[PATCH] D128952: [AMDGPU] Add WMMA clang builtins

[Piotr Sobczak via Phabricator via cfe-commits]

piotr created this revision.
Herald added subscribers: kosarev, kerbowa, t-tye, tpr, dstuttard, yaxunl, jvesely, kzhuravl.
Herald added a project: All.
piotr requested review of this revision.
Herald added subscribers: cfe-commits, wdng.
Herald added a project: clang.

Add WMMA clang builtins and tests. Extra changes in code
are needed to handle function overloads.

WavefrontSize 32:
__builtin_amdgcn_wmma_f32_16x16x16_f16_w32
__builtin_amdgcn_wmma_f32_16x16x16_bf16_w32
__builtin_amdgcn_wmma_f16_16x16x16_f16_w32
__builtin_amdgcn_wmma_bf16_16x16x16_bf16_w32
__builtin_amdgcn_wmma_i32_16x16x16_iu8_w32
__builtin_amdgcn_wmma_i32_16x16x16_iu4_w32

WavefrontSize 64:
__builtin_amdgcn_wmma_f32_16x16x16_f16_w64
__builtin_amdgcn_wmma_f32_16x16x16_bf16_w64
__builtin_amdgcn_wmma_f16_16x16x16_f16_w64
__builtin_amdgcn_wmma_bf16_16x16x16_bf16_w64
__builtin_amdgcn_wmma_i32_16x16x16_iu8_w64
__builtin_amdgcn_wmma_i32_16x16x16_iu4_w64


Repository:
  rG LLVM Github Monorepo

https://reviews.llvm.org/D128952

Files:
  clang/include/clang/Basic/BuiltinsAMDGPU.def
  clang/lib/CodeGen/CGBuiltin.cpp
  clang/test/CodeGenOpenCL/builtins-amdgcn-wmma-w32.cl
  clang/test/CodeGenOpenCL/builtins-amdgcn-wmma-w64.cl

-------------- next part --------------
A non-text attachment was scrubbed...
Name: D128952.441535.patch
Type: text/x-patch
Size: 12385 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20220630/15aff4f7/attachment.bin>