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

use 128 bit loads (i've tried 256->128 to death and its slower)

Browse source
This commit is contained in:
Eve 2024-11-01 21:06:00 -04:00
parent 61408e7fad
commit ad01d31b60
1 changed files with 11 additions and 2 deletions
Download patch file Download diff file Expand all files Collapse all files

13
ggml/src/ggml-quants.c
View file

@ -5750,6 +5750,7 @@ void ggml_vec_dot_q8_0_q8_0(int n, float * restrict s, size_t bs, const void * r
// Main loop
for (; ib < nb; ++ib) {
#if defined(__AVX2__)
// Compute combined scale for the block
const __m256 d = _mm256_set1_ps(GGML_FP16_TO_FP32(x[ib].d) * GGML_FP16_TO_FP32(y[ib].d));
__m256i qx = _mm256_loadu_si256((const __m256i *)x[ib].qs);
@ -5758,10 +5759,18 @@ void ggml_vec_dot_q8_0_q8_0(int n, float * restrict s, size_t bs, const void * r
const __m256 q = mul_sum_i8_pairs_float(qx, qy);
// Multiply q with scale and accumulate
#if defined(__AVX2__)
acc = _mm256_fmadd_ps( d, q, acc );
#else
acc = _mm256_add_ps( _mm256_mul_ps( d, q ), acc );
const __m128i qx_0 = _mm_loadu_si128((const __m128i *)x[ib].qs);
const __m128i qx_1 = _mm_loadu_si128((const __m128i *)x[ib].qs + 1);
const __m128i qy_0 = _mm_loadu_si128((const __m128i *)y[ib].qs);
const __m128i qy_1 = _mm_loadu_si128((const __m128i *)y[ib].qs + 1);
const __m128i p16_0 = mul_add_epi8_sse(qx_0, qy_0);
const __m128i p16_1 = mul_add_epi8_sse(qx_1, qy_1);
const __m128i p_0 = _mm_madd_epi16(p16_0, _mm_set1_epi16(1));
const __m128i p_1 = _mm_madd_epi16(p16_1, _mm_set1_epi16(1));
acc = _mm256_add_ps(_mm256_mul_ps(_mm256_set1_ps(GGML_FP16_TO_FP32(x[ib].d)*GGML_FP16_TO_FP32(y[ib].d)),
_mm256_cvtepi32_ps(MM256_SET_M128I(p_1, p_0))), acc);
#endif
}