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

Restore direct GEMM to previous version

pull/180/head
mcian 2017-07-31 14:06:23 +02:00
parent dc49906542
commit 99afdcd908
1 changed files with 6 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
src/routines/level3/xgemm.cpp
View File

@ -280,11 +280,13 @@ void Xgemm<T>::GemmDirect(const size_t m, const size_t n, const size_t k,
kernel.SetArgument(16, static_cast<int>(b_conjugate));
// Computes the global and local thread sizes
//const auto m_ceiled = Ceil(m, db_["WGD"]);
//const auto n_ceiled = Ceil(n, db_["WGD"]);
const auto m_ceiled = Ceil(m, db_["WGD"]);
const auto n_ceiled = Ceil(n, db_["WGD"]);
const auto global = std::vector<size_t>{
CeilDiv(m * db_["MDIMCD"], db_["WGD"]),
CeilDiv(n * db_["NDIMCD"], db_["WGD"])
// CeilDiv(m * db_["MDIMCD"], db_["WGD"]),
// CeilDiv(n * db_["NDIMCD"], db_["WGD"])
(m_ceiled * db_["MDIMCD"]) / db_["WGD"],
(n_ceiled * db_["NDIMCD"]) / db_["WGD"]
};
const auto local = std::vector<size_t>{db_["MDIMCD"], db_["NDIMCD"]};