/usr/include/fflas-ffpack/fflas/fflas_sparse/sell/sell_utils.inl is in fflas-ffpack-common 2.2.2-5.
This file is owned by root:root, with mode 0o644.
The actual contents of the file can be viewed below.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 | /* -*- mode: C++; tab-width: 8; indent-tabs-mode: t; c-basic-offset: 8 -*- */
// vim:sts=8:sw=8:ts=8:noet:sr:cino=>s,f0,{0,g0,(0,\:0,t0,+0,=s
/*
* Copyright (C) 2014 the FFLAS-FFPACK group
*
* Written by Bastien Vialla <bastien.vialla@lirmm.fr>
*
*
* ========LICENCE========
* This file is part of the library FFLAS-FFPACK.
*
* FFLAS-FFPACK is free software: you can redistribute it and/or modify
* it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with this library; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
* ========LICENCE========
*.
*/
#ifndef __FFLASFFPACK_fflas_sparse_sell_utils_INL
#define __FFLASFFPACK_fflas_sparse_sell_utils_INL
namespace FFLAS {
template <class Field>
inline void fspmv(const Field &F, const Sparse<Field, SparseMatrix_t::SELL_ZO> &A, typename Field::ConstElement_ptr x,
typename Field::Element_ptr y, FieldCategories::ModularTag) {
fspmv(F, A, x, y, FieldCategories::UnparametricTag());
freduce(F, A.m, y, 1);
}
template <class Field> inline void sparse_delete(const Sparse<Field, SparseMatrix_t::SELL> &A) {
fflas_delete(A.dat);
fflas_delete(A.col);
fflas_delete(A.st);
fflas_delete(A.chunkSize);
}
template <class Field> inline void sparse_delete(const Sparse<Field, SparseMatrix_t::SELL_ZO> &A) {
fflas_delete(A.col);
fflas_delete(A.st);
fflas_delete(A.chunkSize);
}
namespace sell_details {
struct Info {
uint64_t size = 0;
uint64_t perm = 0;
uint64_t begin = 0;
Info(uint64_t it, uint64_t s, uint64_t p) : size(s), perm(p), begin(it) {}
Info() = default;
Info(const Info &) = default;
Info(Info &&) = default;
Info &operator=(const Info &) = default;
Info &operator=(Info &&) = default;
};
template <class ValT, class IdxT> struct Coo {
using Self = Coo<ValT, IdxT>;
ValT val = 0;
IdxT row = 0;
IdxT col = 0;
Coo(ValT v, IdxT r, IdxT c) : val(v), row(r), col(c) {}
Coo() = default;
Coo(const Self &) = default;
Coo(Self &&) = default;
Self &operator=(const Self &) = default;
Self &operator=(Self &&) = default;
};
}
template <class Field> inline void sparse_print(const Sparse<Field, SparseMatrix_t::SELL> &A) {
uint64_t it = 0;
for (size_t i = 0; i < A.nChunks; ++i) {
for (size_t k = 0; k < A.chunk; ++k) {
std::cout << i *A.chunk + k << " : ";
for (size_t j = 0; j < A.chunkSize[i]; ++j) {
std::cout << A.dat[it + j * A.chunk + k] << " ";
}
std::cout << std::endl;
}
it += A.chunkSize[i] * A.chunk;
}
}
template <class Field, class IndexT>
inline void sparse_init(const Field &F, Sparse<Field, SparseMatrix_t::SELL> &A, const IndexT *row, const IndexT *col,
typename Field::ConstElement_ptr dat, uint64_t rowdim, uint64_t coldim, uint64_t nnz,
uint64_t sigma = 0) {
using namespace sell_details;
using coo = Coo<typename Field::Element, IndexT>;
if (!sigma)
sigma = rowdim;
A.kmax = Protected::DotProdBoundClassic(F, F.one);
A.m = rowdim;
A.n = coldim;
A.nnz = nnz;
#ifdef __FFLASFFPACK_HAVE_SSE4_1_INSTRUCTIONS
using simd = Simd<typename Field::Element>;
A.chunk = simd::vect_size;
#else
A.chunk = 8;
#endif
index_t m = (A.m % A.chunk == 0) ? A.m : ROUND_DOWN(A.m, A.chunk) + A.chunk;
A.nChunks = (m / A.chunk);
std::vector<coo> data;
std::vector<Info> infos(A.nChunks * A.chunk);
for (uint64_t i = 0; i < nnz; ++i) {
data.emplace_back(dat[i], row[i], col[i]);
}
IndexT currow = row[0];
infos[currow].begin = 0;
for (uint64_t i = 0; i < nnz; ++i) {
if (row[i] != currow) {
currow = row[i];
infos[currow].begin = i;
}
infos[row[i]].size++;
}
A.maxrow = (std::max_element(infos.begin(), infos.end(),
[](const Info &a, const Info &b) { return a.size >= b.size; }))->size;
// cout << "maxrow : " << A.maxrow << endl;
if (A.maxrow < A.kmax)
A.delayed = true;
for (uint64_t i = 0; i < rowdim; ++i) {
infos[i].perm = i;
}
#ifdef SELL_DEBUG
for (auto &x : infos) {
cout << x.size << " ";
}
std::cout << std::endl;
#endif
uint64_t it = 0;
for (; it < ROUND_DOWN(rowdim, sigma); it += sigma) {
std::sort(infos.begin() + it, infos.begin() + it + sigma,
[](const Info &a, const Info &b) { return a.size >= b.size; });
}
if (it != rowdim) {
std::sort(infos.begin() + it, infos.end(), [](Info a, Info b) { return a.size >= b.size; });
}
// cout << "sorted : " << std::is_sorted(infos.begin(), infos.end(), [](Info
// a, Info b){
// return a.size >= b.size;
// }) << endl;
for (size_t i = 0; i < infos.size(); ++i) {
if (infos[i].begin > nnz)
std::cout << "ERROR sort " << i << " size : " << infos[i].size << " begin : " << infos[i].begin
<< " perm : " << infos[i].perm << std::endl;
}
#ifdef SELL_DEBUG
for (auto &x : infos) {
cout << x.size << " ";
}
std::cout << std::endl;
#endif
A.perm = fflas_new<index_t>(rowdim, Alignment::CACHE_LINE);
// cout << "perm : ";
for (uint64_t i = 0; i < rowdim; ++i) {
// cout << "(" << i << " , " << infos[i].perm << ") ";
A.perm[infos[i].perm] = i;
}
// for(size_t i = 0 ; i < A.m ; ++i)
// cout << A.perm[i] << " ";
// cout << endl;
// cout << endl;
// add info if rowdim%chunk != 0, with empty infos (size = 0, begin = 0)
// infos.resize(A.nChunks*A.chunk);
// for(auto & x:infos)
// if(x.begin > nnz)
// cout << "ERROR resize" << endl;
A.chunkSize = fflas_new<index_t>(A.nChunks, Alignment::CACHE_LINE);
for (uint64_t i = 0; i < A.nChunks; ++i)
A.chunkSize[i] = 0;
for (uint64_t i = 0; i < A.nChunks; ++i) {
for (uint64_t j = 0; j < A.chunk; ++j) {
if (infos[i * A.chunk + j].size >= A.chunkSize[i])
A.chunkSize[i] = infos[i * A.chunk + j].size;
}
}
#ifdef SELL_DEBUG
for (uint64_t i = 0; i < A.nChunks; ++i)
cout << "chunk " << i << " : " << A.chunkSize[i] << endl;
;
#endif
uint64_t sum = 0;
for (uint64_t i = 0; i < A.nChunks; ++i)
sum += A.chunkSize[i];
#ifdef SELL_DEBUG
cout << "sum : " << sum << " chunk : " << A.chunk << endl;
#endif
A.col = fflas_new<index_t>(sum * A.chunk, Alignment::CACHE_LINE);
A.dat = fflas_new(F, sum * A.chunk, 1, Alignment::CACHE_LINE);
A.nElements = sum * A.chunk;
for (uint64_t i = 0; i < sum * A.chunk; ++i) {
A.col[i] = 0;
F.assign(A.dat[i], F.zero);
}
it = 0;
for (uint64_t i = 0; i < A.nChunks; ++i) {
for (uint64_t k = 0; k < A.chunk; ++k) {
uint64_t start = infos[i * A.chunk + k].begin;
#ifdef SELL_DEBUG
cout << it << " " << start << " " << infos[i * A.chunk + k].size << endl;
cout << " ";
#endif
for (uint64_t j = 0; j < infos[i * A.chunk + k].size; ++j) {
if (it + k + j * A.chunk >= sum * A.chunk)
std::cout << "error : " << it + k + j *A.chunk << " " << sum *A.chunk << std::endl;
A.dat[it + k + j * A.chunk] = data[start + j].val;
A.col[it + k + j * A.chunk] = data[start + j].col;
#ifdef SELL_DEBUG
cout << data[start + j].val << " ";
#endif
}
#ifdef SELL_DEBUG
cout << endl;
#endif
}
it += A.chunkSize[i] * A.chunk;
}
A.st = fflas_new<uint64_t>(A.nChunks, Alignment::CACHE_LINE);
A.st[0] = 0;
for (uint64_t i = 1; i < A.nChunks; ++i) {
A.st[i] = A.chunkSize[i - 1] * A.chunk;
}
for (uint64_t i = 1; i < A.nChunks; ++i) {
A.st[i] += A.st[i - 1];
}
#ifdef SELL_DEBUG
cout << "st : ";
for (uint64_t i = 0; i < A.nChunks; ++i)
cout << A.st[i] << " ";
cout << endl;
#endif
}
template <class Field, class IndexT>
inline void sparse_init(const Field &F, Sparse<Field, SparseMatrix_t::SELL_ZO> &A, const IndexT *row, const IndexT *col,
typename Field::ConstElement_ptr dat, uint64_t rowdim, uint64_t coldim, uint64_t nnz) {}
} // FFLAS
#endif
|