[PATCH] D106909: [clang] Add clang builtins support for gfx90a

Fri Jul 30 11:10:06 PDT 2021

gandhi21299 marked 7 inline comments as done.
gandhi21299 added inline comments.

================
Comment at: clang/include/clang/Basic/BuiltinsAMDGPU.def:201
+TARGET_BUILTIN(__builtin_amdgcn_global_atomic_fadd_f32, "ff*1fi", "t", "gfx90a-insts")
+TARGET_BUILTIN(__builtin_amdgcn_global_atomic_fadd_2f16, "hh*1hi", "t", "gfx90a-insts")
+TARGET_BUILTIN(__builtin_amdgcn_global_atomic_fmin_f64, "dd*1di", "t", "gfx90a-insts")
----------------
arsenm wrote:
> yaxunl wrote:
> > arsenm wrote:
> > > "_2f16" looks weird to me. The instruction names call it "pk"
> > This is to have a consistent postfix naming convention, since the stem part here are the same. the postfix is for the argument type of the builtin function.
> Just a plain 2 isn't consistent either. The llvm type naming convention would add a v prefix, but the builtins should probably follow the instructions
Yea, v2f16 looks reasonable.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D106909/new/

https://reviews.llvm.org/D106909