Yi DING 8cb8da53c9 [CK_TILE] FMHA BWD Optimization For GFX950 (#2628) ...

* simplify fmha_bwd_kernel MakeKargs & dq_dram_window

* simply duplicate

* trload pipeline

* Try two-stage

* add prefetch

* optimize & iglp

[ROCm/composable_kernel commit: 4fde1646e5]

2025-08-12 11:11:55 +08:00

..

bfloat16.hpp

[CK TILE] GEMM with packed i4 (#1885)

2025-02-20 09:59:49 +01:00

e8m0.hpp

Add e8m0 scaled convert into CK_TILE (#2617)

2025-08-07 21:37:28 +08:00

float8.hpp

[CK_TILE] Fix UB and corner cases in f32/f16 to/from f8 conversion (#2571)

2025-07-31 09:54:17 +05:00

half.hpp

[CK TILE] GEMM with packed i4 (#1885)

2025-02-20 09:59:49 +01:00

int8.hpp

[CK TILE] GEMM with packed i4 (#1885)

2025-02-20 09:59:49 +01:00

integer.hpp

[CK Tile] Int8 Support on CK Tile GEMM (#2267)

2025-06-25 08:20:35 -07:00

integral_constant.hpp

[CK_TILE] FMHA BWD Optimization For GFX950 (#2628)

2025-08-12 11:11:55 +08:00

math.hpp

upgrade from clang-format-12 to clang-format-18 (#2568)

2025-07-28 11:34:07 -07:00

mxfp_convert.hpp

Add e8m0 scaled convert into CK_TILE (#2617)

2025-08-07 21:37:28 +08:00

null_type.hpp

layernorm2d forward (#1339)

2024-06-24 08:45:52 +08:00

numeric.hpp

[CK_TILE] Add pk_fp4 data type (#2422)

2025-07-14 20:35:06 +08:00

pk_fp4.hpp

Add e8m0 scaled convert into CK_TILE (#2617)

2025-08-07 21:37:28 +08:00

pk_int4.hpp

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

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

type_convert.hpp

Add e8m0 scaled convert into CK_TILE (#2617)

2025-08-07 21:37:28 +08:00

vector_type.hpp

[CK_TILE] Enable printing more structures in CK-Tile (#2443)

2025-08-07 15:45:27 +03:00