aboutsummaryrefslogtreecommitdiffhomepage
path: root/bench/btl/generic_bench/bench.hh
blob: 005c3639552b837b5c44679da50f1d3fb8ba8491 (plain)
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
//=====================================================
// File   :  bench.hh
// Author :  L. Plagne <laurent.plagne@edf.fr)>
// Copyright (C) EDF R&D,  lun sep 30 14:23:16 CEST 2002
//=====================================================
//
// This program is free software; you can redistribute it and/or
// modify it under the terms of the GNU General Public License
// as published by the Free Software Foundation; either version 2
// of the License, or (at your option) any later version.
//
// This program 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 General Public License for more details.
// You should have received a copy of the GNU General Public License
// along with this program; if not, write to the Free Software
// Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111-1307, USA.
//
#ifndef BENCH_HH
#define BENCH_HH

#include "btl.hh"
#include "bench_parameter.hh"
#include <iostream>
#include "utilities.h"
#include "size_lin_log.hh"
#include "xy_file.hh"
#include <vector>
#include <string>
#include "timers/portable_perf_analyzer.hh"
// #include "timers/mixed_perf_analyzer.hh"
// #include "timers/x86_perf_analyzer.hh"
// #include "timers/STL_perf_analyzer.hh"
#ifdef HAVE_MKL
extern "C" void cblas_saxpy(const int, const float, const float*, const int, float *, const int);
#endif
using namespace std;

template <template<class> class Perf_Analyzer, class Action>
BTL_DONT_INLINE void bench( int size_min, int size_max, int nb_point )
{
  if (BtlConfig::skipAction(Action::name()))
    return;

  string filename="bench_"+Action::name()+".dat";

  INFOS("starting " <<filename);

  // utilities

  std::vector<double> tab_mflops(nb_point);
  std::vector<int> tab_sizes(nb_point);

  // matrices and vector size calculations
  size_lin_log(nb_point,size_min,size_max,tab_sizes);

  std::vector<int> oldSizes;
  std::vector<double> oldFlops;
  bool hasOldResults = read_xy_file(filename, oldSizes, oldFlops, true);
  int oldi = oldSizes.size() - 1;

  // loop on matrix size
  Perf_Analyzer<Action> perf_action;
  for (int i=nb_point-1;i>=0;i--)
  {
    //INFOS("size=" <<tab_sizes[i]<<"   ("<<nb_point-i<<"/"<<nb_point<<")");
    std::cout << " " << "size = " << tab_sizes[i] << "  " << std::flush;

    BTL_DISABLE_SSE_EXCEPTIONS();
    #ifdef HAVE_MKL
    {
      float dummy;
      cblas_saxpy(1,0,&dummy,1,&dummy,1);
    }
    #endif

    tab_mflops[i] = perf_action.eval_mflops(tab_sizes[i]);
    std::cout << tab_mflops[i];
    
    if (hasOldResults)
    {
      while (oldi>=0 && oldSizes[oldi]>tab_sizes[i])
        --oldi;
      if (oldi>=0 && oldSizes[oldi]==tab_sizes[i])
      {
        if (oldFlops[oldi]<tab_mflops[i])
          std::cout << "\t > ";
        else
          std::cout << "\t < ";
        std::cout << oldFlops[oldi];
      }
      --oldi;
    }
    std::cout << " MFlops    (" << nb_point-i << "/" << nb_point << ")" << std::endl;
  }

  if (!BtlConfig::Instance.overwriteResults)
  {
    if (hasOldResults)
    {
      // merge the two data
      std::vector<int> newSizes;
      std::vector<double> newFlops;
      int i=0;
      int j=0;
      while (i<tab_sizes.size() && j<oldSizes.size())
      {
        if (tab_sizes[i] == oldSizes[j])
        {
          newSizes.push_back(tab_sizes[i]);
          newFlops.push_back(std::max(tab_mflops[i], oldFlops[j]));
          ++i;
          ++j;
        }
        else if (tab_sizes[i] < oldSizes[j])
        {
          newSizes.push_back(tab_sizes[i]);
          newFlops.push_back(tab_mflops[i]);
          ++i;
        }
        else
        {
          newSizes.push_back(oldSizes[j]);
          newFlops.push_back(oldFlops[j]);
          ++j;
        }
      }
      while (i<tab_sizes.size())
      {
        newSizes.push_back(tab_sizes[i]);
        newFlops.push_back(tab_mflops[i]);
        ++i;
      }
      while (j<oldSizes.size())
      {
        newSizes.push_back(oldSizes[j]);
        newFlops.push_back(oldFlops[j]);
        ++j;
      }
      tab_mflops = newFlops;
      tab_sizes = newSizes;
    }
  }

  // dump the result in a file  :
  dump_xy_file(tab_sizes,tab_mflops,filename);

}

// default Perf Analyzer

template <class Action>
BTL_DONT_INLINE void bench( int size_min, int size_max, int nb_point ){

  // if the rdtsc is not available :
  bench<Portable_Perf_Analyzer,Action>(size_min,size_max,nb_point);
  // if the rdtsc is available :
//    bench<Mixed_Perf_Analyzer,Action>(size_min,size_max,nb_point);


  // Only for small problem size. Otherwize it will be too long
//   bench<X86_Perf_Analyzer,Action>(size_min,size_max,nb_point);
//   bench<STL_Perf_Analyzer,Action>(size_min,size_max,nb_point);

}

#endif