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

Improve AVX2 for vec_dot_q4_3_q8_0 (#1138)

Browse source
This commit is contained in:
Stephan Walter 2023-04-23 11:01:03 +00:00 committed by GitHub
parent c6524f46eb
commit 53c8434398
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 5 additions and 9 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

14
ggml.c
View file

@ -2947,6 +2947,7 @@ static void ggml_vec_dot_q4_3_q8_0(const int n, float * restrict s, const void *
#elif defined(__AVX2__)
// Initialize accumulator with zeros
__m256 acc = _mm256_setzero_ps();
float summs = 0.0f;
// Main loop
for (int i = 0; i < nb; i++) {
@ -2954,9 +2955,8 @@ static void ggml_vec_dot_q4_3_q8_0(const int n, float * restrict s, const void *
const __m128 d1 = _mm_set1_ps(GGML_FP16_TO_FP32(x[2*i + 1].d));
const __m256 dx = _mm256_set_m128(d1, d0);
const __m128 m0 = _mm_set1_ps(GGML_FP16_TO_FP32(x[2*i + 0].m));
const __m128 m1 = _mm_set1_ps(GGML_FP16_TO_FP32(x[2*i + 1].m));
const __m256 mx = _mm256_set_m128(m1, m0);
summs += GGML_FP16_TO_FP32(x[2*i + 0].m) * y[i].s0
+ GGML_FP16_TO_FP32(x[2*i + 1].m) * y[i].s1;
const __m128i bx0 = bytes_from_nibbles_16(x[2*i + 0].qs);
const __m128i bx1 = bytes_from_nibbles_16(x[2*i + 1].qs);
@ -2965,16 +2965,12 @@ static void ggml_vec_dot_q4_3_q8_0(const int n, float * restrict s, const void *
const __m256 dy = _mm256_broadcast_ss(&y[i].d);
const __m256i by = _mm256_loadu_si256((const __m256i *)y[i].qs);
const __m256i syi = _mm256_maddubs_epi16(_mm256_set1_epi8(1), by);
const __m256 syf = sum_i16_pairs_float(syi);
const __m256 q = mul_sum_i8_pairs_float(bx, by);
const __m256 sxy = _mm256_fmadd_ps(q, dx, _mm256_mul_ps(mx, syf));
acc = _mm256_fmadd_ps(sxy, dy, acc);
acc = _mm256_fmadd_ps(q, _mm256_mul_ps(dx, dy), acc);
}
*s = hsum_float_8(acc);
*s = hsum_float_8(acc) + summs;
#else
// scalar
float sumf = 0.0;