[TILE ENGINE] Restructure Tile Engine's benchmarking and profiling

[arai713]

Proposed changes

This PR introduces a restructure for the benchmarking and profiling aspects of CK Tile's Tile Engine, expanding on the groundwork from this previous PR and outlined in this design document. The files have been restructured to have a base structure for benchmarks, profiling and problem descriptions that is inherited from for each variation of GEMM. All common functions across the benchmarking and profiling files have been moved into newly added common utility files. This should greatly reduce the duplicated code present in Tile Engine and make it simpler to add in new operations.

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

[Copilot]

Pull request overview

This PR restructures the Tile Engine's benchmarking and profiling infrastructure to reduce code duplication and improve maintainability by introducing a base class architecture for GEMM operations and consolidating common utilities.

Changes:

• Introduced base GemmProfiler and GemmBenchmark classes with template-based inheritance for different GEMM variations
• Consolidated common functionality (metrics, settings, performance results) into shared utility files
• Refactored Python benchmarking scripts to use inheritance and shared utility modules

Reviewed changes

Copilot reviewed 27 out of 27 changed files in this pull request and generated 7 comments.

Show a summary per file

File	Description
tile_engine/ops/gemm/gemm_universal/gemm_universal_profiler.hpp	New profiler class inheriting from base GemmProfiler template
tile_engine/ops/gemm/gemm_universal/gemm_universal_benchmark_single.cpp	Updated to use new profiler architecture and DataTypeTraits from ck_tile namespace
tile_engine/ops/gemm/gemm_universal/gemm_universal_benchmark.py	Refactored to inherit from shared GemmBenchmark base class
tile_engine/ops/gemm/gemm_universal/gemm_universal_benchmark.hpp	New header containing gemm_host_reference function moved from deleted file
tile_engine/ops/gemm/gemm_profiler.hpp	New base profiler template class with common benchmarking logic
tile_engine/ops/gemm/gemm_benchmark.hpp	New base header with GemmProblem struct and comparison utilities
tile_engine/ops/gemm/gemm_benchmark.py	New base Python class for GEMM benchmarking with shared functionality
tile_engine/ops/common/utils.hpp	New common utilities header with shared structs and helper functions
tile_engine/ops/common/benchmark_utils.py	New Python utilities module with shared benchmarking functions
tile_engine/ops/gemm/gemm_preshuffle/*	Updated to use base classes and remove duplicated code
tile_engine/ops/gemm/gemm_multi_d/*	Updated to use base classes and restructured problem definition

---

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

[amd-meskelin]

I'm noting here that #3434 seems to have broken the test target generation in https://github.com/ROCm/composable_kernel/blob/develop/test/ck_tile/gemm_tile_engine/CMakeLists.txt, as the base gemm_instance_builder.py is no longer an executable script. I can see that #3514 is adding the same recipes for gemm_streamk; I'm wondering what the plan is here, since I don't (yet) have access to the design doc.

[arai713]

I'm noting here that #3434 seems to have broken the test target generation in https://github.com/ROCm/composable_kernel/blob/develop/test/ck_tile/gemm_tile_engine/CMakeLists.txt, as the base gemm_instance_builder.py is no longer an executable script. I can see that #3514 is adding the same recipes for gemm_streamk; I'm wondering what the plan is here, since I don't (yet) have access to the design doc.

I will be fixing that script in #3514 to run the newly named script based on the restructure from #3434. gemm_streamk is not part of the Tile Engine restructure in these PRs. When we add gemm_streamk to the restructured format, we will update the corresponding tests to use the correct script.