[PATCH] D130579: AMDGPU: Remove manual selection for atomic fadd

[Petar Avramovic via Phabricator via llvm-commits]

Petar.Avramovic added inline comments.


================
Comment at: llvm/lib/Target/AMDGPU/FLATInstructions.td:1487
 
-let OtherPredicates = [HasAtomicFaddNoRtnInsts] in
+let OtherPredicates = [HasAtomicFaddNoRtnInsts] in {
 defm : GlobalFLATNoRtnAtomicPats <GLOBAL_ATOMIC_ADD_F32,    atomic_load_fadd_global_noret_32, f32>;
----------------
Predicates are edited to match predicates for atomicrmw fadd (_NO_RTN)
see llvm/test/CodeGen/AMDGPU/llvm.amdgcn.atomic.fadd.ll global_atomic_fadd_f32_wrong_subtarget

I left v2f16_rtn, f64_no_rtn, f64_rtn patterns in isGFX90APlus, should each of them get feature for itself? 

Here is the list of predicates and atomic global fadd instructions they have
HasAtomicFaddRtnInsts: f32_rtn
HasAtomicFaddNoRtnInsts: f32_no_rtn
HasAtomicPkFaddNoRtnInsts: v2f16_no_rtn
isGFX90APlus: v2f16_rtn, f64_no_rtn, f64_rtn

GFX90A has all of the above



CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D130579/new/

https://reviews.llvm.org/D130579