Skip to content

Commit 0d669e0

Browse files
authored
Update dgemm_kernel_8x8_skylakex.c
1 parent 17cdd9f commit 0d669e0

File tree

1 file changed

+4
-0
lines changed

1 file changed

+4
-0
lines changed

kernel/x86_64/dgemm_kernel_8x8_skylakex.c

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -1,6 +1,8 @@
11
#include "common.h"
22
#include <stdint.h>
33
#include <immintrin.h>
4+
5+
#define ICOPY_4
46
//register usage: zmm3 for alpha, zmm4-zmm7 for temporary use, zmm8-zmm31 for accumulators.
57
/* row-major c_block */
68
#define INNER_KERNEL_k1m1n8 \
@@ -743,6 +745,7 @@ static void KERNEL_EDGE(double *packed_a, double *packed_b, BLASLONG m, BLASLONG
743745
}
744746
}
745747
}
748+
#ifdef ICOPY_4
746749
static void copy_4_to_8(double *src,double *dst,BLASLONG m,BLASLONG k){
747750
BLASLONG m_count,k_count;double *src1,*dst1,*src2;__m256d tmp;
748751
src1 = src; dst1 = dst; src2 = src1 + 4 * k;
@@ -760,6 +763,7 @@ static void copy_4_to_8(double *src,double *dst,BLASLONG m,BLASLONG k){
760763
}
761764
}
762765
}
766+
#endif
763767
int __attribute__ ((noinline)) CNAME(BLASLONG m, BLASLONG n, BLASLONG k, double alpha, double * __restrict__ A, double * __restrict__ B, double * __restrict__ C, BLASLONG ldc){
764768
if(m==0 || n==0 || k==0 || alpha == 0.0) return 0;
765769
BLASLONG ndiv8 = n/8;double ALPHA = alpha;

0 commit comments

Comments
 (0)