diff options
author | Cedric Nugteren <web@cedricnugteren.nl> | 2016-07-23 10:20:11 +0200 |
---|---|---|
committer | Cedric Nugteren <web@cedricnugteren.nl> | 2016-07-23 10:20:11 +0200 |
commit | 75fe8235f78520fbbfff7c9c035ecd9f1aa3e6f6 (patch) | |
tree | d49e584504b8192f250d96227aa3be01fbac2c92 /src/routines/level2 | |
parent | b33bec4a59d9d4d0b2e6a3d7e5f1d6e23d4279cb (diff) |
Improved the XgemvFastRot kernel by tiled loading of the input matrix A, enabling better memory performance
Diffstat (limited to 'src/routines/level2')
-rw-r--r-- | src/routines/level2/xgemv.cpp | 2 |
1 files changed, 1 insertions, 1 deletions
diff --git a/src/routines/level2/xgemv.cpp b/src/routines/level2/xgemv.cpp index 2842ef07..e4d407c8 100644 --- a/src/routines/level2/xgemv.cpp +++ b/src/routines/level2/xgemv.cpp @@ -122,7 +122,7 @@ StatusCode Xgemv<T>::MatVec(const Layout layout, const Transpose a_transpose, } if (fast_kernel_rot) { kernel_name = "XgemvFastRot"; - global_size = m_real / db_["WPT3"]; + global_size = m_real; local_size = db_["WGS3"]; } |