vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

iq1s_blocks16: fixed Neon

Browse source
This commit is contained in:
Iwan Kawrakow 2024-03-08 16:20:50 +01:00
parent fbb001e698
commit 15acc7923b
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml-quants.c
View file

@ -9598,8 +9598,8 @@ void ggml_vec_dot_iq1_s_q8_K (int n, float * restrict s, size_t bs, const void
q8b = ggml_vld1q_s8_x4(q8); q8 += 64; q8b = ggml_vld1q_s8_x4(q8); q8 += 64;
const int32x4_t p1 = ggml_vdotq_s32(q1b.val[0], q8b.val[0], ggml_vdotq_s32(q1b.val[1], q8b.val[1], vdupq_n_s32(0))); const int32x4_t p1 = ggml_vdotq_s32(ggml_vdotq_s32(vdupq_n_s32(0), q1b.val[0], q8b.val[0]), q1b.val[1], q8b.val[1]);
const int32x4_t p2 = ggml_vdotq_s32(q1b.val[2], q8b.val[2], ggml_vdotq_s32(q1b.val[3], q8b.val[3], vdupq_n_s32(0))); const int32x4_t p2 = ggml_vdotq_s32(ggml_vdotq_s32(vdupq_n_s32(0), q1b.val[2], q8b.val[2]), q1b.val[3], q8b.val[3]);
sumi1 += vaddvq_s32(p1) * (2*(qh[ib+0] >> 12) + 1); sumi1 += vaddvq_s32(p1) * (2*(qh[ib+0] >> 12) + 1);
sumi2 += vaddvq_s32(p2) * (2*(qh[ib+1] >> 12) + 1); sumi2 += vaddvq_s32(p2) * (2*(qh[ib+1] >> 12) + 1);