jdelony/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-06-28 04:30:15 -05:00
Code Issues Packages Projects Releases Wiki Activity

Fix Zen4

Browse Source
This commit is contained in:
Iwan Kawrakow 2024-09-11 19:44:34 +03:00
parent 70f3a8882c
commit 95fe6923ad
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/iqk/iqk_mul_mat.cpp
View File

@ -6769,8 +6769,8 @@ struct HelperQ41 final : public BaseHelper<step> {
auto vm = F16::set1(GGML_FP16_TO_FP32(dl->m));
auto q = _mm_loadu_si128((const __m128i *)dl->qs);
#ifdef HAVE_FANCY_SIMD
auto ql = _mm_add_epi8(_mm_and_si128(q, mask), m8);
auto qh = _mm_add_epi8(_mm_and_si128(_mm_srli_epi16(q, 4), mask), m8);
auto ql = _mm_and_si128(q, mask);
auto qh = _mm_and_si128(_mm_srli_epi16(q, 4), mask);
v1 = _mm512_fmadd_ps(vd, _mm512_cvtepi32_ps(_mm512_cvtepi8_epi32(ql)), vm);
v2 = _mm512_fmadd_ps(vd, _mm512_cvtepi32_ps(_mm512_cvtepi8_epi32(qh)), vm);
#else