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

+7% tg +5% pp compared to master

Browse source
This commit is contained in:
Eve 2024-11-02 11:04:58 -04:00
parent e069375f09
commit fffe7e6204
1 changed files with 11 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

21
ggml/src/ggml-quants.c
View file

@ -4213,9 +4213,7 @@ void ggml_vec_dot_q4_0_q8_0(int n, float * restrict s, size_t bs, const void * r
sumf = hsum_float_8(acc);
#elif defined(__AVX__)
__m256 accum1 = _mm256_setzero_ps();
__m256 accum2 = _mm256_setzero_ps();
__m256 accum = _mm256_setzero_ps();
for (; ib + 1 < nb; ib += 2) {
const __m128i q4bits_1 = _mm_loadu_si128((const __m128i *)x[ib + 0].qs);
const __m128i q4bits_2 = _mm_loadu_si128((const __m128i *)x[ib + 1].qs);
@ -4228,19 +4226,22 @@ void ggml_vec_dot_q4_0_q8_0(int n, float * restrict s, size_t bs, const void * r
const __m128i q4b_1_1 = _mm_sub_epi8(_mm_and_si128(_mm_set1_epi8(15), _mm_srli_epi16(q4bits_1, 4)), _mm_set1_epi8(8));
const __m128i q4b_2_0 = _mm_sub_epi8(_mm_and_si128(_mm_set1_epi8(15), q4bits_2), _mm_set1_epi8(8));
const __m128i q4b_2_1 = _mm_sub_epi8(_mm_and_si128(_mm_set1_epi8(15), _mm_srli_epi16(q4bits_2, 4)), _mm_set1_epi8(8));
const __m128i p16_1_0 = mul_add_epi8_sse(q4b_1_0, q8b_1_0);
const __m128i p16_1_1 = mul_add_epi8_sse(q4b_1_1, q8b_1_1);
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);
const __m128i p_1 = _mm_add_epi16(p16_1_0, p16_1_1);
const __m128i p_2 = _mm_add_epi16(p16_2_0, p16_2_1);
accum1 = _mm256_add_ps(_mm256_mul_ps(_mm256_set1_ps(GGML_FP16_TO_FP32(y[ib + 0].d)*GGML_FP16_TO_FP32(x[ib + 0].d)),
_mm256_cvtepi32_ps(MM256_SET_M128I(_mm_cvtepi16_epi32(_mm_bsrli_si128(p_1, 8)), _mm_cvtepi16_epi32(p_1)))), accum1);
accum2 = _mm256_add_ps(_mm256_mul_ps(_mm256_set1_ps(GGML_FP16_TO_FP32(y[ib + 1].d)*GGML_FP16_TO_FP32(x[ib + 1].d)),
_mm256_cvtepi32_ps(MM256_SET_M128I(_mm_cvtepi16_epi32(_mm_bsrli_si128(p_2, 8)), _mm_cvtepi16_epi32(p_2)))), accum2);
__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));
__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));
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)));
accum = _mm256_add_ps(_mm256_mul_ps(deltas, _mm256_cvtepi32_ps(MM256_SET_M128I(p_2, p_1))), accum);
}
sumf = hsum_float_8(_mm256_add_ps(accum1, accum2));
sumf = hsum_float_8(accum);
#elif defined(__SSSE3__)
// set constants
const __m128i lowMask = _mm_set1_epi8(0xF);