11], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
@@ -294,6 +310,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned(<2 x i32> %A,
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[2:3], v[2:3], v[4:7] neg_lo:[1,0,0]
 ; W64-NEXT:    global_store_b128 v[8:9], v[12:15], off
 ; W64-NEXT:    global_store_b128 v[10:11], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 0)
@@ -310,6 +327,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed(<2 x i32> %A, <2
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[2:3], v[2:3], v[4:7] neg_lo:[1,1,0]
 ; W64-NEXT:    global_store_b128 v[8:9], v[12:15], off
 ; W64-NEXT:    global_store_b128 v[10:11], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 0)
@@ -326,6 +344,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_unsigned_clamp(<2 x i
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[2:3], v[2:3], v[4:7] clamp
 ; W64-NEXT:    global_store_b128 v[8:9], v[12:15], off
 ; W64-NEXT:    global_store_b128 v[10:11], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
@@ -342,6 +361,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_unsigned_signed_clamp(<2 x i32
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[2:3], v[2:3], v[4:7] neg_lo:[0,1,0] clamp
 ; W64-NEXT:    global_store_b128 v[8:9], v[12:15], off
 ; W64-NEXT:    global_store_b128 v[10:11], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 0, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)
@@ -358,6 +378,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_unsigned_clamp(<2 x i32
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[2:3], v[2:3], v[4:7] neg_lo:[1,0,0] clamp
 ; W64-NEXT:    global_store_b128 v[8:9], v[12:15], off
 ; W64-NEXT:    global_store_b128 v[10:11], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 0, <2 x i32> %B, <4 x i32> %C, i1 1)
@@ -374,6 +395,7 @@ define amdgpu_ps void @test_wmma_i32_16x16x16_ui4_signed_signed_clamp(<2 x i32>
 ; W64-NEXT:    v_wmma_i32_16x16x16_iu4 v[4:7], v[2:3], v[2:3], v[4:7] neg_lo:[1,1,0] clamp
 ; W64-NEXT:    global_store_b128 v[8:9], v[12:15], off
 ; W64-NEXT:    global_store_b128 v[10:11], v[4:7], off
+; W64-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
 ; W64-NEXT:    s_endpgm
 bb:
   %res = call <4 x i32> @llvm.amdgcn.wmma.i32.16x16x16.iu4(i1 1, <2 x i32> %A, i1 1, <2 x i32> %B, <4 x i32> %C, i1 1)