composable_kernel

mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-03 05:01:25 +00:00

Files

Cong Ma 245467f359 [CK TILE] Fix bugs in AQuant preshuffle (#2700 )

* [CK TILE] Fix bugs in AQuant preshuffle

- Make Aquant works with block Mx64x256. `M` could be 16, 32, 64
- Make Aquant works with warp 16x16x32 and 32x32x16.

* [CK TILE] Rename Preshuffle to PreshuffleQuant

The new name, PreshuffleQuant, explicitly states the function's purpose:
to preshuffle the quantization matrix.

* [CK TILE Block Scale] Use GemmConfig to save tile properties

- Remove specialization of GemmQuantTypeConfig
- Pass GemmConfig around which contains tile properties. Stop using hard
  coded tile properties in `gemm_calc_aquant()`

* [CK TILE Block Scale] Rename GemmConfig used in block scale

    - Remove unused GemmConfig
    - Rename GemmConfig used in block scale

---------

Co-authored-by: ThomasNing <thomas.ning@amd.com>

2025-08-27 00:05:54 -07:00

CMakeLists.txt

ck_tile kernel for gemm with groupwise quantized A tensor (#2473 )

2025-07-23 00:10:16 -07:00

test_gemm_aquant_basic_bf8.cpp

ck_tile kernel for gemm with groupwise quantized A tensor (#2473 )

2025-07-23 00:10:16 -07:00

test_gemm_aquant_basic_fp8.cpp

ck_tile kernel for gemm with groupwise quantized A tensor (#2473 )

2025-07-23 00:10:16 -07:00

test_gemm_aquant_basic_i4bf8.cpp

ck_tile kernel for gemm with groupwise quantized A tensor (#2473 )