root/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-27 03:44:35 +00:00
Code Issues Actions 4 Packages Projects Releases Wiki Activity

ggml : fix vld1q_s8_x4 32-bit compat

Browse Source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2024-01-08 23:45:24 +02:00
parent 1fc2f265ff
commit 27afe29927
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-quants.c
View File

@ -7261,7 +7261,7 @@ void ggml_vec_dot_iq2_xxs_q8_K(const int n, float * restrict s, const void * res
const int8_t * restrict q8 = y[i].qs;
float sumf1 = 0, sumf2 = 0;
for (int ib32 = 0; ib32 < QK_K/32; ib32 += 2) {
q8b = vld1q_s8_x4(q8); q8 += 64;
q8b = ggml_vld1q_s8_x4(q8); q8 += 64;
memcpy(aux32, q2, 4*sizeof(uint32_t)); q2 += 8;
q2u.val[0] = vcombine_s8(vld1_s8((const void *)(iq2xxs_grid + aux8[ 0])), vld1_s8((const void *)(iq2xxs_grid + aux8[ 1])));
q2u.val[1] = vcombine_s8(vld1_s8((const void *)(iq2xxs_grid + aux8[ 2])), vld1_s8((const void *)(iq2xxs_grid + aux8[ 3])));