[all-commits] [llvm/llvm-project] 3b8f3b: [AMDGPU][Attributor] Update waves-per-eu attribute...

[Shilei Tian via All-commits]

  Branch: refs/heads/users/shiltian/waves-per-eu-range-fix
  Home:   https://github.com/llvm/llvm-project
  Commit: 3b8f3bcdf13073d697a69147d3a1d092b889d5ee
      https://github.com/llvm/llvm-project/commit/3b8f3bcdf13073d697a69147d3a1d092b889d5ee
  Author: Shilei Tian <i at tianshilei.me>
  Date:   2025-08-25 (Mon, 25 Aug 2025)

  Changed paths:
    M llvm/lib/Target/AMDGPU/AMDGPUAttributor.cpp
    M llvm/test/CodeGen/AMDGPU/addrspacecast-constantexpr.ll
    M llvm/test/CodeGen/AMDGPU/amdgpu-attributor-no-agpr.ll
    M llvm/test/CodeGen/AMDGPU/annotate-existing-abi-attributes.ll
    M llvm/test/CodeGen/AMDGPU/annotate-kernel-features-hsa-call.ll
    M llvm/test/CodeGen/AMDGPU/annotate-kernel-features-hsa.ll
    M llvm/test/CodeGen/AMDGPU/annotate-kernel-features.ll
    M llvm/test/CodeGen/AMDGPU/attr-amdgpu-max-num-workgroups-propagate.ll
    M llvm/test/CodeGen/AMDGPU/attributor-flatscratchinit-undefined-behavior.ll
    M llvm/test/CodeGen/AMDGPU/attributor-flatscratchinit.ll
    M llvm/test/CodeGen/AMDGPU/attributor-loop-issue-58639.ll
    M llvm/test/CodeGen/AMDGPU/direct-indirect-call.ll
    M llvm/test/CodeGen/AMDGPU/duplicate-attribute-indirect.ll
    M llvm/test/CodeGen/AMDGPU/implicitarg-offset-attributes.ll
    M llvm/test/CodeGen/AMDGPU/indirect-call-set-from-other-function.ll
    M llvm/test/CodeGen/AMDGPU/inline-attr.ll
    M llvm/test/CodeGen/AMDGPU/issue120256-annotate-constexpr-addrspacecast.ll
    M llvm/test/CodeGen/AMDGPU/pal-simple-indirect-call.ll
    M llvm/test/CodeGen/AMDGPU/propagate-flat-work-group-size.ll
    M llvm/test/CodeGen/AMDGPU/propagate-waves-per-eu.ll
    M llvm/test/CodeGen/AMDGPU/recursive_global_initializer.ll
    M llvm/test/CodeGen/AMDGPU/remove-no-kernel-id-attribute.ll
    M llvm/test/CodeGen/AMDGPU/simple-indirect-call-2.ll
    M llvm/test/CodeGen/AMDGPU/simple-indirect-call.ll
    M llvm/test/CodeGen/AMDGPU/uniform-work-group-attribute-missing.ll
    M llvm/test/CodeGen/AMDGPU/uniform-work-group-multistep.ll
    M llvm/test/CodeGen/AMDGPU/uniform-work-group-nested-function-calls.ll
    M llvm/test/CodeGen/AMDGPU/uniform-work-group-prevent-attribute-propagation.ll
    M llvm/test/CodeGen/AMDGPU/uniform-work-group-propagate-attribute.ll
    M llvm/test/CodeGen/AMDGPU/uniform-work-group-recursion-test.ll
    M llvm/test/CodeGen/AMDGPU/uniform-work-group-test.ll

  Log Message:
  -----------
  [AMDGPU][Attributor] Update waves-per-eu attribute only when it is not the one calculated with flat workgroup size

Right now we update the waves-per-eu attribute if it's not the maximum range. With this approach, most of the added attributes don't really matter because they match the values calculated from the flat workgroup size. That means even if they're missing, the backend will still end up with the same values later in the pipeline. In this PR, we change the logic so the attribute is only updated if it is different from what's calculated from the flat workgroup size.



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