ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-03-02 01:50:01 +00:00
Code Issues Packages Projects Releases Wiki Activity

q3_K: don't scale when all quants in a block are <= 127 when repacking

Browse Source
This commit is contained in:
Iwan Kawrakow
2025-06-15 10:52:26 +03:00
parent e10f7d1f10
commit 9fe58aac13
1 changed files with 27 additions and 15 deletions
Download Patch File Download Diff File Expand all files Collapse all files

42
ggml/src/iqk/iqk_gemm_kquants.cpp
View File

@@ -2383,7 +2383,11 @@ void iqk_convert_q3_k_q8_k_r8(int n, const void * vx, size_t bx, void * vy, int
auto max4 = _mm_cvtepi32_ps(imax4);
max4 = _mm_max_ps(max4, _mm_movehl_ps(max4, max4));
max4 = _mm_max_ss(max4, _mm_movehdup_ps(max4));
float dnew = std::max(1.f, _mm_cvtss_f32(max4) / 127);
bool needs_scaling = true;
float dnew = _mm_cvtss_f32(max4) / 127;
if (dnew < 1.f) {
dnew = 1.f; needs_scaling = false;
}
d *= dnew;
y[i].d[k] = GGML_FP32_TO_FP16(d);
auto scale = _mm256_set1_ps(std::abs(dnew) > 1e-9f ? 1/dnew : 0.f);
@@ -2392,20 +2396,28 @@ void iqk_convert_q3_k_q8_k_r8(int n, const void * vx, size_t bx, void * vy, int
auto q16_h = _mm256_cvtepi8_epi16(_mm256_extracti128_si256(values[ib32], 1));
q16_l = _mm256_mullo_epi16(q16_l, _mm256_set1_epi16(helper.val[2*ib32+0]));
q16_h = _mm256_mullo_epi16(q16_h, _mm256_set1_epi16(helper.val[2*ib32+1]));
auto i0 = _mm256_cvtepi16_epi32(_mm256_castsi256_si128(q16_l));
auto i1 = _mm256_cvtepi16_epi32(_mm256_extracti128_si256(q16_l, 1));
auto i2 = _mm256_cvtepi16_epi32(_mm256_castsi256_si128(q16_h));
auto i3 = _mm256_cvtepi16_epi32(_mm256_extracti128_si256(q16_h, 1));
i0 = _mm256_cvtps_epi32(_mm256_round_ps(_mm256_mul_ps(scale, _mm256_cvtepi32_ps(i0)), _MM_ROUND_NEAREST));
i1 = _mm256_cvtps_epi32(_mm256_round_ps(_mm256_mul_ps(scale, _mm256_cvtepi32_ps(i1)), _MM_ROUND_NEAREST));
i2 = _mm256_cvtps_epi32(_mm256_round_ps(_mm256_mul_ps(scale, _mm256_cvtepi32_ps(i2)), _MM_ROUND_NEAREST));
i3 = _mm256_cvtps_epi32(_mm256_round_ps(_mm256_mul_ps(scale, _mm256_cvtepi32_ps(i3)), _MM_ROUND_NEAREST));
i0 = _mm256_packs_epi32(i0, i1);
i2 = _mm256_packs_epi32(i2, i3);
i0 = _mm256_packs_epi16(i0, i2);
i0 = _mm256_permutevar8x32_epi32(i0, _mm256_setr_epi32(0, 4, 1, 5, 2, 6, 3, 7));
_mm256_storeu_si256((__m256i *)block, i0);
if (needs_scaling) {
auto i0 = _mm256_cvtepi16_epi32(_mm256_castsi256_si128(q16_l));
auto i1 = _mm256_cvtepi16_epi32(_mm256_extracti128_si256(q16_l, 1));
auto i2 = _mm256_cvtepi16_epi32(_mm256_castsi256_si128(q16_h));
auto i3 = _mm256_cvtepi16_epi32(_mm256_extracti128_si256(q16_h, 1));
i0 = _mm256_cvtps_epi32(_mm256_round_ps(_mm256_mul_ps(scale, _mm256_cvtepi32_ps(i0)), _MM_ROUND_NEAREST));
i1 = _mm256_cvtps_epi32(_mm256_round_ps(_mm256_mul_ps(scale, _mm256_cvtepi32_ps(i1)), _MM_ROUND_NEAREST));
i2 = _mm256_cvtps_epi32(_mm256_round_ps(_mm256_mul_ps(scale, _mm256_cvtepi32_ps(i2)), _MM_ROUND_NEAREST));
i3 = _mm256_cvtps_epi32(_mm256_round_ps(_mm256_mul_ps(scale, _mm256_cvtepi32_ps(i3)), _MM_ROUND_NEAREST));
i0 = _mm256_packs_epi32(i0, i1);
i2 = _mm256_packs_epi32(i2, i3);
i0 = _mm256_packs_epi16(i0, i2);
i0 = _mm256_permutevar8x32_epi32(i0, _mm256_setr_epi32(0, 4, 1, 5, 2, 6, 3, 7));
_mm256_storeu_si256((__m256i *)block, i0);
} else {
// 0, 1, 2, 3, 4, 5, 6, 7, 8, 16, 17, 18, 19, 20, 21, 22, 23, 9, 10, 11, 12, 13, 14, 15, 24, 25, 26, 27, 28, 29, 30, 31
auto i0 = _mm256_packs_epi16(q16_l, q16_h);
auto i0_l = _mm256_castsi256_si128(i0);
auto i0_h = _mm256_extracti128_si256(i0, 1);
_mm_storeu_si128((__m128i *)block+0, _mm_unpacklo_epi64(i0_l, i0_h));
_mm_storeu_si128((__m128i *)block+1, _mm_unpackhi_epi64(i0_l, i0_h));
}
auto qs = (uint32_t *)y[i].qs + 64*ib32;
for (int l = 0; l < 8; ++l) {
qs[8*l + k] = block[l];