mirror of
https://github.com/amd/blis.git
synced 2026-05-11 09:39:59 +00:00
Fixed bug in KNC microkernel where k=0 and beta != 1
This commit is contained in:
@@ -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
|
||||
|
||||
@@ -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
|
||||
|
||||
Reference in New Issue
Block a user