ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-05-01 03:41:53 +00:00
Code Issues Packages Projects Releases Wiki Activity

iq4_ks_r4: slightly better Zen4

Browse Source
This commit is contained in:
Iwan Kawrakow
2024-12-18 18:45:08 +02:00
parent e232c1b785
commit 00452d4683
1 changed files with 4 additions and 2 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
ggml/src/iqk/iqk_mul_mat.cpp
View File

@@ -3148,12 +3148,14 @@ static void mul_mat_iq4_ks_r4_q8_k(int n, const void * vx, size_t bx, const Data
} }
} }
#else #else
h_shift.vec = _mm256_slli_epi16(_mm256_and_si256(scales, _mm256_set1_epi8(1)), 1); //h_shift.vec = _mm256_slli_epi16(_mm256_and_si256(scales, _mm256_set1_epi8(1)), 1);
h_shift.vec = _mm256_add_epi8(_mm256_set1_epi8(-64), _mm256_slli_epi16(_mm256_and_si256(scales, _mm256_set1_epi8(1)), 1));
#endif #endif
for (int ib = 0; ib < QK_K/32; ++ib) { for (int ib = 0; ib < QK_K/32; ++ib) {
#ifdef HAVE_FANCY_SIMD #ifdef HAVE_FANCY_SIMD
auto iscales = _mm256_cvtepi8_epi32(_mm_set1_epi32(h.val[ib])); auto iscales = _mm256_cvtepi8_epi32(_mm_set1_epi32(h.val[ib]));
auto ishifts = _mm256_add_epi32(_mm256_set1_epi32(-64), _mm256_cvtepi8_epi32(_mm_set1_epi32(h_shift.val[ib]))); //auto ishifts = _mm256_add_epi32(_mm256_set1_epi32(-64), _mm256_cvtepi8_epi32(_mm_set1_epi32(h_shift.val[ib])));
auto ishifts = _mm256_cvtepi8_epi32(_mm_set1_epi32(h_shift.val[ib]));
auto scales = _mm256_cvtepi32_ps(iscales); auto scales = _mm256_cvtepi32_ps(iscales);
auto scales_m = _mm256_mul_ps(scales, _mm256_cvtepi32_ps(ishifts)); auto scales_m = _mm256_mul_ps(scales, _mm256_cvtepi32_ps(ishifts));
for (int iy = 0; iy < nrc_y; ++iy) { for (int iy = 0; iy < nrc_y; ++iy) {