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
|
// Haney's induction calculation benchmark.
//
// See: Scott W. Haney, Is C++ Fast Enough for Scientific Computing?
// Computers in Physics Vol. 8 No. 6 (1994), p. 690
//
// Arch D. Robison, C++ Gets Faster for Scientific Computing,
// Computers in Physics Vol. 10 No. 5 (1996), p. 458
//
#include <blitz/vector.h>
#include <blitz/rand-uniform.h>
#include <blitz/benchext.h>
#ifdef BZ_HAVE_STD
#include <valarray>
#else
#include <valarray.h>
#endif
BZ_USING_NAMESPACE(blitz)
#ifndef M_PI
#define M_PI 3.14159265358979323846
#endif
#ifdef BZ_FORTRAN_SYMBOLS_WITH_TRAILING_UNDERSCORES
#define vecopsf vecopsf_
#define vecopsfo vecopsfo_
#endif
extern "C"
{
void vecopsf(float *li, const float *R, const float *w, const int &N,
const int& iters);
void vecopsfo(float *li, const float *R, const float *w, const int &N,
const int& iters);
}
inline float sqr(float x)
{
return x*x;
}
const float Mu0 = 4.0 * M_PI * 1.0e-7;
void HaneyCVersion(BenchmarkExt<int>& bench);
void HaneyFortranVersion(BenchmarkExt<int>& bench);
void HaneyBlitzVersion(BenchmarkExt<int>& bench);
int main()
{
BenchmarkExt<int> bench("Haney Inductance Calculation", 3);
bench.setRateDescription("Operations/s");
bench.beginBenchmarking();
HaneyCVersion(bench);
HaneyFortranVersion(bench);
HaneyBlitzVersion(bench);
bench.endBenchmarking();
bench.saveMatlabGraph("haney.m");
return 0;
}
void initializeRandom(float* data, int length)
{
Random<Uniform> unif(1.0, 2.0);
for (int i=0; i < length; ++i)
data[i] = unif.random();
}
void HaneyCVersion(BenchmarkExt<int>& bench)
{
bench.beginImplementation("Inlined C");
while (!bench.doneImplementationBenchmark()) {
int length = bench.getParameter();
long iters = bench.getIterations();
cout << "length = " << length << " iters = " << iters << endl;
float* li = new float[length];
float* R = new float[length];
float* w = new float[length];
initializeRandom(li, length);
initializeRandom(R, length);
initializeRandom(w, length);
// Tickle the cache
for (int i=0; i < length; ++i)
li[i] = R[i] + log(w[i]);
bench.start();
for (long j=0; j < iters; ++j) {
for (int i=0; i < length; ++i) {
li[i] = Mu0 * R[i] *
(0.5 * (1.0 + (1.0/24.0)
* sqr(w[i]/R[i])) * log(32.0 * sqr(R[i]/w[i]))
+ 0.05 * sqr(w[i]/R[i]) - 0.85);
}
}
bench.stop();
// Subtract the loop overhead
bench.startOverhead();
for (long j=0; j < iters; ++j) {}
bench.stopOverhead();
delete [] li;
delete [] w;
delete [] R;
}
bench.endImplementation();
}
void HaneyFortranVersion(BenchmarkExt<int>& bench)
{
bench.beginImplementation("Fortran");
while (!bench.doneImplementationBenchmark()) {
int length = bench.getParameter();
int iters = (int)bench.getIterations();
cout << "length = " << length << " iters = " << iters << endl;
float* li = new float[length];
float* R = new float[length];
float* w = new float[length];
initializeRandom(li, length);
initializeRandom(R, length);
initializeRandom(w, length);
// Tickle
int oneIter = 1;
vecopsf(li, R, w, length, oneIter);
// Time
bench.start();
vecopsf(li, R, w, length, iters);
bench.stop();
// Time overhead
bench.startOverhead();
vecopsfo(li, R, w, length, iters);
bench.stopOverhead();
delete [] li;
delete [] w;
delete [] R;
}
bench.endImplementation();
}
void HaneyBlitzVersion(BenchmarkExt<int>& bench)
{
bench.beginImplementation("Blitz++");
while (!bench.doneImplementationBenchmark()) {
int length = bench.getParameter();
int iters = (int)bench.getIterations();
Vector<float> li(length), R(length), w(length);
initializeRandom(li.data(), length);
initializeRandom(R.data(), length);
initializeRandom(w.data(), length);
cout << "length = " << length << " iters = " << iters << endl;
// Tickle
li = w + log(R);
// Time
bench.start();
for (long i=0; i < iters; ++i) {
#if defined(__GNUC__) && (__GNUC__ < 3)
li = Mu0 * R * ( (0.5 + (0.5/24.0) * sqr(w/R) )
* log(32.0 * sqr(R/w)) + 0.05 * sqr(w/R) - 0.85);
#else
li = Mu0 * R * (0.5 * (1.0 + (1.0/24.0) * sqr(w/R))
* log(32.0 * sqr(R/w)) + 0.05 * sqr(w/R) - 0.85);
#endif
}
bench.stop();
// Time overhead
bench.startOverhead();
for (long i=0; i < iters; ++i) {
}
bench.stopOverhead();
}
bench.endImplementation();
}
|
