mirror of
https://github.com/ROCm/composable_kernel.git
synced 2026-07-17 17:19:12 +00:00
327dd55f46
feat: [CK Tile] Adding gfx1250 wrappers for dense and scale builtins (#7852) ## Motivation This PR is part of the [WMMA/MFMA] unification work. It adds all the necessary dense and scale MMA gfx1250 builtins as amdgcn_mma structs. ## Technical Details This change adds 5 new specializations for WMMA gfx1250 scale builtins and 25 new for dense. On top of that we: - dispatch between scale and scale16 builtin call - add scale abstraction for WMMA - small refactor for scale traits (data type to flag will also be needed for dense builtins with packed data types) - fix layout test (we store scale values in every byte) - create gfx1250 target family ## Test Plan All the new wrappers were added to the test suite in `test_amdgcn_mma_layout.inc`. ## Test Result Test pass locally, waiting for the CI. ## Submission Checklist - [x] Look over the contributing guidelines at https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests. Closes https://github.com/ROCm/rocm-libraries/issues/8957
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