Skip to content

Commit 1c67567

Browse files
authored
improve skylakex paralleled sgemm performance
1 parent 3a100b2 commit 1c67567

File tree

1 file changed

+2
-7
lines changed

1 file changed

+2
-7
lines changed

param.h

Lines changed: 2 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -1690,18 +1690,13 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
16901690

16911691
#else
16921692

1693-
#define SGEMM_DEFAULT_P 768
1693+
#define SGEMM_DEFAULT_P 640
16941694
#define DGEMM_DEFAULT_P 384
16951695
#define CGEMM_DEFAULT_P 384
16961696
#define ZGEMM_DEFAULT_P 256
16971697

1698-
#ifdef WINDOWS_ABI
1699-
#define SGEMM_DEFAULT_Q 192
1700-
#define DGEMM_DEFAULT_Q 168
1701-
#else
1702-
#define SGEMM_DEFAULT_Q 192
1698+
#define SGEMM_DEFAULT_Q 320
17031699
#define DGEMM_DEFAULT_Q 168
1704-
#endif
17051700
#define CGEMM_DEFAULT_Q 192
17061701
#define ZGEMM_DEFAULT_Q 128
17071702

0 commit comments

Comments
 (0)