Mohsen Saffari b161cd94cc Refactore the compute reference batched contraction to manage stride-aware calculation and some code cleanings

2025-10-16 09:24:39 +00:00

..

01_fmha

[CK_TILE] FMHA BWD Add Instance for D48 on GFX950 (#2866)

2025-10-13 15:03:46 +08:00

02_layernorm2d

[CK_TILE] Correct BlockWarps calculation and fix smoke-test in rmsnorm (#2540)

2025-10-13 11:52:37 -07:00

03_gemm

[CK_TILE] Blockwise GEMM pipeline v6 - port of v5 from old CK (#2955)

2025-10-13 13:57:37 +02:00

04_img2col

invoke script directly (#2687)

2025-08-19 00:23:07 -07:00

05_reduce

[CK_TILE] fix example reduces, permute and elementwise on gfx11 & gfx12 (#2810)

2025-09-11 12:41:20 +08:00

06_permute

[CK_TILE] fix example reduces, permute and elementwise on gfx11 & gfx12 (#2810)

2025-09-11 12:41:20 +08:00

09_topk_softmax

Fix latest AITER failure and add more AITER tests in CK CI. (#2782)

2025-09-04 13:44:00 -07:00

10_rmsnorm2d

[CK_TILE] Correct BlockWarps calculation and fix smoke-test in rmsnorm (#2540)

2025-10-13 11:52:37 -07:00

11_add_rmsnorm2d_rdquant

[CK_TILE] Refine Generic2dBlockShape to fix ck_tile example 2,10,11,14 on rdna3 and 4 (#2795)

2025-09-10 08:29:20 +08:00

12_smoothquant

[CK_TILE] Refine Generic2dBlockShape to fix ck_tile example 2,10,11,14 on rdna3 and 4 (#2795)

2025-09-10 08:29:20 +08:00

13_moe_sorting

Felix/opt sorting (#2902)

2025-10-15 09:24:03 +08:00

14_moe_smoothquant

[CK_TILE] Refine Generic2dBlockShape to fix ck_tile example 2,10,11,14 on rdna3 and 4 (#2795)

2025-09-10 08:29:20 +08:00

15_fused_moe

Felix/opt sorting (#2902)

2025-10-15 09:24:03 +08:00

16_batched_gemm

[CK_TILE] Fix example batched_gemm, grouped_gemm, gemm_multi_d, convolution on gfx11 & gfx12 (#2808)

2025-09-11 07:27:33 -07:00

17_grouped_gemm

feat(grouped_gemm_multi_d): add support for bf16

2025-10-14 18:00:43 -04:00

18_flatmm

fix: nil performance results for gemm examples (#2950)

2025-10-06 12:43:23 -07:00

19_gemm_multi_d

[CK_TILE] Fix example batched_gemm, grouped_gemm, gemm_multi_d, convolution on gfx11 & gfx12 (#2808)

2025-09-11 07:27:33 -07:00

20_grouped_convolution

[CK_TILE] Switch into universal gemms for conv bwds (#2981)

2025-10-14 16:09:16 +02:00

21_elementwise

Congma/ck tile/remove cpp 20 code (#2873)

2025-09-25 10:34:28 -07:00

22_gemm_multi_abd

[CK_TILE] Multiple-ABD GEMM example (#2788)

2025-09-19 01:14:11 +02:00

35_batched_transpose

Fix latest AITER failure and add more AITER tests in CK CI. (#2782)

2025-09-04 13:44:00 -07:00

36_pooling

[CK_TILE] Pooling FWD (Lwpck 3683) (#2956)

2025-10-09 16:13:26 +02:00

38_block_scale_gemm

supporting prefill shapes for preshuffle block scale gemm (#2975)

2025-10-10 15:36:24 -07:00

39_copy

fix(copyright header): add header to missing files (#2807)

2025-09-11 12:27:08 -07:00

40_streamk_gemm

Code style clean-up and documentation

2025-09-29 15:46:24 -06:00

41_batched_contraction

Refactore the compute reference batched contraction to manage stride-aware calculation and some code cleanings

2025-10-16 09:24:39 +00:00

CMakeLists.txt

[CK Tile] contraction multi d - kernel & example (#2901)

2025-10-13 12:30:28 +02:00

remod.py

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

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