mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-14 10:09:41 +00:00

Files

Haocong WANG 173427d877 [FMHA FWD] gfx950 Accuracy enhancement & bug fix (#2900 )

* disable cast_tile_pk_fp16_fp32 on gfx950

* fix wrong encoding when hdim is not exponentiation of 2

---------

Co-authored-by: asleepzzz <hanwen.chang@amd.com>

[ROCm/composable_kernel commit: 959df2a155]

2025-09-24 00:59:41 +08:00

algorithm

refactor: use snake_case naming in ck_tile/core components (#2766 )

2025-09-03 09:34:11 +02:00

arch

[CK-Tile] Add the API to load SGPR (#2878 )

2025-09-23 01:23:56 -07:00

container

[CK_TILE] Add permuteN optimization to remove lds operation in c_shuffle (#2764 )

2025-09-08 22:02:48 -07:00

numeric

[CK_TILE] Refine pk_fp4's fill, pack, and unpack (#2845 )

2025-09-17 10:54:06 +08:00

tensor

[FMHA FWD] gfx950 Accuracy enhancement & bug fix (#2900 )

2025-09-24 00:59:41 +08:00

utility

Fix UB caused by reinterpret_cast (#2849 )

2025-09-18 07:12:37 -07:00

config.hpp

Fix formatting problem (#2768 )

2025-09-02 14:14:10 +03:00

README.md

introducing ck_tile! (#1216 )

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

README.md

ck_tile/core

ck_tile/core contains every basic functions and structures to create a GPU kernel using ck_tile. User should only include ck_tile/core.hpp this single header to use all the functionality. Everything is under ck_tile namespace. The coding style under this folder should be similar to std (snake_case for structure/function, Camel for template types...)

algorithm/
    coordinate transform and some other reusable algorithm
arch/
    contains some basic device building block like mma, buffer addressing, etc...
container/
    contains basic container data structure, array/sequence/tuple/...
numeric/
    data type, and data type related math
tensor/
    tensor descriptors and tile level API
utility/
    other utility function for both host/device