ikawrakow/ik_llama.cpp
1
0
Fork 0
mirror of https://github.com/ikawrakow/ik_llama.cpp.git synced 2026-02-24 15:14:10 +00:00
Code Issues Packages Projects Releases Wiki Activity

WIP

Browse Source
This commit is contained in:
Iwan Kawrakow
2025-04-19 09:12:50 +03:00
parent 998c1b2117
commit bcacf33350
1 changed files with 10 additions and 72 deletions
Download Patch File Download Diff File Expand all files Collapse all files

82
ggml/src/iqk/iqk_mul_mat.cpp
View File

@@ -8654,6 +8654,14 @@ void mul_mat_q8_0_q8_x4_2x2(int n, const void * vx, size_t bx, const DataInfo& i
acc[0] = _mm256_add_ps(acc[0], acc[1]);
// x1,y0 x1,y0, x1,y1, x1,y1, x1,y0 x1,y0, x1,y1, x1,y1,
acc[2] = _mm256_add_ps(acc[2], acc[3]);
//auto sum1 = _mm_add_ps(_mm256_castps256_ps128(acc[0]), _mm256_extractf128_ps(acc[0], 1));
//auto sum2 = _mm_add_ps(_mm256_castps256_ps128(acc[2]), _mm256_extractf128_ps(acc[2], 1));
//auto sum = _mm_hadd_ps(sum1, sum2);
//_mm_storeu_ps(daux, sum);
//info.store(ix+0, 0, daux[0]);
//info.store(ix+0, 1, daux[1]);
//info.store(ix+1, 0, daux[2]);
//info.store(ix+1, 1, daux[3]);
auto sum = _mm_add_ps(_mm256_castps256_ps128(acc[0]), _mm256_extractf128_ps(acc[0], 1));
_mm_storeu_ps(daux, sum);
info.store(ix+0, 0, daux[0] + daux[1]);
@@ -8666,76 +8674,6 @@ void mul_mat_q8_0_q8_x4_2x2(int n, const void * vx, size_t bx, const DataInfo& i
}
}
//void mul_mat_q8_0_q8_x4_2x2(int n, const void * vx, size_t bx, const DataInfo& info, int nrc_x) {
// Q8<2, block_q8_0> q8(info);
// int nblock = n/QK8_0;
// auto cx = (const char *)vx;
// auto y4l = (const block_q8_0_x4 *)q8.y[0];
// auto y4h = (const block_q8_0_x4 *)q8.y[1];
// ggml_half d4[4];
// __m256i dot[4];
// for (int ix = 0; ix < nrc_x; ix += 2) {
// auto q8xl = (const block_q8_0 *)cx; cx += bx;
// auto q8xh = (const block_q8_0 *)cx; cx += bx;
// auto accl = _mm256_setzero_ps();
// auto acch = _mm256_setzero_ps();
// for (int ib4 = 0; ib4 < nblock/4; ++ib4) {
// auto d4y_128 = _mm_cvtph_ps(_mm_loadl_epi64((const __m128i *)y4l[ib4].d));
// auto d4yl = _mm256_set_m128(d4y_128, d4y_128);
// d4y_128 = _mm_cvtph_ps(_mm_loadl_epi64((const __m128i *)y4h[ib4].d));
// auto d4yh = _mm256_set_m128(d4y_128, d4y_128);
// for (int k = 0; k < 4; ++k) {
// d4[2*k+0] = q8xl[k].d;
// d4[2*k+1] = q8xh[k].d;
// auto qxl = _mm256_loadu_si256((const __m256i *)q8xl[k].qs);
// auto qxh = _mm256_loadu_si256((const __m256i *)q8xh[k].qs);
// auto uxl = _mm256_sign_epi8(qxl, qxl);
// auto uxh = _mm256_sign_epi8(qxh, qxh);
// auto qyl = _mm256_loadu_si256((const __m256i*)y4l[ib4].qs + k);
// auto qyh = _mm256_loadu_si256((const __m256i*)y4h[ib4].qs + k);
// auto pll = _mm256_madd_epi16(_mm256_set1_epi16(1), _mm256_maddubs_epi16(uxl, _mm256_sign_epi8(qyl, qxl)));
// auto plh = _mm256_madd_epi16(_mm256_set1_epi16(1), _mm256_maddubs_epi16(uxl, _mm256_sign_epi8(qyh, qxl)));
// auto phl = _mm256_madd_epi16(_mm256_set1_epi16(1), _mm256_maddubs_epi16(uxh, _mm256_sign_epi8(qyl, qxh)));
// auto phh = _mm256_madd_epi16(_mm256_set1_epi16(1), _mm256_maddubs_epi16(uxh, _mm256_sign_epi8(qyh, qxh)));
// auto pyl = _mm256_add_epi32(_mm256_unpacklo_epi32(pll, phl), _mm256_unpackhi_epi32(pll, phl));
// auto pyh = _mm256_add_epi32(_mm256_unpacklo_epi32(plh, phh), _mm256_unpackhi_epi32(plh, phh));
// // ll, hl, lh, hh, ll, hl, lh, hh
// dot[k] = _mm256_add_epi32(_mm256_unpacklo_epi64(pyl, pyh), _mm256_unpackhi_epi64(pyl, pyh));
// }
// auto d4x_128 = _mm_cvtph_ps(_mm_loadl_epi64((const __m128i *)d4));
// auto d4x = _mm256_set_m128(d4x_128, d4x_128);
// auto d4xyl = _mm256_mul_ps(d4x, d4yl);
// auto d4xyh = _mm256_mul_ps(d4x, d4yh);
// acc = _mm256_fmadd_ps(_mm256_shuffle_ps(d4xyl, d4xyh, 0x00), _mm256_cvtepi32_ps(dot[0]), acc);
// acc = _mm256_fmadd_ps(_mm256_shuffle_ps(d4xyl, d4xyh, 0x55), _mm256_cvtepi32_ps(dot[1]), acc);
// acc = _mm256_fmadd_ps(_mm256_shuffle_ps(d4xyl, d4xyh, 0xaa), _mm256_cvtepi32_ps(dot[2]), acc);
// acc = _mm256_fmadd_ps(_mm256_shuffle_ps(d4xyl, d4xyh, 0xff), _mm256_cvtepi32_ps(dot[3]), acc);
// q8x += 4;
// }
// if (int ib0 = 4*(nblock/4); ib0 < nblock) {
// auto yl = q8.y[0] + ib0;
// auto yh = q8.y[1] + ib0;
// for (int k = 0; k < nblock - ib0; ++k) {
// auto qx = _mm256_loadu_si256((const __m256i *)q8x[k].qs);
// auto ux = _mm256_sign_epi8(qx, qx);
// auto pl = _mm256_maddubs_epi16(ux, _mm256_sign_epi8(_mm256_loadu_si256((const __m256i*)yl[k].qs), qx));
// auto ph = _mm256_maddubs_epi16(ux, _mm256_sign_epi8(_mm256_loadu_si256((const __m256i*)yh[k].qs), qx));
// pl = _mm256_madd_epi16(_mm256_set1_epi16(1), pl);
// ph = _mm256_madd_epi16(_mm256_set1_epi16(1), ph);
// auto p = _mm256_add_epi32(_mm256_unpacklo_epi64(pl, ph), _mm256_unpackhi_epi64(pl, ph));
// auto d = GGML_FP16_TO_FP32(q8x[k].d);
// auto dxyl = _mm256_set1_ps(d*GGML_FP16_TO_FP32(yl[k].d));
// auto dxyh = _mm256_set1_ps(d*GGML_FP16_TO_FP32(yh[k].d));
// acc = _mm256_fmadd_ps(_mm256_shuffle_ps(dxyl, dxyh, 0x00), _mm256_cvtepi32_ps(p), acc);
// }
// }
// auto sum = _mm_add_ps(_mm256_castps256_ps128(acc), _mm256_extractf128_ps(acc, 1));
// float daux[4]; _mm_storeu_ps(daux, sum);
// info.store(ix, 0, daux[0] + daux[1]);
// info.store(ix, 1, daux[2] + daux[3]);
// }
//}
struct Dequantizer4bit {
const __m256i m4 = _mm256_set1_epi8(0xf);
inline __m256i dequant(const uint8_t * qs) const {
@@ -16877,8 +16815,8 @@ struct FlashQKfp32 {
#ifdef HAVE_FANCY_SIMD
MAKE_FUNCS(mul_mat_qX_1_q8_2_T<Q8_0_1_Unpacker, nq);
#else
if (nq == 1) return std::make_pair(mul_mat_q8_0_q8_x4_1, 1);
if (nq == 2) return D%128 == 0 ? std::make_pair(mul_mat_q8_0_q8_x4_2x2, 2) : std::make_pair(mul_mat_q8_0_q8_x4_2, 2);
//if (nq == 1) return std::make_pair(mul_mat_q8_0_q8_x4_1, 1);
//if (nq == 2) return D%128 == 0 ? std::make_pair(mul_mat_q8_0_q8_x4_2x2, 2) : std::make_pair(mul_mat_q8_0_q8_x4_2, 2);
MAKE_FUNCS(mul_mat_qX_0_q8_0_T<Q8_0_Unpacker, nq);
#endif
#endif