root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-24 10:24:35 +00:00
Code Issues Actions 5 Packages Projects Releases Wiki Activity

[SYCL] Fix DMMV dequantization (#9279)

Browse Source 
Fixed dmmv dequant for ncols== GGML_SYCL_DMMV_X
This commit is contained in:
Ouadie EL FAROUKI 2024-09-04 16:26:33 +01:00 committed by GitHub
parent c8671ae282
commit 5910ea9427
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
1 changed files with 2 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
ggml/src/ggml-sycl/dmmv.cpp
View File

@ -76,8 +76,8 @@ static void dequantize_mul_mat_vec(const void * __restrict__ vx, const dfloat *
}
// sum up partial sums and write back result
#pragma unroll
for (int mask = WARP_SIZE / 2; mask > 0; mask >>= 1) {
const int mask_start = ncols > GGML_SYCL_DMMV_X ? WARP_SIZE >> 1 : WARP_SIZE >> 2;
for (int mask = mask_start; mask > 0; mask >>= 1) {
tmp +=
dpct::permute_sub_group_by_xor(item_ct1.get_sub_group(), tmp, mask);
}