-
Notifications
You must be signed in to change notification settings - Fork 70
Cutlass APIS tests #642
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
base: main
Are you sure you want to change the base?
Cutlass APIS tests #642
Conversation
There was a problem hiding this 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 adds comprehensive test coverage for CUTLASS APIs on Intel Xe architectures, specifically testing grouped GEMM operations and various epilogue fusion operations. The tests validate different precision configurations (FP8, BF16, mixed precision), scaling operations, and per-row bias with element-wise activation.
Key changes include:
- New test files for grouped GEMM operations with different precision types
- Tests for linear combination with per-row bias and element-wise activation
- FP8 scaling tests for E4M3 and E5M2 formats
- Enhanced testbed infrastructure with grouped GEMM support
Reviewed changes
Copilot reviewed 7 out of 7 changed files in this pull request and generated 6 comments.
Show a summary per file
| File | Description |
|---|---|
test/unit/gemm/device/gemm_universal_mainloopintelxexmx16group_precision.cpp |
Tests for grouped GEMM with mixed precision (int8/bf16) covering various batch sizes and parallelization patterns |
test/unit/gemm/device/gemm_universal_mainloopintelxexmx16group.cpp |
FP8 (E4M3) grouped GEMM tests with basic, single-group, and multi-group configurations |
test/unit/gemm/device/gemm_universal_mainloopintelxexmx16_group_fp8.cpp |
Duplicate FP8 grouped GEMM tests (identical to previous file) |
test/unit/gemm/device/gemm_universal_lincomb_per_rowbias_eltact.cpp |
Tests for linear combination epilogue with per-row bias and ReLU activation, covering edge cases and various matrix sizes |
test/unit/gemm/device/gemm_universal_fp_scaling.cpp |
FP8 scaling tests for both E4M3 and E5M2 formats with convert-only operations |
test/unit/gemm/device/gemm_testbed_3x.hpp |
New grouped GEMM test infrastructure including runner class, reference computation, and device synchronization |
test/unit/gemm/device/CMakeLists.txt |
Build configuration for unified test executable combining all new test files |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Pull request overview
Copilot reviewed 7 out of 7 changed files in this pull request and generated 5 comments.
Comments suppressed due to low confidence (1)
test/unit/gemm/device/gemm_universal_mainloopintelxexmx16group.cpp:1
- The file contains duplicate copyright headers and code sections (lines 1-203 are duplicated at lines 204-406). This appears to be an accidental file concatenation. Remove the duplicate section starting at line 204.
/***************************************************************************************************
No description provided.