icuSources/test/perf/collationperf/CollPerf.pl

   1 #!/usr/bin/perl
   2 #  ********************************************************************
   3 #  * COPYRIGHT:
   4 #  * Copyright (c) 2002-2008, International Business Machines Corporation and
   5 #  * others. All Rights Reserved.
   6 #  ********************************************************************
   7
   8 require "../perldriver/Common.pl";
   9
  10 use lib '../perldriver';
  11
  12 my $p;
  13 if ($OnWindows) {
  14     $p = $ICUPathLatest . "/collationperf/Release/collationperf.exe";
  15 }
  16 else {
  17     $p = $ICUPathLatest . "/collationperf/collperf";
  18 }
  19
  20 my @locale = (
  21     "en_US",
  22     "da_DK",
  23     "de_DE",
  24     "fr_FR",
  25     "ja_JP",
  26     "ja_JP",
  27     "ja_JP",
  28     "ja_JP",
  29     "zh_CN",
  30     "zh_CN",
  31     "zh_CN",
  32     "zh_TW",
  33     "zh_TW",
  34     "ko_KR",
  35     "ko_KR",
  36     "ru_RU",
  37     "ru_RU",
  38     "th_TH",
  39     "th_TH"
  40 );
  41
  42 my $filePath   = $CollationDataPath . "/";
  43 my $filePrefix = "TestNames_";
  44 my @data       = (
  45     $filePrefix."Latin.txt",
  46     $filePrefix."Latin.txt",
  47     $filePrefix."Latin.txt",
  48     $filePrefix."Latin.txt",
  49     $filePrefix."Latin.txt",
  50     $filePrefix."Japanese_h.txt",
  51     $filePrefix."Japanese_k.txt",
  52     $filePrefix."Asian.txt",
  53     $filePrefix."Latin.txt",
  54     $filePrefix."Chinese.txt",
  55     $filePrefix."Simplified_Chinese.txt",
  56     $filePrefix."Latin.txt",
  57     $filePrefix."Chinese.txt",
  58     $filePrefix."Latin.txt",
  59     $filePrefix."Korean.txt",
  60     $filePrefix."Latin.txt",
  61     $filePrefix."Russian.txt",
  62     $filePrefix."Latin.txt",
  63     $filePrefix."Thai.txt"
  64 );
  65
  66 my @resultPER;
  67 my @resultFIN;
  68
  69 for ( $n = 0 ; $n < @data ; $n++ ) {
  70     my $resultICU;
  71     my $resultNIX;
  72     $resultICU = @locale[$n].",".@data[$n].",";
  73     $resultNIX = @locale[$n].",".@data[$n].",";
  74     @resultFIN[$n] = @locale[$n].",".@data[$n].",";
  75
  76     #quicksort
  77     my @icu = `$p -locale @locale[$n] -loop 1000 -file $filePath@data[$n] -qsort`;
  78     my @nix = `$p -locale @locale[$n] -unix -loop 1000 -file $filePath@data[$n] -qsort`;
  79
  80     my @icua = split( ' = ', $icu[2] );
  81     my @icub = split( ' ',   $icua[1] );
  82     my @nixa = split( ' = ', $nix[2] );
  83     my @nixb = split( ' ',   $nixa[1] );
  84
  85     $resultICU = $resultICU.$icub[0].",";
  86     $resultNIX = $resultNIX.$nixb[0].",";
  87
  88     #keygen time
  89     @icu = `$p -locale @locale[$n] -loop 1000 -file $filePath@data[$n] -keygen`;
  90     @nix = `$p -locale @locale[$n] -unix -loop 1000 -file $filePath@data[$n] -keygen`;
  91
  92     @icua = split( ' = ', $icu[2] );
  93     @icub = split( ' ',   $icua[1] );
  94     @nixa = split( ' = ', $nix[2] );
  95     @nixb = split( ' ',   $nixa[1] );
  96
  97     $resultICU = $resultICU.$icub[0].",";
  98     $resultNIX = $resultNIX.$nixb[0].",";
  99
 100     #keygen len
 101     @icua = split( ' = ', $icu[3] );
 102     @nixa = split( ' = ', $nix[3] );
 103
 104     chomp( @icua[1] );
 105     chomp( @nixa[1] );
 106
 107     $resultICU = $resultICU.$icua[1].",";
 108     $resultNIX = $resultNIX.$nixa[1].",";
 109
 110     my @resultSplitICU;
 111     my @resultSplitNIX;
 112
 113     #percent
 114     for ( $i = 0 ; $i < 3 ; $i++ ) {
 115         my $percent = 0;
 116         @resultSplitICU = split( ',', $resultICU );
 117         @resultSplitNIX = split( ',', $resultNIX );
 118         if ( @resultSplitICU[ 2 + $i ] > 0 ) {
 119             $percent = substr((((
 120                 @resultSplitNIX[ 2 + $i ] - @resultSplitICU[ 2 + $i ]) / @resultSplitICU[ 2 + $i ]) * 100),
 121                 0, 7);
 122         }
 123         @resultPER[$n] = @resultPER[$n].$percent."%,";
 124     }
 125
 126     #store ICU result
 127     for ( $j = 0 ; $j < 3 ; $j++ ) {
 128         @resultFIN[$n] = @resultFIN[$n].@resultSplitICU[ 2 + $j ].",";
 129     }
 130
 131     #store Unix result
 132     for ( $j = 0 ; $j < 3 ; $j++ ) {
 133         @resultFIN[$n] = @resultFIN[$n].@resultSplitNIX[ 2 + $j ].",";
 134     }
 135
 136     #store Percent result
 137     @resultFIN[$n] = @resultFIN[$n].@resultPER[$n];
 138 }
 139
 140 # Print the results in a HTML page
 141 printOutput();
 142
 143 exit(0);
 144
 145 # This subroutine creates the web page and prints out the results in a table
 146 sub printOutput {
 147     my $title = "Collation:  ICU " . $ICULatestVersion . " vs GLIBC";
 148     my $html  = localtime;
 149     $html =~ s/://g;         # ':' illegal
 150     $html =~ s/\s*\d+$//;    # delete year
 151     $html =~ s/^\w+\s*//;    # delete dow
 152     $html = "CollationPerformance $html.html";
 153     $html = "../results/" . $html;
 154     $html =~ s/ /_/g;
 155     open( HTML, ">$html" ) or die "Can't write to $html: $!";
 156     print HTML <<EOF;
 157 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
 158 "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
 159 <html xmlns="http://www.w3.org/1999/xhtml">
 160 <head>
 161 <meta http-equiv="Content-Type" content="text/html; charset=utf-8" />
 162 <title>Collation: ICU4C vs. glibc</title>
 163 <link rel="stylesheet" href="../icu.css" type="text/css" />
 164 </head>
 165 <body>
 166 <!--#include virtual="../ssi/header.html" -->
 167 EOF
 168
 169     print HTML "<h2>Collation: ICU4C ".$ICULatestVersion." vs. GLIBC</h2>\n";
 170
 171     print HTML <<EOF;
 172 <p>The performance test takes a locale and creates a RuleBasedCollator with
 173 default options. A large list of names is used as data in each test, where the
 174 names vary according to language. Each Collation operation over the whole list
 175 is repeated 1000 times. The percentage values in the final column are the most
 176 useful. They measure differences, where positive is better for ICU4C, and
 177 negative is better for the compared implementation.</p>
 178 <h3>Key</h3>
 179 <table border="1" cellspacing="0" cellpadding="4">
 180 <tr>
 181 <th align="left">Operation</th>
 182 <th align="left">Units</th>
 183 <th align="left">Description</th>
 184 </tr>
 185 <tr>
 186 <td>strcoll</td>
 187 <td>nanosecs</td>
 188 <td>Timing for string collation, an incremental compare of strings.</td>
 189 </tr>
 190 <tr>
 191 <td>keygen</td>
 192 <td>nanosecs</td>
 193 <td>Timing for generation of sort keys, used to 'precompile' information so
 194 that subsequent operations can use binary comparison.</td>
 195 </tr>
 196 <tr>
 197 <td>keylen</td>
 198 <td>bytes/char</td>
 199 <td>The average length of the generated sort keys, in bytes per character
 200 (Unicode/ISO 10646 code point). Generally this is the important field for sort
 201 key performance, since it directly impacts the time necessary for binary
 202 comparison, and the overhead of memory usage and retrieval time for sort
 203 keys.</td>
 204 </tr>
 205 </table>
 206 EOF
 207     printData();
 208
 209     print HTML <<EOF;
 210 <h3><i>Notes</i></h3>
 211 <ol>
 212 <li>As with all performance measurements, the results will vary according to
 213 the hardware and compiler. The strcoll operation is particularly sensitive; we
 214 have found that even slight changes in code alignment can produce 10%
 215 differences.</li>
 216 <li>For more information on incremental vs. sort key comparison, the importance
 217 of multi-level sorting, and other features of collation, see <a href=
 218 "http://www.unicode.org/reports/tr10/">Unicode Collation (UCA)</a>.</li>
 219 <li>For general information on ICU collation see <a href=
 220 "/userguide/Collate_Intro.html">User Guide</a>.</li>
 221 <li>For information on APIs, see <a href="/apiref/icu4c/ucol_8h.html">C</a>,
 222 <a href="/apiref/icu4c/classCollator.html">C++</a>, or <a href=
 223 "/apiref/icu4j/com/ibm/icu/text/Collator.html">Java</a>.</li>
 224 </ol>
 225 <!--#include virtual="../ssi/footer.html" -->
 226 </body>
 227 </html>
 228
 229 EOF
 230
 231     close(HTML) or die "Can't close $html: $!";
 232 }
 233
 234 # This subroutine formats and prints the table.
 235 sub printData() {
 236     print HTML <<EOF;
 237 <h3>Data</h3>
 238 <table border="1" cellspacing="0" cellpadding="4">
 239 <tr>
 240 <td align="left"><b>Locale</b></td>
 241 <td align="left"><b>Data file</b></td>
 242 <td align="left"><b>strcoll</b> <i>(ICU)</i></td>
 243 <td align="left"><b>keygen</b> <i>(ICU)</i></td>
 244 <td align="left"><b>keylen</b> <i>(ICU)</i></td>
 245 <td align="left"><b>strcoll</b> <i>(GLIBC)</i></td>
 246 <td align="left"><b>keygen</b> <i>(GLIBC)</i></td>
 247 <td align="left"><b>keylen</b> <i>(GLIBC)</i></td>
 248 <td align="left"><b>strcoll</b> <i>(GLIBC-ICU)/ICU)</i></td>
 249 <td align="left"><b>keygen</b> <i>(GLIBC-ICU)/ICU)</i></td>
 250 <td align="left"><b>keylen</b> <i>(GLIBC-ICU)/ICU)</i></td>
 251 </tr>
 252 EOF
 253
 254     for ( $n = 0 ; $n < @resultFIN ; $n++ ) {
 255         print HTML "<tr>";
 256         my @parsed = split( ',', @resultFIN[$n] );
 257         for ( $i = 0 ; $i < @parsed ; $i++ ) {
 258             my $value = @parsed[$i];
 259             print HTML "<td align=\"center\">";
 260
 261             if ( $value =~ m/^[-]/ ) {
 262                 print HTML "<font color=\"red\">$value</font>";
 263             }
 264             else {
 265                 print HTML "$value";
 266             }
 267
 268             print HTML "</td>";
 269
 270         }
 271         print HTML "</tr>\n";
 272     }
 273
 274     print HTML<<EOF;
 275 </table>
 276 EOF
 277 }