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
|
// TinyVector<T,N> DAXPY benchmark
//#define BZ_DISABLE_KCC_COPY_PROPAGATION_KLUDGE
#include <blitz/array.h>
#include <blitz/timer.h>
#include <random/uniform.h>
BZ_USING_NAMESPACE(blitz)
ranlib::Uniform<double> rnd;
template<class T>
void optimizationSink(T&);
template<int N_rank>
void tinyDAXPYBenchmark(TinyVector<double,N_rank>, int iters, double a)
{
Timer timer;
TinyVector<double,N_rank> ta, tb, tc, td, te, tf, tg, th, ti, tj;
for (int i=0; i < N_rank; ++i)
{
ta[i] = rnd.random()+1;
tb[i] = rnd.random()+1;
tc[i] = rnd.random()+1;
td[i] = rnd.random()+1;
te[i] = rnd.random()+1;
tf[i] = rnd.random()+1;
tg[i] = rnd.random()+1;
th[i] = rnd.random()+1;
ti[i] = rnd.random()+1;
tj[i] = rnd.random()+1;
}
double b = -a;
double numFlops = 0;
if (N_rank < 20)
{
timer.start();
for (int i=0; i < iters; ++i)
{
ta += a * tb;
tc += a * td;
te += a * tf;
tg += a * th;
ti += a * tj;
tb += b * ta;
td += b * tc;
tf += b * te;
th += b * tg;
tj += b * ti;
ta += a * tb;
tc += a * td;
te += a * tf;
tg += a * th;
ti += a * tj;
tb += b * ta;
td += b * tc;
tf += b * te;
th += b * tg;
tj += b * ti;
}
timer.stop();
numFlops = 40.0 * N_rank * double(iters);
}
else {
timer.start();
for (int i=0; i < iters; ++i)
{
ta += a * tb;
tb += b * ta;
}
timer.stop();
numFlops = 4.0 * N_rank * double(iters);
}
optimizationSink(ta);
optimizationSink(tb);
optimizationSink(tc);
optimizationSink(td);
optimizationSink(te);
optimizationSink(tf);
optimizationSink(tg);
optimizationSink(th);
optimizationSink(ti);
optimizationSink(tj);
timer.stop();
float Gflops = numFlops / (1e9*timer.elapsed());
if (iters > 1)
{
cout << setw(5) << N_rank << '\t' << Gflops << endl;
}
}
double a = 0.3429843;
template<class T>
void optimizationSink(T&)
{
}
int main()
{
cout << "TinyVector<double,N> DAXPY benchmark" << endl
<< setw(5) << "N" << '\t' << "Gflops/" << Timer::indep_var() << endl;
tinyDAXPYBenchmark(TinyVector<double,1>(), 800000, a);
tinyDAXPYBenchmark(TinyVector<double,2>(), 800000, a);
tinyDAXPYBenchmark(TinyVector<double,3>(), 800000, a);
tinyDAXPYBenchmark(TinyVector<double,4>(), 700000, a);
tinyDAXPYBenchmark(TinyVector<double,5>(), 600000, a);
tinyDAXPYBenchmark(TinyVector<double,6>(), 500000, a);
tinyDAXPYBenchmark(TinyVector<double,7>(), 500000, a);
tinyDAXPYBenchmark(TinyVector<double,8>(), 500000, a);
tinyDAXPYBenchmark(TinyVector<double,9>(), 500000, a);
tinyDAXPYBenchmark(TinyVector<double,10>(), 500000, a);
return 0;
}
|
