native support for icopy_4

90% MKL 1-thread performance.

Author: wjc404

kernel/x86_64/KERNEL.SKYLAKEX

@@ -7,7 +7,7 @@ SGEMMITCOPY    =  sgemm_tcopy_16_skylakex.c
 SGEMMONCOPY    =  sgemm_ncopy_4_skylakex.c
 SGEMMOTCOPY    =  ../generic/gemm_tcopy_4.c
 
-DGEMMKERNEL    =  dgemm_kernel_8x8_skylakex.c
+DGEMMKERNEL    =  dgemm_kernel_4x8_skylakex_2.c
 
 DGEMMONCOPY    =  dgemm_ncopy_8_skylakex.c
 DGEMMOTCOPY    =  dgemm_tcopy_8_skylakex.c