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

faster with madd

Browse source
This commit is contained in:
Eve 2024-11-02 17:12:23 -04:00
parent 629befc729
commit 7de0bdc2db
1 changed files with 2 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

4
ggml/src/ggml-quants.c
View file

@ -4232,9 +4232,9 @@ void ggml_vec_dot_q4_0_q8_0(int n, float * restrict s, size_t bs, const void * r
const __m128i p16_2_0 = mul_add_epi8_sse(q4b_2_0, q8b_2_0);
const __m128i p16_2_1 = mul_add_epi8_sse(q4b_2_1, q8b_2_1);
__m128i p_1 = _mm_add_epi16(p16_1_0, p16_1_1);
p_1 = _mm_add_epi32(_mm_cvtepi16_epi32(_mm_bsrli_si128(p_1, 8)), _mm_cvtepi16_epi32(p_1));
p_1 = _mm_madd_epi16(p_1, _mm_set1_epi16(1));
__m128i p_2 = _mm_add_epi16(p16_2_0, p16_2_1);
p_2 = _mm_add_epi32(_mm_cvtepi16_epi32(_mm_bsrli_si128(p_2, 8)), _mm_cvtepi16_epi32(p_2));
p_2 = _mm_madd_epi16(p_2, _mm_set1_epi16(1));
const __m256 deltas = _mm256_set_m128(_mm_set1_ps(GGML_FP16_TO_FP32(x[ib + 1].d) * GGML_FP16_TO_FP32(y[ib + 1].d)),
_mm_set1_ps(GGML_FP16_TO_FP32(x[ib].d) * GGML_FP16_TO_FP32(y[ib].d)));