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

ggml : fix row condition for i8mm kernels (#10561)

Browse source 
ggml-ci
This commit is contained in:
Georgi Gerganov 2024-11-28 14:56:37 +02:00 • committed by GitHub
parent eea986f215
commit 76b27d29c2
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 13 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

6
ggml/src/ggml-cpu/ggml-cpu-quants.c
View file

@ -1813,11 +1813,13 @@ void ggml_vec_dot_q4_0_q8_0(int n, float * restrict s, size_t bs, const void * r
sumv0 = vmlaq_f32(sumv0,(vcvtq_f32_s32(vmmlaq_s32((vmmlaq_s32((vmmlaq_s32((vmmlaq_s32(vdupq_n_s32(0), l0, r0)),
l1, r1)), l2, r2)), l3, r3))), scale);
}
float32x4_t sumv1 = vextq_f32(sumv0, sumv0, 2);
float32x4_t sumv1 = vextq_f32 (sumv0, sumv0, 2);
float32x4_t sumv2 = vzip1q_f32(sumv0, sumv1);
vst1_f32(s, vget_low_f32(sumv2));
vst1_f32(s, vget_low_f32 (sumv2));
vst1_f32(s + bs, vget_high_f32(sumv2));
return;
}
#endif