[PATCH] D147159: [AMDGPU][GFX940] Simplify definitions of GLOBAL_ATOMIC

Thu Mar 30 12:02:04 PDT 2023

rampitec added inline comments.

================
Comment at: llvm/lib/Target/AMDGPU/FLATInstructions.td:730

+defm GLOBAL_ATOMIC_ADD_F64 : FLAT_Global_Atomic_Pseudo<"global_atomic_add_f64", VReg_64, f64>;
+defm GLOBAL_ATOMIC_MIN_F64 : FLAT_Global_Atomic_Pseudo<"global_atomic_min_f64", VReg_64, f64>;
----------------
mariusz-sikora-at-amd wrote:
> rampitec wrote:
> > I do not understand this. These were only added in gfx90a.
> This patch is only removing 'complexity' from gfx940. For gfx90A we still have Real Instruction under SubtargetPredicate = IsGFX90APlus.
> I will add test for gfx908.
> 
> This patch is only removing 'complexity' from gfx940. For gfx90A we still have Real Instruction under SubtargetPredicate = IsGFX90APlus.
> I will add test for gfx908.

But isn't pseudo still need a predicate too, not just real instruction?

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D147159/new/

https://reviews.llvm.org/D147159