Mirrors/llama.cpp
1
0
Fork 0
mirror of https://github.com/ggerganov/llama.cpp.git synced 2024-12-28 15:18:26 +01:00
Code Issues Packages Projects Releases Wiki Activity

ggml : IQ4_NL sgemm + Q4_0 AVX optimization (#9422)

Browse Source 
* squashed

readd my iq4_nl sgemm PR https://github.com/ggerganov/llama.cpp/pull/8049

have ggml_vec_dot_q4_0 do two blocks per loop for avx

try out f16c ggml_vec_dot_iq4_nl, but it's not really faster. as per https://github.com/ggerganov/llama.cpp/pull/8549 we can calculate several blocks at a time with no issue

* shuffle

* remove f16c iq4_nl as i cant make it faster than before
This commit is contained in:
Eve 2024-09-16 06:48:24 +00:00 committed by GitHub
parent 0aadac10c7
commit 5c3d0f1824
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
2 changed files with 71 additions and 36 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

69
ggml/src/ggml-quants.c
View File

@ -230,6 +230,12 @@ static inline __m128i packNibbles( __m128i bytes1, __m128i bytes2 )
return _mm_packus_epi16( bytes1, bytes2); return _mm_packus_epi16( bytes1, bytes2);
} }
static inline __m128i mul_add_epi8_sse(const __m128i x, const __m128i y) {
const __m128i ax = _mm_sign_epi8(x, x);
const __m128i sy = _mm_sign_epi8(y, x);
return _mm_maddubs_epi16(ax, sy);
}
#endif #endif
#elif defined(__SSSE3__) #elif defined(__SSSE3__)
// horizontally add 4x4 floats // horizontally add 4x4 floats
@ -4206,37 +4212,37 @@ void ggml_vec_dot_q4_0_q8_0(int n, float * restrict s, size_t bs, const void * r
sumf = hsum_float_8(acc); sumf = hsum_float_8(acc);
#elif defined(__AVX__) #elif defined(__AVX__)
// Initialize accumulator with zeros const __m128i mone = _mm_set1_epi16(1);
__m256 acc = _mm256_setzero_ps();
// Main loop __m256 accum1 = _mm256_setzero_ps();
for (; ib < nb; ++ib) { __m256 accum2 = _mm256_setzero_ps();
// Compute combined scale for the block for (; ib + 1 < nb; ib += 2) {
const __m256 d = _mm256_set1_ps( GGML_FP16_TO_FP32(x[ib].d) * GGML_FP16_TO_FP32(y[ib].d) ); 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);
const __m128i q8b_1_0 = _mm_loadu_si128((const __m128i *)y[ib + 0].qs);
const __m128i q8b_1_1 = _mm_loadu_si128((const __m128i *)y[ib + 0].qs + 1);
const __m128i q8b_2_0 = _mm_loadu_si128((const __m128i *)y[ib + 1].qs);
const __m128i q8b_2_1 = _mm_loadu_si128((const __m128i *)y[ib + 1].qs + 1);
const __m128i lowMask = _mm_set1_epi8(0xF); const __m128i q4b_1_0 = _mm_sub_epi8(_mm_and_si128(_mm_set1_epi8(15), q4bits_1), _mm_set1_epi8(8));
const __m128i off = _mm_set1_epi8(8); 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 tmp = _mm_loadu_si128((const __m128i *)x[ib].qs); 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);
__m128i bx_0 = _mm_and_si128(lowMask, tmp); const __m128i p16_1_1 = mul_add_epi8_sse(q4b_1_1, q8b_1_1);
__m128i by_0 = _mm_loadu_si128((const __m128i *)y[ib].qs); const __m128i p16_2_0 = mul_add_epi8_sse(q4b_2_0, q8b_2_0);
bx_0 = _mm_sub_epi8(bx_0, off); const __m128i p16_2_1 = mul_add_epi8_sse(q4b_2_1, q8b_2_1);
const __m128i i32_0 = mul_sum_i8_pairs(bx_0, by_0); const __m128i p_1_0 = _mm_madd_epi16(p16_1_0, mone);
const __m128i p_1_1 = _mm_madd_epi16(p16_1_1, mone);
bx_0 = _mm_and_si128(lowMask, _mm_srli_epi64(tmp, 4)); const __m128i p_2_0 = _mm_madd_epi16(p16_2_0, mone);
by_0 = _mm_loadu_si128((const __m128i *)(y[ib].qs + 16)); const __m128i p_2_1 = _mm_madd_epi16(p16_2_1, mone);
bx_0 = _mm_sub_epi8(bx_0, off); 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)),
const __m128i i32_1 = mul_sum_i8_pairs(bx_0, by_0); _mm256_cvtepi32_ps(MM256_SET_M128I(p_1_1, p_1_0))), 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)),
// Convert int32_t to float _mm256_cvtepi32_ps(MM256_SET_M128I(p_2_1, p_2_0))), accum2);
__m256 p = _mm256_cvtepi32_ps(MM256_SET_M128I(i32_0, i32_1));
// Apply the scale, and accumulate
acc = _mm256_add_ps(_mm256_mul_ps( d, p ), acc);
} }
sumf = hsum_float_8(acc); sumf = hsum_float_8(_mm256_add_ps(accum1, accum2));
#elif defined(__SSSE3__) #elif defined(__SSSE3__)
// set constants // set constants
const __m128i lowMask = _mm_set1_epi8(0xF); const __m128i lowMask = _mm_set1_epi8(0xF);
@ -11819,15 +11825,6 @@ void ggml_vec_dot_iq3_s_q8_K (int n, float * restrict s, size_t bs, const void *
#endif #endif
} }
#if defined(__AVX__)
static inline __m128i mul_add_epi8_sse(const __m128i x, const __m128i y) {
const __m128i ax = _mm_sign_epi8(x, x);
const __m128i sy = _mm_sign_epi8(y, x);
return _mm_maddubs_epi16(ax, sy);
}
#endif
#if defined(__AVX2__) #if defined(__AVX2__)
static inline __m256i mul_add_epi8(const __m256i x, const __m256i y) { static inline __m256i mul_add_epi8(const __m256i x, const __m256i y) {
const __m256i ax = _mm256_sign_epi8(x, x); const __m256i ax = _mm256_sign_epi8(x, x);

38
ggml/src/llamafile/sgemm.cpp
View File

@ -235,6 +235,14 @@ template <> inline __m512 load(const ggml_fp16_t *p) {
} }
#endif // __AVX512F__ #endif // __AVX512F__
////////////////////////////////////////////////////////////////////////////////////////////////////
// CONSTANTS
#if defined(__AVX__) || defined(__AVX2__) || defined(__AVX512F__)
static const int8_t kvalues_iq4nl[16] = {-127, -104, -83, -65, -49, -35, -22, -10, 1, 13, 25, 38, 53, 69, 89, 113};
static const __m128i iq4nlt = _mm_loadu_si128((const __m128i *) kvalues_iq4nl);
#endif
//////////////////////////////////////////////////////////////////////////////////////////////////// ////////////////////////////////////////////////////////////////////////////////////////////////////
// FLOATING POINT MATRIX MULTIPLICATION // FLOATING POINT MATRIX MULTIPLICATION
@ -933,6 +941,20 @@ class tinyBLAS_Q0_AVX {
return _mm_sub_epi8(_mm_and_si128(_mm_set1_epi8(15), _mm_srli_epi16(x, 4)), _mm_set1_epi8(8)); return _mm_sub_epi8(_mm_and_si128(_mm_set1_epi8(15), _mm_srli_epi16(x, 4)), _mm_set1_epi8(8));
} }
inline __m256i load(const block_iq4_nl *b) {
return MM256_SET_M128I(load1(b), load0(b));
}
inline __m128i load0(const block_iq4_nl *b) {
const __m128i x = _mm_loadu_si128((const __m128i *)(b->qs));
return _mm_shuffle_epi8(iq4nlt, _mm_and_si128(_mm_set1_epi8(15), x));
}
inline __m128i load1(const block_iq4_nl *b) {
const __m128i x = _mm_loadu_si128((const __m128i *)(b->qs));
return _mm_shuffle_epi8(iq4nlt, _mm_and_si128(_mm_set1_epi8(15), _mm_srli_epi16(x, 4)));
}
inline __m256 updot(__m256i u, __m256i s) { inline __m256 updot(__m256i u, __m256i s) {
__m256i res; __m256i res;
#if defined(__AVXVNNI__) || (defined(__AVX512VNNI__) && defined(__AVX512VL__)) #if defined(__AVXVNNI__) || (defined(__AVX512VNNI__) && defined(__AVX512VL__))
@ -1159,6 +1181,22 @@ bool llamafile_sgemm(int64_t m, int64_t n, int64_t k, const void *A, int64_t lda
#endif #endif
} }
case GGML_TYPE_IQ4_NL: {
if (Btype != GGML_TYPE_Q8_0)
return false;
#if defined(__AVX2__) || defined(__AVX512F__) || defined(__AVX__)
tinyBLAS_Q0_AVX<block_iq4_nl, block_q8_0, float> tb{
k, (const block_iq4_nl *)A, lda,
(const block_q8_0 *)B, ldb,
(float *)C, ldc,
ith, nth};
tb.matmul(m, n);
return true;
#else
return false;
#endif
}
default: default:
return false; return false;
} }