[all-commits] [llvm/llvm-project] 907e73: AMDGPU: Don't use amdgpu_ps for some atomic tests

[Matt Arsenault via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 907e73936fcf76b856348aa27bf0c85151d32563
      https://github.com/llvm/llvm-project/commit/907e73936fcf76b856348aa27bf0c85151d32563
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2024-06-16 (Sun, 16 Jun 2024)

  Changed paths:
    M llvm/test/CodeGen/AMDGPU/llvm.amdgcn.raw.ptr.buffer.atomic.fadd.ll

  Log Message:
  -----------
  AMDGPU: Don't use amdgpu_ps for some atomic tests

The default calling convention requires fewer shuffling moves and shrinks
the test output.


  Commit: 55696dba2353bcc5a99165134b1d7d15abc51577
      https://github.com/llvm/llvm-project/commit/55696dba2353bcc5a99165134b1d7d15abc51577
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2024-06-16 (Sun, 16 Jun 2024)

  Changed paths:
    R llvm/test/CodeGen/AMDGPU/llvm.amdgcn.raw.ptr.buffer.atomic.fadd.ll
    A llvm/test/CodeGen/AMDGPU/llvm.amdgcn.raw.ptr.buffer.atomic.fadd_nortn.ll
    A llvm/test/CodeGen/AMDGPU/llvm.amdgcn.raw.ptr.buffer.atomic.fadd_rtn.ll

  Log Message:
  -----------
  AMDGPU: Expand raw ptr buffer atomic fadd tests

These were only checking the no return versions on gfx908


  Commit: 570f36229aab50d2bc433bfb120869b33c8db076
      https://github.com/llvm/llvm-project/commit/570f36229aab50d2bc433bfb120869b33c8db076
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2024-06-16 (Sun, 16 Jun 2024)

  Changed paths:
    M llvm/lib/Target/AMDGPU/SIISelLowering.cpp
    A llvm/test/CodeGen/AMDGPU/llvm.amdgcn.raw.ptr.buffer.atomic.fadd.v2bf16.ll

  Log Message:
  -----------
  AMDGPU: Fix legalization for llvm.amdgcn.raw.buffer.atomic.fadd.v2bf16

We somehow ended up with llvm.amdgcn.raw.ptr.buffer.atomic.fadd, and
llvm.amdgcn.raw.buffer.ptr.atomic.fadd.v2bf16 intrinsic definitions, despite
the second being the canonical mangling for the first intrinsic with v2bf16.
This requires us to handle it as a separate case. Surprisingly, cases
generating the regular 1st intrinsic with bfloat end up getting ID'd as
the second. The selection is still broken for the gfx9 cases.


Compare: https://github.com/llvm/llvm-project/compare/1644a31ae931...570f36229aab

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