root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-28 12:24:35 +00:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

metal : fix NSG1 > 1

Browse Source
This commit is contained in:
Georgi Gerganov 2024-02-08 16:39:38 +02:00
parent 845876d012
commit e8b00e2941
No known key found for this signature in database
GPG Key ID: 449E073F9DC10735
1 changed files with 1 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
ggml-metal.metal
View File

@ -4980,7 +4980,7 @@ void kernel_mul_mm2_impl(
for (int is = 0; is < NSG1; is++) { for (int is = 0; is < NSG1; is++) {
const int i1 = i11 + is*8; const int i1 = i11 + is*8;
const int nstore = MIN(8*NSG1, ne1 - i1) * (8*NSG0); const int nstore = MIN(8, ne1 - i1) * (8*NSG0);
for (int i = tiisg; i < nstore; i += NW) { for (int i = tiisg; i < nstore; i += NW) {
const int ic = i%(8*NSG0); const int ic = i%(8*NSG0);