amd/blis
1
0
Fork 0
mirror of https://github.com/amd/blis.git synced 2026-05-11 17:50:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

Fixed bug in KNC microkernel where k=0 and beta != 1

Browse Source
This commit is contained in:
Tyler Smith
2014-10-22 16:30:16 -05:00
parent e64dba5633
commit 95cdae65d6
2 changed files with 4 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
kernels/mic/3/bli_dgemm_opt_30x8.c
View File

@@ -264,8 +264,6 @@ void bli_dgemm_opt_30x8(
auxinfo_t* data
)
{
if( k == 0) return;
double * a_next = bli_auxinfo_next_a( data );
double * b_next = bli_auxinfo_next_b( data );
@@ -382,6 +380,8 @@ void bli_dgemm_opt_30x8(
//Used when <= 40 iterations
CONSIDER_UNDER_40:
mov rsi, k
test rsi, rsi
je POSTACCUM
LOOP_UNDER_40:
ONE_ITER_MAIN_LOOP(rcx, rsi)
jne LOOP_UNDER_40

4
kernels/mic/3/bli_sgemm_opt_30x16.c
View File

@@ -264,8 +264,6 @@ void bli_sgemm_opt_30x16(
auxinfo_t* data
)
{
if( k == 0 ) return;
float * a_next = bli_auxinfo_next_a( data );
float * b_next = bli_auxinfo_next_b( data );
@@ -382,6 +380,8 @@ void bli_sgemm_opt_30x16(
//Used when <= 40 iterations
CONSIDER_UNDER_40:
mov rsi, k
test rsi, rsi
je POSTACCUM
LOOP_UNDER_40:
ONE_ITER_MAIN_LOOP(rcx, rsi)
jne LOOP_UNDER_40