[CK_TILE] Introduces a new GEMM API that splits the existing basic GEMM class into multiple specialized classes. (#2520)

* Init commit new API

* apply clang-format

* PreShuffle preapring

* Apply Preshuffle condition to universal_gemm

* Fix: convert size_t to index_t

* Review changes

* Mode 100755 -> 100644

---------

Co-authored-by: Adam Osewski <19374865+aosewski@users.noreply.github.com>
This commit is contained in:
Mateusz Ozga
2025-07-24 20:39:56 +02:00
committed by GitHub
parent 1e84fdaca7
commit b507d889c1
28 changed files with 2094 additions and 1519 deletions

View File

@@ -24,7 +24,7 @@ template <typename GemmConfig,
typename CLayout,
bool Persistent,
typename CDEElementWise>
float gemm(const ck_tile::GemmHostArgs</*NumDTensor = 0*/>& args, const ck_tile::stream_config& s)
float gemm(const ck_tile::GemmHostArgs& args, const ck_tile::stream_config& s)
{
if constexpr(Persistent)