[all-commits] [llvm/llvm-project] 447183: [mlir][amdgpu] Remove shared memory optimization p...

[Jakub Kuderski via All-commits]

  Branch: refs/heads/main
  Home:   https://github.com/llvm/llvm-project
  Commit: 44718311dee486f1823876e8af9100afcc50041b
      https://github.com/llvm/llvm-project/commit/44718311dee486f1823876e8af9100afcc50041b
  Author: Jakub Kuderski <jakub at nod-labs.com>
  Date:   2024-04-11 (Thu, 11 Apr 2024)

  Changed paths:
    M mlir/include/mlir/Dialect/AMDGPU/CMakeLists.txt
    M mlir/include/mlir/Dialect/AMDGPU/IR/AMDGPU.td
    R mlir/include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.h
    R mlir/include/mlir/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.td
    R mlir/include/mlir/Dialect/AMDGPU/TransformOps/CMakeLists.txt
    M mlir/include/mlir/Dialect/AMDGPU/Transforms/Passes.h
    M mlir/include/mlir/Dialect/AMDGPU/Transforms/Passes.td
    R mlir/include/mlir/Dialect/AMDGPU/Transforms/Transforms.h
    R mlir/include/mlir/Dialect/AMDGPU/Transforms/Utils.h
    M mlir/include/mlir/InitAllExtensions.h
    M mlir/lib/Dialect/AMDGPU/CMakeLists.txt
    M mlir/lib/Dialect/AMDGPU/IR/AMDGPUDialect.cpp
    R mlir/lib/Dialect/AMDGPU/TransformOps/AMDGPUTransformOps.cpp
    R mlir/lib/Dialect/AMDGPU/TransformOps/CMakeLists.txt
    M mlir/lib/Dialect/AMDGPU/Transforms/CMakeLists.txt
    R mlir/lib/Dialect/AMDGPU/Transforms/OptimizeSharedMemory.cpp
    R mlir/lib/Dialect/AMDGPU/Transforms/Utils.cpp
    R mlir/test/Dialect/AMDGPU/optimize_shmem_reads_writes.mlir
    R mlir/test/Dialect/AMDGPU/transform_optimize_shmem_reads_writes.mlir
    M utils/bazel/llvm-project-overlay/mlir/BUILD.bazel

  Log Message:
  -----------
  [mlir][amdgpu] Remove shared memory optimization pass (#88225)

This implementation has a number of issues and ultimately does not work
on gfx9.
* It does not reduce bank conflicts with wide memory accesses.
* It does not correctly account for when LDS bank conflicts occur on
amdgpu.
* The implementation is too fragile to be used on real-world code. For
example, the code bails out on any `memref.subview` in the root op, even
when the subview is not a user of any of the `memref.alloc` ops.

I do not see how these can be easily fixed, therefore I think it's
better to delete this code.



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