1 /* Copyright (c) FFLAS-FFPACK
2 * Written by Philippe LEDENT <philippe.ledent@etu.univ-grenoble-alpes.fr>
3 * ========LICENCE========
4 * This file is part of the library FFLAS-FFPACK.
5 *
6 * FFLAS-FFPACK is free software: you can redistribute it and/or modify
7 * it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * This library is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with this library; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 * ========LICENCE========
20 */
21
22 // declare that the call to openblas_set_numthread will be made here, hence don't do it
23 // everywhere in the call stack
24 #define __FFLASFFPACK_OPENBLAS_NT_ALREADY_SET 1
25
26 #include "fflas-ffpack/fflas-ffpack-config.h"
27 #include <iostream>
28 #include <givaro/modular.h>
29 #include <givaro/givrational.h>
30
31 #include "fflas-ffpack/fflas-ffpack.h"
32 #include "fflas-ffpack/utils/timer.h"
33 #include "fflas-ffpack/utils/fflas_io.h"
34 #include "fflas-ffpack/paladin/parallel.h"
35 #include "fflas-ffpack/paladin/fflas_plevel1.h"
36 #include "fflas-ffpack/utils/args-parser.h"
37
38
39 using namespace std;
40 using namespace FFLAS;
41 using namespace FFPACK;
42
43 template<class Field>
run_with_field(int q,size_t iter,size_t N,const size_t BS,const size_t p,const size_t threads)44 typename Field::Element run_with_field(int q, size_t iter, size_t N, const size_t BS, const size_t p, const size_t threads){
45 Field F(q);
46 typename Field::RandIter G(F, BS);
47
48 typename Field::Element_ptr A, B;
49 typename Field::Element d; F.init(d);
50
51 #ifdef __GIVARO_USE_OPENMP
52 Givaro::OMPTimer chrono, time;
53 #else
54 Givaro::Timer chrono, time;
55 #endif
56 time.clear();
57
58 for (size_t i=0;i<iter;++i){
59 A = fflas_new(F, N);
60 B = fflas_new(F, N);
61
62 PAR_BLOCK { pfrand(F, G, N, 1, A); pfrand(F, G, N, 1, B); }
63
64 // FFLAS::WriteMatrix(std::cerr, F, 1, N, A, 1);
65 // FFLAS::WriteMatrix(std::cerr, F, 1, N, B, 1);
66
67 F.assign(d, F.zero);
68
69
70 FFLAS::ParSeqHelper::Parallel<
71 FFLAS::CuttingStrategy::Block,
72 FFLAS::StrategyParameter::Threads> ParHelper(threads);
73
74 chrono.clear();
75 if (p){
76 chrono.start();
77 F.assign(d, fdot(F, N, A, 1U, B, 1U, ParHelper));
78 chrono.stop();
79 } else {
80 chrono.start();
81 F.assign(d, fdot(F, N, A, 1U, B, 1U, FFLAS::ParSeqHelper::Sequential()));
82 chrono.stop();
83 }
84
85
86 // std::cerr << chrono
87 // << " Gfops: " << ((double(2*N)/1000.)/1000.)/(1000.*chrono.realtime())
88 // << std::endl;
89
90 time+=chrono;
91 FFLAS::fflas_delete(A);
92 FFLAS::fflas_delete(B);
93 }
94 // -----------
95 // Standard output for benchmark
96 std::cout << "Time: " << time.realtime()/iter
97 << " Gfops: " << ((double(2*N)/1000.)/1000.)/(1000.*time.realtime())* double(iter);
98
99 // F.write(std::cerr, d) << std::endl;
100 return d;
101 }
102
main(int argc,char ** argv)103 int main(int argc, char** argv) {
104
105 #ifdef __FFLASFFPACK_OPENBLAS_NUM_THREADS
106 openblas_set_num_threads(__FFLASFFPACK_OPENBLAS_NUM_THREADS);
107 #endif
108
109 size_t iter = 20; // to get nonzero time
110 size_t N = 5000;
111 size_t BS = 5000;
112 int q = 131071101;
113 size_t p =0;
114 size_t maxallowed_threads; PAR_BLOCK { maxallowed_threads=NUM_THREADS; }
115 size_t threads=maxallowed_threads;
116
117 Argument as[] = {
118 { 'n', "-n N", "Set the dimension of the matrix C.",TYPE_INT , &N },
119 { 'q', "-q Q", "Set the field characteristic (0 for the integers).", TYPE_INT , &q },
120 { 'i', "-i R", "Set number of repetitions.", TYPE_INT , &iter },
121 { 'b', "-b B", "Set the bitsize of the random elements.", TYPE_INT , &BS},
122 { 'p', "-p P", "0 for sequential, 1 for parallel.", TYPE_INT , &p },
123 { 't', "-t T", "number of virtual threads to drive the partition.", TYPE_INT , &threads },
124 END_OF_ARGUMENTS
125 };
126
127 FFLAS::parseArguments(argc,argv,as);
128
129 if (q > 0){
130 BS = Givaro::Integer(q).bitsize();
131 double d = run_with_field<Givaro::ModularBalanced<double> >(q, iter, N, BS, p, threads);
132 std::cout << " d: " << d;
133 } else {
134 auto d = run_with_field<Givaro::ZRing<Givaro::Integer> > (q, iter, N, BS, p, threads);
135 std::cout << " size: " << logtwo(d>0?d:-d);
136 }
137
138 FFLAS::writeCommandString(std::cout, as) << std::endl;
139 return 0;
140 }
141 /* -*- mode: C++; tab-width: 4; indent-tabs-mode: nil; c-basic-offset: 4 -*- */
142 // vim:sts=4:sw=4:ts=4:et:sr:cino=>s,f0,{0,g0,(0,\:0,t0,+0,=s
143