[all-commits] [llvm/llvm-project] f78687: AMDGPU: Don't assert on misaligned DS read2/write2...

[Matt Arsenault via All-commits]

  Branch: refs/heads/master
  Home:   https://github.com/llvm/llvm-project
  Commit: f78687df9b790b4f4177a72cbd25b49d14c437b4
      https://github.com/llvm/llvm-project/commit/f78687df9b790b4f4177a72cbd25b49d14c437b4
  Author: Matt Arsenault <Matthew.Arsenault at amd.com>
  Date:   2020-08-26 (Wed, 26 Aug 2020)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUISelDAGToDAG.cpp
    M llvm/test/CodeGen/AMDGPU/ds_read2.ll
    M llvm/test/CodeGen/AMDGPU/ds_write2.ll

  Log Message:
  -----------
  AMDGPU: Don't assert on misaligned DS read2/write2 offsets

This would assert with unaligned DS access enabled. The offset may not
be aligned. Theoretically the pattern predicate should check the
memory alignment, although it is possible to have the memory be
aligned but not the immediate offset.

In this case I would expect it to use ds_{read|write}_b64 with
unaligned access, but am not clear if there's a reason it doesn't.