undo slow dgemm/skylake microoptimization

the compare is more costly than the work
This commit is contained in:
Arjan van de Ven 2018-10-06 14:00:37 +00:00
parent d74dc39b0f
commit 6d43c51ccf
1 changed files with 3 additions and 5 deletions

View File

@ -647,11 +647,9 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
#define SAVE2x2(ALPHA) \
if (ALPHA != 1.0) { \
xmm0 = _mm_set1_pd(ALPHA); \
xmm4 *= xmm0; \
xmm6 *= xmm0; \
} \
xmm0 = _mm_set1_pd(ALPHA); \
xmm4 *= xmm0; \
xmm6 *= xmm0; \
\
xmm4 += _mm_loadu_pd(CO1); \
xmm6 += _mm_loadu_pd(CO1 + ldc); \