diff --git a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
index 6dcbced010a5a..5731de82f462e 100644
--- a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
@@ -646,6 +646,8 @@ class WaitcntBrackets {
 
   void applyWaitcnt(const AMDGPU::Waitcnt &Wait);
   void applyWaitcnt(InstCounterType T, unsigned Count);
+  bool hasRedundantXCntWithKmCnt(const AMDGPU::Waitcnt &Wait);
+  bool canOptimizeXCntWithLoadCnt(const AMDGPU::Waitcnt &Wait);
   void applyXcnt(const AMDGPU::Waitcnt &Wait);
   void updateByEvent(WaitEventType E, MachineInstr &MI);
 
@@ -1287,20 +1289,26 @@ void WaitcntBrackets::applyWaitcnt(InstCounterType T, unsigned Count) {
   }
 }
 
-void WaitcntBrackets::applyXcnt(const AMDGPU::Waitcnt &Wait) {
+bool WaitcntBrackets::hasRedundantXCntWithKmCnt(const AMDGPU::Waitcnt &Wait) {
   // Wait on XCNT is redundant if we are already waiting for a load to complete.
   // SMEM can return out of order, so only omit XCNT wait if we are waiting till
   // zero.
-  if (Wait.KmCnt == 0 && hasPendingEvent(SMEM_GROUP))
-    return applyWaitcnt(X_CNT, 0);
+  return Wait.KmCnt == 0 && hasPendingEvent(SMEM_GROUP);
+}
 
+bool WaitcntBrackets::canOptimizeXCntWithLoadCnt(const AMDGPU::Waitcnt &Wait) {
   // If we have pending store we cannot optimize XCnt because we do not wait for
   // stores. VMEM loads retun in order, so if we only have loads XCnt is
   // decremented to the same number as LOADCnt.
-  if (Wait.LoadCnt != ~0u && hasPendingEvent(VMEM_GROUP) &&
-      !hasPendingEvent(STORE_CNT))
-    return applyWaitcnt(X_CNT, std::min(Wait.XCnt, Wait.LoadCnt));
+  return Wait.LoadCnt != ~0u && hasPendingEvent(VMEM_GROUP) &&
+         !hasPendingEvent(STORE_CNT);
+}
 
+void WaitcntBrackets::applyXcnt(const AMDGPU::Waitcnt &Wait) {
+  if (hasRedundantXCntWithKmCnt(Wait))
+    return applyWaitcnt(X_CNT, 0);
+  if (canOptimizeXCntWithLoadCnt(Wait))
+    return applyWaitcnt(X_CNT, std::min(Wait.XCnt, Wait.LoadCnt));
   applyWaitcnt(X_CNT, Wait.XCnt);
 }
 
@@ -1636,6 +1644,8 @@ bool WaitcntGeneratorGFX12Plus::applyPreexistingWaitcnt(
     }
   }
 
+  // Save the pre combine waitcnt in order to make xcnt checks.
+  AMDGPU::Waitcnt PreCombine = Wait;
   if (CombinedLoadDsCntInstr) {
     // Only keep an S_WAIT_LOADCNT_DSCNT if both counters actually need
     // to be waited for. Otherwise, let the instruction be deleted so
@@ -1726,6 +1736,12 @@ bool WaitcntGeneratorGFX12Plus::applyPreexistingWaitcnt(
   }
 
   for (auto CT : inst_counter_types(NUM_EXTENDED_INST_CNTS)) {
+    if ((CT == KM_CNT && ScoreBrackets.hasRedundantXCntWithKmCnt(PreCombine)) ||
+        (CT == LOAD_CNT &&
+         ScoreBrackets.canOptimizeXCntWithLoadCnt(PreCombine)))
+      // Xcnt may need to be updated depending on a pre-existing KM/LOAD_CNT
+      // due to taking the backedge of a block.
+      ScoreBrackets.applyXcnt(PreCombine);
     if (!WaitInstrs[CT])
       continue;
 
diff --git a/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll b/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll
index b5b2655246c3f..365cdc20cb3f9 100644
--- a/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll
+++ b/llvm/test/CodeGen/AMDGPU/flat-saddr-load.ll
@@ -2115,7 +2115,7 @@ define amdgpu_ps void @flat_addr_64bit_lsr_iv(ptr inreg %arg) {
 ; GFX1250-SDAG-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX1250-SDAG-NEXT:  .LBB116_1: ; %bb3
 ; GFX1250-SDAG-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
+; GFX1250-SDAG-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-SDAG-NEXT:    s_add_nc_u64 s[4:5], s[2:3], s[0:1]
 ; GFX1250-SDAG-NEXT:    s_add_nc_u64 s[0:1], s[0:1], 4
 ; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
@@ -2134,7 +2134,6 @@ define amdgpu_ps void @flat_addr_64bit_lsr_iv(ptr inreg %arg) {
 ; GFX1250-GISEL-NEXT:  .LBB116_1: ; %bb3
 ; GFX1250-GISEL-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX1250-GISEL-NEXT:    s_wait_dscnt 0x0
-; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_add_co_u32 v4, vcc_lo, v0, v2
 ; GFX1250-GISEL-NEXT:    v_add_co_ci_u32_e64 v5, null, v1, v3, vcc_lo
@@ -2170,7 +2169,7 @@ define amdgpu_ps void @flat_addr_64bit_lsr_iv_multiload(ptr inreg %arg, ptr inre
 ; GFX1250-SDAG-NEXT:    s_mov_b64 s[0:1], 0
 ; GFX1250-SDAG-NEXT:  .LBB117_1: ; %bb3
 ; GFX1250-SDAG-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1250-SDAG-NEXT:    s_wait_xcnt 0x0
+; GFX1250-SDAG-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-SDAG-NEXT:    s_add_nc_u64 s[4:5], s[2:3], s[0:1]
 ; GFX1250-SDAG-NEXT:    s_add_nc_u64 s[0:1], s[0:1], 4
 ; GFX1250-SDAG-NEXT:    s_wait_dscnt 0x0
@@ -2191,7 +2190,6 @@ define amdgpu_ps void @flat_addr_64bit_lsr_iv_multiload(ptr inreg %arg, ptr inre
 ; GFX1250-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[0:1]
 ; GFX1250-GISEL-NEXT:  .LBB117_1: ; %bb3
 ; GFX1250-GISEL-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1250-GISEL-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-GISEL-NEXT:    v_add_co_u32 v4, vcc_lo, v0, v2
 ; GFX1250-GISEL-NEXT:    v_add_co_ci_u32_e64 v5, null, v1, v3, vcc_lo
diff --git a/llvm/test/CodeGen/AMDGPU/fmin3.ll b/llvm/test/CodeGen/AMDGPU/fmin3.ll
index 6a6f232c55e24..2756472652bc9 100644
--- a/llvm/test/CodeGen/AMDGPU/fmin3.ll
+++ b/llvm/test/CodeGen/AMDGPU/fmin3.ll
@@ -1233,7 +1233,6 @@ define amdgpu_kernel void @test_fmin3_olt_0_f64(ptr addrspace(1) %out, ptr addrs
 ; GFX1250-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-NEXT:    buffer_load_b64 v[2:3], off, s[16:19], null scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_loadcnt 0x0
-; GFX1250-NEXT:    s_wait_xcnt 0x1
 ; GFX1250-NEXT:    s_mov_b32 s4, s14
 ; GFX1250-NEXT:    s_mov_b32 s5, s15
 ; GFX1250-NEXT:    s_mov_b32 s0, s8
@@ -1443,7 +1442,6 @@ define amdgpu_kernel void @test_fmin3_olt_1_f64(ptr addrspace(1) %out, ptr addrs
 ; GFX1250-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-NEXT:    buffer_load_b64 v[2:3], off, s[16:19], null scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_loadcnt 0x0
-; GFX1250-NEXT:    s_wait_xcnt 0x1
 ; GFX1250-NEXT:    s_mov_b32 s4, s14
 ; GFX1250-NEXT:    s_mov_b32 s5, s15
 ; GFX1250-NEXT:    s_mov_b32 s0, s8
diff --git a/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll b/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll
index a50791e10f5a2..ed565ca43f9a3 100644
--- a/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-atomicrmw-fadd.ll
@@ -8814,7 +8814,7 @@ define half @global_agent_atomic_fadd_ret_f16__amdgpu_no_fine_grained_memory(ptr
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-TRUE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_cbranch_execnz .LBB44_1
 ; GFX1250-TRUE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -8857,7 +8857,7 @@ define half @global_agent_atomic_fadd_ret_f16__amdgpu_no_fine_grained_memory(ptr
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-FAKE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_cbranch_execnz .LBB44_1
 ; GFX1250-FAKE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -9322,7 +9322,7 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_gra
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-TRUE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_cbranch_execnz .LBB45_1
 ; GFX1250-TRUE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -9365,7 +9365,7 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_gra
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-FAKE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_cbranch_execnz .LBB45_1
 ; GFX1250-FAKE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -9844,7 +9844,7 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_neg__amdgpu_no_fine_gra
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-TRUE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_cbranch_execnz .LBB46_1
 ; GFX1250-TRUE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -9888,7 +9888,7 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_neg__amdgpu_no_fine_gra
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-FAKE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_cbranch_execnz .LBB46_1
 ; GFX1250-FAKE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -10365,7 +10365,6 @@ define void @global_agent_atomic_fadd_noret_f16__amdgpu_no_fine_grained_memory(p
 ; GFX1250-TRUE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-TRUE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -10407,7 +10406,6 @@ define void @global_agent_atomic_fadd_noret_f16__amdgpu_no_fine_grained_memory(p
 ; GFX1250-FAKE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-FAKE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -10857,7 +10855,6 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_g
 ; GFX1250-TRUE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-TRUE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -10899,7 +10896,6 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_g
 ; GFX1250-FAKE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-FAKE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -11363,7 +11359,6 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b_neg__amdgpu_no_fine_g
 ; GFX1250-TRUE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-TRUE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -11406,7 +11401,6 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b_neg__amdgpu_no_fine_g
 ; GFX1250-FAKE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-FAKE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -11861,7 +11855,7 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_pos__align4__amdgpu_no_
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v5
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-TRUE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_cbranch_execnz .LBB50_1
 ; GFX1250-TRUE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -11893,7 +11887,7 @@ define half @global_agent_atomic_fadd_ret_f16__offset12b_pos__align4__amdgpu_no_
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v5
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-FAKE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_cbranch_execnz .LBB50_1
 ; GFX1250-FAKE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -12245,7 +12239,6 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b__align4_pos__amdgpu_n
 ; GFX1250-TRUE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v5
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-TRUE16-NEXT:    v_mov_b32_e32 v5, v3
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -12276,7 +12269,6 @@ define void @global_agent_atomic_fadd_noret_f16__offset12b__align4_pos__amdgpu_n
 ; GFX1250-FAKE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v5
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-FAKE16-NEXT:    v_mov_b32_e32 v5, v3
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -12631,7 +12623,7 @@ define half @global_system_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_gr
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-TRUE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_cbranch_execnz .LBB52_1
 ; GFX1250-TRUE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -12674,7 +12666,7 @@ define half @global_system_atomic_fadd_ret_f16__offset12b_pos__amdgpu_no_fine_gr
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-FAKE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_cbranch_execnz .LBB52_1
 ; GFX1250-FAKE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -13154,7 +13146,6 @@ define void @global_system_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_
 ; GFX1250-TRUE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-TRUE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -13196,7 +13187,6 @@ define void @global_system_atomic_fadd_noret_f16__offset12b_pos__amdgpu_no_fine_
 ; GFX1250-FAKE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-FAKE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -13676,7 +13666,7 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__amdgpu_no_fine_grained_memory(
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-TRUE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_cbranch_execnz .LBB54_1
 ; GFX1250-TRUE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -13722,7 +13712,7 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__amdgpu_no_fine_grained_memory(
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-FAKE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_cbranch_execnz .LBB54_1
 ; GFX1250-FAKE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -14273,7 +14263,7 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__offset12b_pos__amdgpu_no_fine_
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-TRUE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_cbranch_execnz .LBB55_1
 ; GFX1250-TRUE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -14319,7 +14309,7 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__offset12b_pos__amdgpu_no_fine_
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-FAKE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_cbranch_execnz .LBB55_1
 ; GFX1250-FAKE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -14888,7 +14878,7 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__offset12b_neg__amdgpu_no_fine_
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-TRUE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_cbranch_execnz .LBB56_1
 ; GFX1250-TRUE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -14936,7 +14926,7 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__offset12b_neg__amdgpu_no_fine_
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-FAKE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_cbranch_execnz .LBB56_1
 ; GFX1250-FAKE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -15502,7 +15492,6 @@ define void @global_agent_atomic_fadd_noret_bf16__amdgpu_no_fine_grained_memory(
 ; GFX1250-TRUE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-TRUE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -15547,7 +15536,6 @@ define void @global_agent_atomic_fadd_noret_bf16__amdgpu_no_fine_grained_memory(
 ; GFX1250-FAKE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-FAKE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -16081,7 +16069,6 @@ define void @global_agent_atomic_fadd_noret_bf16__offset12b_pos__amdgpu_no_fine_
 ; GFX1250-TRUE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-TRUE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -16126,7 +16113,6 @@ define void @global_agent_atomic_fadd_noret_bf16__offset12b_pos__amdgpu_no_fine_
 ; GFX1250-FAKE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v2, v3
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-FAKE16-NEXT:    v_mov_b32_e32 v3, v2
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -16678,7 +16664,6 @@ define void @global_agent_atomic_fadd_noret_bf16__offset12b_neg__amdgpu_no_fine_
 ; GFX1250-TRUE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-TRUE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -16725,7 +16710,6 @@ define void @global_agent_atomic_fadd_noret_bf16__offset12b_neg__amdgpu_no_fine_
 ; GFX1250-FAKE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v2, v3
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-FAKE16-NEXT:    v_mov_b32_e32 v3, v2
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -17269,7 +17253,7 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__offset12b_pos__align4__amdgpu_
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v5
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-TRUE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_cbranch_execnz .LBB60_1
 ; GFX1250-TRUE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -17305,7 +17289,7 @@ define bfloat @global_agent_atomic_fadd_ret_bf16__offset12b_pos__align4__amdgpu_
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v5
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-FAKE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_cbranch_execnz .LBB60_1
 ; GFX1250-FAKE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -17753,7 +17737,6 @@ define void @global_agent_atomic_fadd_noret_bf16__offset12b__align4_pos__amdgpu_
 ; GFX1250-TRUE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v5
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-TRUE16-NEXT:    v_mov_b32_e32 v5, v3
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -17788,7 +17771,6 @@ define void @global_agent_atomic_fadd_noret_bf16__offset12b__align4_pos__amdgpu_
 ; GFX1250-FAKE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_DEV
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v2, v3
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-FAKE16-NEXT:    v_mov_b32_e32 v3, v2
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -18238,7 +18220,7 @@ define bfloat @global_system_atomic_fadd_ret_bf16__offset12b_pos__amdgpu_no_fine
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-TRUE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_cbranch_execnz .LBB62_1
 ; GFX1250-TRUE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -18284,7 +18266,7 @@ define bfloat @global_system_atomic_fadd_ret_bf16__offset12b_pos__amdgpu_no_fine
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v5, v7
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
+; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
 ; GFX1250-FAKE16-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_cbranch_execnz .LBB62_1
 ; GFX1250-FAKE16-NEXT:  ; %bb.2: ; %atomicrmw.end
@@ -18854,7 +18836,6 @@ define void @global_system_atomic_fadd_noret_bf16__offset12b_pos__amdgpu_no_fine
 ; GFX1250-TRUE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-TRUE16-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX1250-TRUE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1250-TRUE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-TRUE16-NEXT:    v_mov_b32_e32 v5, v4
 ; GFX1250-TRUE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-TRUE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
@@ -18899,7 +18880,6 @@ define void @global_system_atomic_fadd_noret_bf16__offset12b_pos__amdgpu_no_fine
 ; GFX1250-FAKE16-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-FAKE16-NEXT:    global_inv scope:SCOPE_SYS
 ; GFX1250-FAKE16-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v2, v3
-; GFX1250-FAKE16-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-FAKE16-NEXT:    v_mov_b32_e32 v3, v2
 ; GFX1250-FAKE16-NEXT:    s_or_b32 s0, vcc_lo, s0
 ; GFX1250-FAKE16-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.is.shared.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.is.shared.ll
index 355d0026091d9..7aecae901becf 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.is.shared.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.is.shared.ll
@@ -101,7 +101,6 @@ define amdgpu_kernel void @is_local_vgpr(ptr addrspace(1) %ptr.ptr) {
 ; GFX1250-NEXT:    global_load_b64 v[0:1], v0, s[2:3] scale_offset scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_loadcnt 0x0
 ; GFX1250-NEXT:    v_cmp_eq_u32_e32 vcc_lo, s1, v1
-; GFX1250-NEXT:    s_wait_xcnt 0x0
 ; GFX1250-NEXT:    v_cndmask_b32_e64 v0, 0, 1, vcc_lo
 ; GFX1250-NEXT:    global_store_b32 v[0:1], v0, off
 ; GFX1250-NEXT:    s_endpgm