1<!doctype html PUBLIC "-//W3C//DTD HTML 3.2 Final//EN"> 2<html> 3<head> 4<meta http-equiv="content-type" content="text/html; charset=iso-8859-1"> 5<meta http-equiv="content-style-type" content="text/css"> 6<link rel="stylesheet" type="text/css" href="style.css"> 7<title>ProGuard Results</title> 8</head> 9<body> 10 11<h2>Results</h2> 12 13<b>ProGuard</b> successfully processes any Java bytecode, ranging from small 14midlets to entire run-time libraries. It primarily reduces the size of the 15processed code, with some potential increase in efficiency as an added bonus. 16The improvements obviously depend on the original code. The table below 17presents some typical results: 18<p> 19 20<table> 21 22<tr> 23<th width="28%">Input Program</th> 24<th width="12%">Original size</th> 25<th width="12%">After shrinking</th> 26<th width="12%">After optim.</th> 27<th width="12%">After obfusc.</th> 28<th width="12%">Total reduction</th> 29<th width="12%">Time</th> 30<th width="12%">Memory usage</th> 31</tr> 32 33<tr> 34<td><a target="other" href="http://java.sun.com/j2me/">Worm</a>, a sample midlet from Sun's JME</td> 35<td align="center">10.3 K</td> 36<td align="center">9.8 K</td> 37<td align="center">9.6 K</td> 38<td align="center">8.5 K</td> 39<td align="center">18 %</td> 40<td align="center">2 s</td> 41<td align="center">19 M</td> 42</tr> 43 44<tr> 45<td><a target="other" href="http://www.javadocking.com/">Javadocking</a>, a docking library</td> 46<td align="center">290 K</td> 47<td align="center">281 K</td> 48<td align="center">270 K</td> 49<td align="center">201 K</td> 50<td align="center">30 %</td> 51<td align="center">12 s</td> 52<td align="center">32 M</td> 53</tr> 54 55<tr> 56<td><b>ProGuard</b> itself</td> 57<td align="center">648 K</td> 58<td align="center">579 K</td> 59<td align="center">557 K</td> 60<td align="center">348 K</td> 61<td align="center">46 %</td> 62<td align="center">28 s</td> 63<td align="center">66 M</td> 64</tr> 65 66<tr> 67<td><a target="other" href="http://www.clarkware.com/software/JDepend.html">JDepend</a>, a Java quality metrics tool</td> 68<td align="center">57 K</td> 69<td align="center">36 K</td> 70<td align="center">33 K</td> 71<td align="center">28 K</td> 72<td align="center">51 %</td> 73<td align="center">6 s</td> 74<td align="center">24 M</td> 75</tr> 76 77<tr> 78<td><a target="other" href="http://java.sun.com/javase/6/">the run-time classes</a> from Sun's Java 6</td> 79<td align="center">53 M</td> 80<td align="center">23 M</td> 81<td align="center">22 M</td> 82<td align="center">18 M</td> 83<td align="center">66 %</td> 84<td align="center">16 min</td> 85<td align="center">270 M</td> 86</tr> 87 88<tr> 89<td><a target="other" href="http://jakarta.apache.org/tomcat/index.html">Tomcat</a>, the Apache servlet container</td> 90<td align="center">1.1 M</td> 91<td align="center">466 K</td> 92<td align="center">426 K</td> 93<td align="center">295 K</td> 94<td align="center">74 %</td> 95<td align="center">17 s</td> 96<td align="center">44 M</td> 97</tr> 98 99<tr> 100<td><a target="other" href="http://www.kclee.com/clemens/java/javancss/">JavaNCSS</a>, a Java source metrics tool</td> 101<td align="center">632 K</td> 102<td align="center">242 K</td> 103<td align="center">212 K</td> 104<td align="center">152 K</td> 105<td align="center">75 %</td> 106<td align="center">20 s</td> 107<td align="center">36 M</td> 108</tr> 109 110<tr> 111<td><a target="other" href="http://ant.apache.org/">Ant</a>, the Apache build tool</td> 112<td align="center">2.4 M</td> 113<td align="center">401 K</td> 114<td align="center">325 K</td> 115<td align="center">242 K</td> 116<td align="center">90 %</td> 117<td align="center">23 s</td> 118<td align="center">61 M</td> 119</tr> 120 121</table> 122<p> 123Results were measured with ProGuard 4.0 on a 2.6 GHz Pentium 4 with 512 MB 124of memory, using Sun JDK 1.5.0 in Fedora Core 3 Linux. 125<p> 126The program sizes include companion libraries. The shrinking step produces the 127best results for programs that use only small parts of their libraries. The 128obfuscation step can significantly shrink large programs even further, since 129the identifiers of their many internal references can be replaced by short 130identifiers. 131<p> 132The Java 6 run-time classes are the most complex example. The classes perform 133a lot of introspection, interacting with the native code of the virtual 134machine. The 1500+ lines of configuration were largely composed by automated 135analysis, complemented by a great deal of trial and error. The configuration 136is probably not complete, but the resulting library successfully serves as a 137run-time environment for running applications like ProGuard and the ProGuard 138GUI. 139<p> 140For small inputs, timings are governed by the reading and parsing of the jars. 141For large inputs, the optimization step becomes more important. For instance, 142processing the Java 6 run-time classes without optimization only takes 2 143minutes. 144<p> 145Memory usage (the amount of physical memory used by ProGuard while processing) 146is governed by the basic java virtual machine and by the total size of the 147library jars and program jars. 148<hr> 149<address> 150Copyright © 2002-2009 151<a href="http://www.graphics.cornell.edu/~eric/">Eric Lafortune</a>. 152</address> 153 154</body> 155</html> 156