[BE] Add Complex Kernel Test #38
Closed
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
Summary: - Updated `parse_single_trace_content` to only process compilation events, improving efficiency by skipping non-compilation events. - Enhanced the handling of IR file mappings and bidirectional mappings between different IR types. - Improved the organization of output lines for final writing, ensuring accurate grouping of compilation and launch events by kernel hash. - Added new utility functions for flattening and unflattening dictionaries, and for generating launch diffs, enhancing the overall functionality of the module. These changes aim to streamline the parsing process and improve the clarity and maintainability of the codebase.
Summary: - Introduced `ArgumentViewer` to display argument distributions and differences in a structured format. - Added `DiffViewer` to handle and present differing fields, integrating `ArgumentViewer` for extracted arguments. - Created `StackDiffViewer` to visualize stack traces with collapsible sections for better user experience. - Updated `KernelOverview` to incorporate the new components for enhanced launch analysis and metadata display. - Enhanced `dataLoader` to support launch_diff events, improving kernel data processing. These changes aim to improve the visualization and analysis of kernel launch data, enhancing the overall functionality of the application.
Summary: - Added comments for clarity, including a note for the dropdown arrow icon in both `ArgumentViewer` and `StackDiffViewer`. - Updated the `renderSimpleDiff` function in `DiffViewer` to use a more descriptive parameter name. These changes improve code readability and maintainability across the components.
Summary: - Introduced a new test file `test_complex_kernels.py` to validate the functionality of two distinct Triton kernels: an autotuned matrix multiplication kernel and a fused element-wise operation kernel. - Implemented a comprehensive test plan that includes multiple launches with varied parameters for both kernels, ensuring thorough testing of their performance and correctness. - Enhanced logging capabilities during the tests to facilitate debugging and analysis of kernel launches. These changes aim to improve the testing framework for Triton kernels, ensuring robust validation of their functionality and performance.
meta-cla
bot
added
the
CLA Signed
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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 introduces a new, comprehensive test case (
tests/test_complex_kernels.py) to validate the trace parsing logic, especially for the recently addedlaunch_difffeature.Motivation
The existing tests were insufficient to fully validate the parser's ability to handle more realistic scenarios involving multiple, distinct kernels and varied launch parameters within a single run. This test was created to provide a robust, end-to-end validation for the kernel grouping and launch diffing functionalities.
What the Test Does
Defines Two Kernels:
matmul_kernel: An autotuned matrix multiplication kernel designed to test the parser's handling of autotuner-generated configurations.fused_op_kernel: A simpler element-wise kernel used to test basic launch parameter variations.Multiple, Varied Launches:
matmulkernel is launched three times with different input tensor shapes.fused_opkernel is launched four times with different scalar arguments (scale_factor) and compile-time constants (activation).End-to-End Parsing:
tritonparse.utils.unified_parse.How it Improves Testing
launch_diffgeneration, ensuring that parameter variations are correctly identified and summarized.