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
|
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.1 plus MathML 2.0//EN" "http://www.w3.org/Math/DTD/mathml2/xhtml-math11-f.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" lang="en-US">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=utf-8"/>
<meta name="viewport" content="width=device-width, initial-scale=0.4"/>
<meta name="google" content="notranslate" />
<link rel="canonical" href="https://sleef.org/benchmark.xhtml" />
<link rel="icon" href="favicon.png" />
<link rel="stylesheet" type="text/css" href="texlike.css"/>
<link rel="stylesheet" type="text/css" href="sleef.css"/>
<title>SLEEF - Benchmark Results</title>
</head>
<body translate="no" class="notranslate">
<h1>SLEEF Documentation - Benchmark Results</h1>
<h2>Table of contents</h2>
<ul class="none" style="font-family: arial, sansserif; padding-left: 0.5cm;">
<li><a class="underlined" href="index.xhtml">Introduction</a></li>
<li><a class="underlined" href="compile.xhtml">Compiling and installing the library</a></li>
<li><a class="underlined" href="purec.xhtml">Math library reference</a></li>
<li><a class="underlined" href="dft.xhtml">DFT library reference</a></li>
<li><a class="underlined" href="misc.xhtml">Other tools included in the package</a></li>
<li> </li>
<li><a class="underlined" href="benchmark.xhtml">Benchmark results</a></li>
<li> </li>
<li><a class="underlined" href="additional.xhtml">Additional notes</a></li>
</ul>
<h2>Benchmark results</h2>
<p class="noindent" style="font-size:1em; margin-top:0.5cm;">
These graphs show comparison of the execution time between
<a class="undecorated"
href="https://github.com/shibatch/sleef">SLEEF</a>-3.2 compiled
with GCC-7.2 and Intel SVML included in Intel C Compiler 18.0.1.
</p>
<p style="font-size:1em;">
The execution time of each function is measured by executing each
function 10^8 times and taking the average time. Each time a
function is executed, a uniformly distributed random number is set
to each element of the argument vector(each element is set a
different value.) The ranges of the random number for each
function are shown below. Argument vectors are generated before
the measurement, and the time to generate random argument vectors
is not included in the execution time.
</p>
<br/>
<ul>
<li>Trigonometric functions : [0, 6.28] and [0, 10^6] for
double-precision functions. [0, 6.28] and [0, 30000] for
single-precision functions.</li>
<li>Log : [0, 10^300] and [0, 10^38] for double-precision
functions and single-precision functions, respectively.</li>
<li>Exp : [-700, 700] and [-100, 100] for double-precision
functions and single-precision functions, respectively.</li>
<li>Pow : [-30, 30] for both the first and the second
arguments.</li>
<li>Asin : [-1, 1]</li>
<li>Atan : [-10, 10]</li>
<li>Atan2 : [-10, 10] for both the first and the second
arguments.</li>
</ul>
<br/>
<p style="font-size:1em;">
The accuracy of SVML functions can be chosen by compiler options,
not the function names. "-fimf-max-error=1.0" option is specified
to icc to obtain the 1-ulp-accuracy results, and
"-fimf-max-error=5.0" option is used for the 5-ulp-accuracy
results.
</p>
<p style="font-size:1em;">
Those results are measured on a PC with Intel Core i7-6700 CPU @
3.40GHz with Turbo Boost turned off. The CPU should be always
running at 3.4GHz during the measurement.
</p>
<p style="font-size:1.2em; margin-top:1.0cm;">
<b style="color:#0050a0;">Click graphs to magnify</b>.
</p>
<p style="margin-bottom:1cm;"> </p>
<p style="text-align:center; margin-bottom:2cm;">
<a class="nothing" href="trigdp.png">
<img src="trigdp.png" alt="Performance graph for DP trigonometric functions" width="50%" height="50%"/>
</a>
<br />
Fig. 6.1: Execution time of double precision trigonometric functions
</p>
<p style="text-align:center; margin-bottom:2cm;">
<a class="nothing" href="trigsp.png">
<img src="trigsp.png" alt="Performance graph for SP trigonometric functions" width="50%" height="50%"/>
</a>
<br />
Fig. 6.2: Execution time of single precision trigonometric functions
</p>
<p style="text-align:center; margin-bottom:2cm;">
<a class="nothing" href="nontrigdp.png">
<img src="nontrigdp.png" alt="Performance graph for other DP functions" width="50%" height="50%"/>
</a>
<br />
Fig. 6.3: Execution time of double precision log, exp, pow and inverse trigonometric functions
</p>
<p style="text-align:center; margin-bottom:2cm;">
<a class="nothing" href="nontrigsp.png">
<img src="nontrigsp.png" alt="Performance graph for other SP functions" width="50%" height="50%"/>
</a>
<br />
Fig. 6.4: Execution time of single precision log, exp, pow and inverse trigonometric functions
</p>
<p class="footer">
Copyright © <!--YEAR--> SLEEF Project.<br/>
SLEEF is open-source software and is distributed under the Boost Software License, Version 1.0.
</p>
<!--TEST-->
</body>
</html>
|
