Skip to content

[AMDGPU] Use v_mov_b64 in codegen on gfx1250 #148272

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 1 commit into from
Jul 12, 2025
Merged

[AMDGPU] Use v_mov_b64 in codegen on gfx1250 #148272

merged 1 commit into from
Jul 12, 2025

Conversation

rampitec
Copy link
Collaborator

No description provided.

@rampitec Graphite App
Copy link
Collaborator Author

This stack of pull requests is managed by Graphite. Learn more about stacking.

@rampitec rampitec requested a review from shiltian July 11, 2025 18:29
@rampitec rampitec marked this pull request as ready for review July 11, 2025 18:29
@llvmbot
Copy link
Member

llvmbot commented Jul 11, 2025

@llvm/pr-subscribers-backend-amdgpu

Author: Stanislav Mekhanoshin (rampitec)

Changes

Patch is 31.35 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/148272.diff

5 Files Affected:

  • (modified) llvm/lib/Target/AMDGPU/GCNSubtarget.h (+1-1)
  • (modified) llvm/lib/Target/AMDGPU/SIInstrInfo.cpp (+1-1)
  • (modified) llvm/test/CodeGen/AMDGPU/copy_phys_vgpr64.mir (+103)
  • (modified) llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll (+45-52)
  • (modified) llvm/test/CodeGen/AMDGPU/v_mov_b64_expansion.mir (+18-9)
diff --git a/llvm/lib/Target/AMDGPU/GCNSubtarget.h b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
index 8ea60871b6613..7d3b821825e82 100644
--- a/llvm/lib/Target/AMDGPU/GCNSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
@@ -1154,7 +1154,7 @@ class GCNSubtarget final : public AMDGPUGenSubtargetInfo,
 
   bool hasMadF16() const;
 
-  bool hasMovB64() const { return GFX940Insts; }
+  bool hasMovB64() const { return GFX940Insts || GFX1250Insts; }
 
   bool hasLshlAddU64Inst() const { return HasLshlAddU64Inst; }
 
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
index ec7ef66f2c1aa..1bf152bb34851 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -2214,7 +2214,7 @@ bool SIInstrInfo::expandPostRAPseudo(MachineInstr &MI) const {
     if (ST.hasMovB64()) {
       MI.setDesc(get(AMDGPU::V_MOV_B64_e32));
       if (SrcOp.isReg() || isInlineConstant(MI, 1) ||
-          isUInt<32>(SrcOp.getImm()))
+          isUInt<32>(SrcOp.getImm()) || ST.has64BitLiterals())
         break;
     }
     if (SrcOp.isImm()) {
diff --git a/llvm/test/CodeGen/AMDGPU/copy_phys_vgpr64.mir b/llvm/test/CodeGen/AMDGPU/copy_phys_vgpr64.mir
index d5dfb5dd0848d..cc976fe13c47c 100644
--- a/llvm/test/CodeGen/AMDGPU/copy_phys_vgpr64.mir
+++ b/llvm/test/CodeGen/AMDGPU/copy_phys_vgpr64.mir
@@ -4,6 +4,7 @@
 # RUN: llc -mtriple=amdgcn -mcpu=gfx942 -run-pass postrapseudos -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX942 %s
 # RUN: llc -mtriple=amdgcn -mcpu=gfx1010 -run-pass postrapseudos -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX10 %s
 # RUN: llc -mtriple=amdgcn -mcpu=gfx1100 -run-pass postrapseudos -verify-machineinstrs %s -o - | FileCheck -check-prefix=GFX10 %s
+# RUN: llc -mtriple=amdgcn -mcpu=gfx1250 -run-pass postrapseudos %s -o - | FileCheck -check-prefix=GFX1250 %s
 
 ---
 name: copy_v64_to_v64
@@ -32,6 +33,10 @@ body: |
     ; GFX10-NEXT: {{  $}}
     ; GFX10-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr2, implicit $exec, implicit-def $vgpr0_vgpr1, implicit $vgpr2_vgpr3
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr3, implicit $exec, implicit killed $vgpr2_vgpr3, implicit $exec
+    ; GFX1250-LABEL: name: copy_v64_to_v64
+    ; GFX1250: liveins: $vgpr2_vgpr3
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr0_vgpr1 = V_MOV_B64_e32 killed $vgpr2_vgpr3, implicit $exec, implicit $exec
     $vgpr0_vgpr1 = COPY killed $vgpr2_vgpr3, implicit $exec
 ...
 
@@ -62,6 +67,10 @@ body: |
     ; GFX10-NEXT: {{  $}}
     ; GFX10-NEXT: $vgpr0 = V_MOV_B32_e32 $sgpr2, implicit $exec, implicit-def $vgpr0_vgpr1, implicit $sgpr2_sgpr3
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $sgpr3, implicit $exec, implicit killed $sgpr2_sgpr3, implicit $exec
+    ; GFX1250-LABEL: name: copy_s64_to_v64
+    ; GFX1250: liveins: $sgpr2_sgpr3
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr0_vgpr1 = V_MOV_B64_e32 killed $sgpr2_sgpr3, implicit $exec, implicit $exec
     $vgpr0_vgpr1 = COPY killed $sgpr2_sgpr3, implicit $exec
 ...
 
@@ -94,6 +103,11 @@ body: |
     ; GFX10-NEXT: {{  $}}
     ; GFX10-NEXT: $vgpr0 = V_ACCVGPR_READ_B32_e64 $agpr2, implicit $exec, implicit-def $vgpr0_vgpr1, implicit $agpr2_agpr3
     ; GFX10-NEXT: $vgpr1 = V_ACCVGPR_READ_B32_e64 $agpr3, implicit $exec, implicit killed $agpr2_agpr3, implicit $exec
+    ; GFX1250-LABEL: name: copy_a64_to_v64
+    ; GFX1250: liveins: $agpr2_agpr3
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr0 = V_ACCVGPR_READ_B32_e64 $agpr2, implicit $exec, implicit-def $vgpr0_vgpr1, implicit $agpr2_agpr3
+    ; GFX1250-NEXT: $vgpr1 = V_ACCVGPR_READ_B32_e64 $agpr3, implicit $exec, implicit killed $agpr2_agpr3, implicit $exec
     $vgpr0_vgpr1 = COPY killed $agpr2_agpr3, implicit $exec
 ...
 
@@ -130,6 +144,11 @@ body: |
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr3, implicit $exec, implicit $vgpr2_vgpr3_vgpr4_vgpr5
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr4, implicit $exec, implicit $vgpr2_vgpr3_vgpr4_vgpr5
     ; GFX10-NEXT: $vgpr3 = V_MOV_B32_e32 $vgpr5, implicit $exec, implicit $vgpr2_vgpr3_vgpr4_vgpr5, implicit $exec
+    ; GFX1250-LABEL: name: copy_v128_to_v128_fwd
+    ; GFX1250: liveins: $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr0_vgpr1 = V_MOV_B64_e32 $vgpr2_vgpr3, implicit $exec, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3, implicit $vgpr2_vgpr3_vgpr4_vgpr5
+    ; GFX1250-NEXT: $vgpr2_vgpr3 = V_MOV_B64_e32 $vgpr4_vgpr5, implicit $exec, implicit $vgpr2_vgpr3_vgpr4_vgpr5, implicit $exec
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY killed $vgpr2_vgpr3_vgpr4_vgpr5, implicit $exec
 ...
 
@@ -166,6 +185,11 @@ body: |
     ; GFX10-NEXT: $vgpr4 = V_MOV_B32_e32 $vgpr2, implicit $exec, implicit $vgpr0_vgpr1_vgpr2_vgpr3
     ; GFX10-NEXT: $vgpr3 = V_MOV_B32_e32 $vgpr1, implicit $exec, implicit $vgpr0_vgpr1_vgpr2_vgpr3
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr0, implicit $exec, implicit $vgpr0_vgpr1_vgpr2_vgpr3, implicit $exec
+    ; GFX1250-LABEL: name: copy_v128_to_v128_back
+    ; GFX1250: liveins: $vgpr0_vgpr1_vgpr2_vgpr3
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr4_vgpr5 = V_MOV_B64_e32 $vgpr2_vgpr3, implicit $exec, implicit-def $vgpr2_vgpr3_vgpr4_vgpr5, implicit $vgpr0_vgpr1_vgpr2_vgpr3
+    ; GFX1250-NEXT: $vgpr2_vgpr3 = V_MOV_B64_e32 $vgpr0_vgpr1, implicit $exec, implicit $vgpr0_vgpr1_vgpr2_vgpr3, implicit $exec
     $vgpr2_vgpr3_vgpr4_vgpr5 = COPY killed $vgpr0_vgpr1_vgpr2_vgpr3, implicit $exec
 ...
 
@@ -202,6 +226,12 @@ body: |
     ; GFX10-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr4, implicit $exec, implicit-def $vgpr0_vgpr1_vgpr2, implicit $vgpr4_vgpr5_vgpr6
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr5, implicit $exec, implicit $vgpr4_vgpr5_vgpr6
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr6, implicit $exec, implicit killed $vgpr4_vgpr5_vgpr6, implicit $exec
+    ; GFX1250-LABEL: name: copy_v96_to_v96
+    ; GFX1250: liveins: $vgpr4_vgpr5_vgpr6
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr4, implicit $exec, implicit-def $vgpr0_vgpr1_vgpr2, implicit $vgpr4_vgpr5_vgpr6
+    ; GFX1250-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr5, implicit $exec, implicit $vgpr4_vgpr5_vgpr6
+    ; GFX1250-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr6, implicit $exec, implicit killed $vgpr4_vgpr5_vgpr6, implicit $exec
     $vgpr0_vgpr1_vgpr2 = COPY killed $vgpr4_vgpr5_vgpr6, implicit $exec
 ...
 
@@ -232,6 +262,10 @@ body: |
     ; GFX10-NEXT: {{  $}}
     ; GFX10-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr2, implicit $exec, implicit-def $vgpr0_vgpr1, implicit $vgpr2_vgpr3
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr3, implicit $exec, implicit killed $vgpr2_vgpr3, implicit $exec
+    ; GFX1250-LABEL: name: copy_v64_to_v64_undef_sub0
+    ; GFX1250: liveins: $vgpr3
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr0_vgpr1 = V_MOV_B64_e32 killed $vgpr2_vgpr3, implicit $exec, implicit $exec
     $vgpr0_vgpr1 = COPY killed $vgpr2_vgpr3, implicit $exec
 ...
 
@@ -262,6 +296,10 @@ body: |
     ; GFX10-NEXT: {{  $}}
     ; GFX10-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr2, implicit $exec, implicit-def $vgpr0_vgpr1, implicit $vgpr2_vgpr3
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr3, implicit $exec, implicit killed $vgpr2_vgpr3, implicit $exec
+    ; GFX1250-LABEL: name: copy_v64_to_v64_undef_sub1
+    ; GFX1250: liveins: $vgpr2
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr0_vgpr1 = V_MOV_B64_e32 killed $vgpr2_vgpr3, implicit $exec, implicit $exec
     $vgpr0_vgpr1 = COPY killed $vgpr2_vgpr3, implicit $exec
 ...
 
@@ -298,6 +336,11 @@ body: |
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $sgpr5, implicit $exec, implicit $sgpr4_sgpr5_sgpr6_sgpr7
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $sgpr6, implicit $exec, implicit $sgpr4_sgpr5_sgpr6_sgpr7
     ; GFX10-NEXT: $vgpr3 = V_MOV_B32_e32 $sgpr7, implicit $exec, implicit killed $sgpr4_sgpr5_sgpr6_sgpr7
+    ; GFX1250-LABEL: name: copy_s128_to_v128_killed
+    ; GFX1250: liveins: $sgpr4_sgpr5_sgpr6_sgpr7
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr0_vgpr1 = V_MOV_B64_e32 $sgpr4_sgpr5, implicit $exec, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3, implicit $sgpr4_sgpr5_sgpr6_sgpr7
+    ; GFX1250-NEXT: $vgpr2_vgpr3 = V_MOV_B64_e32 $sgpr6_sgpr7, implicit $exec, implicit killed $sgpr4_sgpr5_sgpr6_sgpr7
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY killed $sgpr4_sgpr5_sgpr6_sgpr7
 ...
 
@@ -330,6 +373,11 @@ body: |
     ; GFX10-NEXT: {{  $}}
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr2, implicit $exec, implicit-def $vgpr1_vgpr2, implicit $vgpr2_vgpr3
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr3, implicit $exec, implicit $vgpr2_vgpr3, implicit $exec
+    ; GFX1250-LABEL: name: copy_v64_to_v64_unaligned
+    ; GFX1250: liveins: $vgpr2_vgpr3
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr2, implicit $exec, implicit-def $vgpr1_vgpr2, implicit $vgpr2_vgpr3
+    ; GFX1250-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr3, implicit $exec, implicit $vgpr2_vgpr3, implicit $exec
     $vgpr1_vgpr2 = COPY killed $vgpr2_vgpr3, implicit $exec
 ...
 
@@ -362,6 +410,11 @@ body: |
     ; GFX10-NEXT: {{  $}}
     ; GFX10-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr3, implicit $exec, implicit-def $vgpr0_vgpr1, implicit $vgpr3_vgpr4
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr4, implicit $exec, implicit killed $vgpr3_vgpr4, implicit $exec
+    ; GFX1250-LABEL: name: copy_v64_unaligned_to_v64
+    ; GFX1250: liveins: $vgpr3_vgpr4
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr3, implicit $exec, implicit-def $vgpr0_vgpr1, implicit $vgpr3_vgpr4
+    ; GFX1250-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr4, implicit $exec, implicit killed $vgpr3_vgpr4, implicit $exec
     $vgpr0_vgpr1 = COPY killed $vgpr3_vgpr4, implicit $exec
 ...
 
@@ -402,6 +455,13 @@ body: |
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr9, implicit $exec, implicit $vgpr8_vgpr9_vgpr10_vgpr11
     ; GFX10-NEXT: $vgpr3 = V_MOV_B32_e32 $vgpr10, implicit $exec, implicit $vgpr8_vgpr9_vgpr10_vgpr11
     ; GFX10-NEXT: $vgpr4 = V_MOV_B32_e32 $vgpr11, implicit $exec, implicit killed $vgpr8_vgpr9_vgpr10_vgpr11, implicit $exec
+    ; GFX1250-LABEL: name: copy_v128_to_v128_unaligned
+    ; GFX1250: liveins: $vgpr8_vgpr9_vgpr10_vgpr11
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr8, implicit $exec, implicit-def $vgpr1_vgpr2_vgpr3_vgpr4, implicit $vgpr8_vgpr9_vgpr10_vgpr11
+    ; GFX1250-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr9, implicit $exec, implicit $vgpr8_vgpr9_vgpr10_vgpr11
+    ; GFX1250-NEXT: $vgpr3 = V_MOV_B32_e32 $vgpr10, implicit $exec, implicit $vgpr8_vgpr9_vgpr10_vgpr11
+    ; GFX1250-NEXT: $vgpr4 = V_MOV_B32_e32 $vgpr11, implicit $exec, implicit killed $vgpr8_vgpr9_vgpr10_vgpr11, implicit $exec
     $vgpr1_vgpr2_vgpr3_vgpr4 = COPY killed $vgpr8_vgpr9_vgpr10_vgpr11, implicit $exec
 ...
 
@@ -442,6 +502,13 @@ body: |
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr8, implicit $exec, implicit $vgpr7_vgpr8_vgpr9_vgpr10
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr9, implicit $exec, implicit $vgpr7_vgpr8_vgpr9_vgpr10
     ; GFX10-NEXT: $vgpr3 = V_MOV_B32_e32 $vgpr10, implicit $exec, implicit killed $vgpr7_vgpr8_vgpr9_vgpr10, implicit $exec
+    ; GFX1250-LABEL: name: copy_v128_unaligned_to_v128
+    ; GFX1250: liveins: $vgpr7_vgpr8_vgpr9_vgpr10
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr7, implicit $exec, implicit-def $vgpr0_vgpr1_vgpr2_vgpr3, implicit $vgpr7_vgpr8_vgpr9_vgpr10
+    ; GFX1250-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr8, implicit $exec, implicit $vgpr7_vgpr8_vgpr9_vgpr10
+    ; GFX1250-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr9, implicit $exec, implicit $vgpr7_vgpr8_vgpr9_vgpr10
+    ; GFX1250-NEXT: $vgpr3 = V_MOV_B32_e32 $vgpr10, implicit $exec, implicit killed $vgpr7_vgpr8_vgpr9_vgpr10, implicit $exec
     $vgpr0_vgpr1_vgpr2_vgpr3 = COPY killed $vgpr7_vgpr8_vgpr9_vgpr10, implicit $exec
 ...
 
@@ -474,6 +541,11 @@ body: |
     ; GFX10-NEXT: {{  $}}
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $sgpr8, implicit $exec, implicit-def $vgpr1_vgpr2, implicit $sgpr8_sgpr9
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $sgpr9, implicit $exec, implicit killed $sgpr8_sgpr9, implicit $exec
+    ; GFX1250-LABEL: name: copy_s64_to_v64_unaligned
+    ; GFX1250: liveins: $sgpr8_sgpr9
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr1 = V_MOV_B32_e32 $sgpr8, implicit $exec, implicit-def $vgpr1_vgpr2, implicit $sgpr8_sgpr9
+    ; GFX1250-NEXT: $vgpr2 = V_MOV_B32_e32 $sgpr9, implicit $exec, implicit killed $sgpr8_sgpr9, implicit $exec
     $vgpr1_vgpr2 = COPY killed $sgpr8_sgpr9, implicit $exec
 ...
 
@@ -514,6 +586,13 @@ body: |
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $sgpr9, implicit $exec, implicit $sgpr8_sgpr9_sgpr10_sgpr11
     ; GFX10-NEXT: $vgpr3 = V_MOV_B32_e32 $sgpr10, implicit $exec, implicit $sgpr8_sgpr9_sgpr10_sgpr11
     ; GFX10-NEXT: $vgpr4 = V_MOV_B32_e32 $sgpr11, implicit $exec, implicit killed $sgpr8_sgpr9_sgpr10_sgpr11, implicit $exec
+    ; GFX1250-LABEL: name: copy_s128_to_v128_unaligned
+    ; GFX1250: liveins: $sgpr8_sgpr9_sgpr10_sgpr11
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr1 = V_MOV_B32_e32 $sgpr8, implicit $exec, implicit-def $vgpr1_vgpr2_vgpr3_vgpr4, implicit $sgpr8_sgpr9_sgpr10_sgpr11
+    ; GFX1250-NEXT: $vgpr2 = V_MOV_B32_e32 $sgpr9, implicit $exec, implicit $sgpr8_sgpr9_sgpr10_sgpr11
+    ; GFX1250-NEXT: $vgpr3 = V_MOV_B32_e32 $sgpr10, implicit $exec, implicit $sgpr8_sgpr9_sgpr10_sgpr11
+    ; GFX1250-NEXT: $vgpr4 = V_MOV_B32_e32 $sgpr11, implicit $exec, implicit killed $sgpr8_sgpr9_sgpr10_sgpr11, implicit $exec
     $vgpr1_vgpr2_vgpr3_vgpr4 = COPY killed $sgpr8_sgpr9_sgpr10_sgpr11, implicit $exec
 ...
 
@@ -550,6 +629,12 @@ body: |
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr8, implicit $exec, implicit-def $vgpr1_vgpr2_vgpr3, implicit $vgpr8_vgpr9_vgpr10
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr9, implicit $exec, implicit $vgpr8_vgpr9_vgpr10
     ; GFX10-NEXT: $vgpr3 = V_MOV_B32_e32 $vgpr10, implicit $exec, implicit killed $vgpr8_vgpr9_vgpr10, implicit $exec
+    ; GFX1250-LABEL: name: copy_v96_to_v96_unaligned
+    ; GFX1250: liveins: $vgpr8_vgpr9_vgpr10
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr8, implicit $exec, implicit-def $vgpr1_vgpr2_vgpr3, implicit $vgpr8_vgpr9_vgpr10
+    ; GFX1250-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr9, implicit $exec, implicit $vgpr8_vgpr9_vgpr10
+    ; GFX1250-NEXT: $vgpr3 = V_MOV_B32_e32 $vgpr10, implicit $exec, implicit killed $vgpr8_vgpr9_vgpr10, implicit $exec
     $vgpr1_vgpr2_vgpr3 = COPY killed $vgpr8_vgpr9_vgpr10, implicit $exec
 ...
 
@@ -586,6 +671,12 @@ body: |
     ; GFX10-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr7, implicit $exec, implicit-def $vgpr0_vgpr1_vgpr2, implicit $vgpr7_vgpr8_vgpr9
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr8, implicit $exec, implicit $vgpr7_vgpr8_vgpr9
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr9, implicit $exec, implicit killed $vgpr7_vgpr8_vgpr9, implicit $exec
+    ; GFX1250-LABEL: name: copy_v96_unaligned_to_v96
+    ; GFX1250: liveins: $vgpr7_vgpr8_vgpr9
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr0 = V_MOV_B32_e32 $vgpr7, implicit $exec, implicit-def $vgpr0_vgpr1_vgpr2, implicit $vgpr7_vgpr8_vgpr9
+    ; GFX1250-NEXT: $vgpr1 = V_MOV_B32_e32 $vgpr8, implicit $exec, implicit $vgpr7_vgpr8_vgpr9
+    ; GFX1250-NEXT: $vgpr2 = V_MOV_B32_e32 $vgpr9, implicit $exec, implicit killed $vgpr7_vgpr8_vgpr9, implicit $exec
     $vgpr0_vgpr1_vgpr2 = COPY killed $vgpr7_vgpr8_vgpr9, implicit $exec
 ...
 
@@ -622,6 +713,12 @@ body: |
     ; GFX10-NEXT: $vgpr0 = V_MOV_B32_e32 $sgpr0, implicit $exec, implicit-def $vgpr0_vgpr1_vgpr2, implicit $sgpr0_sgpr1_sgpr2
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $sgpr1, implicit $exec, implicit $sgpr0_sgpr1_sgpr2
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $sgpr2, implicit $exec, implicit killed $sgpr0_sgpr1_sgpr2, implicit $exec
+    ; GFX1250-LABEL: name: copy_s96_to_v96
+    ; GFX1250: liveins: $sgpr0_sgpr1_sgpr2
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr0 = V_MOV_B32_e32 $sgpr0, implicit $exec, implicit-def $vgpr0_vgpr1_vgpr2, implicit $sgpr0_sgpr1_sgpr2
+    ; GFX1250-NEXT: $vgpr1 = V_MOV_B32_e32 $sgpr1, implicit $exec, implicit $sgpr0_sgpr1_sgpr2
+    ; GFX1250-NEXT: $vgpr2 = V_MOV_B32_e32 $sgpr2, implicit $exec, implicit killed $sgpr0_sgpr1_sgpr2, implicit $exec
     $vgpr0_vgpr1_vgpr2 = COPY killed $sgpr0_sgpr1_sgpr2, implicit $exec
 ...
 
@@ -658,5 +755,11 @@ body: |
     ; GFX10-NEXT: $vgpr3 = V_MOV_B32_e32 $sgpr2, implicit $exec, implicit-def $vgpr1_vgpr2_vgpr3, implicit $sgpr0_sgpr1_sgpr2
     ; GFX10-NEXT: $vgpr2 = V_MOV_B32_e32 $sgpr1, implicit $exec, implicit $sgpr0_sgpr1_sgpr2
     ; GFX10-NEXT: $vgpr1 = V_MOV_B32_e32 $sgpr0, implicit $exec, implicit killed $sgpr0_sgpr1_sgpr2, implicit $exec
+    ; GFX1250-LABEL: name: copy_s96_to_v96_unaligned
+    ; GFX1250: liveins: $sgpr0_sgpr1_sgpr2
+    ; GFX1250-NEXT: {{  $}}
+    ; GFX1250-NEXT: $vgpr3 = V_MOV_B32_e32 $sgpr2, implicit $exec, implicit-def $vgpr1_vgpr2_vgpr3, implicit $sgpr0_sgpr1_sgpr2
+    ; GFX1250-NEXT: $vgpr2 = V_MOV_B32_e32 $sgpr1, implicit $exec, implicit $sgpr0_sgpr1_sgpr2
+    ; GFX1250-NEXT: $vgpr1 = V_MOV_B32_e32 $sgpr0, implicit $exec, implicit killed $sgpr0_sgpr1_sgpr2, implicit $exec
     $vgpr1_vgpr2_vgpr3 = COPY killed $sgpr0_sgpr1_sgpr2, implicit $exec
 ...
diff --git a/llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll b/llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll
index d8079651787ad..5d35adc8cbe0a 100644
--- a/llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll
@@ -65,15 +65,15 @@ define i16 @test_v7i16_load_store(ptr addrspace(1) %ptr1, ptr addrspace(1) %ptr2
 ; GCN-SDAG-NEXT:    s_wait_kmcnt 0x0
 ; GCN-SDAG-NEXT:    global_load_b128 v[4:7], v[0:1], off
 ; GCN-SDAG-NEXT:    global_load_b128 v[0:3], v[2:3], off
+; GCN-SDAG-NEXT:    v_mov_b64_e32 v[8:9], 0
 ; GCN-SDAG-NEXT:    s_wait_loadcnt 0x0
 ; GCN-SDAG-NEXT:    v_pk_add_u16 v10, v6, v2
 ; GCN-SDAG-NEXT:    v_pk_add_u16 v11, v7, v3
 ; GCN-SDAG-NEXT:    s_wait_xcnt 0x0
-; GCN-SDAG-NEXT:    v_dual_mov_b32 v2, 12 :: v_dual_mov_b32 v6, 8
-; GCN-SDAG-NEXT:    v_dual_mov_b32 v3, 0 :: v_dual_mov_b32 v7, 0
-; GCN-SDAG-NEXT:    v_mov_b32_e32 v8, 0
+; GCN-SDAG-NEXT:    v_mov_b64_e32 v[2:3], 12
+; GCN-SDAG-NEXT:    v_mov_b64_e32 v[6:7], 8
 ; GCN-SDAG-NEXT:    v_pk_add_u16 v4, v4, v0
-; GCN-SDAG-NEXT:    v_dual_mov_b32 v9, 0 :: v_dual_lshrrev_b32 v0, 16, v10
+; GCN-SDAG-NEXT:    v_lshrrev_b32_e32 v0, 16, v10
 ; GCN-SDAG-NEXT:    v_pk_add_u16 v5, v5, v1
 ; GCN-SDAG-NEXT:    s_clause 0x2
 ; GCN-SDAG-NEXT:    global_store_b16 v[2:3], v11, off
@@ -87,19 +87,17 @@ define i16 @test_v7i16_load_store(ptr addrspace(1) %ptr1, ptr addrspace(1) %ptr2
 ; GCN-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GCN-GISEL-NEXT:    global_load_b128 v[4:7], v[0:1], off
 ; GCN-GISEL-NEXT:    global_load_b128 v[0:3], v[2:3], off
-; GCN-GISEL-NEXT:    v_dual_mov_b32 v8, 0 :: v_dual_mov_b32 v10, 2
-; GCN-GISEL-NEXT:    v_dual_mov_b32 v9, 0 :: v_dual_mov_b32 v11, 0
-; GCN-GISEL-NEXT:    v_dual_mov_b32 v12, 4 :: v_dual_mov_b32 v14, 6
-; GCN-GISEL-NEXT:    v_dual_mov_b32 v13, 0 :: v_dual_mov_b32 v15, 0
-; GCN-GISEL-NEXT:    v_dual_mov_b32 v16, 8 :: v_dual_mov_b32 v18, 10
-; GCN-GISEL-NEXT:    v_dual_mov_b32 v17, 0 :: v_dual_mov_b32 v19, 0
+; GCN-GISEL-NEXT:    v_mov_b64_e32 v[8:9], 0
+; GCN-GISEL-NEXT:    v_mov_b64_e32 v[10:11], 2
+; GCN-GISEL-NEXT:    v_mov_b64_e32 v[12:13], 4
+; GCN-GISEL-NEXT:    v_mov_b64_e32 v[14:15], 6
+; GCN-GISEL-NEXT:    v_mov_b64_e32 v[16:17], 8
+; GCN-GISEL-NEXT:    v_mov_b64_e32 v[18:19], 10
+; GCN-GISEL-NEXT:    v_mov_b64_e32 v[20:21], 12
 ; GCN-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GCN-GISEL-NEXT:    v_pk_add_u16 v2, v6, v2
 ; GCN-GISEL-NEXT:    v_pk_add_u16 v4, v4, v0
-; GCN-GISEL-NEXT:    v_mov_b32_e32 v20, 12
 ; GCN-GISEL-NEXT:    v_pk_add_u16 v1, v5, v1
-; GCN-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_4)
-; GCN-GISEL-NEXT:    v_dual_mov_b32 v21, 0 :: v_dual_lshrrev_b32 v0, 16, v2
 ; GCN-GISEL-NEXT:    v_pk_add_u16 v3, v7, v3
 ; GCN-GISEL-NEXT:    s_clause 0x6
 ; GCN-GISEL-NEXT:    global_store_b16 v[8:9], v4, off
@@ -109,6 +107,7 @@ define i16 @test_v7i16_load_store(ptr addrspace(1) %ptr1, ptr addrspace(1) %ptr2
 ; GCN-GISEL-NEXT:    global_store_b16 v[16:17], v2, off
 ; GCN-GISEL-NEXT:    global_store_d16_hi_b16 v[18:19], v2, off
 ; GCN-GISEL-NEXT:    global_store_b16 v[20:21], v3, off
+; GCN-GISEL-NEXT:    v_lshrrev_b32_e32 v0, 16, v2
 ; GCN-GISEL-NEXT:    s_set_pc_i64 s[30:31]
   %vec1 = load <7 x i16>, ptr addrspace...
[truncated]

@rampitec rampitec requested a review from changpeng July 11, 2025 22:00
@rampitec rampitec merged commit d1e3ab9 into main Jul 12, 2025
13 checks passed
@rampitec rampitec deleted the users/rampitec/07-10-_amdgpu_use_v_mov_b64_in_codegen_on_gfx1250 branch July 12, 2025 05:16
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@changpeng changpeng changpeng approved these changes

@shiltian shiltian Awaiting requested review from shiltian

Assignees
No one assigned
Labels
backend:AMDGPU
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@rampitec @llvmbot @changpeng