[PATCH] D128205: [AMDGPU] gfx11 Select on Buffer Atomic FAdd Rtn type

[Joe Nash via Phabricator via llvm-commits]

Joe_Nash added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/SIISelLowering.cpp:12636
     if ((AS == AMDGPUAS::GLOBAL_ADDRESS || AS == AMDGPUAS::FLAT_ADDRESS) &&
-         Subtarget->hasAtomicFaddInsts()) {
+        Subtarget->hasAtomicFaddNoRtnInsts()) {
       if (Subtarget->hasGFX940Insts())
----------------
foad wrote:
> Is this change intentional? What is the effect of it?
I think this is NFC. hasAtomicFaddInsts definition was changed in D124538. There are no current subtargets that have FeatureAtomicFaddRtnInsts but not FeatureAtomicFaddNoRtnInsts. So they are both correct, but this is more precise. I would need @Petar.Avramovic to confirm.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D128205/new/

https://reviews.llvm.org/D128205