[Dev] Remove Redundant Dynamic Shared Memory sync #107
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…ability and maintainability
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This pull request includes significant updates to the benchmarking workflow and enhancements to the benchmarking operators. The most important changes include restructuring the GitHub Actions workflow for benchmarks, updating submodules, and adding new operator configurations and methods for benchmarking.
Workflow Improvements:
.github/workflows/benchmark.yml: Split the benchmark job intobenchmark_baseandbenchmark_comparejobs, added steps to upload and download commit IDs, and updated the Python setup to version 3.9. [1] [2] [3] [4]Submodule Update:
3rdparty/tvm: Updated the submodule commit to the latest version.Benchmarking Enhancements:
benchmark/operators/benchmark_ops_matmul.py: Added new operator configurations and methodsprepare_set_group_4xandprepare_set_group_llmto generate benchmark sets.benchmark/operators/compare_benchmark.py: Added error handling for missing benchmark sets and logging to compare base and head commits. [1] [2]Codebase Simplification:
bitblas/base/utils.py: Removed unusedprofile_tensorsattribute and updated theprofilemethod to include data distribution. [1] [2] [3] [4]bitblas/ops/operator.py: Removedprofile_tensorsand added memory cleanup after profiling latency. [1] [2] [3]These changes collectively enhance the benchmarking process by making it more modular, adding new functionalities, and simplifying the codebase.