ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-03-13 23:40:09 +00:00
Code Issues Packages Projects Releases Wiki Activity

iq3_ks: AVX2

Browse Source
This commit is contained in:
Iwan Kawrakow
2024-10-10 10:48:42 +03:00
parent a1a3ce6886
commit e734e888e1
1 changed files with 42 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

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

@@ -1664,6 +1664,48 @@ struct DequantizerIQ3K final : public BaseDequantizer<block_iq3_k> {
constexpr static uint8_t k_shuff[16] = {0, 4, 8, 12, 1, 5, 9, 13, 2, 6, 10, 14, 3, 7, 11, 15};
};
struct DequantizerIQ3KS final : public BaseDequantizer<block_iq3_ks, true> {
DequantizerIQ3KS(const void * vx, size_t bx) : BaseDequantizer(vx, bx), values(load_values()) {}
template <typename Q8>
inline __m256i new_block(int i, const Q8& q8, __m256 * accm) {
hbits = _mm256_loadu_si256((const __m256i *)x[i].qh);
auto scales128 = _mm_cvtepu8_epi16(_mm_loadl_epi64((const __m128i *)x[i].scales));
auto shifts = _mm_and_si128(_mm_cmpeq_epi16(_mm_and_si128(scales128, m1), m1), m4);
scales128 = _mm_add_epi16(_mm_and_si128(scales128, mask), m127);
auto scales_s = _mm_mullo_epi16(scales128, _mm_add_epi16(m64, shifts));
s8k.accum_mins(scales_s, q8, i, d, accm);
return MM256_SET_M128I(scales128, scales128);
}
inline void prepare(int i, int j) {
bits.prepare(x[i].qs, j);
auto h256 = j == 0 ? hbits : _mm256_srli_epi16(hbits, 4);
bits.values[0] = _mm256_or_si256(bits.values[0], _mm256_and_si256(_mm256_slli_epi16(h256, 2), hmask));
bits.values[1] = _mm256_or_si256(bits.values[1], _mm256_and_si256(_mm256_slli_epi16(h256, 1), hmask));
bits.values[2] = _mm256_or_si256(bits.values[2], _mm256_and_si256(h256, hmask));
bits.values[3] = _mm256_or_si256(bits.values[3], _mm256_and_si256(_mm256_srli_epi16(h256, 1), hmask));
bits.values[0] = _mm256_shuffle_epi8(values, bits.values[0]);
bits.values[1] = _mm256_shuffle_epi8(values, bits.values[1]);
bits.values[2] = _mm256_shuffle_epi8(values, bits.values[2]);
bits.values[3] = _mm256_shuffle_epi8(values, bits.values[3]);
}
static inline __m256i load_values() {
static const uint8_t kvalues_iq3nl[16] = {1, 24, 41, 54, 65, 77, 92, 111, 5, 28, 45, 58, 69, 81, 96, 115};
auto val128 = _mm_loadu_si128((const __m128i *)kvalues_iq3nl);
return MM256_SET_M128I(val128, val128);
}
Q2Bits bits;
const Scales8KBase s8k;
const __m256i values;
__m256i hbits;
const __m128i mask = _mm_set1_epi16(254);
const __m128i m127 = _mm_set1_epi16(-127);
const __m128i m64 = _mm_set1_epi16(-64);
const __m128i m1 = _mm_set1_epi16(1);
const __m128i m4 = _mm_set1_epi16(4);
const __m256i hmask = _mm256_set1_epi8(4);
};
struct DequantizerIQ4K final : public BaseDequantizer<block_iq4_k> {
DequantizerIQ4K(const void * vx, size_t bx) : BaseDequantizer(vx, bx), iqxk(4, -128), values(load_iq4nl_values_256()) {}
template <typename Q8>