PR Import Test

[jayhawk-commits]

Summary

This PR updates the matrix multiply kernel to use an optimized path for small matrix sizes. This improves performance on select benchmarks by up to 15%.

Rationale

Small matrices (e.g., 32x32 or smaller) were previously handled by a general-purpose kernel. Profiling showed excessive thread divergence and shared memory underutilization. This PR adds a specialized kernel to improve efficiency.

Changes

• Added optimized_small_mm_kernel.cpp with tuning for small matrices
• Updated kernel selection logic in kernel_dispatcher.cpp
• Added unit tests in test_small_matrices.cpp

Testing

• Ran all existing unit tests locally with ctest: ✅ Passed
• Verified correctness against CPU reference implementation
• Benchmarked on MI300 with small matrix sizes — observed 10–15% performance improvement
• CI pipelines should pass on merge

Notes

• No changes to public APIs
• Should not affect large matrix workloads