Skip to content

Enable AVX-VNNI 256-bit path for IQ4_NL R4 matmul #1467

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
ikawrakow merged 1 commit into from
Mar 20, 2026
Merged

Enable AVX-VNNI 256-bit path for IQ4_NL R4 matmul #1467

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 13 additions & 0 deletions ggml/src/iqk/iqk_gemm_legacy_quants.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -892,7 +892,9 @@ static void mul_mat_iq4_nl_r4_q8_2(int n, const void * vx, size_t bx, const Data
GGML_ASSERT(nrc_x%4 == 0);
Q8<nrc_y, block_q8_2_x4> q8(info);
auto m4 = _mm256_set1_epi8(0xf);
#ifndef HAVE_VNNI256
auto m1 = _mm256_set1_epi16(1);
#endif
auto values128 = _mm_loadu_si128((const __m128i *)iq4k_values);
auto values = MM256_SET_M128I(values128, values128);
int nb = n / QK4_NL;
Expand All @@ -910,6 +912,16 @@ static void mul_mat_iq4_nl_r4_q8_2(int n, const void * vx, size_t bx, const Data
qs[3] = _mm256_shuffle_epi8(values, _mm256_and_si256(_mm256_srli_epi16(bits2, 4), m4));
return scales;
};
#ifdef HAVE_VNNI256
auto dot = [&qs] (__m256i y) {
auto sumi = _mm256_setzero_si256();
sumi = _mm256_dpbusd_epi32(sumi, _mm256_sign_epi8(qs[0], qs[0]), _mm256_sign_epi8(_mm256_shuffle_epi32(y, 0x00), qs[0]));
sumi = _mm256_dpbusd_epi32(sumi, _mm256_sign_epi8(qs[1], qs[1]), _mm256_sign_epi8(_mm256_shuffle_epi32(y, 0x55), qs[1]));
sumi = _mm256_dpbusd_epi32(sumi, _mm256_sign_epi8(qs[2], qs[2]), _mm256_sign_epi8(_mm256_shuffle_epi32(y, 0xaa), qs[2]));
sumi = _mm256_dpbusd_epi32(sumi, _mm256_sign_epi8(qs[3], qs[3]), _mm256_sign_epi8(_mm256_shuffle_epi32(y, 0xff), qs[3]));
return sumi;
};
#else
auto dot = [&qs, &m1] (__m256i y) {
auto u1 = _mm256_sign_epi8(qs[0], qs[0]);
auto u2 = _mm256_sign_epi8(qs[1], qs[1]);
Expand All @@ -923,6 +935,7 @@ static void mul_mat_iq4_nl_r4_q8_2(int n, const void * vx, size_t bx, const Data
_mm256_madd_epi16(m1, _mm256_maddubs_epi16(u2, _mm256_sign_epi8(_mm256_shuffle_epi32(y, 0xff), qs[3]))));
return _mm256_add_epi32(sumi1, sumi2);
};
#endif
for (int ix = 0; ix < nrc_x; ix += 4) {
const block_iq4_nl_r4 * iq4 = (const block_iq4_nl_r4 *)((const char *)vx + ix*bx);
for (int ib4 = 0; ib4 < nb/4; ++ib4) {
Expand Down