LLM-Inference/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-08-12 22:29:39 +04:00
Code Issues Packages Projects Releases Wiki Activity

Fix IQ3_XXS CPU GEMM (#2224)

Browse Source
This commit is contained in:
Kawrakow
2026-08-01 08:51:59 +03:00
committed by GitHub
parent 3f53a05902
commit a8ae4fb36f
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
ggml/src/iqk/iqk_gemm_iquants.cpp
+1 -1
View File
@@ -550,7 +550,7 @@ struct DequantizerIQ3XXS final : public BaseDequantizer<block_iq3_xxs> {
make4_unsigned(qs, bits.values);
sign_2_values(signs+0, bits.values+0);
sign_2_values(signs+4, bits.values+2);
for (int k = 0; k < 4; ++k) bits.values[k] = _mm256_add_epi32(bits.values[k], min_value);
for (int k = 0; k < 4; ++k) bits.values[k] = _mm256_add_epi8(bits.values[k], min_value);
}
inline void prepare(int i, int j, const Q8<1>& q8, __m256i * q8_quants) {
for (int k = 0; k < 4; ++k) q8_quants[k] = q8.load_quants(0, i, 4*j+k);