Skip to content
This repository has been archived by the owner on Aug 30, 2024. It is now read-only.

Commit

Permalink
comment add
Browse files Browse the repository at this point in the history
  • Loading branch information
luoyu-intel committed May 30, 2024
1 parent 4e99482 commit 18250c1
Showing 1 changed file with 2 additions and 2 deletions.
4 changes: 2 additions & 2 deletions neural_speed/core/ne_layers.c
Original file line number Diff line number Diff line change
Expand Up @@ -4397,13 +4397,13 @@ static void ne_compute_forward_add_f32(const struct ne_compute_params* params, c
NE_ASSERT(nb0 == sizeof(float));
NE_ASSERT(nb00 == sizeof(float));
NE_ASSERT(ne00 == ne10);
if ((ne_nrows(src1) == 1 || ne_nrows(src0) == ne_nrows(src1)) && ne10 == ne00) {
/*if ((ne_nrows(src1) == 1 || ne_nrows(src0) == ne_nrows(src1)) && ne10 == ne00) {
if (nb10 == sizeof(float)) {
int step1 = ne11 == 1 ? 0 : ne10;
bestla_add(nr, ne00, (const float*)src0->data, (const float*)src1->data, step1, (float*)dst->data);
return;
}
}
}*/
if (nb10 == sizeof(float)) {
for (int64_t ir = ith; ir < nr; ir += nth) {
// src0 and dst are same shape => same indices
Expand Down

0 comments on commit 18250c1

Please sign in to comment.