composable_kernel

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

Files

Emily Martins 352dee5225 Fix CK Tile Stream-K BF16 Validation Errors (#3039 )

Prior to this change, the number of accumulations passed into
calculate_rtol_atol was 1. That said, in most cases, this is not correct
when there are multiple workgroups contributing to the same macro tile
in C.

This change ensures uses the function estimate_num_wgs_per_tile, which
was extracted into a common file and generalized, to estimate the number
of workgroups per macro tile. This estimate is passed into
calculate_rtol_atol to ensure we get a better relative and absolute
tolerance.

2025-10-17 09:33:38 -07:00

generic_2d_block_shape.hpp

[CK_TILE][REGRESSION] Correct blockSize in Generic2dBlockShape (c254f… (#2837 )

2025-09-16 08:47:55 -07:00

load_interleaved_pk_type.hpp

Update include path to break the remod's cyclic dep issue (#2978 )

2025-10-13 13:24:47 +02:00

README.md

introducing ck_tile! (#1216 )

2024-04-15 19:27:12 -05:00

streamk_common.hpp

Fix CK Tile Stream-K BF16 Validation Errors (#3039 )

2025-10-17 09:33:38 -07:00

tensor_layout.hpp

introducing ck_tile! (#1216 )

2024-04-15 19:27:12 -05:00

utils.hpp

chore: unset executable permission (#2303 )

2025-06-10 09:13:59 -07:00

README.md

common

this folder is designed not to be included directly by use, e.g. if use include ck_tile/ops/fmha.hpp, then everything under common should also be included.

to achieve this we will duplicate the header include path under common to other module under ops/* inside remod.py. for internal developer, you can also include ck_tile/ops/common.hpp for convenience. (and so does external users...)