[PATCH] D131560: AMDGPU: Improve atomicrmw fadd selection

[Brian Sumner via Phabricator via llvm-commits]

b-sumner added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/SIISelLowering.cpp:12781
+      // Global fadd f32 no-rtn for gfx908 (and gfx11+).
+      if (!Subtarget->hasGFX90AInsts() && AS == AMDGPUAS::GLOBAL_ADDRESS &&
+          Ty->isFloatTy() && RMW->use_empty())
----------------
rampitec wrote:
> I think both are still unsafe even when supported, on both targets.
I agree, these are unsafe.


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D131560/new/

https://reviews.llvm.org/D131560