floraachy
/
OpenBLAS
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Add files via upload

This commit is contained in:
wjc404 2019-11-04 20:10:12 +08:00 committed by GitHub
parent fbacd2605d
commit 430c11e135
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 5 additions and 3 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

8
kernel/x86_64/sgemm_kernel_16x4_skylakex_2.c
View File

@ -54,15 +54,17 @@
#define COMPUTE_m16(ndim) \ #define COMPUTE_m16(ndim) \
INIT_m16n##ndim\ INIT_m16n##ndim\
"movq %%r13,%4; movq %%r14,%1; leaq (%1,%%r12,2),%%r15; addq %%r12,%%r15; movq %2,%5;"\ "movq %%r13,%4; movq %%r14,%1; leaq (%1,%%r12,2),%%r15; addq %%r12,%%r15; movq %2,%5;"\
"cmpq $16,%4; jb "#ndim"016162f;"\ "cmpq $18,%4; jb "#ndim"016162f;"\
#ndim"016161:\n\t"\ #ndim"016161:\n\t"\
KERNEL_k1m16n##ndim\ KERNEL_k1m16n##ndim\
KERNEL_k1m16n##ndim\ KERNEL_k1m16n##ndim\
KERNEL_k1m16n##ndim\
"prefetcht1 (%5); prefetcht1 63(%5); addq %3,%5;"\ "prefetcht1 (%5); prefetcht1 63(%5); addq %3,%5;"\
KERNEL_k1m16n##ndim\ KERNEL_k1m16n##ndim\
KERNEL_k1m16n##ndim\ KERNEL_k1m16n##ndim\
"prefetcht1 (%8); addq $"#ndim",%8;"\ KERNEL_k1m16n##ndim\
"subq $4,%4; cmpq $16,%4; jnb "#ndim"016161b;"\ "prefetcht1 (%8); addq $32,%8;"\
"subq $6,%4; cmpq $18,%4; jnb "#ndim"016161b;"\
"movq %2,%5;"\ "movq %2,%5;"\
#ndim"016162:\n\t"\ #ndim"016162:\n\t"\
"testq %4,%4; jz "#ndim"016163f;"\ "testq %4,%4; jz "#ndim"016163f;"\