• Lei Wang's avatar
    [Bugfix] Fix Transposed Fragment Layout for amd GEMM_RS matrix core (#346) · 0acb8586
    Lei Wang authored
    * [Refactor] Update GEMM Fragment Layout and Improve Matrix Multiplication Functionality
    
    - Adjusted the layout configuration in `gemm_layouts.cc` to correct the repetition parameters for warp and block layouts, enhancing the efficiency of the GEMM fragment generation.
    - Refactored the `matmul_rs` function in `test_tilelang_test_amd.py` to improve readability by restructuring the function signature and ensuring consistent formatting.
    - Updated the test execution call to run the new `test_gemm_rs_f16f32f32_nt` function, enhancing test coverage for the GEMM functionality.
    
    * lint fix
    
    * bugfix
    0acb8586
test_tilelang_test_amd.py 6.07 KB