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 2019-07-19 23:47:58 +08:00 committed by GitHub
parent 9b04baeaee
commit 9c89757562
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
1 changed files with 28 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

29
kernel/x86_64/dgemm_kernel_4x8_haswell.S
View File

@ -1865,6 +1865,14 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
SAVE4x12
salq $3, K
prefetcht2 32(B)
prefetcht2 32(B, K, 8)
prefetcht2 96(B)
prefetcht2 96(B, K, 8)
addq $128, B
sarq $3, K
decq I # i --
jne .L12_11
ALIGN_4
@ -1872,6 +1880,11 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
/**************************************************************************
* Rest of M
***************************************************************************/
movq M, I
sarq $2, I
salq $7, I
subq I, B
.L12_20:
// Test rest of M
@ -2102,7 +2115,7 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
jmp .L13_16
PREFETCHT0_C
.L13_13:
test $1, %rax
@ -2147,6 +2160,14 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
SAVE4x12
salq $3, K
prefetcht2 (B)
prefetcht2 (B, K, 8)
prefetcht2 64(B)
prefetcht2 64(B, K, 8)
addq $128, B
sarq $3, K
decq I # i --
jne .L13_11
ALIGN_4
@ -2154,6 +2175,12 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
/**************************************************************************
* Rest of M
***************************************************************************/
movq M, I
sarq $2, I
salq $7, I
subq I, B
.L13_20:
// Test rest of M