AMDGPU: Force sc0 and sc1 on stores for gfx940 and gfx941

[arsenm]

Should this be a feature set by default in the subtarget constructor instead? Should you be able to turn this off?

[kzhuravl]

In D149986#4329302, @arsenm wrote:

Should this be a feature set by default in the subtarget constructor instead? Should you be able to turn this off?

Users should not be able to turn this off. If user wants it off, user can use gfx942. Should it be moved to the subtarget constructor?

[jmmartinez]

llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp

524

NIT: Is the use of the bitwise or " | " intended? I'd use the logical or " || " instead.

[kzhuravl]

llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp

524

It is intentional, we need both SC0 and SC1 bits set. If I switch this to || it will short circuit and not invoke enableSC1Bit.

[Pierre-vh]

I think that if this is a new property of the GFX940/941 targets, and turning it off shouldn't be possible, we shouldn't even bother with a feature and just set a bool in the ST for those targets

llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp

524

IMHO then it needs a comment to explain that it's intentional, otherwise some innocent maintainer in the future could think it's a typo and change it

[kzhuravl]

In D149986#4334274, @Pierre-vh wrote:

I think that if this is a new property of the GFX940/941 targets, and turning it off shouldn't be possible, we shouldn't even bother with a feature and just set a bool in the ST for those targets

AFAIK, the subtarget only knows the generation and has access to features. The subtarget cannot differentiate between gfx940, gfx941, gfx942.

[kzhuravl]

Address review comments.

llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp

524

Added comment.

[scott.linder]

llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp

517–529

It may be a bit more verbose, but I would suggest just having a bool Changed variable, like other (admittedly more complicated) functions in the unit do. I don't think any future maintainer will mis-interpret this, even without the comment

[arsenm]

llvm/lib/TargetParser/TargetParser.cpp

330–334

I don't see a reason to set this here. There's no need to expose this to the IR. We already leak too many target details here as it is

[kzhuravl]

Address review feedback

[kzhuravl]

Address @arsenm's comment

clang/test/CodeGenOpenCL/amdgpu-features.cl

 // GFX900: "target-features"="+16-bit-insts,+ci-insts,+dpp,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX902: "target-features"="+16-bit-insts,+ci-insts,+dpp,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX904: "target-features"="+16-bit-insts,+ci-insts,+dpp,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX906: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot7-insts,+dpp,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX908: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx8-insts,+gfx9-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX909: "target-features"="+16-bit-insts,+ci-insts,+dpp,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX90A: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-fadd-rtn-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX90C: "target-features"="+16-bit-insts,+ci-insts,+dpp,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
-// GFX940: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
-// GFX941: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX940: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+force-store-sc0-sc1,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
+// GFX941: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+force-store-sc0-sc1,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX942: "target-features"="+16-bit-insts,+atomic-buffer-global-pk-add-f16-insts,+atomic-ds-pk-add-16-insts,+atomic-fadd-rtn-insts,+atomic-flat-pk-add-16-insts,+atomic-global-pk-add-bf16-inst,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+fp8-insts,+gfx8-insts,+gfx9-insts,+gfx90a-insts,+gfx940-insts,+mai-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize64"
 // GFX1010: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
 // GFX1011: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
 // GFX1012: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
 // GFX1013: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dpp,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
 // GFX1030: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
 // GFX1031: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"
 // GFX1032: "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot10-insts,+dot2-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+gfx10-3-insts,+gfx10-insts,+gfx8-insts,+gfx9-insts,+s-memrealtime,+s-memtime-inst,+wavefrontsize32"

llvm/lib/Target/AMDGPU/AMDGPU.td

 >;
 
 def FeatureVALUTransUseHazard : SubtargetFeature<"valu-trans-use-hazard",
   "HasVALUTransUseHazard",
   "true",
   "Hazard when TRANS instructions are closely followed by a use of the result"
 >;
 
+def FeatureForceStoreSC0SC1 : SubtargetFeature<"force-store-sc0-sc1",
+  "HasForceStoreSC0SC1",
+  "true",
+  "Has SC0 and SC1 on stores"
+>;
+
 //===------------------------------------------------------------===//
 // Subtarget Features (options and debugging)
 //===------------------------------------------------------------===//
 
 class FeatureMaxPrivateElementSize<int size> : SubtargetFeature<
   "max-private-element-size-"#size,
   "MaxPrivateElementSize",
   !cast<string>(size),
def FeatureISAVersion9_4_Common : FeatureSet<
    FeatureSupportsSRAMECC,
    FeaturePackedTID,
    FeatureArchitectedFlatScratch,
    FullRate64Ops,
    FeatureBackOffBarrier]>;
 
 def FeatureISAVersion9_4_0 : FeatureSet<
   !listconcat(FeatureISAVersion9_4_Common.Features,
-    [])>;
+    [FeatureForceStoreSC0SC1])>;
 
 def FeatureISAVersion9_4_1 : FeatureSet<
   !listconcat(FeatureISAVersion9_4_Common.Features,
-    [])>;
+    [FeatureForceStoreSC0SC1])>;
 
 def FeatureISAVersion9_4_2 : FeatureSet<
   !listconcat(FeatureISAVersion9_4_Common.Features,
     [])>;
 
 // TODO: Organize more features into groups.
 def FeatureGroup {
   // Bugs present on gfx10.1.

llvm/lib/Target/AMDGPU/GCNSubtarget.h

protected:
   bool HasOffset3fBug = false;
   bool HasFlatSegmentOffsetBug = false;
   bool HasImageStoreD16Bug = false;
   bool HasImageGather4D16Bug = false;
   bool HasGFX11FullVGPRs = false;
   bool HasMADIntraFwdBug = false;
   bool HasVOPDInsts = false;
   bool HasVALUTransUseHazard = false;
+  bool HasForceStoreSC0SC1 = false;
 
   // Dummy feature to use for assembler in tablegen.
   bool FeatureDisable = false;
 
   SelectionDAGTargetInfo TSInfo;
 private:
   SIInstrInfo InstrInfo;
   SITargetLowering TLInfo;
public:
   bool hasLdsDirect() const { return getGeneration() >= GFX11; }
 
   bool hasVALUPartialForwardingHazard() const {
     return getGeneration() >= GFX11;
   }
 
   bool hasVALUTransUseHazard() const { return HasVALUTransUseHazard; }
 
+  bool hasForceStoreSC0SC1() const { return HasForceStoreSC0SC1; }
+
   bool hasVALUMaskWriteHazard() const { return getGeneration() >= GFX11; }
 
   /// Return if operations acting on VGPR tuples require even alignment.
   bool needsAlignedVGPRs() const { return GFX90AInsts; }
 
   /// Return true if the target has the S_PACK_HL_B32_B16 instruction.
   bool hasSPackHL() const { return GFX11Insts; }
 

llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp

virtual bool insertRelease(MachineBasicBlock::iterator &MI,
                              SIAtomicScope Scope,
                              SIAtomicAddrSpace AddrSpace,
                              bool IsCrossAddrSpaceOrdering,
                              Position Pos) const = 0;
 
   /// Virtual destructor to allow derivations to be deleted.
   virtual ~SICacheControl() = default;
 
+  virtual bool tryForceStoreSC0SC1(const SIMemOpInfo &MOI,
+                                   MachineBasicBlock::iterator &MI) const {
+    return false;
+  }
 };
 
 class SIGfx6CacheControl : public SICacheControl {
 protected:
 
   /// Sets GLC bit to "true" if present in \p MI. Returns true if \p MI
   /// is modified, false otherwise.
   bool enableGLCBit(const MachineBasicBlock::iterator &MI) const {
bool enableVolatileAndOrNonTemporal(MachineBasicBlock::iterator &MI,
                                       bool IsNonTemporal) const override;
 
   bool insertAcquire(MachineBasicBlock::iterator &MI, SIAtomicScope Scope,
                      SIAtomicAddrSpace AddrSpace, Position Pos) const override;
 
   bool insertRelease(MachineBasicBlock::iterator &MI, SIAtomicScope Scope,
                      SIAtomicAddrSpace AddrSpace, bool IsCrossAddrSpaceOrdering,
                      Position Pos) const override;
+
+  bool tryForceStoreSC0SC1(const SIMemOpInfo &MOI,
+                           MachineBasicBlock::iterator &MI) const override {
+    if (ST.hasForceStoreSC0SC1() &&
+        (MOI.getInstrAddrSpace() & (SIAtomicAddrSpace::SCRATCH |
+                                    SIAtomicAddrSpace::GLOBAL |
+                                    SIAtomicAddrSpace::OTHER)) !=
+         SIAtomicAddrSpace::NONE)
+      return enableSC0Bit(MI) | enableSC1Bit(MI);
+    return false;
+  }
 };
 
 class SIGfx10CacheControl : public SIGfx7CacheControl {
 protected:
 
   /// Sets DLC bit to "true" if present in \p MI. Returns true if \p MI
   /// is modified, false otherwise.
   bool enableDLCBit(const MachineBasicBlock::iterator &MI) const {
     return enableNamedBit(MI, AMDGPU::CPol::DLC);
   }
 
for (auto MI = MBB.begin(); MI != MBB.end(); ++MI) {
         MI = II->getIterator();
       }
 
       if (!(MI->getDesc().TSFlags & SIInstrFlags::maybeAtomic))
         continue;
 
       if (const auto &MOI = MOA.getLoadInfo(MI))
         Changed |= expandLoad(*MOI, MI);
-      else if (const auto &MOI = MOA.getStoreInfo(MI))
+      else if (const auto &MOI = MOA.getStoreInfo(MI)) {
         Changed |= expandStore(*MOI, MI);
-      else if (const auto &MOI = MOA.getAtomicFenceInfo(MI))
+        Changed |= CC->tryForceStoreSC0SC1(*MOI, MI);
+      } else if (const auto &MOI = MOA.getAtomicFenceInfo(MI))
         Changed |= expandAtomicFence(*MOI, MI);
       else if (const auto &MOI = MOA.getAtomicCmpxchgOrRmwInfo(MI))
         Changed |= expandAtomicCmpxchgOrRmw(*MOI, MI);
     }
   }
 
   Changed |= removeAtomicPseudoMIs();
   return Changed;

llvm/lib/TargetParser/TargetParser.cpp

case GK_GFX1010:
       Features["16-bit-insts"] = true;
       Features["dpp"] = true;
       Features["gfx8-insts"] = true;
       Features["gfx9-insts"] = true;
       Features["gfx10-insts"] = true;
       Features["s-memrealtime"] = true;
       Features["s-memtime-inst"] = true;
       break;
-    case GK_GFX942:
     case GK_GFX941:
     case GK_GFX940:
+      Features["force-store-sc0-sc1"] = true;
+      [[fallthrough]];
+    case GK_GFX942:
       Features["gfx940-insts"] = true;
       Features["fp8-insts"] = true;
       Features["atomic-ds-pk-add-16-insts"] = true;
       Features["atomic-flat-pk-add-16-insts"] = true;
       Features["atomic-global-pk-add-bf16-inst"] = true;
       [[fallthrough]];
     case GK_GFX90A:
       Features["gfx90a-insts"] = true;
       Features["atomic-buffer-global-pk-add-f16-insts"] = true;

llvm/test/CodeGen/AMDGPU/force-store-sc0-sc1.ll

+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
+; RUN: llc -march=amdgcn -mcpu=gfx940 -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,FORCESC0SC1 %s
+; RUN: llc -march=amdgcn -mcpu=gfx941 -verify-machineinstrs < %s | FileCheck --check-prefixes=GCN,FORCESC0SC1 %s
+; RUN: llc -march=amdgcn -mcpu=gfx942 -verify-machineinstrs -mattr=-forcestoresc1 < %s | FileCheck --check-prefixes=GCN,NOSC0SC1 %s
+
+define amdgpu_kernel void @store_global(ptr addrspace(1) %ptr) {
+; FORCESC0SC1-LABEL: store_global:
+; FORCESC0SC1:       ; %bb.0: ; %entry
+; FORCESC0SC1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; FORCESC0SC1-NEXT:    v_mov_b32_e32 v0, 0
+; FORCESC0SC1-NEXT:    v_mov_b32_e32 v1, 1.0
+; FORCESC0SC1-NEXT:    s_waitcnt lgkmcnt(0)
+; FORCESC0SC1-NEXT:    global_store_dword v0, v1, s[0:1] sc0 sc1
+; FORCESC0SC1-NEXT:    s_endpgm
+;
+; NOSC0SC1-LABEL: store_global:
+; NOSC0SC1:       ; %bb.0: ; %entry
+; NOSC0SC1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; NOSC0SC1-NEXT:    v_mov_b32_e32 v0, 0
+; NOSC0SC1-NEXT:    v_mov_b32_e32 v1, 1.0
+; NOSC0SC1-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSC0SC1-NEXT:    global_store_dword v0, v1, s[0:1]
+; NOSC0SC1-NEXT:    s_endpgm
+entry:
+  store float 1.000000e+00, ptr addrspace(1) %ptr, align 4
+  ret void
+}
+
+define amdgpu_kernel void @store_flat(ptr addrspace(0) %ptr) {
+; FORCESC0SC1-LABEL: store_flat:
+; FORCESC0SC1:       ; %bb.0: ; %entry
+; FORCESC0SC1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; FORCESC0SC1-NEXT:    v_mov_b32_e32 v2, 1.0
+; FORCESC0SC1-NEXT:    s_waitcnt lgkmcnt(0)
+; FORCESC0SC1-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; FORCESC0SC1-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
+; FORCESC0SC1-NEXT:    s_endpgm
+;
+; NOSC0SC1-LABEL: store_flat:
+; NOSC0SC1:       ; %bb.0: ; %entry
+; NOSC0SC1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; NOSC0SC1-NEXT:    v_mov_b32_e32 v2, 1.0
+; NOSC0SC1-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSC0SC1-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; NOSC0SC1-NEXT:    flat_store_dword v[0:1], v2
+; NOSC0SC1-NEXT:    s_endpgm
+entry:
+  store float 1.000000e+00, ptr addrspace(0) %ptr, align 4
+  ret void
+}
+
+define amdgpu_kernel void @store_lds(ptr addrspace(3) %ptr) {
+; GCN-LABEL: store_lds:
+; GCN:       ; %bb.0: ; %entry
+; GCN-NEXT:    s_load_dword s0, s[0:1], 0x24
+; GCN-NEXT:    v_mov_b32_e32 v0, 1.0
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    v_mov_b32_e32 v1, s0
+; GCN-NEXT:    ds_write_b32 v1, v0
+; GCN-NEXT:    s_endpgm
+entry:
+  store float 1.000000e+00, ptr addrspace(3) %ptr, align 4
+  ret void
+}
+
+define amdgpu_kernel void @store_scratch(ptr addrspace(5) %ptr) {
+; FORCESC0SC1-LABEL: store_scratch:
+; FORCESC0SC1:       ; %bb.0: ; %entry
+; FORCESC0SC1-NEXT:    s_load_dword s0, s[0:1], 0x24
+; FORCESC0SC1-NEXT:    v_mov_b32_e32 v0, 1.0
+; FORCESC0SC1-NEXT:    s_waitcnt lgkmcnt(0)
+; FORCESC0SC1-NEXT:    scratch_store_dword off, v0, s0 sc0 sc1
+; FORCESC0SC1-NEXT:    s_endpgm
+;
+; NOSC0SC1-LABEL: store_scratch:
+; NOSC0SC1:       ; %bb.0: ; %entry
+; NOSC0SC1-NEXT:    s_load_dword s0, s[0:1], 0x24
+; NOSC0SC1-NEXT:    v_mov_b32_e32 v0, 1.0
+; NOSC0SC1-NEXT:    s_waitcnt lgkmcnt(0)
+; NOSC0SC1-NEXT:    scratch_store_dword off, v0, s0
+; NOSC0SC1-NEXT:    s_endpgm
+entry:
+  store float 1.000000e+00, ptr addrspace(5) %ptr, align 4
+  ret void
+}
+
+define amdgpu_ps void @store_buffer(<4 x i32> inreg %rsrc, float %data, i32 %index) {
+; FORCESC0SC1-LABEL: store_buffer:
+; FORCESC0SC1:       ; %bb.0: ; %main_body
+; FORCESC0SC1-NEXT:    buffer_store_dword v0, v1, s[0:3], 0 idxen sc0 sc1
+; FORCESC0SC1-NEXT:    s_endpgm
+;
+; NOSC0SC1-LABEL: store_buffer:
+; NOSC0SC1:       ; %bb.0: ; %main_body
+; NOSC0SC1-NEXT:    buffer_store_dword v0, v1, s[0:3], 0 idxen
+; NOSC0SC1-NEXT:    s_endpgm
+main_body:
+  call void @llvm.amdgcn.buffer.store.f32(float %data, <4 x i32> %rsrc, i32 %index, i32 0, i1 0, i1 0)
+  ret void
+}
+
+define amdgpu_kernel void @store_global_atomic(ptr addrspace(1) %ptr) {
+; FORCESC0SC1-LABEL: store_global_atomic:
+; FORCESC0SC1:       ; %bb.0: ; %entry
+; FORCESC0SC1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; FORCESC0SC1-NEXT:    v_mov_b32_e32 v0, 0
+; FORCESC0SC1-NEXT:    v_mov_b32_e32 v1, 1.0
+; FORCESC0SC1-NEXT:    buffer_wbl2 sc1
+; FORCESC0SC1-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; FORCESC0SC1-NEXT:    global_store_dword v0, v1, s[0:1] sc0 sc1
+; FORCESC0SC1-NEXT:    s_endpgm
+;
+; NOSC0SC1-LABEL: store_global_atomic:
+; NOSC0SC1:       ; %bb.0: ; %entry
+; NOSC0SC1-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; NOSC0SC1-NEXT:    v_mov_b32_e32 v0, 0
+; NOSC0SC1-NEXT:    v_mov_b32_e32 v1, 1.0
+; NOSC0SC1-NEXT:    buffer_wbl2 sc1
+; NOSC0SC1-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; NOSC0SC1-NEXT:    global_store_dword v0, v1, s[0:1] sc1
+; NOSC0SC1-NEXT:    s_endpgm
+entry:
+  store atomic float 1.000000e+00, ptr addrspace(1) %ptr syncscope("agent-one-as") seq_cst, align 4
+  ret void
+}
+
+define amdgpu_kernel void @store_global_atomic_system(ptr addrspace(1) %ptr) {
+; GCN-LABEL: store_global_atomic_system:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
+; GCN-NEXT:    v_mov_b32_e32 v0, 0
+; GCN-NEXT:    v_mov_b32_e32 v1, 1.0
+; GCN-NEXT:    s_waitcnt lgkmcnt(0)
+; GCN-NEXT:    global_store_dword v0, v1, s[0:1] sc0 sc1
+; GCN-NEXT:    s_endpgm
+  store atomic float 1.000000e+00, ptr addrspace(1) %ptr monotonic, align 4
+  ret void
+}
+
+
+declare void @llvm.amdgcn.buffer.store.f32(float, <4 x i32>, i32, i32, i1, i1)

llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx940.ll

-; RUN: llc -march=amdgcn -mcpu=gfx940 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX940,VGPRCD %s
-; RUN: llc -march=amdgcn -mcpu=gfx940 -global-isel -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GISEL,VGPRCD %s
-; RUN: llc -march=amdgcn -mcpu=gfx940 -stress-regalloc=10 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX940,AGPRCD %s
-; RUN: llc -march=amdgcn -mcpu=gfx940 -stress-regalloc=10 -global-isel -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GISEL,AGPRCD %s
+; RUN: llc -march=amdgcn -mcpu=gfx942 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX940,VGPRCD %s
+; RUN: llc -march=amdgcn -mcpu=gfx942 -global-isel -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GISEL,VGPRCD %s
+; RUN: llc -march=amdgcn -mcpu=gfx942 -stress-regalloc=10 -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX940,AGPRCD %s
+; RUN: llc -march=amdgcn -mcpu=gfx942 -stress-regalloc=10 -global-isel -verify-machineinstrs < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GISEL,AGPRCD %s
 
 declare <4 x i32> @llvm.amdgcn.mfma.i32.16x16x32.i8(i64, i64, <4 x i32>, i32, i32, i32)
 declare <16 x i32> @llvm.amdgcn.mfma.i32.32x32x16.i8(i64, i64, <16 x i32>, i32, i32, i32)
 declare <4 x float> @llvm.amdgcn.mfma.f32.16x16x8.xf32(<2 x float>, <2 x float>, <4 x float>, i32, i32, i32)
 declare <16 x float> @llvm.amdgcn.mfma.f32.32x32x4.xf32(<2 x float>, <2 x float>, <16 x float>, i32, i32, i32)
 declare <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.bf8.bf8(i64, i64, <4 x float>, i32, i32, i32)
 declare <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.bf8.fp8(i64, i64, <4 x float>, i32, i32, i32)
 declare <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.fp8.bf8(i64, i64, <4 x float>, i32, i32, i32)

llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-agent.ll

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_unordered_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_unordered_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    flat_load_b32 v2, v[0:1]

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc1
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_monotonic_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc1
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_monotonic_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    flat_load_b32 v2, v[0:1] glc

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_acquire_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_acquire_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    flat_load_b32 v2, v[0:1] glc

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_seq_cst_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_seq_cst_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)

 ;
 ; GFX940-NOTTGSPLIT-LABEL: flat_agent_unordered_store:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_unordered_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_unordered_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ;
 ; GFX940-NOTTGSPLIT-LABEL: flat_agent_monotonic_store:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc1
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_monotonic_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc1
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_monotonic_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc1
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_release_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc1
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_release_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc1
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_seq_cst_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc1
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_seq_cst_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_swap v2, v[0:1], v2 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_acquire_ret_atomicrmw:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_swap v2, v[0:1], v2 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_acquire_ret_atomicrmw:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_swap v2, v[0:1], v2 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_acq_rel_ret_atomicrmw:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_swap v2, v[0:1], v2 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_acq_rel_ret_atomicrmw:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_swap v2, v[0:1], v2 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_seq_cst_ret_atomicrmw:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_swap v2, v[0:1], v2 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_seq_cst_ret_atomicrmw:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-LABEL: flat_agent_monotonic_monotonic_ret_cmpxchg:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_monotonic_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_monotonic_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_acquire_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_acquire_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_release_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_release_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_acq_rel_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_acq_rel_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_seq_cst_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_seq_cst_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_monotonic_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_monotonic_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_acquire_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_acquire_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_release_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_release_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_acq_rel_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_acq_rel_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_seq_cst_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_seq_cst_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_monotonic_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_monotonic_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_acquire_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_acquire_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_release_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_release_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_acq_rel_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_acq_rel_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_unordered_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_unordered_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    flat_load_b32 v2, v[0:1]

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc1
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_monotonic_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc1
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_monotonic_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    flat_load_b32 v2, v[0:1] glc

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_acquire_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_acquire_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    flat_load_b32 v2, v[0:1] glc

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_seq_cst_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1] sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_seq_cst_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0)

 ;
 ; GFX940-NOTTGSPLIT-LABEL: flat_agent_one_as_unordered_store:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_unordered_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_unordered_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ;
 ; GFX940-NOTTGSPLIT-LABEL: flat_agent_one_as_monotonic_store:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc1
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_monotonic_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc1
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_monotonic_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc1
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_release_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc1
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_release_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc1
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_seq_cst_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc1
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_seq_cst_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_swap v2, v[0:1], v2 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_acquire_ret_atomicrmw:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_swap v2, v[0:1], v2 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_acquire_ret_atomicrmw:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_swap v2, v[0:1], v2 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_acq_rel_ret_atomicrmw:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_swap v2, v[0:1], v2 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_acq_rel_ret_atomicrmw:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_swap v2, v[0:1], v2 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_seq_cst_ret_atomicrmw:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_swap v2, v[0:1], v2 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_seq_cst_ret_atomicrmw:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-LABEL: flat_agent_one_as_monotonic_monotonic_ret_cmpxchg:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_monotonic_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_monotonic_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_acquire_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_acquire_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_release_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_release_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_acq_rel_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_acq_rel_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_seq_cst_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_seq_cst_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_monotonic_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_monotonic_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_acquire_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_acquire_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_release_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_release_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_acq_rel_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_acq_rel_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_seq_cst_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_seq_cst_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_monotonic_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_monotonic_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_acquire_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_acquire_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_release_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_release_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_acq_rel_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_acq_rel_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_agent_one_as_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    flat_atomic_cmpswap v2, v[0:1], v[2:3] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_agent_one_as_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v2, s2 :: v_dual_mov_b32 v3, s3

llvm/test/CodeGen/AMDGPU/memory-legalizer-flat-nontemporal.ll

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1] nt
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_nontemporal_load_0:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1] nt
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_nontemporal_load_0:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    flat_load_b32 v2, v[0:1] slc dlc

 ; GFX940-NOTTGSPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_lshl_add_u64 v[0:1], s[0:1], 0, v[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1] nt
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_nontemporal_load_1:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_lshl_add_u64 v[0:1], s[0:1], 0, v[0:1]
 ; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1] nt
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_nontemporal_load_1:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 nt
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 nt sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_nontemporal_store_0:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s1
 ; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[0:1]
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s2
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s3
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 nt
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 nt sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_nontemporal_store_0:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s0 :: v_dual_mov_b32 v1, s1
 ; GFX11-WGP-NEXT:    flat_load_b32 v2, v[0:1]

 ; GFX940-NOTTGSPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX940-NOTTGSPLIT-NEXT:    flat_load_dword v2, v[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    v_lshl_add_u64 v[0:1], s[2:3], 0, v[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 nt
+; GFX940-NOTTGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 nt sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: flat_nontemporal_store_1:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, s0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX940-TGSPLIT-NEXT:    flat_load_dword v2, v[2:3]
 ; GFX940-TGSPLIT-NEXT:    v_lshl_add_u64 v[0:1], s[2:3], 0, v[0:1]
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 nt
+; GFX940-TGSPLIT-NEXT:    flat_store_dword v[0:1], v2 sc0 nt sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: flat_nontemporal_store_1:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v1, s0 :: v_dual_lshlrev_b32 v0, 2, v0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, s1

llvm/test/CodeGen/AMDGPU/memory-legalizer-global-agent.ll

 ;
 ; GFX940-NOTTGSPLIT-LABEL: global_agent_unordered_load:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_unordered_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1]
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_unordered_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]

 ;
 ; GFX940-NOTTGSPLIT-LABEL: global_agent_monotonic_load:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_monotonic_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_monotonic_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] glc

 ; GFX940-NOTTGSPLIT-LABEL: global_agent_acquire_load:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_acquire_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_acquire_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] glc

 ; GFX940-NOTTGSPLIT-LABEL: global_agent_seq_cst_load:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_seq_cst_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_seq_cst_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0

 ;
 ; GFX940-NOTTGSPLIT-LABEL: global_agent_unordered_store:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_unordered_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_unordered_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ;
 ; GFX940-NOTTGSPLIT-LABEL: global_agent_monotonic_store:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc1
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_monotonic_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc1
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_monotonic_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc1
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_release_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc1
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_release_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc1
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_seq_cst_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc1
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_seq_cst_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_swap v1, v0, v1, s[2:3] sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_acquire_ret_atomicrmw:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-TGSPLIT-NEXT:    global_atomic_swap v1, v0, v1, s[2:3] sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_acquire_ret_atomicrmw:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_swap v1, v0, v1, s[2:3] sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_acq_rel_ret_atomicrmw:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_swap v1, v0, v1, s[2:3] sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_acq_rel_ret_atomicrmw:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_swap v1, v0, v1, s[2:3] sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_seq_cst_ret_atomicrmw:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_swap v1, v0, v1, s[2:3] sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_seq_cst_ret_atomicrmw:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-LABEL: global_agent_monotonic_monotonic_ret_cmpxchg:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_monotonic_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_monotonic_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_acquire_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_acquire_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_release_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_release_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_acq_rel_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_acq_rel_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_seq_cst_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_seq_cst_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_monotonic_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_monotonic_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_acquire_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_acquire_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_release_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_release_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_acq_rel_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_acq_rel_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_seq_cst_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_seq_cst_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_monotonic_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_monotonic_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_acquire_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_acquire_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_release_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_release_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_acq_rel_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_acq_rel_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ;
 ; GFX940-NOTTGSPLIT-LABEL: global_agent_one_as_unordered_load:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1]
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_unordered_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1]
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_unordered_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1]

 ;
 ; GFX940-NOTTGSPLIT-LABEL: global_agent_one_as_monotonic_load:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_monotonic_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_monotonic_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] glc

 ; GFX940-NOTTGSPLIT-LABEL: global_agent_one_as_acquire_load:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_acquire_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_acquire_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    global_load_b32 v1, v0, s[0:1] glc

 ; GFX940-NOTTGSPLIT-LABEL: global_agent_one_as_seq_cst_load:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_seq_cst_load:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_load_dword v1, v0, s[0:1] sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_seq_cst_load:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-WGP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_waitcnt_vscnt null, 0x0

 ;
 ; GFX940-NOTTGSPLIT-LABEL: global_agent_one_as_unordered_store:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_unordered_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_unordered_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ;
 ; GFX940-NOTTGSPLIT-LABEL: global_agent_one_as_monotonic_store:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc1
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_monotonic_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc1
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_monotonic_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc1
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_release_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc1
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_release_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc1
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_seq_cst_store:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc1
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_seq_cst_store:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_swap v1, v0, v1, s[2:3] sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_acquire_ret_atomicrmw:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-TGSPLIT-NEXT:    global_atomic_swap v1, v0, v1, s[2:3] sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_acquire_ret_atomicrmw:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_swap v1, v0, v1, s[2:3] sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_acq_rel_ret_atomicrmw:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_swap v1, v0, v1, s[2:3] sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_acq_rel_ret_atomicrmw:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_swap v1, v0, v1, s[2:3] sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_seq_cst_ret_atomicrmw:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s4
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_swap v1, v0, v1, s[2:3] sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_seq_cst_ret_atomicrmw:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-LABEL: global_agent_one_as_monotonic_monotonic_ret_cmpxchg:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_monotonic_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_monotonic_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_acquire_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_acquire_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_acq_rel_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_acq_rel_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_seq_cst_monotonic_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_seq_cst_monotonic_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_monotonic_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_monotonic_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_acquire_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_acquire_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_release_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_release_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_acq_rel_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_acq_rel_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_seq_cst_acquire_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_seq_cst_acquire_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_monotonic_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_monotonic_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_acquire_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_acquire_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_release_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_release_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_acq_rel_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_acq_rel_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_agent_one_as_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
 ; GFX940-TGSPLIT-NEXT:    buffer_wbl2 sc1
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] offset:16 sc0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    buffer_inv sc1
-; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v2, v0, s[0:1] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_agent_one_as_seq_cst_seq_cst_ret_cmpxchg:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v0, s2 :: v_dual_mov_b32 v1, s3

llvm/test/CodeGen/AMDGPU/memory-legalizer-global-nontemporal.ll

 ; GFX940-NOTTGSPLIT-LABEL: global_nontemporal_load_0:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s0, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s0
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_nontemporal_load_0:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s0, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s0
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_nontemporal_load_0:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-LABEL: global_nontemporal_load_1:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    global_load_dword v0, v0, s[0:1] nt
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v1, v0, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_nontemporal_load_1:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    global_load_dword v0, v0, s[0:1] nt
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v1, v0, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_nontemporal_load_1:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    v_dual_mov_b32 v1, 0 :: v_dual_lshlrev_b32 v0, 2, v0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    global_load_b32 v0, v0, s[0:1] slc dlc

 ; GFX940-NOTTGSPLIT-LABEL: global_nontemporal_store_0:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s0, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s0
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] nt
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 nt sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_nontemporal_store_0:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s0, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s0
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] nt
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 nt sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_nontemporal_store_0:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT-LABEL: global_nontemporal_store_1:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s0, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s0
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] nt
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 nt sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: global_nontemporal_store_1:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx4 s[0:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    v_lshlrev_b32_e32 v0, 2, v0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s0, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s0
-; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] nt
+; GFX940-TGSPLIT-NEXT:    global_store_dword v0, v1, s[2:3] sc0 nt sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: global_nontemporal_store_1:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_load_b128 s[0:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

llvm/test/CodeGen/AMDGPU/memory-legalizer-private-nontemporal.ll

 ; GFX940-NOTTGSPLIT-LABEL: private_nontemporal_load_0:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    scratch_load_dword v0, off, s4 nt
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v1, v0, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: private_nontemporal_load_0:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    scratch_load_dword v0, off, s4 nt
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v1, v0, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: private_nontemporal_load_0:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, 0

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_lshl_add_u32 v0, v0, 2, s4
 ; GFX940-NOTTGSPLIT-NEXT:    scratch_load_dword v0, v0, off nt
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX940-NOTTGSPLIT-NEXT:    global_store_dword v1, v0, s[2:3] sc0 sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: private_nontemporal_load_1:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_lshl_add_u32 v0, v0, 2, s4
 ; GFX940-TGSPLIT-NEXT:    scratch_load_dword v0, v0, off nt
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt vmcnt(0)
-; GFX940-TGSPLIT-NEXT:    global_store_dword v1, v0, s[2:3]
+; GFX940-TGSPLIT-NEXT:    global_store_dword v1, v0, s[2:3] sc0 sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: private_nontemporal_load_1:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b32 s2, s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x8
 ; GFX11-WGP-NEXT:    v_mov_b32_e32 v1, 0

 ; GFX940-NOTTGSPLIT-LABEL: private_nontemporal_store_0:
 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s0, s[2:3], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
-; GFX940-NOTTGSPLIT-NEXT:    scratch_store_dword off, v0, s4 nt
+; GFX940-NOTTGSPLIT-NEXT:    scratch_store_dword off, v0, s4 sc0 nt sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: private_nontemporal_store_0:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s0, s[2:3], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v0, s0
-; GFX940-TGSPLIT-NEXT:    scratch_store_dword off, v0, s4 nt
+; GFX940-TGSPLIT-NEXT:    scratch_store_dword off, v0, s4 sc0 nt sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: private_nontemporal_store_0:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)

 ; GFX940-NOTTGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_lshl_add_u32 v0, v0, 2, s4
 ; GFX940-NOTTGSPLIT-NEXT:    s_load_dword s0, s[2:3], 0x0
 ; GFX940-NOTTGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-NOTTGSPLIT-NEXT:    v_mov_b32_e32 v1, s0
-; GFX940-NOTTGSPLIT-NEXT:    scratch_store_dword v0, v1, off nt
+; GFX940-NOTTGSPLIT-NEXT:    scratch_store_dword v0, v1, off sc0 nt sc1
 ; GFX940-NOTTGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX940-TGSPLIT-LABEL: private_nontemporal_store_1:
 ; GFX940-TGSPLIT:       ; %bb.0: ; %entry
 ; GFX940-TGSPLIT-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s4, s[0:1], 0x8
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_lshl_add_u32 v0, v0, 2, s4
 ; GFX940-TGSPLIT-NEXT:    s_load_dword s0, s[2:3], 0x0
 ; GFX940-TGSPLIT-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX940-TGSPLIT-NEXT:    v_mov_b32_e32 v1, s0
-; GFX940-TGSPLIT-NEXT:    scratch_store_dword v0, v1, off nt
+; GFX940-TGSPLIT-NEXT:    scratch_store_dword v0, v1, off sc0 nt sc1
 ; GFX940-TGSPLIT-NEXT:    s_endpgm
 ;
 ; GFX11-WGP-LABEL: private_nontemporal_store_1:
 ; GFX11-WGP:       ; %bb.0: ; %entry
 ; GFX11-WGP-NEXT:    s_clause 0x1
 ; GFX11-WGP-NEXT:    s_load_b64 s[2:3], s[0:1], 0x0
 ; GFX11-WGP-NEXT:    s_load_b32 s0, s[0:1], 0x8
 ; GFX11-WGP-NEXT:    s_waitcnt lgkmcnt(0)