haraldwolff
/
CLBlast
mirror of https://github.com/CNugteren/CLBlast.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Made the batched routines also chose direct/indirect kernel like the main GEMM routine

pull/244/head
Cedric Nugteren 2018-01-18 19:41:02 +01:00
parent 0e5eaa6eb9
commit bc54411d19
2 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
src/routines/levelx/xgemmbatched.cpp
View File

@ -96,7 +96,7 @@ void XgemmBatched<T>::DoGemmBatched(const Layout layout, const Transpose a_trans
}
// Selects which version of the batched GEMM to run
const auto do_gemm_direct = true;
const auto do_gemm_direct = Xgemm<T>::UseDirectKernel(m, n, k, db_["XGEMM_MIN_INDIRECT_SIZE"]);
if (do_gemm_direct) { // single generic kernel
BatchedGemmDirect(m, n, k, alphas_device,
a_buffer, a_offsets_int, a_ld, b_buffer, b_offsets_int, b_ld,

2
src/routines/levelx/xgemmstridedbatched.cpp
View File

@ -76,7 +76,7 @@ void XgemmStridedBatched<T>::DoGemmStridedBatched(const Layout layout, const Tra
}
// Selects which version of the batched GEMM to run
const auto do_gemm_direct = true;
const auto do_gemm_direct = Xgemm<T>::UseDirectKernel(m, n, k, db_["XGEMM_MIN_INDIRECT_SIZE"]);;
if (do_gemm_direct) { // single generic kernel
BatchedGemmDirect(m, n, k, alpha,
a_buffer, a_offset, a_ld, a_stride,