[llvm] c66777e - AMDGPU: Generalize atomicrmw handling in custom expansion

[Matt Arsenault via llvm-commits]

Author: Matt Arsenault
Date: 2024-08-08T12:30:24+04:00
New Revision: c66777ee1b55ba2349b71d35fd03149803c8c819

URL: https://github.com/llvm/llvm-project/commit/c66777ee1b55ba2349b71d35fd03149803c8c819
DIFF: https://github.com/llvm/llvm-project/commit/c66777ee1b55ba2349b71d35fd03149803c8c819.diff

LOG: AMDGPU: Generalize atomicrmw handling in custom expansion

Use the utility function instead of assuming fadd. No change
as-is, but will soon be used for other expansions.

Added: 
    

Modified: 
    llvm/lib/Target/AMDGPU/SIISelLowering.cpp

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 4d58e51908182..4e9c271197613 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -42,6 +42,7 @@
 #include "llvm/Support/CommandLine.h"
 #include "llvm/Support/KnownBits.h"
 #include "llvm/Support/ModRef.h"
+#include "llvm/Transforms/Utils/LowerAtomic.h"
 #include <optional>
 
 using namespace llvm;
@@ -16645,7 +16646,9 @@ void SITargetLowering::emitExpandAtomicRMW(AtomicRMWInst *AI) const {
       Addr, PointerType::get(Ctx, AMDGPUAS::PRIVATE_ADDRESS));
   Value *LoadedPrivate =
       Builder.CreateLoad(ValTy, CastToPrivate, "loaded.private");
-  Value *NewVal = Builder.CreateFAdd(LoadedPrivate, Val, "val.new");
+
+  Value *NewVal = buildAtomicRMWValue(Op, Builder, LoadedPrivate, Val);
+
   Builder.CreateStore(NewVal, CastToPrivate);
   Builder.CreateBr(PhiBB);