diff --git a/llvm/include/llvm/CodeGen/AtomicExpandUtils.h b/llvm/include/llvm/CodeGen/AtomicExpandUtils.h
--- a/llvm/include/llvm/CodeGen/AtomicExpandUtils.h
+++ b/llvm/include/llvm/CodeGen/AtomicExpandUtils.h
@@ -10,6 +10,7 @@
 #define LLVM_CODEGEN_ATOMICEXPANDUTILS_H
 
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/Analysis/TargetTransformInfo.h"
 #include "llvm/IR/IRBuilder.h"
 #include "llvm/Support/AtomicOrdering.h"
 
@@ -57,7 +58,9 @@
 ///     [...]
 ///
 /// Returns true if the containing function was modified.
-bool expandAtomicRMWToCmpXchg(AtomicRMWInst *AI, CreateCmpXchgInstFun CreateCmpXchg);
+bool expandAtomicRMWToCmpXchg(AtomicRMWInst *AI,
+                              CreateCmpXchgInstFun CreateCmpXchg,
+                              SmallVector<BasicBlock *> &CmpXchgLoopBlocks);
 
 } // end namespace llvm
 
diff --git a/llvm/lib/CodeGen/AtomicExpandPass.cpp b/llvm/lib/CodeGen/AtomicExpandPass.cpp
--- a/llvm/lib/CodeGen/AtomicExpandPass.cpp
+++ b/llvm/lib/CodeGen/AtomicExpandPass.cpp
@@ -17,6 +17,7 @@
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/STLFunctionalExtras.h"
 #include "llvm/ADT/SmallVector.h"
+#include "llvm/Analysis/DomTreeUpdater.h"
 #include "llvm/Analysis/InstSimplifyFolder.h"
 #include "llvm/Analysis/OptimizationRemarkEmitter.h"
 #include "llvm/CodeGen/AtomicExpandUtils.h"
@@ -48,6 +49,7 @@
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/Target/TargetMachine.h"
+#include "llvm/Transforms/Utils/Local.h"
 #include "llvm/Transforms/Utils/LowerAtomic.h"
 #include <cassert>
 #include <cstdint>
@@ -62,6 +64,7 @@
 class AtomicExpand : public FunctionPass {
   const TargetLowering *TLI = nullptr;
   const DataLayout *DL = nullptr;
+  SmallVector<BasicBlock *> CmpXchgLoopBlocks;
 
 public:
   static char ID; // Pass identification, replacement for typeid
@@ -72,6 +75,11 @@
 
   bool runOnFunction(Function &F) override;
 
+  void getAnalysisUsage(AnalysisUsage &AU) const override {
+    AU.addRequired<DominatorTreeWrapperPass>();
+    AU.addRequired<TargetTransformInfoWrapperPass>();
+  }
+
 private:
   bool bracketInstWithFences(Instruction *I, AtomicOrdering Order);
   IntegerType *getCorrespondingIntegerType(Type *T, const DataLayout &DL);
@@ -104,7 +112,8 @@
       IRBuilderBase &Builder, Type *ResultType, Value *Addr, Align AddrAlign,
       AtomicOrdering MemOpOrder, SyncScope::ID SSID,
       function_ref<Value *(IRBuilderBase &, Value *)> PerformOp,
-      CreateCmpXchgInstFun CreateCmpXchg);
+      CreateCmpXchgInstFun CreateCmpXchg,
+      SmallVector<BasicBlock *> &CmpXchgLoopBlocks);
   bool tryExpandAtomicCmpXchg(AtomicCmpXchgInst *CI);
 
   bool expandAtomicCmpXchg(AtomicCmpXchgInst *CI);
@@ -123,7 +132,8 @@
 
   friend bool
   llvm::expandAtomicRMWToCmpXchg(AtomicRMWInst *AI,
-                                 CreateCmpXchgInstFun CreateCmpXchg);
+                                 CreateCmpXchgInstFun CreateCmpXchg,
+                                 SmallVector<BasicBlock *> &CmpXchgLoopBlocks);
 };
 
 // IRBuilder to be used for replacement atomic instructions.
@@ -142,9 +152,12 @@
 
 char &llvm::AtomicExpandID = AtomicExpand::ID;
 
-INITIALIZE_PASS(AtomicExpand, DEBUG_TYPE, "Expand Atomic instructions", false,
-                false)
-
+INITIALIZE_PASS_BEGIN(AtomicExpand, DEBUG_TYPE, "Expand Atomic instructions",
+                      false, false)
+INITIALIZE_PASS_DEPENDENCY(DominatorTreeWrapperPass)
+INITIALIZE_PASS_DEPENDENCY(TargetTransformInfoWrapperPass)
+INITIALIZE_PASS_END(AtomicExpand, DEBUG_TYPE, "Expand Atomic instructions",
+                    false, false)
 FunctionPass *llvm::createAtomicExpandPass() { return new AtomicExpand(); }
 
 // Helper functions to retrieve the size of atomic instructions.
@@ -190,6 +203,7 @@
     return false;
   TLI = Subtarget->getTargetLowering();
   DL = &F.getParent()->getDataLayout();
+  CmpXchgLoopBlocks.clear();
 
   SmallVector<Instruction *, 1> AtomicInsts;
 
@@ -337,6 +351,22 @@
     } else if (CASI)
       MadeChange |= tryExpandAtomicCmpXchg(CASI);
   }
+
+  DominatorTreeWrapperPass *const DTW =
+      getAnalysisIfAvailable<DominatorTreeWrapperPass>();
+  DomTreeUpdater DTU(DTW ? &DTW->getDomTree() : nullptr,
+                     DomTreeUpdater::UpdateStrategy::Eager);
+  auto TTI = &getAnalysis<TargetTransformInfoWrapperPass>().getTTI(F);
+  for (BasicBlock *BB : CmpXchgLoopBlocks) {
+    simplifyCFG(BB, *TTI, RequireAndPreserveDomTree ? &DTU : nullptr,
+                SimplifyCFGOptions()
+                    .forwardSwitchCondToPhi(true)
+                    .convertSwitchRangeToICmp(true)
+                    .convertSwitchToLookupTable(true)
+                    .needCanonicalLoops(false)
+                    .hoistCommonInsts(true)
+                    .sinkCommonInsts(true));
+  }
   return MadeChange;
 }
 
@@ -602,7 +632,7 @@
                << AI->getOperationName(AI->getOperation()) << " operation at "
                << MemScope << " memory scope";
       });
-      expandAtomicRMWToCmpXchg(AI, createCmpXchgInstFun);
+      expandAtomicRMWToCmpXchg(AI, createCmpXchgInstFun, CmpXchgLoopBlocks);
     }
     return true;
   }
@@ -873,7 +903,8 @@
   if (ExpansionKind == TargetLoweringBase::AtomicExpansionKind::CmpXChg) {
     OldResult = insertRMWCmpXchgLoop(Builder, PMV.WordType, PMV.AlignedAddr,
                                      PMV.AlignedAddrAlignment, MemOpOrder, SSID,
-                                     PerformPartwordOp, createCmpXchgInstFun);
+                                     PerformPartwordOp, createCmpXchgInstFun,
+                                     CmpXchgLoopBlocks);
   } else {
     assert(ExpansionKind == TargetLoweringBase::AtomicExpansionKind::LLSC);
     OldResult = insertRMWLLSCLoop(Builder, PMV.WordType, PMV.AlignedAddr,
@@ -1479,7 +1510,8 @@
     IRBuilderBase &Builder, Type *ResultTy, Value *Addr, Align AddrAlign,
     AtomicOrdering MemOpOrder, SyncScope::ID SSID,
     function_ref<Value *(IRBuilderBase &, Value *)> PerformOp,
-    CreateCmpXchgInstFun CreateCmpXchg) {
+    CreateCmpXchgInstFun CreateCmpXchg,
+    SmallVector<BasicBlock *> &CmpXchgLoopBlocks) {
   LLVMContext &Ctx = Builder.getContext();
   BasicBlock *BB = Builder.GetInsertBlock();
   Function *F = BB->getParent();
@@ -1501,8 +1533,9 @@
   //     [...]
   BasicBlock *ExitBB =
       BB->splitBasicBlock(Builder.GetInsertPoint(), "atomicrmw.end");
+  CmpXchgLoopBlocks.push_back(ExitBB);
   BasicBlock *LoopBB = BasicBlock::Create(Ctx, "atomicrmw.start", F, ExitBB);
-
+  CmpXchgLoopBlocks.push_back(LoopBB);
   // The split call above "helpfully" added a branch at the end of BB (to the
   // wrong place), but we want a load. It's easiest to just remove
   // the branch entirely.
@@ -1559,8 +1592,9 @@
 }
 
 // Note: This function is exposed externally by AtomicExpandUtils.h
-bool llvm::expandAtomicRMWToCmpXchg(AtomicRMWInst *AI,
-                                    CreateCmpXchgInstFun CreateCmpXchg) {
+bool llvm::expandAtomicRMWToCmpXchg(
+    AtomicRMWInst *AI, CreateCmpXchgInstFun CreateCmpXchg,
+    SmallVector<BasicBlock *> &CmpXchgLoopBlocks) {
   ReplacementIRBuilder Builder(AI, AI->getModule()->getDataLayout());
   Builder.setIsFPConstrained(
       AI->getFunction()->hasFnAttribute(Attribute::StrictFP));
@@ -1574,7 +1608,7 @@
         return buildAtomicRMWValue(AI->getOperation(), Builder, Loaded,
                                    AI->getValOperand());
       },
-      CreateCmpXchg);
+      CreateCmpXchg, CmpXchgLoopBlocks);
 
   AI->replaceAllUsesWith(Loaded);
   AI->eraseFromParent();
@@ -1722,9 +1756,10 @@
   // CAS libcall, via a CAS loop, instead.
   if (!Success) {
     expandAtomicRMWToCmpXchg(
-        I, [this](IRBuilderBase &Builder, Value *Addr, Value *Loaded,
-                  Value *NewVal, Align Alignment, AtomicOrdering MemOpOrder,
-                  SyncScope::ID SSID, Value *&Success, Value *&NewLoaded) {
+        I,
+        [this](IRBuilderBase &Builder, Value *Addr, Value *Loaded,
+               Value *NewVal, Align Alignment, AtomicOrdering MemOpOrder,
+               SyncScope::ID SSID, Value *&Success, Value *&NewLoaded) {
           // Create the CAS instruction normally...
           AtomicCmpXchgInst *Pair = Builder.CreateAtomicCmpXchg(
               Addr, Loaded, NewVal, Alignment, MemOpOrder,
@@ -1734,7 +1769,8 @@
 
           // ...and then expand the CAS into a libcall.
           expandAtomicCASToLibcall(Pair);
-        });
+        },
+        CmpXchgLoopBlocks);
   }
 }
 
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fadd.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fadd.ll
deleted file mode 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fadd.ll
+++ /dev/null
@@ -1,2929 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -march=amdgcn -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GFX7LESS %s
-; RUN: llc -march=amdgcn -mcpu=tonga -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX8 %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1064 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1032 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1164 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1132 %s
-; RUN: llc -march=amdgcn -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GFX7LESS %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX9-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1064-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1032-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1164-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1132-DPP %s
-
-declare i32 @llvm.amdgcn.workitem.id.x()
-define amdgpu_kernel void @global_atomic_fadd_uni_value(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_uni_value:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_mov_b64 s[2:3], exec
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX7LESS-NEXT:    s_cbranch_execz .LBB0_3
-; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s6, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX7LESS-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s6
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    buffer_wbinvl1
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX7LESS-NEXT:  .LBB0_3:
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_uni_value:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX8-NEXT:    s_cbranch_execz .LBB0_3
-; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX8-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX8-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    buffer_wbinvl1_vol
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX8-NEXT:  .LBB0_3:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_uni_value:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX9-NEXT:    s_cbranch_execz .LBB0_3
-; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_bcnt1_i32_b64 s5, s[2:3]
-; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v0, s5
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    buffer_wbinvl1_vol
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX9-NEXT:  .LBB0_3:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_uni_value:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX1064-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1064-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    buffer_gl0_inv
-; GFX1064-NEXT:    buffer_gl1_inv
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1064-NEXT:  .LBB0_3:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_uni_value:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s4, vcc_lo
-; GFX1032-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1032-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1032-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    buffer_gl0_inv
-; GFX1032-NEXT:    buffer_gl1_inv
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1032-NEXT:  .LBB0_3:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_uni_value:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1164-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    buffer_gl0_inv
-; GFX1164-NEXT:    buffer_gl1_inv
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1164-NEXT:  .LBB0_3:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_uni_value:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1132-NEXT:    s_mov_b32 s4, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_dual_mul_f32 v2, 4.0, v0 :: v_dual_mov_b32 v1, s4
-; GFX1132-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    buffer_gl0_inv
-; GFX1132-NEXT:    buffer_gl1_inv
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1132-NEXT:  .LBB0_3:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_bcnt1_i32_b64 s5, s[2:3]
-; GFX9-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s5
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    buffer_wbinvl1_vol
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX9-DPP-NEXT:  .LBB0_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1064-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    buffer_gl0_inv
-; GFX1064-DPP-NEXT:    buffer_gl1_inv
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1064-DPP-NEXT:  .LBB0_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s4, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1032-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1032-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    buffer_gl0_inv
-; GFX1032-DPP-NEXT:    buffer_gl1_inv
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1032-DPP-NEXT:  .LBB0_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1164-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    buffer_gl0_inv
-; GFX1164-DPP-NEXT:    buffer_gl1_inv
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1164-DPP-NEXT:  .LBB0_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s4, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_dual_mul_f32 v2, 4.0, v0 :: v_dual_mov_b32 v1, s4
-; GFX1132-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    buffer_gl0_inv
-; GFX1132-DPP-NEXT:    buffer_gl1_inv
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1132-DPP-NEXT:  .LBB0_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %result = atomicrmw fadd ptr addrspace(1) %ptr, float 4.0 seq_cst
-  ret void
-}
-
-define amdgpu_kernel void @global_atomic_fadd_div_value(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_div_value:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB1_1: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_add_f32_e32 v1, v2, v0
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v2
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    buffer_wbinvl1
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v2
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB1_1
-; GFX7LESS-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_div_value:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mov_b32_e32 v2, 0
-; GFX8-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX8-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX8-NEXT:    s_add_i32 s4, s4, 32
-; GFX8-NEXT:    s_min_u32 s4, s5, s4
-; GFX8-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX8-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX8-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX8-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX8-NEXT:    v_add_f32_e32 v2, s6, v2
-; GFX8-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX8-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execz .LBB1_5
-; GFX8-NEXT:  ; %bb.3:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    buffer_wbinvl1_vol
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX8-NEXT:  .LBB1_5:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_div_value:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX9-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX9-NEXT:    s_add_i32 s4, s4, 32
-; GFX9-NEXT:    s_min_u32 s4, s5, s4
-; GFX9-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX9-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX9-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX9-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX9-NEXT:    v_add_f32_e32 v2, s6, v2
-; GFX9-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX9-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execz .LBB1_5
-; GFX9-NEXT:  ; %bb.3:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    buffer_wbinvl1_vol
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX9-NEXT:  .LBB1_5:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_div_value:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX1064-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX1064-NEXT:    s_add_i32 s4, s4, 32
-; GFX1064-NEXT:    s_min_u32 s4, s5, s4
-; GFX1064-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX1064-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX1064-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1064-NEXT:    v_add_f32_e32 v2, s6, v2
-; GFX1064-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1064-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1064-NEXT:  ; %bb.3:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    buffer_gl0_inv
-; GFX1064-NEXT:    buffer_gl1_inv
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1064-NEXT:  .LBB1_5:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_div_value:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1032-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    s_ff1_i32_b32 s3, s2
-; GFX1032-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1032-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1032-NEXT:    s_andn2_b32 s2, s2, s3
-; GFX1032-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1032-NEXT:    v_add_f32_e32 v2, s4, v2
-; GFX1032-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1032-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1032-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1032-NEXT:  ; %bb.3:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    buffer_gl0_inv
-; GFX1032-NEXT:    buffer_gl1_inv
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1032-NEXT:  .LBB1_5:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_div_value:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_ctz_i32_b32 s4, s3
-; GFX1164-NEXT:    s_ctz_i32_b32 s5, s2
-; GFX1164-NEXT:    s_add_i32 s4, s4, 32
-; GFX1164-NEXT:    s_min_u32 s4, s5, s4
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX1164-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1164-NEXT:    s_and_not1_b64 s[2:3], s[2:3], s[4:5]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1164-NEXT:    v_add_f32_e32 v2, s6, v2
-; GFX1164-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1164-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1164-NEXT:  ; %bb.3:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    buffer_gl0_inv
-; GFX1164-NEXT:    buffer_gl1_inv
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1164-NEXT:  .LBB1_5:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_div_value:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_ctz_i32_b32 s3, s2
-; GFX1132-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1132-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 s2, s2, s3
-; GFX1132-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_add_f32_e32 v2, s4, v2
-; GFX1132-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1132-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1132-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1132-NEXT:  ; %bb.3:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    buffer_gl0_inv
-; GFX1132-NEXT:    buffer_gl1_inv
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1132-NEXT:  .LBB1_5:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, 0
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-DPP-NEXT:    s_nop 0
-; GFX9-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_bcast:15 row_mask:0xa bank_mask:0xf
-; GFX9-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX9-DPP-NEXT:    v_readlane_b32 s4, v1, 63
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s5, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v4, s5
-; GFX9-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_add_f32_e32 v3, s4, v4
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v5, v[3:4], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    buffer_wbinvl1_vol
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v4, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX9-DPP-NEXT:  .LBB1_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1064-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1064-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1064-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1064-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1064-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_readlane_b32 s4, v1, 0
-; GFX1064-DPP-NEXT:    v_readlane_b32 s5, v1, 32
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_add_f32_e64 v1, s4, s5
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v0, 0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, v1
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v4, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_add_f32_e32 v3, v4, v5
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v3, v0, v[3:4], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    buffer_gl0_inv
-; GFX1064-DPP-NEXT:    buffer_gl1_inv
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v4, v3
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1064-DPP-NEXT:  .LBB1_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1032-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1032-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1032-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1032-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v0, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, v1
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v3
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s2, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v4, s2
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_add_f32_e32 v3, v4, v5
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v3, v0, v[3:4], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    buffer_gl0_inv
-; GFX1032-DPP-NEXT:    buffer_gl1_inv
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v4
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v4, v3
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1032-DPP-NEXT:  .LBB1_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1164-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1164-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1164-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_permlane64_b32 v2, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v0
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v0, 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v3
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v4, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_e32 v3, v4, v5
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v3, v0, v[3:4], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    buffer_gl0_inv
-; GFX1164-DPP-NEXT:    buffer_gl1_inv
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v4, v3
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1164-DPP-NEXT:  .LBB1_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1132-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1132-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1132-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v5, v1
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v3
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v4, s2
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_add_f32_e32 v3, v4, v5
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v3, v0, v[3:4], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    buffer_gl0_inv
-; GFX1132-DPP-NEXT:    buffer_gl1_inv
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v4
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v4, v3
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1132-DPP-NEXT:  .LBB1_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %id.x = call i32 @llvm.amdgcn.workitem.id.x()
-  %divValue = bitcast i32 %id.x to float
-  %result = atomicrmw fadd ptr addrspace(1) %ptr, float %divValue seq_cst
-  ret void
-}
-
-define amdgpu_kernel void @global_atomic_fadd_uni_value_scope_defalut(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_mov_b64 s[2:3], exec
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX7LESS-NEXT:    s_cbranch_execz .LBB2_3
-; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s6, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX7LESS-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s6
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX7LESS-NEXT:  .LBB2_3:
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX8-NEXT:    s_cbranch_execz .LBB2_3
-; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX8-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX8-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX8-NEXT:  .LBB2_3:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX9-NEXT:    s_cbranch_execz .LBB2_3
-; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_bcnt1_i32_b64 s5, s[2:3]
-; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v0, s5
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX9-NEXT:  .LBB2_3:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX1064-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1064-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1064-NEXT:  .LBB2_3:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s4, vcc_lo
-; GFX1032-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1032-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1032-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1032-NEXT:  .LBB2_3:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1164-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1164-NEXT:  .LBB2_3:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1132-NEXT:    s_mov_b32 s4, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_dual_mul_f32 v2, 4.0, v0 :: v_dual_mov_b32 v1, s4
-; GFX1132-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1132-NEXT:  .LBB2_3:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_bcnt1_i32_b64 s5, s[2:3]
-; GFX9-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s5
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX9-DPP-NEXT:  .LBB2_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1064-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1064-DPP-NEXT:  .LBB2_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s4, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1032-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1032-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1032-DPP-NEXT:  .LBB2_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1164-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1164-DPP-NEXT:  .LBB2_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s4, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_dual_mul_f32 v2, 4.0, v0 :: v_dual_mov_b32 v1, s4
-; GFX1132-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1132-DPP-NEXT:  .LBB2_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %result = atomicrmw fadd ptr addrspace(1) %ptr, float 4.0 monotonic, align 4
-  ret void
-}
-
-define amdgpu_kernel void @global_atomic_fadd_div_value_scope_defalut(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB3_1: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_add_f32_e32 v1, v2, v0
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v2
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v2
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB3_1
-; GFX7LESS-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mov_b32_e32 v2, 0
-; GFX8-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX8-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX8-NEXT:    s_add_i32 s4, s4, 32
-; GFX8-NEXT:    s_min_u32 s4, s5, s4
-; GFX8-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX8-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX8-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX8-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX8-NEXT:    v_add_f32_e32 v2, s6, v2
-; GFX8-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX8-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execz .LBB3_5
-; GFX8-NEXT:  ; %bb.3:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX8-NEXT:  .LBB3_5:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX9-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX9-NEXT:    s_add_i32 s4, s4, 32
-; GFX9-NEXT:    s_min_u32 s4, s5, s4
-; GFX9-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX9-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX9-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX9-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX9-NEXT:    v_add_f32_e32 v2, s6, v2
-; GFX9-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX9-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execz .LBB3_5
-; GFX9-NEXT:  ; %bb.3:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX9-NEXT:  .LBB3_5:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX1064-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX1064-NEXT:    s_add_i32 s4, s4, 32
-; GFX1064-NEXT:    s_min_u32 s4, s5, s4
-; GFX1064-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX1064-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX1064-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1064-NEXT:    v_add_f32_e32 v2, s6, v2
-; GFX1064-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1064-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1064-NEXT:  ; %bb.3:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1064-NEXT:  .LBB3_5:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1032-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    s_ff1_i32_b32 s3, s2
-; GFX1032-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1032-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1032-NEXT:    s_andn2_b32 s2, s2, s3
-; GFX1032-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1032-NEXT:    v_add_f32_e32 v2, s4, v2
-; GFX1032-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1032-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1032-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1032-NEXT:  ; %bb.3:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1032-NEXT:  .LBB3_5:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_ctz_i32_b32 s4, s3
-; GFX1164-NEXT:    s_ctz_i32_b32 s5, s2
-; GFX1164-NEXT:    s_add_i32 s4, s4, 32
-; GFX1164-NEXT:    s_min_u32 s4, s5, s4
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX1164-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1164-NEXT:    s_and_not1_b64 s[2:3], s[2:3], s[4:5]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1164-NEXT:    v_add_f32_e32 v2, s6, v2
-; GFX1164-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1164-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1164-NEXT:  ; %bb.3:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1164-NEXT:  .LBB3_5:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_ctz_i32_b32 s3, s2
-; GFX1132-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1132-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 s2, s2, s3
-; GFX1132-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_add_f32_e32 v2, s4, v2
-; GFX1132-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1132-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1132-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1132-NEXT:  ; %bb.3:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1132-NEXT:  .LBB3_5:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, 0
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-DPP-NEXT:    s_nop 0
-; GFX9-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_bcast:15 row_mask:0xa bank_mask:0xf
-; GFX9-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX9-DPP-NEXT:    v_readlane_b32 s4, v1, 63
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s5, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v4, s5
-; GFX9-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_add_f32_e32 v3, s4, v4
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v5, v[3:4], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v4, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX9-DPP-NEXT:  .LBB3_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1064-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1064-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1064-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1064-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1064-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_readlane_b32 s4, v1, 0
-; GFX1064-DPP-NEXT:    v_readlane_b32 s5, v1, 32
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_add_f32_e64 v1, s4, s5
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v0, 0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, v1
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v4, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_add_f32_e32 v3, v4, v5
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v3, v0, v[3:4], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v4, v3
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1064-DPP-NEXT:  .LBB3_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1032-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1032-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1032-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1032-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v0, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, v1
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v3
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s2, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v4, s2
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_add_f32_e32 v3, v4, v5
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v3, v0, v[3:4], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v4
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v4, v3
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1032-DPP-NEXT:  .LBB3_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1164-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1164-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1164-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_permlane64_b32 v2, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v0
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v0, 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v3
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v4, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_e32 v3, v4, v5
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v3, v0, v[3:4], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v4, v3
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1164-DPP-NEXT:  .LBB3_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1132-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1132-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1132-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v5, v1
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v3
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v4, s2
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_add_f32_e32 v3, v4, v5
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v3, v0, v[3:4], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v4
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v4, v3
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1132-DPP-NEXT:  .LBB3_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %id.x = call i32 @llvm.amdgcn.workitem.id.x()
-  %divValue = bitcast i32 %id.x to float
-  %result = atomicrmw fadd ptr addrspace(1) %ptr, float %divValue monotonic, align 4
-  ret void
-}
-
-
-define amdgpu_kernel void @global_atomic_fadd_uni_value_scope_agent(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_mov_b64 s[2:3], exec
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX7LESS-NEXT:    s_cbranch_execz .LBB4_3
-; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s6, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX7LESS-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s6
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX7LESS-NEXT:  .LBB4_3:
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX8-NEXT:    s_cbranch_execz .LBB4_3
-; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX8-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX8-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX8-NEXT:  .LBB4_3:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX9-NEXT:    s_cbranch_execz .LBB4_3
-; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_bcnt1_i32_b64 s5, s[2:3]
-; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v0, s5
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX9-NEXT:  .LBB4_3:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX1064-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1064-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1064-NEXT:  .LBB4_3:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s4, vcc_lo
-; GFX1032-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1032-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1032-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1032-NEXT:  .LBB4_3:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1164-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1164-NEXT:  .LBB4_3:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1132-NEXT:    s_mov_b32 s4, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_dual_mul_f32 v2, 4.0, v0 :: v_dual_mov_b32 v1, s4
-; GFX1132-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1132-NEXT:  .LBB4_3:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_bcnt1_i32_b64 s5, s[2:3]
-; GFX9-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s5
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX9-DPP-NEXT:  .LBB4_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1064-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1064-DPP-NEXT:  .LBB4_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s4, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1032-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1032-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1032-DPP-NEXT:  .LBB4_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1164-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1164-DPP-NEXT:  .LBB4_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s4, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_dual_mul_f32 v2, 4.0, v0 :: v_dual_mov_b32 v1, s4
-; GFX1132-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1132-DPP-NEXT:  .LBB4_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %result = atomicrmw fadd ptr addrspace(1) %ptr, float 4.0 syncscope("agent") monotonic, align 4
-  ret void
-}
-
-
-define amdgpu_kernel void @global_atomic_fadd_div_value_scope_agent(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB5_1: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_add_f32_e32 v1, v2, v0
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v2
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v2
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB5_1
-; GFX7LESS-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mov_b32_e32 v2, 0
-; GFX8-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX8-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX8-NEXT:    s_add_i32 s4, s4, 32
-; GFX8-NEXT:    s_min_u32 s4, s5, s4
-; GFX8-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX8-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX8-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX8-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX8-NEXT:    v_add_f32_e32 v2, s6, v2
-; GFX8-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX8-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execz .LBB5_5
-; GFX8-NEXT:  ; %bb.3:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX8-NEXT:  .LBB5_5:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX9-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX9-NEXT:    s_add_i32 s4, s4, 32
-; GFX9-NEXT:    s_min_u32 s4, s5, s4
-; GFX9-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX9-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX9-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX9-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX9-NEXT:    v_add_f32_e32 v2, s6, v2
-; GFX9-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX9-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execz .LBB5_5
-; GFX9-NEXT:  ; %bb.3:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX9-NEXT:  .LBB5_5:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX1064-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX1064-NEXT:    s_add_i32 s4, s4, 32
-; GFX1064-NEXT:    s_min_u32 s4, s5, s4
-; GFX1064-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX1064-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX1064-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1064-NEXT:    v_add_f32_e32 v2, s6, v2
-; GFX1064-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1064-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1064-NEXT:  ; %bb.3:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1064-NEXT:  .LBB5_5:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1032-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    s_ff1_i32_b32 s3, s2
-; GFX1032-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1032-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1032-NEXT:    s_andn2_b32 s2, s2, s3
-; GFX1032-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1032-NEXT:    v_add_f32_e32 v2, s4, v2
-; GFX1032-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1032-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1032-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1032-NEXT:  ; %bb.3:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1032-NEXT:  .LBB5_5:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_ctz_i32_b32 s4, s3
-; GFX1164-NEXT:    s_ctz_i32_b32 s5, s2
-; GFX1164-NEXT:    s_add_i32 s4, s4, 32
-; GFX1164-NEXT:    s_min_u32 s4, s5, s4
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX1164-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1164-NEXT:    s_and_not1_b64 s[2:3], s[2:3], s[4:5]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1164-NEXT:    v_add_f32_e32 v2, s6, v2
-; GFX1164-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1164-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1164-NEXT:  ; %bb.3:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1164-NEXT:  .LBB5_5:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_ctz_i32_b32 s3, s2
-; GFX1132-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1132-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 s2, s2, s3
-; GFX1132-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_add_f32_e32 v2, s4, v2
-; GFX1132-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1132-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1132-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1132-NEXT:  ; %bb.3:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_add_f32_e32 v0, v1, v2
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1132-NEXT:  .LBB5_5:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, 0
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_shr:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-DPP-NEXT:    s_nop 0
-; GFX9-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_shr:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_shr:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_shr:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_bcast:15 row_mask:0xa bank_mask:0xf
-; GFX9-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX9-DPP-NEXT:    v_readlane_b32 s4, v1, 63
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s5, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v4, s5
-; GFX9-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_add_f32_e32 v3, s4, v4
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v5, v[3:4], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v4
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v4, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX9-DPP-NEXT:  .LBB5_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1064-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1064-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1064-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1064-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1064-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_readlane_b32 s4, v1, 0
-; GFX1064-DPP-NEXT:    v_readlane_b32 s5, v1, 32
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_add_f32_e64 v1, s4, s5
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v0, 0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, v1
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v3
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v4, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_add_f32_e32 v3, v4, v5
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v3, v0, v[3:4], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v4, v3
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1064-DPP-NEXT:  .LBB5_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1032-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1032-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1032-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1032-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v0, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, v1
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v3
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s2, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v4, s2
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_add_f32_e32 v3, v4, v5
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v3, v0, v[3:4], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v4
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v4, v3
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1032-DPP-NEXT:  .LBB5_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1164-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1164-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1164-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_permlane64_b32 v2, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v3, exec_hi, v0
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v0, 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(SKIP_1) | instid1(VALU_DEP_3)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v3
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v4, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_add_f32_e32 v3, v4, v5
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v3, v0, v[3:4], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v4
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v4, v3
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1164-DPP-NEXT:  .LBB5_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, 0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:1 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1132-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:2 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:4 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1132-DPP-NEXT:    v_add_f32_dpp v1, v1, v1 row_xmask:8 row_mask:0xf bank_mask:0xf bound_ctrl:1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1132-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    v_add_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v3, exec_lo, 0
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_dual_mov_b32 v0, 0 :: v_dual_mov_b32 v5, v1
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v3
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v4, s2
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_add_f32_e32 v3, v4, v5
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v3, v0, v[3:4], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v3, v4
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v4, v3
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1132-DPP-NEXT:  .LBB5_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %id.x = call i32 @llvm.amdgcn.workitem.id.x()
-  %divValue = bitcast i32 %id.x to float
-  %result = atomicrmw fadd ptr addrspace(1) %ptr, float %divValue syncscope("agent") monotonic, align 4
-  ret void
-}
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmax.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmax.ll
deleted file mode 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmax.ll
+++ /dev/null
@@ -1,3199 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -march=amdgcn -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GFX7LESS %s
-; RUN: llc -march=amdgcn -mcpu=tonga -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX8 %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1064 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1032 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1164 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1132 %s
-; RUN: llc -march=amdgcn -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GFX7LESS %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX9-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1064-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1032-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1164-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1132-DPP %s
-
-declare i32 @llvm.amdgcn.workitem.id.x()
-define amdgpu_kernel void @global_atomic_fadd_uni_value(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_uni_value:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS-NEXT:    s_cbranch_execz .LBB0_3
-; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    buffer_wbinvl1
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v2
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX7LESS-NEXT:  .LBB0_3:
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_uni_value:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_cbranch_execz .LBB0_3
-; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s0
-; GFX8-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[2:3], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    buffer_wbinvl1_vol
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX8-NEXT:  .LBB0_3:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_uni_value:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_cbranch_execz .LBB0_3
-; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    buffer_wbinvl1_vol
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX9-NEXT:  .LBB0_3:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_uni_value:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    buffer_gl0_inv
-; GFX1064-NEXT:    buffer_gl1_inv
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1064-NEXT:  .LBB0_3:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_uni_value:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    buffer_gl0_inv
-; GFX1032-NEXT:    buffer_gl1_inv
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1032-NEXT:  .LBB0_3:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_uni_value:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    buffer_gl0_inv
-; GFX1164-NEXT:    buffer_gl1_inv
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1164-NEXT:  .LBB0_3:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_uni_value:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    buffer_gl0_inv
-; GFX1132-NEXT:    buffer_gl1_inv
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1132-NEXT:  .LBB0_3:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    buffer_wbinvl1_vol
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX9-DPP-NEXT:  .LBB0_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    buffer_gl0_inv
-; GFX1064-DPP-NEXT:    buffer_gl1_inv
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1064-DPP-NEXT:  .LBB0_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    buffer_gl0_inv
-; GFX1032-DPP-NEXT:    buffer_gl1_inv
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1032-DPP-NEXT:  .LBB0_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    buffer_gl0_inv
-; GFX1164-DPP-NEXT:    buffer_gl1_inv
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1164-DPP-NEXT:  .LBB0_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    buffer_gl0_inv
-; GFX1132-DPP-NEXT:    buffer_gl1_inv
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1132-DPP-NEXT:  .LBB0_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %result = atomicrmw fmax ptr addrspace(1) %ptr, float 4.0 seq_cst
-  ret void
-}
-
-define amdgpu_kernel void @global_atomic_fadd_div_value(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_div_value:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    v_mul_f32_e32 v2, 1.0, v0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB1_1: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    buffer_wbinvl1
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB1_1
-; GFX7LESS-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_div_value:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX8-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX8-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX8-NEXT:    s_add_i32 s4, s4, 32
-; GFX8-NEXT:    s_min_u32 s4, s5, s4
-; GFX8-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX8-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX8-NEXT:    v_mul_f32_e32 v1, 1.0, v2
-; GFX8-NEXT:    v_mul_f32_e64 v2, 1.0, s6
-; GFX8-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX8-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX8-NEXT:    v_max_f32_e32 v2, v1, v2
-; GFX8-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX8-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execz .LBB1_5
-; GFX8-NEXT:  ; %bb.3:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    v_mul_f32_e32 v2, 1.0, v2
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    buffer_wbinvl1_vol
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX8-NEXT:  .LBB1_5:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_div_value:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX9-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX9-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX9-NEXT:    s_add_i32 s4, s4, 32
-; GFX9-NEXT:    s_min_u32 s4, s5, s4
-; GFX9-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX9-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX9-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX9-NEXT:    v_max_f32_e64 v2, s6, s6
-; GFX9-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX9-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX9-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX9-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX9-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execz .LBB1_5
-; GFX9-NEXT:  ; %bb.3:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    buffer_wbinvl1_vol
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX9-NEXT:  .LBB1_5:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_div_value:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX1064-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX1064-NEXT:    s_add_i32 s4, s4, 32
-; GFX1064-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1064-NEXT:    s_min_u32 s4, s5, s4
-; GFX1064-NEXT:    v_readlane_b32 s5, v0, s4
-; GFX1064-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1064-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX1064-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1064-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1064-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1064-NEXT:  ; %bb.3:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    buffer_gl0_inv
-; GFX1064-NEXT:    buffer_gl1_inv
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1064-NEXT:  .LBB1_5:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_div_value:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1032-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    s_ff1_i32_b32 s3, s2
-; GFX1032-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1032-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1032-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1032-NEXT:    s_andn2_b32 s2, s2, s3
-; GFX1032-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1032-NEXT:    v_max_f32_e64 v2, s4, s4
-; GFX1032-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1032-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1032-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1032-NEXT:  ; %bb.3:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    buffer_gl0_inv
-; GFX1032-NEXT:    buffer_gl1_inv
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1032-NEXT:  .LBB1_5:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_div_value:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_ctz_i32_b32 s4, s3
-; GFX1164-NEXT:    s_ctz_i32_b32 s5, s2
-; GFX1164-NEXT:    s_add_i32 s4, s4, 32
-; GFX1164-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1164-NEXT:    s_min_u32 s4, s5, s4
-; GFX1164-NEXT:    v_readlane_b32 s5, v0, s4
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1164-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1164-NEXT:    s_and_not1_b64 s[2:3], s[2:3], s[4:5]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1164-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1164-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1164-NEXT:  ; %bb.3:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    buffer_gl0_inv
-; GFX1164-NEXT:    buffer_gl1_inv
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1164-NEXT:  .LBB1_5:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_div_value:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    s_ctz_i32_b32 s3, s2
-; GFX1132-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1132-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1132-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 s2, s2, s3
-; GFX1132-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e64 v2, s4, s4
-; GFX1132-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1132-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1132-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1132-NEXT:  ; %bb.3:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_dual_max_f32 v2, v1, v1 :: v_dual_mov_b32 v1, s3
-; GFX1132-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    buffer_gl0_inv
-; GFX1132-NEXT:    buffer_gl1_inv
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1132-NEXT:  .LBB1_5:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v4
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:2 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:4 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:8 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX9-DPP-NEXT:    v_readlane_b32 s4, v1, 63
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_max_f32_e64 v0, s4, s4
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s5, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, s5
-; GFX9-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX9-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    buffer_wbinvl1_vol
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX9-DPP-NEXT:  .LBB1_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1064-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_readlane_b32 s4, v1, 32
-; GFX1064-DPP-NEXT:    v_readlane_b32 s5, v1, 0
-; GFX1064-DPP-NEXT:    v_max_f32_e64 v1, s4, s4
-; GFX1064-DPP-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    buffer_gl0_inv
-; GFX1064-DPP-NEXT:    buffer_gl1_inv
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1064-DPP-NEXT:  .LBB1_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1032-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v4
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, s3
-; GFX1032-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    buffer_gl0_inv
-; GFX1032-DPP-NEXT:    buffer_gl1_inv
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1032-DPP-NEXT:  .LBB1_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_waitcnt_depctr 0xfff
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_permlane64_b32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    buffer_gl0_inv
-; GFX1164-DPP-NEXT:    buffer_gl1_inv
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1164-DPP-NEXT:  .LBB1_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v2, v2 :: v_dual_mov_b32 v3, 0xff800000
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v3, v3 :: v_dual_mov_b32 v3, 0xff800000
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v3, v3 :: v_dual_mov_b32 v3, 0xff800000
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1132-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v0, v0, v0 :: v_dual_mov_b32 v5, s3
-; GFX1132-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    buffer_gl0_inv
-; GFX1132-DPP-NEXT:    buffer_gl1_inv
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1132-DPP-NEXT:  .LBB1_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %id.x = call i32 @llvm.amdgcn.workitem.id.x()
-  %divValue = bitcast i32 %id.x to float
-  %result = atomicrmw fmax ptr addrspace(1) %ptr, float %divValue seq_cst
-  ret void
-}
-
-define amdgpu_kernel void @global_atomic_fadd_uni_value_scope_defalut(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS-NEXT:    s_cbranch_execz .LBB2_3
-; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v2
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX7LESS-NEXT:  .LBB2_3:
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_cbranch_execz .LBB2_3
-; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s0
-; GFX8-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[2:3], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX8-NEXT:  .LBB2_3:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_cbranch_execz .LBB2_3
-; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX9-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX9-NEXT:  .LBB2_3:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1064-NEXT:  .LBB2_3:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1032-NEXT:  .LBB2_3:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1164-NEXT:  .LBB2_3:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1132-NEXT:  .LBB2_3:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX9-DPP-NEXT:  .LBB2_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1064-DPP-NEXT:  .LBB2_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1032-DPP-NEXT:  .LBB2_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1164-DPP-NEXT:  .LBB2_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1132-DPP-NEXT:  .LBB2_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %result = atomicrmw fmax ptr addrspace(1) %ptr, float 4.0 monotonic, align 4
-  ret void
-}
-
-define amdgpu_kernel void @global_atomic_fadd_div_value_scope_defalut(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    v_mul_f32_e32 v2, 1.0, v0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB3_1: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB3_1
-; GFX7LESS-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX8-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX8-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX8-NEXT:    s_add_i32 s4, s4, 32
-; GFX8-NEXT:    s_min_u32 s4, s5, s4
-; GFX8-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX8-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX8-NEXT:    v_mul_f32_e32 v1, 1.0, v2
-; GFX8-NEXT:    v_mul_f32_e64 v2, 1.0, s6
-; GFX8-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX8-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX8-NEXT:    v_max_f32_e32 v2, v1, v2
-; GFX8-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX8-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execz .LBB3_5
-; GFX8-NEXT:  ; %bb.3:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    v_mul_f32_e32 v2, 1.0, v2
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX8-NEXT:  .LBB3_5:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX9-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX9-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX9-NEXT:    s_add_i32 s4, s4, 32
-; GFX9-NEXT:    s_min_u32 s4, s5, s4
-; GFX9-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX9-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX9-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX9-NEXT:    v_max_f32_e64 v2, s6, s6
-; GFX9-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX9-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX9-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX9-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX9-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execz .LBB3_5
-; GFX9-NEXT:  ; %bb.3:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX9-NEXT:  .LBB3_5:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX1064-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX1064-NEXT:    s_add_i32 s4, s4, 32
-; GFX1064-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1064-NEXT:    s_min_u32 s4, s5, s4
-; GFX1064-NEXT:    v_readlane_b32 s5, v0, s4
-; GFX1064-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1064-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX1064-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1064-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1064-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1064-NEXT:  ; %bb.3:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1064-NEXT:  .LBB3_5:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1032-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    s_ff1_i32_b32 s3, s2
-; GFX1032-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1032-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1032-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1032-NEXT:    s_andn2_b32 s2, s2, s3
-; GFX1032-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1032-NEXT:    v_max_f32_e64 v2, s4, s4
-; GFX1032-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1032-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1032-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1032-NEXT:  ; %bb.3:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1032-NEXT:  .LBB3_5:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_ctz_i32_b32 s4, s3
-; GFX1164-NEXT:    s_ctz_i32_b32 s5, s2
-; GFX1164-NEXT:    s_add_i32 s4, s4, 32
-; GFX1164-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1164-NEXT:    s_min_u32 s4, s5, s4
-; GFX1164-NEXT:    v_readlane_b32 s5, v0, s4
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1164-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1164-NEXT:    s_and_not1_b64 s[2:3], s[2:3], s[4:5]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1164-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1164-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1164-NEXT:  ; %bb.3:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1164-NEXT:  .LBB3_5:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    s_ctz_i32_b32 s3, s2
-; GFX1132-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1132-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1132-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 s2, s2, s3
-; GFX1132-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e64 v2, s4, s4
-; GFX1132-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1132-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1132-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1132-NEXT:  ; %bb.3:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_dual_max_f32 v2, v1, v1 :: v_dual_mov_b32 v1, s3
-; GFX1132-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1132-NEXT:  .LBB3_5:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v4
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:2 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:4 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:8 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX9-DPP-NEXT:    v_readlane_b32 s4, v1, 63
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_max_f32_e64 v0, s4, s4
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s5, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, s5
-; GFX9-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX9-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX9-DPP-NEXT:  .LBB3_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1064-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_readlane_b32 s4, v1, 32
-; GFX1064-DPP-NEXT:    v_readlane_b32 s5, v1, 0
-; GFX1064-DPP-NEXT:    v_max_f32_e64 v1, s4, s4
-; GFX1064-DPP-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1064-DPP-NEXT:  .LBB3_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1032-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v4
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, s3
-; GFX1032-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1032-DPP-NEXT:  .LBB3_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_waitcnt_depctr 0xfff
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_permlane64_b32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1164-DPP-NEXT:  .LBB3_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v2, v2 :: v_dual_mov_b32 v3, 0xff800000
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v3, v3 :: v_dual_mov_b32 v3, 0xff800000
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v3, v3 :: v_dual_mov_b32 v3, 0xff800000
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1132-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v0, v0, v0 :: v_dual_mov_b32 v5, s3
-; GFX1132-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1132-DPP-NEXT:  .LBB3_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %id.x = call i32 @llvm.amdgcn.workitem.id.x()
-  %divValue = bitcast i32 %id.x to float
-  %result = atomicrmw fmax ptr addrspace(1) %ptr, float %divValue monotonic, align 4
-  ret void
-}
-
-
-define amdgpu_kernel void @global_atomic_fadd_uni_value_scope_agent(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS-NEXT:    s_cbranch_execz .LBB4_3
-; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v2
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX7LESS-NEXT:  .LBB4_3:
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_cbranch_execz .LBB4_3
-; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s0
-; GFX8-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[2:3], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX8-NEXT:  .LBB4_3:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_cbranch_execz .LBB4_3
-; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX9-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX9-NEXT:  .LBB4_3:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1064-NEXT:  .LBB4_3:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1032-NEXT:  .LBB4_3:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1164-NEXT:  .LBB4_3:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1132-NEXT:  .LBB4_3:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX9-DPP-NEXT:  .LBB4_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1064-DPP-NEXT:  .LBB4_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1032-DPP-NEXT:  .LBB4_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1164-DPP-NEXT:  .LBB4_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v0, 4.0, v0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1132-DPP-NEXT:  .LBB4_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %result = atomicrmw fmax ptr addrspace(1) %ptr, float 4.0 syncscope("agent") monotonic, align 4
-  ret void
-}
-
-
-define amdgpu_kernel void @global_atomic_fadd_div_value_scope_agent(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    v_mul_f32_e32 v2, 1.0, v0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB5_1: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB5_1
-; GFX7LESS-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX8-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX8-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX8-NEXT:    s_add_i32 s4, s4, 32
-; GFX8-NEXT:    s_min_u32 s4, s5, s4
-; GFX8-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX8-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX8-NEXT:    v_mul_f32_e32 v1, 1.0, v2
-; GFX8-NEXT:    v_mul_f32_e64 v2, 1.0, s6
-; GFX8-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX8-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX8-NEXT:    v_max_f32_e32 v2, v1, v2
-; GFX8-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX8-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execz .LBB5_5
-; GFX8-NEXT:  ; %bb.3:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    v_mul_f32_e32 v2, 1.0, v2
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX8-NEXT:  .LBB5_5:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX9-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX9-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX9-NEXT:    s_add_i32 s4, s4, 32
-; GFX9-NEXT:    s_min_u32 s4, s5, s4
-; GFX9-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX9-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX9-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX9-NEXT:    v_max_f32_e64 v2, s6, s6
-; GFX9-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX9-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX9-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX9-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX9-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execz .LBB5_5
-; GFX9-NEXT:  ; %bb.3:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX9-NEXT:  .LBB5_5:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX1064-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX1064-NEXT:    s_add_i32 s4, s4, 32
-; GFX1064-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1064-NEXT:    s_min_u32 s4, s5, s4
-; GFX1064-NEXT:    v_readlane_b32 s5, v0, s4
-; GFX1064-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1064-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX1064-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1064-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1064-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1064-NEXT:  ; %bb.3:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1064-NEXT:  .LBB5_5:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1032-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    s_ff1_i32_b32 s3, s2
-; GFX1032-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1032-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1032-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1032-NEXT:    s_andn2_b32 s2, s2, s3
-; GFX1032-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1032-NEXT:    v_max_f32_e64 v2, s4, s4
-; GFX1032-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1032-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1032-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1032-NEXT:  ; %bb.3:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1032-NEXT:  .LBB5_5:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_ctz_i32_b32 s4, s3
-; GFX1164-NEXT:    s_ctz_i32_b32 s5, s2
-; GFX1164-NEXT:    s_add_i32 s4, s4, 32
-; GFX1164-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1164-NEXT:    s_min_u32 s4, s5, s4
-; GFX1164-NEXT:    v_readlane_b32 s5, v0, s4
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1164-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1164-NEXT:    s_and_not1_b64 s[2:3], s[2:3], s[4:5]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1164-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1164-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1164-NEXT:  ; %bb.3:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1164-NEXT:  .LBB5_5:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    s_ctz_i32_b32 s3, s2
-; GFX1132-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1132-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1132-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 s2, s2, s3
-; GFX1132-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e64 v2, s4, s4
-; GFX1132-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1132-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1132-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1132-NEXT:  ; %bb.3:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_dual_max_f32 v2, v1, v1 :: v_dual_mov_b32 v1, s3
-; GFX1132-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-NEXT:    v_max_f32_e32 v0, v0, v2
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1132-NEXT:  .LBB5_5:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v4
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:2 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:4 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:8 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX9-DPP-NEXT:    v_readlane_b32 s4, v1, 63
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_max_f32_e64 v0, s4, s4
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s5, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, s5
-; GFX9-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX9-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX9-DPP-NEXT:  .LBB5_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1064-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_readlane_b32 s4, v1, 32
-; GFX1064-DPP-NEXT:    v_readlane_b32 s5, v1, 0
-; GFX1064-DPP-NEXT:    v_max_f32_e64 v1, s4, s4
-; GFX1064-DPP-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1064-DPP-NEXT:  .LBB5_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1032-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v4
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, s3
-; GFX1032-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1032-DPP-NEXT:  .LBB5_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_waitcnt_depctr 0xfff
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0xff800000
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_permlane64_b32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1164-DPP-NEXT:  .LBB5_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, 0xff800000
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, 0xff800000
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v2, v2 :: v_dual_mov_b32 v3, 0xff800000
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v2, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v3, v3 :: v_dual_mov_b32 v3, 0xff800000
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v3, v3 :: v_dual_mov_b32 v3, 0xff800000
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1132-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v0, v0, v0 :: v_dual_mov_b32 v5, s3
-; GFX1132-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v4, v4, v0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1132-DPP-NEXT:  .LBB5_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %id.x = call i32 @llvm.amdgcn.workitem.id.x()
-  %divValue = bitcast i32 %id.x to float
-  %result = atomicrmw fmax ptr addrspace(1) %ptr, float %divValue syncscope("agent") monotonic, align 4
-  ret void
-}
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmin.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmin.ll
deleted file mode 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fmin.ll
+++ /dev/null
@@ -1,3199 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -march=amdgcn -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GFX7LESS %s
-; RUN: llc -march=amdgcn -mcpu=tonga -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX8 %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1064 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1032 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1164 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1132 %s
-; RUN: llc -march=amdgcn -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GFX7LESS %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX9-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1064-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1032-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1164-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1132-DPP %s
-
-declare i32 @llvm.amdgcn.workitem.id.x()
-define amdgpu_kernel void @global_atomic_fadd_uni_value(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_uni_value:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS-NEXT:    s_cbranch_execz .LBB0_3
-; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    buffer_wbinvl1
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v2
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX7LESS-NEXT:  .LBB0_3:
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_uni_value:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_cbranch_execz .LBB0_3
-; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s0
-; GFX8-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[2:3], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    buffer_wbinvl1_vol
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX8-NEXT:  .LBB0_3:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_uni_value:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_cbranch_execz .LBB0_3
-; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    buffer_wbinvl1_vol
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX9-NEXT:  .LBB0_3:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_uni_value:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    buffer_gl0_inv
-; GFX1064-NEXT:    buffer_gl1_inv
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1064-NEXT:  .LBB0_3:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_uni_value:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    buffer_gl0_inv
-; GFX1032-NEXT:    buffer_gl1_inv
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1032-NEXT:  .LBB0_3:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_uni_value:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    buffer_gl0_inv
-; GFX1164-NEXT:    buffer_gl1_inv
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1164-NEXT:  .LBB0_3:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_uni_value:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    buffer_gl0_inv
-; GFX1132-NEXT:    buffer_gl1_inv
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1132-NEXT:  .LBB0_3:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    buffer_wbinvl1_vol
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX9-DPP-NEXT:  .LBB0_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    buffer_gl0_inv
-; GFX1064-DPP-NEXT:    buffer_gl1_inv
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1064-DPP-NEXT:  .LBB0_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    buffer_gl0_inv
-; GFX1032-DPP-NEXT:    buffer_gl1_inv
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1032-DPP-NEXT:  .LBB0_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    buffer_gl0_inv
-; GFX1164-DPP-NEXT:    buffer_gl1_inv
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1164-DPP-NEXT:  .LBB0_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    buffer_gl0_inv
-; GFX1132-DPP-NEXT:    buffer_gl1_inv
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1132-DPP-NEXT:  .LBB0_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %result = atomicrmw fmin ptr addrspace(1) %ptr, float 4.0 seq_cst
-  ret void
-}
-
-define amdgpu_kernel void @global_atomic_fadd_div_value(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_div_value:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    v_mul_f32_e32 v2, 1.0, v0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB1_1: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    buffer_wbinvl1
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB1_1
-; GFX7LESS-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_div_value:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX8-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX8-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX8-NEXT:    s_add_i32 s4, s4, 32
-; GFX8-NEXT:    s_min_u32 s4, s5, s4
-; GFX8-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX8-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX8-NEXT:    v_mul_f32_e32 v1, 1.0, v2
-; GFX8-NEXT:    v_mul_f32_e64 v2, 1.0, s6
-; GFX8-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX8-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX8-NEXT:    v_min_f32_e32 v2, v1, v2
-; GFX8-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX8-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execz .LBB1_5
-; GFX8-NEXT:  ; %bb.3:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    v_mul_f32_e32 v2, 1.0, v2
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    buffer_wbinvl1_vol
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX8-NEXT:  .LBB1_5:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_div_value:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX9-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX9-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX9-NEXT:    s_add_i32 s4, s4, 32
-; GFX9-NEXT:    s_min_u32 s4, s5, s4
-; GFX9-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX9-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX9-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX9-NEXT:    v_max_f32_e64 v2, s6, s6
-; GFX9-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX9-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX9-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX9-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX9-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execz .LBB1_5
-; GFX9-NEXT:  ; %bb.3:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    buffer_wbinvl1_vol
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX9-NEXT:  .LBB1_5:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_div_value:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX1064-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX1064-NEXT:    s_add_i32 s4, s4, 32
-; GFX1064-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1064-NEXT:    s_min_u32 s4, s5, s4
-; GFX1064-NEXT:    v_readlane_b32 s5, v0, s4
-; GFX1064-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1064-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX1064-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1064-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1064-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1064-NEXT:  ; %bb.3:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    buffer_gl0_inv
-; GFX1064-NEXT:    buffer_gl1_inv
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1064-NEXT:  .LBB1_5:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_div_value:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1032-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    s_ff1_i32_b32 s3, s2
-; GFX1032-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1032-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1032-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1032-NEXT:    s_andn2_b32 s2, s2, s3
-; GFX1032-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1032-NEXT:    v_max_f32_e64 v2, s4, s4
-; GFX1032-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1032-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1032-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1032-NEXT:  ; %bb.3:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    buffer_gl0_inv
-; GFX1032-NEXT:    buffer_gl1_inv
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1032-NEXT:  .LBB1_5:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_div_value:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_ctz_i32_b32 s4, s3
-; GFX1164-NEXT:    s_ctz_i32_b32 s5, s2
-; GFX1164-NEXT:    s_add_i32 s4, s4, 32
-; GFX1164-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1164-NEXT:    s_min_u32 s4, s5, s4
-; GFX1164-NEXT:    v_readlane_b32 s5, v0, s4
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1164-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1164-NEXT:    s_and_not1_b64 s[2:3], s[2:3], s[4:5]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1164-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1164-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1164-NEXT:  ; %bb.3:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    buffer_gl0_inv
-; GFX1164-NEXT:    buffer_gl1_inv
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1164-NEXT:  .LBB1_5:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_div_value:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    s_ctz_i32_b32 s3, s2
-; GFX1132-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1132-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1132-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 s2, s2, s3
-; GFX1132-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e64 v2, s4, s4
-; GFX1132-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1132-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1132-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1132-NEXT:  ; %bb.3:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_dual_max_f32 v2, v1, v1 :: v_dual_mov_b32 v1, s3
-; GFX1132-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    buffer_gl0_inv
-; GFX1132-NEXT:    buffer_gl1_inv
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1132-NEXT:  .LBB1_5:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v4
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:2 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:4 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:8 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX9-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX9-DPP-NEXT:    v_readlane_b32 s4, v1, 63
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_max_f32_e64 v0, s4, s4
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s5, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, s5
-; GFX9-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX9-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    buffer_wbinvl1_vol
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX9-DPP-NEXT:  .LBB1_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1064-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_readlane_b32 s4, v1, 32
-; GFX1064-DPP-NEXT:    v_readlane_b32 s5, v1, 0
-; GFX1064-DPP-NEXT:    v_max_f32_e64 v1, s4, s4
-; GFX1064-DPP-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    buffer_gl0_inv
-; GFX1064-DPP-NEXT:    buffer_gl1_inv
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1064-DPP-NEXT:  .LBB1_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1032-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v4
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, s3
-; GFX1032-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    buffer_gl0_inv
-; GFX1032-DPP-NEXT:    buffer_gl1_inv
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1032-DPP-NEXT:  .LBB1_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_waitcnt_depctr 0xfff
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_permlane64_b32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    buffer_gl0_inv
-; GFX1164-DPP-NEXT:    buffer_gl1_inv
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1164-DPP-NEXT:  .LBB1_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v2, v2 :: v_dual_mov_b32 v3, 0x7f800000
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v3, v3 :: v_dual_mov_b32 v3, 0x7f800000
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v3, v3 :: v_dual_mov_b32 v3, 0x7f800000
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1132-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v0, v0, v0 :: v_dual_mov_b32 v5, s3
-; GFX1132-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    buffer_gl0_inv
-; GFX1132-DPP-NEXT:    buffer_gl1_inv
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1132-DPP-NEXT:  .LBB1_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %id.x = call i32 @llvm.amdgcn.workitem.id.x()
-  %divValue = bitcast i32 %id.x to float
-  %result = atomicrmw fmin ptr addrspace(1) %ptr, float %divValue seq_cst
-  ret void
-}
-
-define amdgpu_kernel void @global_atomic_fadd_uni_value_scope_defalut(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS-NEXT:    s_cbranch_execz .LBB2_3
-; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v2
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX7LESS-NEXT:  .LBB2_3:
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_cbranch_execz .LBB2_3
-; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s0
-; GFX8-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[2:3], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX8-NEXT:  .LBB2_3:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_cbranch_execz .LBB2_3
-; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX9-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX9-NEXT:  .LBB2_3:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1064-NEXT:  .LBB2_3:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1032-NEXT:  .LBB2_3:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1164-NEXT:  .LBB2_3:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1132-NEXT:  .LBB2_3:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX9-DPP-NEXT:  .LBB2_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1064-DPP-NEXT:  .LBB2_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1032-DPP-NEXT:  .LBB2_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1164-DPP-NEXT:  .LBB2_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1132-DPP-NEXT:  .LBB2_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %result = atomicrmw fmin ptr addrspace(1) %ptr, float 4.0 monotonic, align 4
-  ret void
-}
-
-define amdgpu_kernel void @global_atomic_fadd_div_value_scope_defalut(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    v_mul_f32_e32 v2, 1.0, v0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB3_1: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB3_1
-; GFX7LESS-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX8-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX8-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX8-NEXT:    s_add_i32 s4, s4, 32
-; GFX8-NEXT:    s_min_u32 s4, s5, s4
-; GFX8-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX8-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX8-NEXT:    v_mul_f32_e32 v1, 1.0, v2
-; GFX8-NEXT:    v_mul_f32_e64 v2, 1.0, s6
-; GFX8-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX8-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX8-NEXT:    v_min_f32_e32 v2, v1, v2
-; GFX8-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX8-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execz .LBB3_5
-; GFX8-NEXT:  ; %bb.3:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    v_mul_f32_e32 v2, 1.0, v2
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX8-NEXT:  .LBB3_5:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX9-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX9-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX9-NEXT:    s_add_i32 s4, s4, 32
-; GFX9-NEXT:    s_min_u32 s4, s5, s4
-; GFX9-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX9-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX9-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX9-NEXT:    v_max_f32_e64 v2, s6, s6
-; GFX9-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX9-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX9-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX9-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX9-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execz .LBB3_5
-; GFX9-NEXT:  ; %bb.3:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX9-NEXT:  .LBB3_5:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX1064-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX1064-NEXT:    s_add_i32 s4, s4, 32
-; GFX1064-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1064-NEXT:    s_min_u32 s4, s5, s4
-; GFX1064-NEXT:    v_readlane_b32 s5, v0, s4
-; GFX1064-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1064-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX1064-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1064-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1064-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1064-NEXT:  ; %bb.3:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1064-NEXT:  .LBB3_5:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1032-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    s_ff1_i32_b32 s3, s2
-; GFX1032-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1032-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1032-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1032-NEXT:    s_andn2_b32 s2, s2, s3
-; GFX1032-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1032-NEXT:    v_max_f32_e64 v2, s4, s4
-; GFX1032-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1032-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1032-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1032-NEXT:  ; %bb.3:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1032-NEXT:  .LBB3_5:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_ctz_i32_b32 s4, s3
-; GFX1164-NEXT:    s_ctz_i32_b32 s5, s2
-; GFX1164-NEXT:    s_add_i32 s4, s4, 32
-; GFX1164-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1164-NEXT:    s_min_u32 s4, s5, s4
-; GFX1164-NEXT:    v_readlane_b32 s5, v0, s4
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1164-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1164-NEXT:    s_and_not1_b64 s[2:3], s[2:3], s[4:5]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1164-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1164-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1164-NEXT:  ; %bb.3:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1164-NEXT:  .LBB3_5:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    s_ctz_i32_b32 s3, s2
-; GFX1132-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1132-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1132-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 s2, s2, s3
-; GFX1132-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e64 v2, s4, s4
-; GFX1132-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1132-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1132-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1132-NEXT:  ; %bb.3:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_dual_max_f32 v2, v1, v1 :: v_dual_mov_b32 v1, s3
-; GFX1132-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1132-NEXT:  .LBB3_5:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v4
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:2 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:4 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:8 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX9-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX9-DPP-NEXT:    v_readlane_b32 s4, v1, 63
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_max_f32_e64 v0, s4, s4
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s5, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, s5
-; GFX9-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX9-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX9-DPP-NEXT:  .LBB3_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1064-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_readlane_b32 s4, v1, 32
-; GFX1064-DPP-NEXT:    v_readlane_b32 s5, v1, 0
-; GFX1064-DPP-NEXT:    v_max_f32_e64 v1, s4, s4
-; GFX1064-DPP-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1064-DPP-NEXT:  .LBB3_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1032-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v4
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, s3
-; GFX1032-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1032-DPP-NEXT:  .LBB3_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_waitcnt_depctr 0xfff
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_permlane64_b32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1164-DPP-NEXT:  .LBB3_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v2, v2 :: v_dual_mov_b32 v3, 0x7f800000
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v3, v3 :: v_dual_mov_b32 v3, 0x7f800000
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v3, v3 :: v_dual_mov_b32 v3, 0x7f800000
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1132-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v0, v0, v0 :: v_dual_mov_b32 v5, s3
-; GFX1132-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1132-DPP-NEXT:  .LBB3_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %id.x = call i32 @llvm.amdgcn.workitem.id.x()
-  %divValue = bitcast i32 %id.x to float
-  %result = atomicrmw fmin ptr addrspace(1) %ptr, float %divValue monotonic, align 4
-  ret void
-}
-
-
-define amdgpu_kernel void @global_atomic_fadd_uni_value_scope_agent(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, exec_lo, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, exec_hi, v0
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX7LESS-NEXT:    s_cbranch_execz .LBB4_3
-; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, v0
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[2:3], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v2, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v2
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX7LESS-NEXT:  .LBB4_3:
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_cbranch_execz .LBB4_3
-; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s1
-; GFX8-NEXT:    v_mov_b32_e32 v2, s0
-; GFX8-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[2:3], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX8-NEXT:  .LBB4_3:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_cbranch_execz .LBB4_3
-; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX9-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX9-NEXT:  .LBB4_3:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1064-NEXT:  .LBB4_3:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1032-NEXT:  .LBB4_3:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1164-NEXT:  .LBB4_3:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1132-NEXT:  .LBB4_3:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX9-DPP-NEXT:  .LBB4_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1064-DPP-NEXT:  .LBB4_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v0, v2, v[0:1], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1032-DPP-NEXT:  .LBB4_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1164-DPP-NEXT:  .LBB4_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v0, 4.0, v0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v2, v[0:1], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1132-DPP-NEXT:  .LBB4_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %result = atomicrmw fmin ptr addrspace(1) %ptr, float 4.0 syncscope("agent") monotonic, align 4
-  ret void
-}
-
-
-define amdgpu_kernel void @global_atomic_fadd_div_value_scope_agent(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    v_mul_f32_e32 v2, 1.0, v0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB5_1: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX7LESS-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB5_1
-; GFX7LESS-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX8-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX8-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX8-NEXT:    s_add_i32 s4, s4, 32
-; GFX8-NEXT:    s_min_u32 s4, s5, s4
-; GFX8-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX8-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX8-NEXT:    v_mul_f32_e32 v1, 1.0, v2
-; GFX8-NEXT:    v_mul_f32_e64 v2, 1.0, s6
-; GFX8-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX8-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX8-NEXT:    v_min_f32_e32 v2, v1, v2
-; GFX8-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX8-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execz .LBB5_5
-; GFX8-NEXT:  ; %bb.3:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    v_mul_f32_e32 v2, 1.0, v2
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mul_f32_e32 v0, 1.0, v1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX8-NEXT:  .LBB5_5:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX9-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX9-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX9-NEXT:    s_add_i32 s4, s4, 32
-; GFX9-NEXT:    s_min_u32 s4, s5, s4
-; GFX9-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX9-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX9-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX9-NEXT:    v_max_f32_e64 v2, s6, s6
-; GFX9-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX9-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX9-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX9-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX9-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execz .LBB5_5
-; GFX9-NEXT:  ; %bb.3:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX9-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX9-NEXT:  .LBB5_5:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX1064-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX1064-NEXT:    s_add_i32 s4, s4, 32
-; GFX1064-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1064-NEXT:    s_min_u32 s4, s5, s4
-; GFX1064-NEXT:    v_readlane_b32 s5, v0, s4
-; GFX1064-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1064-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX1064-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1064-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1064-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1064-NEXT:  ; %bb.3:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1064-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1064-NEXT:  .LBB5_5:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1032-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    s_ff1_i32_b32 s3, s2
-; GFX1032-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1032-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1032-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1032-NEXT:    s_andn2_b32 s2, s2, s3
-; GFX1032-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1032-NEXT:    v_max_f32_e64 v2, s4, s4
-; GFX1032-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1032-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1032-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1032-NEXT:  ; %bb.3:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1032-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1032-NEXT:  .LBB5_5:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_4) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_ctz_i32_b32 s4, s3
-; GFX1164-NEXT:    s_ctz_i32_b32 s5, s2
-; GFX1164-NEXT:    s_add_i32 s4, s4, 32
-; GFX1164-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1164-NEXT:    s_min_u32 s4, s5, s4
-; GFX1164-NEXT:    v_readlane_b32 s5, v0, s4
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1164-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1164-NEXT:    s_and_not1_b64 s[2:3], s[2:3], s[4:5]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1164-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1164-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1164-NEXT:  ; %bb.3:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_max_f32_e32 v2, v1, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1164-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1164-NEXT:  .LBB5_5:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    s_ctz_i32_b32 s3, s2
-; GFX1132-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1132-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1132-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 s2, s2, s3
-; GFX1132-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e64 v2, s4, s4
-; GFX1132-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1132-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1132-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1132-NEXT:  ; %bb.3:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_dual_max_f32 v2, v1, v1 :: v_dual_mov_b32 v1, s3
-; GFX1132-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-NEXT:    v_max_f32_e32 v0, v1, v1
-; GFX1132-NEXT:    v_min_f32_e32 v0, v0, v2
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1132-NEXT:  .LBB5_5:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v4
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:2 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:4 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:8 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v3, v3, v3
-; GFX9-DPP-NEXT:    v_min_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX9-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX9-DPP-NEXT:    v_readlane_b32 s4, v1, 63
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_max_f32_e64 v0, s4, s4
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s5, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, s5
-; GFX9-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX9-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX9-DPP-NEXT:  .LBB5_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1064-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_readlane_b32 s4, v1, 32
-; GFX1064-DPP-NEXT:    v_readlane_b32 s5, v1, 0
-; GFX1064-DPP-NEXT:    v_max_f32_e64 v1, s4, s4
-; GFX1064-DPP-NEXT:    v_max_f32_e64 v2, s5, s5
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1064-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1064-DPP-NEXT:  .LBB5_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1032-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v4
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, s3
-; GFX1032-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1032-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1032-DPP-NEXT:  .LBB5_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_waitcnt_depctr 0xfff
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0x7f800000
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_permlane64_b32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v0, v0, v0
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1164-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1164-DPP-NEXT:  .LBB5_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, 0x7f800000
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, 0x7f800000
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v2, v2 :: v_dual_mov_b32 v3, 0x7f800000
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v1, v1, v1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v2, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v3, v3 :: v_dual_mov_b32 v3, 0x7f800000
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v2, v3, v3 :: v_dual_mov_b32 v3, 0x7f800000
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v2, v3, v3
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1132-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v2, v2, v2
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_dual_max_f32 v0, v0, v0 :: v_dual_mov_b32 v5, s3
-; GFX1132-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_max_f32_e32 v4, v5, v5
-; GFX1132-DPP-NEXT:    v_min_f32_e32 v4, v4, v0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1132-DPP-NEXT:  .LBB5_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %id.x = call i32 @llvm.amdgcn.workitem.id.x()
-  %divValue = bitcast i32 %id.x to float
-  %result = atomicrmw fmin ptr addrspace(1) %ptr, float %divValue syncscope("agent") monotonic, align 4
-  ret void
-}
diff --git a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fsub.ll b/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fsub.ll
deleted file mode 100644
--- a/llvm/test/CodeGen/AMDGPU/global_atomics_scan_fsub.ll
+++ /dev/null
@@ -1,3100 +0,0 @@
-; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -march=amdgcn -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GFX7LESS %s
-; RUN: llc -march=amdgcn -mcpu=tonga -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX8 %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX9 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1064 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1032 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1164 %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=Iterative -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1132 %s
-; RUN: llc -march=amdgcn -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefix=GFX7LESS %s
-; RUN: llc -march=amdgcn -mcpu=gfx900 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX9-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1064-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1010 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1032-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=-wavefrontsize32,+wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1164-DPP %s
-; RUN: llc -march=amdgcn -mcpu=gfx1100 -mattr=+wavefrontsize32,-wavefrontsize64 -amdgpu-atomic-optimizer-strategy=DPP -verify-machineinstrs < %s | FileCheck -enable-var-scope -check-prefixes=GFX1132-DPP %s
-
-declare i32 @llvm.amdgcn.workitem.id.x()
-define amdgpu_kernel void @global_atomic_fadd_uni_value(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_uni_value:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_mov_b64 s[2:3], exec
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX7LESS-NEXT:    s_cbranch_execz .LBB0_3
-; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s6, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX7LESS-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s6
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    buffer_wbinvl1
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX7LESS-NEXT:  .LBB0_3:
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_uni_value:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX8-NEXT:    s_cbranch_execz .LBB0_3
-; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX8-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX8-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    buffer_wbinvl1_vol
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX8-NEXT:  .LBB0_3:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_uni_value:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX9-NEXT:    s_cbranch_execz .LBB0_3
-; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_bcnt1_i32_b64 s5, s[2:3]
-; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v0, s5
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    buffer_wbinvl1_vol
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX9-NEXT:  .LBB0_3:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_uni_value:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX1064-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1064-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    buffer_gl0_inv
-; GFX1064-NEXT:    buffer_gl1_inv
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1064-NEXT:  .LBB0_3:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_uni_value:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s4, vcc_lo
-; GFX1032-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1032-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1032-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    buffer_gl0_inv
-; GFX1032-NEXT:    buffer_gl1_inv
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1032-NEXT:  .LBB0_3:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_uni_value:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1164-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    buffer_gl0_inv
-; GFX1164-NEXT:    buffer_gl1_inv
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1164-NEXT:  .LBB0_3:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_uni_value:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1132-NEXT:    s_mov_b32 s4, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_dual_mul_f32 v2, 4.0, v0 :: v_dual_mov_b32 v1, s4
-; GFX1132-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    buffer_gl0_inv
-; GFX1132-NEXT:    buffer_gl1_inv
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1132-NEXT:  .LBB0_3:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_bcnt1_i32_b64 s5, s[2:3]
-; GFX9-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s5
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    buffer_wbinvl1_vol
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX9-DPP-NEXT:  .LBB0_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1064-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    buffer_gl0_inv
-; GFX1064-DPP-NEXT:    buffer_gl1_inv
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1064-DPP-NEXT:  .LBB0_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s4, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1032-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1032-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    buffer_gl0_inv
-; GFX1032-DPP-NEXT:    buffer_gl1_inv
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1032-DPP-NEXT:  .LBB0_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1164-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    buffer_gl0_inv
-; GFX1164-DPP-NEXT:    buffer_gl1_inv
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1164-DPP-NEXT:  .LBB0_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_uni_value:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s4, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB0_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_dual_mul_f32 v2, 4.0, v0 :: v_dual_mov_b32 v1, s4
-; GFX1132-DPP-NEXT:  .LBB0_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    buffer_gl0_inv
-; GFX1132-DPP-NEXT:    buffer_gl1_inv
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB0_2
-; GFX1132-DPP-NEXT:  .LBB0_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %result = atomicrmw fsub ptr addrspace(1) %ptr, float 4.0 seq_cst
-  ret void
-}
-
-define amdgpu_kernel void @global_atomic_fadd_div_value(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_div_value:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB1_1: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_sub_f32_e32 v1, v2, v0
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v2
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    buffer_wbinvl1
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v2
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB1_1
-; GFX7LESS-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_div_value:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX8-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX8-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX8-NEXT:    s_add_i32 s4, s4, 32
-; GFX8-NEXT:    s_min_u32 s4, s5, s4
-; GFX8-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX8-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX8-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX8-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX8-NEXT:    v_subrev_f32_e32 v2, s6, v2
-; GFX8-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX8-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execz .LBB1_5
-; GFX8-NEXT:  ; %bb.3:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    buffer_wbinvl1_vol
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX8-NEXT:  .LBB1_5:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_div_value:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX9-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX9-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX9-NEXT:    s_add_i32 s4, s4, 32
-; GFX9-NEXT:    s_min_u32 s4, s5, s4
-; GFX9-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX9-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX9-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX9-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX9-NEXT:    v_subrev_f32_e32 v2, s6, v2
-; GFX9-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX9-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execz .LBB1_5
-; GFX9-NEXT:  ; %bb.3:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX9-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    buffer_wbinvl1_vol
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX9-NEXT:  .LBB1_5:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_div_value:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX1064-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX1064-NEXT:    s_add_i32 s4, s4, 32
-; GFX1064-NEXT:    s_min_u32 s4, s5, s4
-; GFX1064-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX1064-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX1064-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1064-NEXT:    v_subrev_f32_e32 v2, s6, v2
-; GFX1064-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1064-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1064-NEXT:  ; %bb.3:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1064-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    buffer_gl0_inv
-; GFX1064-NEXT:    buffer_gl1_inv
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1064-NEXT:  .LBB1_5:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_div_value:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1032-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    s_ff1_i32_b32 s3, s2
-; GFX1032-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1032-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1032-NEXT:    s_andn2_b32 s2, s2, s3
-; GFX1032-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1032-NEXT:    v_subrev_f32_e32 v2, s4, v2
-; GFX1032-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1032-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1032-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1032-NEXT:  ; %bb.3:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1032-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    buffer_gl0_inv
-; GFX1032-NEXT:    buffer_gl1_inv
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1032-NEXT:  .LBB1_5:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_div_value:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_ctz_i32_b32 s4, s3
-; GFX1164-NEXT:    s_ctz_i32_b32 s5, s2
-; GFX1164-NEXT:    s_add_i32 s4, s4, 32
-; GFX1164-NEXT:    s_min_u32 s4, s5, s4
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX1164-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1164-NEXT:    s_and_not1_b64 s[2:3], s[2:3], s[4:5]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1164-NEXT:    v_subrev_f32_e32 v2, s6, v2
-; GFX1164-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1164-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1164-NEXT:  ; %bb.3:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1164-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    buffer_gl0_inv
-; GFX1164-NEXT:    buffer_gl1_inv
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1164-NEXT:  .LBB1_5:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_div_value:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:  .LBB1_1: ; %ComputeLoop
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_ctz_i32_b32 s3, s2
-; GFX1132-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1132-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 s2, s2, s3
-; GFX1132-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_subrev_f32_e32 v2, s4, v2
-; GFX1132-NEXT:    s_cbranch_scc1 .LBB1_1
-; GFX1132-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1132-NEXT:    s_cbranch_execz .LBB1_5
-; GFX1132-NEXT:  ; %bb.3:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-NEXT:  .LBB1_4: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1132-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    buffer_gl0_inv
-; GFX1132-NEXT:    buffer_gl1_inv
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB1_4
-; GFX1132-NEXT:  .LBB1_5:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v4
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:2 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:4 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:8 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX9-DPP-NEXT:    v_readlane_b32 s4, v1, 63
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v0, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s5, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, s5
-; GFX9-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_subrev_f32_e32 v4, s4, v5
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v4, v0, v[4:5], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    buffer_wbinvl1_vol
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX9-DPP-NEXT:  .LBB1_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1064-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_readlane_b32 s4, v1, 0
-; GFX1064-DPP-NEXT:    v_readlane_b32 s5, v1, 32
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_sub_f32_e64 v1, s4, s5
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v4, v5, v0
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    buffer_gl0_inv
-; GFX1064-DPP-NEXT:    buffer_gl1_inv
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1064-DPP-NEXT:  .LBB1_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1032-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v4
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, s3
-; GFX1032-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v4, v5, v0
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    buffer_gl0_inv
-; GFX1032-DPP-NEXT:    buffer_gl1_inv
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1032-DPP-NEXT:  .LBB1_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_waitcnt_depctr 0xfff
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_permlane64_b32 v2, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v4, v5, v0
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    buffer_gl0_inv
-; GFX1164-DPP-NEXT:    buffer_gl1_inv
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1164-DPP-NEXT:  .LBB1_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_div_value:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1132-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB1_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v5, s3
-; GFX1132-DPP-NEXT:  .LBB1_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v4, v5, v0
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_waitcnt_vscnt null, 0x0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    buffer_gl0_inv
-; GFX1132-DPP-NEXT:    buffer_gl1_inv
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB1_2
-; GFX1132-DPP-NEXT:  .LBB1_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %id.x = call i32 @llvm.amdgcn.workitem.id.x()
-  %divValue = bitcast i32 %id.x to float
-  %result = atomicrmw fsub ptr addrspace(1) %ptr, float %divValue seq_cst
-  ret void
-}
-
-define amdgpu_kernel void @global_atomic_fadd_uni_value_scope_defalut(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_mov_b64 s[2:3], exec
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX7LESS-NEXT:    s_cbranch_execz .LBB2_3
-; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s6, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX7LESS-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s6
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX7LESS-NEXT:  .LBB2_3:
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX8-NEXT:    s_cbranch_execz .LBB2_3
-; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX8-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX8-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX8-NEXT:  .LBB2_3:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX9-NEXT:    s_cbranch_execz .LBB2_3
-; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_bcnt1_i32_b64 s5, s[2:3]
-; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v0, s5
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX9-NEXT:  .LBB2_3:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX1064-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1064-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1064-NEXT:  .LBB2_3:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s4, vcc_lo
-; GFX1032-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1032-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1032-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1032-NEXT:  .LBB2_3:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1164-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1164-NEXT:  .LBB2_3:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1132-NEXT:    s_mov_b32 s4, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_dual_mul_f32 v2, 4.0, v0 :: v_dual_mov_b32 v1, s4
-; GFX1132-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1132-NEXT:  .LBB2_3:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_bcnt1_i32_b64 s5, s[2:3]
-; GFX9-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s5
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX9-DPP-NEXT:  .LBB2_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1064-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1064-DPP-NEXT:  .LBB2_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s4, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1032-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1032-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1032-DPP-NEXT:  .LBB2_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1164-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1164-DPP-NEXT:  .LBB2_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_uni_value_scope_defalut:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s4, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB2_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_dual_mul_f32 v2, 4.0, v0 :: v_dual_mov_b32 v1, s4
-; GFX1132-DPP-NEXT:  .LBB2_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB2_2
-; GFX1132-DPP-NEXT:  .LBB2_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %result = atomicrmw fsub ptr addrspace(1) %ptr, float 4.0 monotonic, align 4
-  ret void
-}
-
-define amdgpu_kernel void @global_atomic_fadd_div_value_scope_defalut(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB3_1: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_sub_f32_e32 v1, v2, v0
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v2
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v2
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB3_1
-; GFX7LESS-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX8-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX8-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX8-NEXT:    s_add_i32 s4, s4, 32
-; GFX8-NEXT:    s_min_u32 s4, s5, s4
-; GFX8-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX8-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX8-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX8-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX8-NEXT:    v_subrev_f32_e32 v2, s6, v2
-; GFX8-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX8-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execz .LBB3_5
-; GFX8-NEXT:  ; %bb.3:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX8-NEXT:  .LBB3_5:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX9-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX9-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX9-NEXT:    s_add_i32 s4, s4, 32
-; GFX9-NEXT:    s_min_u32 s4, s5, s4
-; GFX9-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX9-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX9-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX9-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX9-NEXT:    v_subrev_f32_e32 v2, s6, v2
-; GFX9-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX9-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execz .LBB3_5
-; GFX9-NEXT:  ; %bb.3:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX9-NEXT:  .LBB3_5:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX1064-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX1064-NEXT:    s_add_i32 s4, s4, 32
-; GFX1064-NEXT:    s_min_u32 s4, s5, s4
-; GFX1064-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX1064-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX1064-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1064-NEXT:    v_subrev_f32_e32 v2, s6, v2
-; GFX1064-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1064-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1064-NEXT:  ; %bb.3:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1064-NEXT:  .LBB3_5:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1032-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    s_ff1_i32_b32 s3, s2
-; GFX1032-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1032-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1032-NEXT:    s_andn2_b32 s2, s2, s3
-; GFX1032-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1032-NEXT:    v_subrev_f32_e32 v2, s4, v2
-; GFX1032-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1032-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1032-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1032-NEXT:  ; %bb.3:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1032-NEXT:  .LBB3_5:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_ctz_i32_b32 s4, s3
-; GFX1164-NEXT:    s_ctz_i32_b32 s5, s2
-; GFX1164-NEXT:    s_add_i32 s4, s4, 32
-; GFX1164-NEXT:    s_min_u32 s4, s5, s4
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX1164-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1164-NEXT:    s_and_not1_b64 s[2:3], s[2:3], s[4:5]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1164-NEXT:    v_subrev_f32_e32 v2, s6, v2
-; GFX1164-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1164-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1164-NEXT:  ; %bb.3:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1164-NEXT:  .LBB3_5:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:  .LBB3_1: ; %ComputeLoop
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_ctz_i32_b32 s3, s2
-; GFX1132-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1132-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 s2, s2, s3
-; GFX1132-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_subrev_f32_e32 v2, s4, v2
-; GFX1132-NEXT:    s_cbranch_scc1 .LBB3_1
-; GFX1132-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1132-NEXT:    s_cbranch_execz .LBB3_5
-; GFX1132-NEXT:  ; %bb.3:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-NEXT:  .LBB3_4: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB3_4
-; GFX1132-NEXT:  .LBB3_5:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v4
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:2 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:4 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:8 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX9-DPP-NEXT:    v_readlane_b32 s4, v1, 63
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v0, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s5, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, s5
-; GFX9-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_subrev_f32_e32 v4, s4, v5
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v4, v0, v[4:5], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX9-DPP-NEXT:  .LBB3_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1064-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_readlane_b32 s4, v1, 0
-; GFX1064-DPP-NEXT:    v_readlane_b32 s5, v1, 32
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_sub_f32_e64 v1, s4, s5
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v4, v5, v0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1064-DPP-NEXT:  .LBB3_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1032-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v4
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, s3
-; GFX1032-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v4, v5, v0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1032-DPP-NEXT:  .LBB3_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_waitcnt_depctr 0xfff
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_permlane64_b32 v2, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v4, v5, v0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1164-DPP-NEXT:  .LBB3_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_div_value_scope_defalut:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1132-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB3_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v5, s3
-; GFX1132-DPP-NEXT:  .LBB3_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v4, v5, v0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB3_2
-; GFX1132-DPP-NEXT:  .LBB3_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %id.x = call i32 @llvm.amdgcn.workitem.id.x()
-  %divValue = bitcast i32 %id.x to float
-  %result = atomicrmw fsub ptr addrspace(1) %ptr, float %divValue monotonic, align 4
-  ret void
-}
-
-
-define amdgpu_kernel void @global_atomic_fadd_uni_value_scope_agent(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_mov_b64 s[2:3], exec
-; GFX7LESS-NEXT:    v_mbcnt_lo_u32_b32_e64 v0, s2, 0
-; GFX7LESS-NEXT:    v_mbcnt_hi_u32_b32_e32 v0, s3, v0
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX7LESS-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX7LESS-NEXT:    s_cbranch_execz .LBB4_3
-; GFX7LESS-NEXT:  ; %bb.1:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s6, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX7LESS-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, s6
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v1
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v0
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v1
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v1, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX7LESS-NEXT:  .LBB4_3:
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX8-NEXT:    s_cbranch_execz .LBB4_3
-; GFX8-NEXT:  ; %bb.1:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX8-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX8-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX8-NEXT:  .LBB4_3:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX9-NEXT:    s_cbranch_execz .LBB4_3
-; GFX9-NEXT:  ; %bb.1:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_bcnt1_i32_b64 s5, s[2:3]
-; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v0, s5
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX9-NEXT:  .LBB4_3:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX1064-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1064-NEXT:  ; %bb.1:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1064-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1064-NEXT:  .LBB4_3:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s4, vcc_lo
-; GFX1032-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1032-NEXT:  ; %bb.1:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1032-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1032-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1032-NEXT:  .LBB4_3:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1164-NEXT:  ; %bb.1:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1164-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1164-NEXT:  .LBB4_3:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1132-NEXT:    s_mov_b32 s4, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1132-NEXT:  ; %bb.1:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_dual_mul_f32 v2, 4.0, v0 :: v_dual_mov_b32 v1, s4
-; GFX1132-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1132-NEXT:  .LBB4_3:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_bcnt1_i32_b64 s5, s[2:3]
-; GFX9-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s5
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX9-DPP-NEXT:  .LBB4_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[4:5], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1064-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1064-DPP-NEXT:  .LBB4_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s4, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1032-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1032-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1032-DPP-NEXT:  .LBB4_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    s_mov_b64 s[4:5], exec
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s2, 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v0, s3, v0
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    s_bcnt1_i32_b64 s2, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_mul_f32_e32 v2, 4.0, v0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, s4
-; GFX1164-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1164-DPP-NEXT:  .LBB4_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_uni_value_scope_agent:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, s3, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s4, exec_lo
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB4_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    s_bcnt1_i32_b32 s3, s3
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-DPP-NEXT:    v_cvt_f32_ubyte0_e32 v0, s3
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s4, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_dual_mul_f32 v2, 4.0, v0 :: v_dual_mov_b32 v1, s4
-; GFX1132-DPP-NEXT:  .LBB4_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB4_2
-; GFX1132-DPP-NEXT:  .LBB4_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %result = atomicrmw fsub ptr addrspace(1) %ptr, float 4.0 syncscope("agent") monotonic, align 4
-  ret void
-}
-
-
-define amdgpu_kernel void @global_atomic_fadd_div_value_scope_agent(ptr addrspace(1) %ptr) {
-; GFX7LESS-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX7LESS:       ; %bb.0:
-; GFX7LESS-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x9
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX7LESS-NEXT:    s_mov_b64 s[4:5], 0
-; GFX7LESS-NEXT:    s_mov_b32 s3, 0xf000
-; GFX7LESS-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, s2
-; GFX7LESS-NEXT:    s_mov_b32 s2, -1
-; GFX7LESS-NEXT:  .LBB5_1: ; %atomicrmw.start
-; GFX7LESS-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX7LESS-NEXT:    v_sub_f32_e32 v1, v2, v0
-; GFX7LESS-NEXT:    s_waitcnt expcnt(0)
-; GFX7LESS-NEXT:    v_mov_b32_e32 v4, v2
-; GFX7LESS-NEXT:    v_mov_b32_e32 v3, v1
-; GFX7LESS-NEXT:    buffer_atomic_cmpswap v[3:4], off, s[0:3], 0 glc
-; GFX7LESS-NEXT:    s_waitcnt vmcnt(0)
-; GFX7LESS-NEXT:    v_cmp_eq_u32_e32 vcc, v3, v2
-; GFX7LESS-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
-; GFX7LESS-NEXT:    v_mov_b32_e32 v2, v3
-; GFX7LESS-NEXT:    s_andn2_b64 exec, exec, s[4:5]
-; GFX7LESS-NEXT:    s_cbranch_execnz .LBB5_1
-; GFX7LESS-NEXT:  ; %bb.2: ; %atomicrmw.end
-; GFX7LESS-NEXT:    s_endpgm
-;
-; GFX8-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX8:       ; %bb.0:
-; GFX8-NEXT:    s_mov_b64 s[2:3], exec
-; GFX8-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX8-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX8-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX8-NEXT:    s_add_i32 s4, s4, 32
-; GFX8-NEXT:    s_min_u32 s4, s5, s4
-; GFX8-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX8-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX8-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX8-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX8-NEXT:    v_subrev_f32_e32 v2, s6, v2
-; GFX8-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX8-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX8-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX8-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX8-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX8-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execz .LBB5_5
-; GFX8-NEXT:  ; %bb.3:
-; GFX8-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX8-NEXT:    s_mov_b64 s[2:3], 0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX8-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX8-NEXT:    v_mov_b32_e32 v1, s4
-; GFX8-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX8-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX8-NEXT:    v_mov_b32_e32 v4, s1
-; GFX8-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX8-NEXT:    v_mov_b32_e32 v3, s0
-; GFX8-NEXT:    flat_atomic_cmpswap v0, v[3:4], v[0:1] glc
-; GFX8-NEXT:    s_waitcnt vmcnt(0)
-; GFX8-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX8-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX8-NEXT:    v_mov_b32_e32 v1, v0
-; GFX8-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX8-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX8-NEXT:  .LBB5_5:
-; GFX8-NEXT:    s_endpgm
-;
-; GFX9-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_mov_b64 s[2:3], exec
-; GFX9-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX9-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX9-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX9-NEXT:    s_add_i32 s4, s4, 32
-; GFX9-NEXT:    s_min_u32 s4, s5, s4
-; GFX9-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX9-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX9-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX9-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX9-NEXT:    v_subrev_f32_e32 v2, s6, v2
-; GFX9-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX9-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX9-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX9-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX9-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execz .LBB5_5
-; GFX9-NEXT:  ; %bb.3:
-; GFX9-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-NEXT:    v_mov_b32_e32 v3, 0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    s_load_dword s4, s[0:1], 0x0
-; GFX9-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v1, s4
-; GFX9-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX9-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX9-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-NEXT:    v_mov_b32_e32 v1, v0
-; GFX9-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX9-NEXT:  .LBB5_5:
-; GFX9-NEXT:    s_endpgm
-;
-; GFX1064-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1064:       ; %bb.0:
-; GFX1064-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1064-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1064-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    s_ff1_i32_b32 s4, s3
-; GFX1064-NEXT:    s_ff1_i32_b32 s5, s2
-; GFX1064-NEXT:    s_add_i32 s4, s4, 32
-; GFX1064-NEXT:    s_min_u32 s4, s5, s4
-; GFX1064-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX1064-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1064-NEXT:    s_andn2_b64 s[2:3], s[2:3], s[4:5]
-; GFX1064-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1064-NEXT:    v_subrev_f32_e32 v2, s6, v2
-; GFX1064-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1064-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1064-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v0
-; GFX1064-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1064-NEXT:  ; %bb.3:
-; GFX1064-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1064-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1064-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1064-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1064-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1064-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1064-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1064-NEXT:  .LBB5_5:
-; GFX1064-NEXT:    s_endpgm
-;
-; GFX1032-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1032:       ; %bb.0:
-; GFX1032-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1032-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1032-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    s_ff1_i32_b32 s3, s2
-; GFX1032-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1032-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1032-NEXT:    s_andn2_b32 s2, s2, s3
-; GFX1032-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1032-NEXT:    v_subrev_f32_e32 v2, s4, v2
-; GFX1032-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1032-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1032-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1032-NEXT:    s_mov_b32 s2, 0
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v0
-; GFX1032-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1032-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1032-NEXT:  ; %bb.3:
-; GFX1032-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1032-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1032-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1032-NEXT:    global_atomic_cmpswap v0, v3, v[0:1], s[0:1] glc
-; GFX1032-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1032-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1032-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1032-NEXT:  .LBB5_5:
-; GFX1032-NEXT:    s_endpgm
-;
-; GFX1164-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1164:       ; %bb.0:
-; GFX1164-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_ctz_i32_b32 s4, s3
-; GFX1164-NEXT:    s_ctz_i32_b32 s5, s2
-; GFX1164-NEXT:    s_add_i32 s4, s4, 32
-; GFX1164-NEXT:    s_min_u32 s4, s5, s4
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-NEXT:    v_readlane_b32 s6, v0, s4
-; GFX1164-NEXT:    s_lshl_b64 s[4:5], 1, s4
-; GFX1164-NEXT:    s_and_not1_b64 s[2:3], s[2:3], s[4:5]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    s_cmp_lg_u64 s[2:3], 0
-; GFX1164-NEXT:    v_subrev_f32_e32 v2, s6, v2
-; GFX1164-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1164-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1164-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-NEXT:    v_mbcnt_hi_u32_b32 v0, exec_hi, v0
-; GFX1164-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1164-NEXT:    s_xor_b64 s[2:3], exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1164-NEXT:  ; %bb.3:
-; GFX1164-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-NEXT:    v_mov_b32_e32 v1, s2
-; GFX1164-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1164-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1164-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1164-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-NEXT:    v_cmp_eq_u32_e32 vcc, v0, v1
-; GFX1164-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1164-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1164-NEXT:  .LBB5_5:
-; GFX1164-NEXT:    s_endpgm
-;
-; GFX1132-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1132:       ; %bb.0:
-; GFX1132-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1132-NEXT:    s_mov_b32 s2, exec_lo
-; GFX1132-NEXT:  .LBB5_1: ; %ComputeLoop
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_ctz_i32_b32 s3, s2
-; GFX1132-NEXT:    v_readlane_b32 s4, v0, s3
-; GFX1132-NEXT:    s_lshl_b32 s3, 1, s3
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1) | instskip(NEXT) | instid1(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 s2, s2, s3
-; GFX1132-NEXT:    s_cmp_lg_u32 s2, 0
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_subrev_f32_e32 v2, s4, v2
-; GFX1132-NEXT:    s_cbranch_scc1 .LBB5_1
-; GFX1132-NEXT:  ; %bb.2: ; %ComputeEnd
-; GFX1132-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1132-NEXT:    s_mov_b32 s2, 0
-; GFX1132-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_cmpx_eq_u32_e32 0, v0
-; GFX1132-NEXT:    s_xor_b32 s3, exec_lo, s3
-; GFX1132-NEXT:    s_cbranch_execz .LBB5_5
-; GFX1132-NEXT:  ; %bb.3:
-; GFX1132-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-NEXT:    v_mov_b32_e32 v3, 0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-NEXT:    v_mov_b32_e32 v1, s3
-; GFX1132-NEXT:  .LBB5_4: ; %atomicrmw.start
-; GFX1132-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-NEXT:    v_sub_f32_e32 v0, v1, v2
-; GFX1132-NEXT:    global_atomic_cmpswap_b32 v0, v3, v[0:1], s[0:1] glc
-; GFX1132-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v0, v1
-; GFX1132-NEXT:    v_mov_b32_e32 v1, v0
-; GFX1132-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-NEXT:    s_cbranch_execnz .LBB5_4
-; GFX1132-NEXT:  .LBB5_5:
-; GFX1132-NEXT:    s_endpgm
-;
-; GFX9-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX9-DPP:       ; %bb.0:
-; GFX9-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX9-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v4
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX9-DPP-NEXT:    s_not_b64 exec, exec
-; GFX9-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:1 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:2 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:4 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_shr:8 row_mask:0xf bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v3, v2 row_bcast:15 row_mask:0xa bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v2, v2, v3
-; GFX9-DPP-NEXT:    s_nop 1
-; GFX9-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_bcast:31 row_mask:0xc bank_mask:0xf
-; GFX9-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX9-DPP-NEXT:    v_readlane_b32 s4, v1, 63
-; GFX9-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX9-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX9-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX9-DPP-NEXT:  ; %bb.1:
-; GFX9-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX9-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v0, 0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    s_load_dword s5, s[0:1], 0x0
-; GFX9-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, s5
-; GFX9-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX9-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX9-DPP-NEXT:    v_subrev_f32_e32 v4, s4, v5
-; GFX9-DPP-NEXT:    global_atomic_cmpswap v4, v0, v[4:5], s[0:1] glc
-; GFX9-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX9-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX9-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX9-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX9-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX9-DPP-NEXT:  .LBB5_3:
-; GFX9-DPP-NEXT:    s_endpgm
-;
-; GFX1064-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1064-DPP:       ; %bb.0:
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1064-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1064-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1064-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1064-DPP-NEXT:    v_readlane_b32 s4, v1, 0
-; GFX1064-DPP-NEXT:    v_readlane_b32 s5, v1, 32
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1064-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1064-DPP-NEXT:    v_sub_f32_e64 v1, s4, s5
-; GFX1064-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1064-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v4
-; GFX1064-DPP-NEXT:    s_and_saveexec_b64 s[2:3], vcc
-; GFX1064-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1064-DPP-NEXT:  ; %bb.1:
-; GFX1064-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    s_load_dword s2, s[0:1], 0x0
-; GFX1064-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1064-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1064-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1064-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1064-DPP-NEXT:    v_sub_f32_e32 v4, v5, v0
-; GFX1064-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1064-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1064-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1064-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1064-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1064-DPP-NEXT:    s_andn2_b64 exec, exec, s[2:3]
-; GFX1064-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1064-DPP-NEXT:  .LBB5_3:
-; GFX1064-DPP-NEXT:    s_endpgm
-;
-; GFX1032-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1032-DPP:       ; %bb.0:
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1032-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1032-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1032-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1032-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1032-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1032-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1032-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 0, v4
-; GFX1032-DPP-NEXT:    s_and_saveexec_b32 s3, vcc_lo
-; GFX1032-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1032-DPP-NEXT:  ; %bb.1:
-; GFX1032-DPP-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x24
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    s_load_dword s3, s[0:1], 0x0
-; GFX1032-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, s3
-; GFX1032-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1032-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1032-DPP-NEXT:    v_sub_f32_e32 v4, v5, v0
-; GFX1032-DPP-NEXT:    global_atomic_cmpswap v4, v6, v[4:5], s[0:1] glc
-; GFX1032-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1032-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1032-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1032-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1032-DPP-NEXT:    s_andn2_b32 exec_lo, exec_lo, s2
-; GFX1032-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1032-DPP-NEXT:  .LBB5_3:
-; GFX1032-DPP-NEXT:    s_endpgm
-;
-; GFX1164-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1164-DPP:       ; %bb.0:
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1164-DPP-NEXT:    s_not_b64 exec, exec
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_waitcnt_depctr 0xfff
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1164-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_permlane64_b32 v2, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    v_mbcnt_lo_u32_b32 v0, exec_lo, 0
-; GFX1164-DPP-NEXT:    s_or_saveexec_b64 s[2:3], -1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1164-DPP-NEXT:    s_mov_b64 exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instid1(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    v_mbcnt_hi_u32_b32 v4, exec_hi, v0
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], exec
-; GFX1164-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1164-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1164-DPP-NEXT:  ; %bb.1:
-; GFX1164-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    s_load_b32 s2, s[0:1], 0x0
-; GFX1164-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, s2
-; GFX1164-DPP-NEXT:    s_mov_b64 s[2:3], 0
-; GFX1164-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1164-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1164-DPP-NEXT:    v_sub_f32_e32 v4, v5, v0
-; GFX1164-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1164-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1164-DPP-NEXT:    v_cmp_eq_u32_e32 vcc, v4, v5
-; GFX1164-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1164-DPP-NEXT:    s_or_b64 s[2:3], vcc, s[2:3]
-; GFX1164-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1164-DPP-NEXT:    s_and_not1_b64 exec, exec, s[2:3]
-; GFX1164-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1164-DPP-NEXT:  .LBB5_3:
-; GFX1164-DPP-NEXT:    s_endpgm
-;
-; GFX1132-DPP-LABEL: global_atomic_fadd_div_value_scope_agent:
-; GFX1132-DPP:       ; %bb.0:
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v1, 1
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v0
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1132-DPP-NEXT:    s_not_b32 exec_lo, exec_lo
-; GFX1132-DPP-NEXT:    s_or_saveexec_b32 s2, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v1, v2 row_xmask:1 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v2, v1
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v2, 1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:2 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1132-DPP-NEXT:    v_bfrev_b32_e32 v3, 1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v2, v1 row_xmask:4 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_dpp v3, v1 row_xmask:8 row_mask:0xf bank_mask:0xf
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v1, v3
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v2, v1
-; GFX1132-DPP-NEXT:    v_permlanex16_b32 v2, v2, -1, -1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v1, v1, v2
-; GFX1132-DPP-NEXT:    s_mov_b32 exec_lo, s2
-; GFX1132-DPP-NEXT:    v_mbcnt_lo_u32_b32 v4, exec_lo, 0
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(SKIP_2) | instid1(VALU_DEP_2)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v0, v1
-; GFX1132-DPP-NEXT:    s_mov_b32 s2, 0
-; GFX1132-DPP-NEXT:    s_mov_b32 s3, exec_lo
-; GFX1132-DPP-NEXT:    v_cmpx_eq_u32_e32 0, v4
-; GFX1132-DPP-NEXT:    s_cbranch_execz .LBB5_3
-; GFX1132-DPP-NEXT:  ; %bb.1:
-; GFX1132-DPP-NEXT:    s_load_b64 s[0:1], s[0:1], 0x24
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v6, 0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    s_load_b32 s3, s[0:1], 0x0
-; GFX1132-DPP-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v5, s3
-; GFX1132-DPP-NEXT:  .LBB5_2: ; %atomicrmw.start
-; GFX1132-DPP-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1132-DPP-NEXT:    v_sub_f32_e32 v4, v5, v0
-; GFX1132-DPP-NEXT:    global_atomic_cmpswap_b32 v4, v6, v[4:5], s[0:1] glc
-; GFX1132-DPP-NEXT:    s_waitcnt vmcnt(0)
-; GFX1132-DPP-NEXT:    v_cmp_eq_u32_e32 vcc_lo, v4, v5
-; GFX1132-DPP-NEXT:    v_mov_b32_e32 v5, v4
-; GFX1132-DPP-NEXT:    s_or_b32 s2, vcc_lo, s2
-; GFX1132-DPP-NEXT:    s_delay_alu instid0(SALU_CYCLE_1)
-; GFX1132-DPP-NEXT:    s_and_not1_b32 exec_lo, exec_lo, s2
-; GFX1132-DPP-NEXT:    s_cbranch_execnz .LBB5_2
-; GFX1132-DPP-NEXT:  .LBB5_3:
-; GFX1132-DPP-NEXT:    s_endpgm
-  %id.x = call i32 @llvm.amdgcn.workitem.id.x()
-  %divValue = bitcast i32 %id.x to float
-  %result = atomicrmw fsub ptr addrspace(1) %ptr, float %divValue syncscope("agent") monotonic, align 4
-  ret void
-}
diff --git a/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-simplify-cfg-CAS-block.ll b/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-simplify-cfg-CAS-block.ll
new file mode 100644
--- /dev/null
+++ b/llvm/test/Transforms/AtomicExpand/AMDGPU/expand-atomic-simplify-cfg-CAS-block.ll
@@ -0,0 +1,62 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
+; RUN: opt -S -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a -atomic-expand %s | FileCheck -check-prefix=GFX90A %s
+
+declare i32 @llvm.amdgcn.workitem.id.x()
+
+define amdgpu_kernel void @divergent_cfg(ptr addrspace(1) %out, float %in) #0 {
+; GFX90A-LABEL: @divergent_cfg(
+; GFX90A-NEXT:  entry:
+; GFX90A-NEXT:    [[TID:%.*]] = call i32 @llvm.amdgcn.workitem.id.x()
+; GFX90A-NEXT:    [[D_CMP:%.*]] = icmp ult i32 [[TID]], 16
+; GFX90A-NEXT:    br i1 [[D_CMP]], label [[IF:%.*]], label [[ELSE:%.*]]
+; GFX90A:       if:
+; GFX90A-NEXT:    [[TMP0:%.*]] = load float, ptr addrspace(1) [[OUT:%.*]], align 4
+; GFX90A-NEXT:    br label [[ATOMICRMW_START:%.*]]
+; GFX90A:       atomicrmw.start:
+; GFX90A-NEXT:    [[LOADED:%.*]] = phi float [ [[TMP0]], [[IF]] ], [ [[TMP4:%.*]], [[ATOMICRMW_START]] ]
+; GFX90A-NEXT:    [[NEW:%.*]] = fadd float [[LOADED]], [[IN:%.*]]
+; GFX90A-NEXT:    [[TMP1:%.*]] = bitcast float [[NEW]] to i32
+; GFX90A-NEXT:    [[TMP2:%.*]] = bitcast float [[LOADED]] to i32
+; GFX90A-NEXT:    [[TMP3:%.*]] = cmpxchg ptr addrspace(1) [[OUT]], i32 [[TMP2]], i32 [[TMP1]] seq_cst seq_cst, align 4
+; GFX90A-NEXT:    [[SUCCESS:%.*]] = extractvalue { i32, i1 } [[TMP3]], 1
+; GFX90A-NEXT:    [[NEWLOADED:%.*]] = extractvalue { i32, i1 } [[TMP3]], 0
+; GFX90A-NEXT:    [[TMP4]] = bitcast i32 [[NEWLOADED]] to float
+; GFX90A-NEXT:    br i1 [[SUCCESS]], label [[ENDIF:%.*]], label [[ATOMICRMW_START]]
+; GFX90A:       else:
+; GFX90A-NEXT:    [[TMP5:%.*]] = load float, ptr addrspace(1) [[OUT]], align 4
+; GFX90A-NEXT:    br label [[ATOMICRMW_START2:%.*]]
+; GFX90A:       atomicrmw.start2:
+; GFX90A-NEXT:    [[LOADED3:%.*]] = phi float [ [[TMP5]], [[ELSE]] ], [ [[TMP9:%.*]], [[ATOMICRMW_START2]] ]
+; GFX90A-NEXT:    [[NEW4:%.*]] = fadd float [[LOADED3]], [[IN]]
+; GFX90A-NEXT:    [[TMP6:%.*]] = bitcast float [[NEW4]] to i32
+; GFX90A-NEXT:    [[TMP7:%.*]] = bitcast float [[LOADED3]] to i32
+; GFX90A-NEXT:    [[TMP8:%.*]] = cmpxchg ptr addrspace(1) [[OUT]], i32 [[TMP7]], i32 [[TMP6]] seq_cst seq_cst, align 4
+; GFX90A-NEXT:    [[SUCCESS5:%.*]] = extractvalue { i32, i1 } [[TMP8]], 1
+; GFX90A-NEXT:    [[NEWLOADED6:%.*]] = extractvalue { i32, i1 } [[TMP8]], 0
+; GFX90A-NEXT:    [[TMP9]] = bitcast i32 [[NEWLOADED6]] to float
+; GFX90A-NEXT:    br i1 [[SUCCESS5]], label [[ENDIF]], label [[ATOMICRMW_START2]]
+; GFX90A:       endif:
+; GFX90A-NEXT:    [[COMBINE:%.*]] = phi float [ [[TMP4]], [[ATOMICRMW_START]] ], [ [[TMP9]], [[ATOMICRMW_START2]] ]
+; GFX90A-NEXT:    store float [[COMBINE]], ptr addrspace(1) [[OUT]], align 4
+; GFX90A-NEXT:    ret void
+;
+entry:
+  %tid = call i32 @llvm.amdgcn.workitem.id.x()
+  %d_cmp = icmp ult i32 %tid, 16
+  br i1 %d_cmp, label %if, label %else
+
+if:
+  %res_if = atomicrmw fadd ptr  addrspace(1) %out, float %in seq_cst
+  br label %endif
+
+else:
+  %res_else = atomicrmw fadd ptr  addrspace(1) %out, float %in seq_cst
+  br label %endif
+
+endif:
+  %combine = phi float [%res_if, %if], [%res_else, %else]
+  store float %combine, ptr addrspace(1) %out
+  ret void
+}
+
+attributes #0 = { "denormal-fp-math-f32"="preserve-sign,preserve-sign" "amdgpu-unsafe-fp-atomics"="true" }