[all-commits] [llvm/llvm-project] 20c764: AMDGPU: Give v_cvt_sr_pk_bf16_f32 its own subtarge...

[Matt Arsenault via All-commits]

  Branch: refs/heads/users/arsenm/amdgpu/cvt-sr-pk-bf16-distinct-feature
  Home:   https://github.com/llvm/llvm-project
  Commit: 20c764a9994aa62e00e1ee87591fb8b9bccd8a26
      https://github.com/llvm/llvm-project/commit/20c764a9994aa62e00e1ee87591fb8b9bccd8a26
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2026-07-30 (Thu, 30 Jul 2026)

  Changed paths:
    M clang/include/clang/Basic/BuiltinsAMDGPU.td
    A clang/test/SemaOpenCL/builtins-amdgcn-cvt-sr-pk-bf16-f32-err.cl
    M llvm/lib/Target/AMDGPU/AMDGPU.td
    M llvm/lib/Target/AMDGPU/VOP3Instructions.td
    M llvm/lib/TargetParser/AMDGPUTargetParser.cpp
    M llvm/test/CodeGen/AMDGPU/llvm.amdgcn.cvt.sr.pk.bf16.ll
    M llvm/test/MC/AMDGPU/gfx950-unsupported.s

  Log Message:
  -----------
  AMDGPU: Give v_cvt_sr_pk_bf16_f32 its own subtarget feature

v_cvt_sr_pk_bf16_f32 was gated on bf16-cvt-insts, but that feature is
also present on gfx950 where the (non-sr) v_cvt_pk_bf16_f32 was first
added. The stochastic-rounding v_cvt_sr_pk_bf16_f32 was only added
for gfx1250 and has no gfx950 encoding, so it would mis-select and
later hit the "Invalid opcode" assert. Introduce cvt-sr-pk-bf16-f32-inst,
currently added to gfx13 and 125*

Co-authored-by: Claude (Claude-Opus-4.8)



To unsubscribe from these emails, change your notification settings at https://github.com/llvm/llvm-project/settings/notifications