Skip to content

Conversation

@AviralGoelAMD
Copy link
Collaborator

@AviralGoelAMD AviralGoelAMD commented Jan 10, 2026

Proposed changes

The universal GEMM's memory pipeline (software pipeline) has a "default" scheduler implementation for scheduling data computation instructions. This implementation is no longer used and are most likely remnants of early version of the example.

This PR removes the dead code.

Checklist

Please put an x into the boxes that apply. You can also fill these out after creating the PR. If you're not sure, please don't hesitate to ask.

  • I have added tests relevant to the introduced functionality, and the unit tests are passing locally
  • I have added the test to REGRESSION_TESTS list defined at the top of CMakeLists.txt in tests/CMakeLists.txt, IF the test takes more than 30 seconds to run.
  • I have added inline documentation which enables the maintainers with understanding the motivation
  • I have removed the stale documentation which is no longer relevant after this pull request
  • (If this change is user-facing) I have added release notes which provide the end users with a brief summary of the improvement from this pull request
  • I have run clang-format on all changed files
  • Any dependent changes have been merged

Discussion

If this is a relatively large or complex change, feel free to start a discussion by explaining why you chose the solution you did and what alternatives you considered

@AviralGoelAMD AviralGoelAMD force-pushed the aviralgoel/memory_pipeline_refactor branch from 7a7f392 to 09aa536 Compare January 10, 2026 21:07
@AviralGoelAMD AviralGoelAMD force-pushed the aviralgoel/memory_pipeline_refactor branch from b9146fd to edc6b82 Compare January 10, 2026 21:51
Copy link
Contributor

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 unused "Default" scheduler implementation from the universal GEMM's memory pipeline software pipeline. The Default scheduler was likely a remnant from early development and is no longer used in the codebase.

Changes:

  • Removed the Default scheduler implementation from block-level GEMM operations
  • Updated the default scheduler to Intrawave in the base GEMM pipeline problem
  • Removed unused template parameter and associated conditional logic from the universal GEMM kernel
  • Fixed a template specialization mismatch in the AQuant pipeline (was incorrectly labeled as Interwave, now correctly Intrawave)
  • Added helpful documentation comments explaining barrier instructions

Reviewed changes

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

File Description
include/ck_tile/ops/gemm/block/block_universal_gemm_as_bs_cr.hpp Removed the BlockGemmImpl<GemmPipelineScheduler::Default> specialization (83 lines) and added clarifying comments for barrier instructions
include/ck_tile/ops/gemm/pipeline/gemm_pipeline_problem.hpp Changed the default scheduler from Default to Intrawave in GemmPipelineProblemBase
include/ck_tile/ops/gemm/kernel/universal_gemm_kernel.hpp Removed unused UseDefaultScheduler template parameter and associated conditional logic from RunGemm
include/ck_tile/ops/gemm_quant/pipeline/gemm_aquant_pipeline_ag_bg_cr_mem.hpp Fixed template specialization from Interwave to Intrawave to match actual usage

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

@ThomasNing ThomasNing merged commit e809861 into develop Jan 12, 2026
27 of 29 checks passed
@ThomasNing ThomasNing deleted the aviralgoel/memory_pipeline_refactor branch January 12, 2026 17:51
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.

3 participants