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
|
// Array DAXPY benchmark
#include <blitz/benchext.h>
#include <blitz/array.h>
#include <blitz/vector2.h>
#include <random/uniform.h>
BZ_NAMESPACE(blitz)
extern void sink();
BZ_NAMESPACE_END
BZ_USING_NAMESPACE(blitz)
#ifdef BZ_FORTRAN_SYMBOLS_WITH_TRAILING_UNDERSCORES
#define arrdaxpyf arrdaxpyf_
#elif defined(BZ_FORTRAN_SYMBOLS_WITH_DOUBLE_TRAILING_UNDERSCORES)
#define arrdaxpyf arrdaxpyf__
#endif
extern "C" {
void arrdaxpyf(double* A, double* B, int& N, double& a);
}
void arrdaxpyFortran77Version(BenchmarkExt<int>& bench);
void arrdaxpyBlitzVersion(BenchmarkExt<int>& bench);
int main()
{
BenchmarkExt<int> bench("Array DAXPY", 2);
const int numSizes = 8;
bench.setNumParameters(numSizes);
bench.setDependentVariable("flops");
Vector<int> parameters(numSizes);
Vector<long> iters(numSizes);
Vector<double> flops(numSizes);
parameters = pow(2.,tensor::i);
cout << parameters;
iters = 100*16*32*8*8*8/pow3(parameters);
cout << iters;
flops = pow3(parameters) * 2 * 2;
cout << flops;
bench.setParameterVector(parameters);
bench.setParameterDescription("3D Array size");
bench.setIterations(iters);
bench.setOpsPerIteration(flops);
bench.beginBenchmarking();
arrdaxpyBlitzVersion(bench);
arrdaxpyFortran77Version(bench);
bench.endBenchmarking();
bench.saveMatlabGraph("arrdaxpy.m");
return 0;
}
void initializeRandomDouble(double* data, int numElements)
{
ranlib::Uniform<double> rnd;
for (int i=0; i < numElements; ++i)
data[i] = rnd.random();
}
void arrdaxpyBlitzVersion(BenchmarkExt<int>& bench)
{
bench.beginImplementation("Blitz++");
while (!bench.doneImplementationBenchmark())
{
int N = bench.getParameter();
cout << "Blitz++: N = " << N << endl;
cout.flush();
long iters = bench.getIterations();
Array<double,3> A(N,N,N), B(N,N,N);
initializeRandomDouble(A.data(), N*N*N);
initializeRandomDouble(B.data(), N*N*N);
TinyVector<int,2> size = N-2;
double a = 0.34928313;
double b = - a;
bench.start();
for (long i=0; i < iters; ++i)
{
A += a * B;
A += b * B;
sink();
}
bench.stop();
bench.startOverhead();
for (long i=0; i < iters; ++i) {
sink();
}
bench.stopOverhead();
}
bench.endImplementation();
}
void arrdaxpyFortran77Version(BenchmarkExt<int>& bench)
{
bench.beginImplementation("Fortran 77");
while (!bench.doneImplementationBenchmark())
{
int N = bench.getParameter();
cout << "Fortran 77: N = " << N << endl;
cout.flush();
int iters = (int)bench.getIterations();
size_t arraySize = size_t(N) * size_t(N) * N;
double* A = new double[arraySize];
double* B = new double[arraySize];
initializeRandomDouble(A, arraySize);
initializeRandomDouble(B, arraySize);
double a = 0.34928313;
for (long i=0; i < iters; ++i)
{
arrdaxpyf(A,B,N,a);
sink();
}
bench.stop();
bench.startOverhead();
for (long i=0; i < iters; ++i) {
sink();
}
bench.stopOverhead();
delete [] A;
delete [] B;
}
bench.endImplementation();
}
|
