We read every piece of feedback, and take your input very seriously.
To see all available qualifiers, see our documentation.
1 parent 2515e11 commit 1b98000Copy full SHA for 1b98000
1 file changed
kernel/x86_64/zgemm_kernel_4x2_haswell.c
@@ -50,7 +50,7 @@
50
"vmovupd (%0),%%ymm0; vmovupd 32(%0),%%ymm1; prefetcht0 512(%0); addq $64,%0;"\
51
acc_m4n2_con(0,1,4,5,6,7,0,16,%1) acc_m4n2_con(0,1,8,9,10,11,0,16,%1,%%r12,1)
52
#define KERNEL_2_k1m4n4 \
53
- "vpermilpd $5,%%ymm0,%%ymm0; vpermilpd $5,%%ymm1,%%ymm1;"\
+ "vpermilpd $5,-64(%0),%%ymm0; vpermilpd $5,-32(%0),%%ymm1;"\
54
acc_m4n2_con(0,1,4,5,6,7,8,24,%1) acc_m4n2_con(0,1,8,9,10,11,8,24,%1,%%r12,1)
55
#define KERNEL_1_k1m4n6 KERNEL_1_k1m4n4 acc_m4n2_con(0,1,12,13,14,15,0,16,%1,%%r12,2)
56
#define KERNEL_2_k1m4n6 KERNEL_2_k1m4n4 acc_m4n2_con(0,1,12,13,14,15,8,24,%1,%%r12,2)
0 commit comments