Skip to content

Remove option support-regular-ptr #4076

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 14 commits into from
May 2, 2025

Conversation

etiotto
Copy link
Contributor

@etiotto etiotto commented May 2, 2025

This PR removes the now obsolete option "support-regular-ptr" from the pipeline pass registration and related transformation functions. Key changes include:

  • Updating the pipeline pass registration macro in triton_xpu.cc to drop the extra parameter.
  • Removing the bool supportRegularPtr parameter from function signatures and calls in SoftwarePipeliner.cpp and MatmulLoopPipeline.cpp.
  • Adjusting the declaration in Schedule.h to match the updated function signature.

@etiotto etiotto marked this pull request as draft May 2, 2025 17:40
@etiotto etiotto self-assigned this May 2, 2025
@etiotto etiotto requested a review from Copilot May 2, 2025 17:40
Copy link

@Copilot Copilot AI left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Pull Request Overview

This PR removes the now obsolete option "support-regular-ptr" from the pipeline pass registration and related transformation functions. Key changes include:

  • Updating the pipeline pass registration macro in triton_xpu.cc to drop the extra parameter.
  • Removing the bool supportRegularPtr parameter from function signatures and calls in SoftwarePipeliner.cpp and MatmulLoopPipeline.cpp.
  • Adjusting the declaration in Schedule.h to match the updated function signature.

Reviewed Changes

Copilot reviewed 4 out of 9 changed files in this pull request and generated no comments.

File Description
third_party/intel/triton_xpu.cc Updated the ADD_PASS_WRAPPER call to remove the extra bool parameter for pipeline creation.
third_party/intel/lib/TritonIntelGPUTransforms/Pipeliner/SoftwarePipeliner.cpp Removed the supportRegularPtr parameter from pipelineLoop and updated its call sites.
third_party/intel/lib/TritonIntelGPUTransforms/Pipeliner/MatmulLoopPipeline.cpp Removed the supportRegularPtr parameter from collectOpsToPipeline and preProcessLoopAndGetSchedule calls.
third_party/intel/lib/TritonIntelGPUTransforms/Pipeliner/Schedule.h Removed the supportRegularPtr parameter from the function declaration.
Files not reviewed (5)
  • test/TritonIntelGPU/loop-pipeline.mlir: Language not supported
  • test/TritonIntelGPU/split-barrier.mlir: Language not supported
  • third_party/intel/cmake/3122.diff: Language not supported
  • third_party/intel/cmake/3138.diff: Language not supported
  • third_party/intel/include/Dialect/TritonIntelGPU/Transforms/Passes.td: Language not supported
Comments suppressed due to low confidence (3)

third_party/intel/triton_xpu.cc:91

  • Verify that the updated macro call now requiring only an int parameter (instead of int, bool) for createTritonIntelGPUPipeline is consistent with downstream expectations.
ADD_PASS_WRAPPER_OPT_2("add_pipeline",

third_party/intel/lib/TritonIntelGPUTransforms/Pipeliner/SoftwarePipeliner.cpp:42

  • Ensure that all call sites for pipelineLoop have been updated to omit the removed supportRegularPtr parameter.
pipelineLoop(scf::ForOp forOp, int numStages,

third_party/intel/lib/TritonIntelGPUTransforms/Pipeliner/MatmulLoopPipeline.cpp:120

  • Confirm that the removal of the supportRegularPtr argument from both collectOpsToPipeline and preProcessLoopAndGetSchedule aligns with the overall pipeline transformation logic.
SmallVectorImpl<LoadDotOperand> &loadOps) {

etiotto added 2 commits May 2, 2025 17:52
Signed-off-by: Tiotto, Ettore <[email protected]>
Signed-off-by: Tiotto, Ettore <[email protected]>
@whitneywhtsang whitneywhtsang marked this pull request as ready for review May 2, 2025 18:25
@whitneywhtsang whitneywhtsang enabled auto-merge (squash) May 2, 2025 18:31
@etiotto etiotto linked an issue May 2, 2025 that may be closed by this pull request
@whitneywhtsang whitneywhtsang merged commit aabfb98 into main May 2, 2025
9 checks passed
@whitneywhtsang whitneywhtsang deleted the etiotto.prefetch_tensor_of_ptrs.2 branch May 2, 2025 19:04
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

[Performance] Enable prefetching for tt.load with tensor of pointer
3 participants