[all-commits] [llvm/llvm-project] 88aced: AMDGPU: Fix computation for getOccupancyWithLocalM...

[Matt Arsenault via All-commits]

  Branch: refs/heads/master
  Home:   https://github.com/llvm/llvm-project
  Commit: 88aced1e454195e038560abb3a0732d020aa4295
      https://github.com/llvm/llvm-project/commit/88aced1e454195e038560abb3a0732d020aa4295
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2020-03-03 (Tue, 03 Mar 2020)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUSubtarget.cpp
    M llvm/test/CodeGen/AMDGPU/occupancy-levels.ll

  Log Message:
  -----------
  AMDGPU: Fix computation for getOccupancyWithLocalMemSize

The computation here didn't really make sense to me, and reported
wildy different results depending on the flat work group size
attribute.

I think this should really report a range derived from the possible
work group size bounds, and only allow an occupancy that is a multiple
of the group size.


  Commit: f9047ede589bce390e3115c75656523e085809a4
      https://github.com/llvm/llvm-project/commit/f9047ede589bce390e3115c75656523e085809a4
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2020-03-03 (Tue, 03 Mar 2020)

  Changed paths:
    M llvm/lib/Transforms/Scalar/LICM.cpp

  Log Message:
  -----------
  LICM: Reorder condition checks

Check the fast math flag before the more expensive loop check.


Compare: https://github.com/llvm/llvm-project/compare/27a3ecee4558...f9047ede589b