[all-commits] [llvm/llvm-project] 59c311: AMDGPU: Add more tests for f16 fdiv lowering

[Matt Arsenault via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 59c311c5d4a04a6a4f8c4abf140a63af1079e34c
      https://github.com/llvm/llvm-project/commit/59c311c5d4a04a6a4f8c4abf140a63af1079e34c
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2023-07-05 (Wed, 05 Jul 2023)

  Changed paths:
    M llvm/test/CodeGen/AMDGPU/GlobalISel/fdiv.f16.ll
    M llvm/test/CodeGen/AMDGPU/fdiv.f16.ll
    M llvm/test/CodeGen/AMDGPU/fdiv.ll

  Log Message:
  -----------
  AMDGPU: Add more tests for f16 fdiv lowering

Probably should merge the DAG and gisel tests.


  Commit: 9c82dc6a6ba1f3d75b5547680e0a8532684879c9
      https://github.com/llvm/llvm-project/commit/9c82dc6a6ba1f3d75b5547680e0a8532684879c9
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2023-07-05 (Wed, 05 Jul 2023)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUCodeGenPrepare.cpp
    M llvm/lib/Target/AMDGPU/AMDGPULegalizerInfo.cpp
    M llvm/lib/Target/AMDGPU/SIISelLowering.cpp
    M llvm/test/CodeGen/AMDGPU/GlobalISel/fdiv.f16.ll
    M llvm/test/CodeGen/AMDGPU/GlobalISel/legalize-fdiv.mir
    M llvm/test/CodeGen/AMDGPU/fdiv.f16.ll
    M llvm/test/CodeGen/AMDGPU/fneg-combines.f16.ll

  Log Message:
  -----------
  AMDGPU: Always use v_rcp_f16 and v_rsq_f16

These inherited the fast math checks from f32, but the manual suggests
these should be accurate enough for unconditional use. The definition
of correctly rounded is 0.5ulp, but the manual says "0.51ulp". I've
been a bit nervous about changing this as the OpenCL conformance test
does not cover half. Brute force produces identical values compared to
a reference host implementation for all values.


Compare: https://github.com/llvm/llvm-project/compare/cf08c103266b...9c82dc6a6ba1