floraachy/OpenBLAS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

sbgemm: spr: enlarge P to 256 for performance

Browse Source
This commit is contained in:
Wangyang Guo
2021-09-15 20:29:49 -07:00
parent 10d52646e2
commit 7b2f5cb3b7
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
param.h
View File

@@ -1781,7 +1781,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
// But for AMX, they are not the same, set UNROLL_M = 32 to workaround
#define SBGEMM_DEFAULT_UNROLL_N 16
#define SBGEMM_DEFAULT_UNROLL_M 32
#define SBGEMM_DEFAULT_P 192
#define SBGEMM_DEFAULT_P 256
#define SBGEMM_DEFAULT_Q 1024
#define SBGEMM_DEFAULT_R sbgemm_r