Codebase list findbugs / 1598850e-d78d-47dc-9ac1-7bac0d7ec657/main src / doc / performance.html
1598850e-d78d-47dc-9ac1-7bac0d7ec657/main

Tree @1598850e-d78d-47dc-9ac1-7bac0d7ec657/main (Download .tar.gz)

performance.html @1598850e-d78d-47dc-9ac1-7bac0d7ec657/mainraw · history · blame

<html>
<head>
<title>FindBugs Performance Improvements and Regressions</title>
<link rel="stylesheet" type="text/css" href="findbugs.css">
@GOOGLE_ANALYTICS@
</head>
<body>

    <table width="100%">
        <tr>

            @HTML_SIDEBAR@

            <td align="left" valign="top">

                <h1>FindBugs Performance Improvements and Regressions</h1> I did a performance check against 179
                benchmarks applications I regularly test against. Overall (total the total time to analyze all 179
                benchmarks), FindBugs 2.0 gives a 9% performance improvement over 1.3.9. 154 of the 179 benchmarks saw
                performance improvements; 24 saw regressions. All of the benchmarks that saw regressions of more than
                10% were small benchmarks (analyzed in less than 60 seconds), which makes consistent benchmarking
                particularly difficult. I'm working to repeat the benchmarks, see if the results are consistent. I took
                a look, and couldn't find anything that stood out as being a performance glitch in FindBugs. I haven't
                yet done benchmarking with constrained memory. It is possible that you may need to increase the heap
                size for FindBugs 2.0.

                <h2>Important Request</h2>
                <p> If you are seeing any significant performance regressions in FindBugs 2.0,
                I very much need your help. Please either email <a href="mailto:findbugs@cs.umd.edu">findbugs@cs.umd.edu</a>
                or file <a href="http://sourceforge.net/p/findbugs/_list/tickets?source=navbar">a
                    bug report</a>.&nbsp;with the following information from the xml file for your project (from both the
                1.3.9 and 2.0.0 version if possible). Sending me your code or pointing me to a open source repository
                would be great, but I know that isn't feasible for a lot of projects. The information I'm requesting
                doesn't include any information about the code being analyzed other than the total size of the code
                being analyzed and the total number of issues found at the different confidence levels. The
                &lt;FindBugsSummary ... &gt; start tag. For example: <quote> <pre>
   &lt;FindBugsSummary timestamp="Tue, 30 Dec 2008 21:29:52 -0500"
      total_classes="206" referenced_classes="325" total_bugs="72" total_size="7654" num_packages="21"
      vm_version="20.4-b02-402" cpu_seconds="62.52" clock_seconds="22.01"
      peak_mbytes="112.21" alloc_mbytes="1683.38" gc_seconds="1.19"
      priority_3="56" priority_2="14" priority_1="2"&gt;
</pre> </quote> The &lt;FindBugsProfile&gt;...&lt;/FindBugsProfile&gt; element. For example: <quote>
                <pre>
   &lt;FindBugsProfile&gt;
      &lt;ClassProfile name="edu.umd.cs.findbugs.detect.IncompatMask" totalMilliseconds="11"
        invocations="206" avgMicrosecondsPerInvocation="55" maxMicrosecondsPerInvocation="475"
        standardDeviationMircosecondsPerInvocation="75"/&gt;
      &lt;ClassProfile name="edu.umd.cs.findbugs.detect.FindFinalizeInvocations" totalMilliseconds="11"
        invocations="206" avgMicrosecondsPerInvocation="55" maxMicrosecondsPerInvocation="402"
        standardDeviationMircosecondsPerInvocation="69"/&gt;
      &lt;ClassProfile name="edu.umd.cs.findbugs.classfile.engine.bcel.LockDataflowFactory" totalMilliseconds="11"
        invocations="23" avgMicrosecondsPerInvocation="515" maxMicrosecondsPerInvocation="2637"
        standardDeviationMircosecondsPerInvocation="639"/&gt;
   ...
 &lt;/FindBugsProfile&gt;
</pre> </quote> @HTML_FOOTER@

            </td>

        </tr>
    </table>

</body>
</html>