blob: 128d9f78091227f7d988a6d98232cfa816c4a3b9 [file] [log] [blame]
<?xml version="1.0" encoding="ISO-8859-1"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd">
<html xmlns="http://www.w3.org/1999/xhtml">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=ISO-8859-1" />
<title>OProfile manual</title>
<meta name="generator" content="DocBook XSL Stylesheets V1.75.2" />
</head>
<body>
<div class="book" title="OProfile manual">
<div class="titlepage">
<div>
<div>
<h1 class="title"><a id="oprofile-guide"></a>OProfile manual</h1>
</div>
<div>
<div class="authorgroup">
<div class="author">
<h3 class="author"><span class="firstname">John</span> <span class="surname">Levon</span></h3>
<div class="affiliation">
<div class="address">
<p>
<code class="email">&lt;<a class="email" href="mailto:levon@movementarian.org">levon@movementarian.org</a>&gt;</code>
</p>
</div>
</div>
</div>
</div>
</div>
<div>
<p class="copyright">Copyright © 2000-2004 Victoria University of Manchester, John Levon and others</p>
</div>
</div>
<hr />
</div>
<div class="toc">
<p>
<b>Table of Contents</b>
</p>
<dl>
<dt>
<span class="chapter">
<a href="#introduction">1. Introduction</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect1">
<a href="#applications">1. Applications of OProfile</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#jitsupport">1.1. Support for dynamically compiled (JIT) code</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#requirements">2. System requirements</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#resources">3. Internet resources</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#install">4. Installation</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#uninstall">5. Uninstalling OProfile</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="chapter">
<a href="#overview">2. Overview</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect1">
<a href="#getting-started">1. Getting started</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#tools-overview">2. Tools summary</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="chapter">
<a href="#controlling">3. Controlling the profiler</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect1">
<a href="#controlling-daemon">1. Using <span class="command"><strong>opcontrol</strong></span></a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#opcontrolexamples">1.1. Examples</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#eventspec">1.2. Specifying performance counter events</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#setup-jit">2. Setting up the JIT profiling feature</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#setup-jit-jvm">2.1. JVM instrumentation</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#oprofile-gui">3. Using <span class="command"><strong>oprof_start</strong></span></a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#detailed-parameters">4. Configuration details</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#hardware-counters">4.1. Hardware performance counters</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#rtc">4.2. OProfile in RTC mode</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#timer">4.3. OProfile in timer interrupt mode</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#p4">4.4. Pentium 4 support</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#ia64">4.5. Intel Itanium 2 support</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#ppc64">4.6. PowerPC64 support</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#cell-be">4.7. Cell Broadband Engine support</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#amd-ibs-support">4.8. AMD64 (x86_64) Instruction-Based Sampling (IBS) support</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#misuse">4.9. Dangerous counter settings</a>
</span>
</dt>
</dl>
</dd>
</dl>
</dd>
<dt>
<span class="chapter">
<a href="#results">4. Obtaining results</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect1">
<a href="#profile-spec">1. Profile specifications</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#profile-spec-examples">1.1. Examples</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#profile-spec-details">1.2. Profile specification parameters</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#locating-and-managing-binary-images">1.3. Locating and managing binary images</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#no-results">1.4. What to do when you don't get any results</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#opreport">2. Image summaries and symbol summaries (<span class="command"><strong>opreport</strong></span>)</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#opreport-merging">2.1. Merging separate profiles</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opreport-comparison">2.2. Side-by-side multiple results</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opreport-callgraph">2.3. Callgraph output</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opreport-diff">2.4. Differential profiles with <span class="command"><strong>opreport</strong></span></a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opreport-anon">2.5. Anonymous executable mappings</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opreport-xml">2.6. XML formatted output</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opreport-options">2.7. Options for <span class="command"><strong>opreport</strong></span></a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#opannotate">3. Outputting annotated source (<span class="command"><strong>opannotate</strong></span>)</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#opannotate-finding-source">3.1. Locating source files</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opannotate-details">3.2. Usage of <span class="command"><strong>opannotate</strong></span></a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#getting-jit-reports">4. OProfile results with JIT samples</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#opgprof">5. <span class="command"><strong>gprof</strong></span>-compatible output (<span class="command"><strong>opgprof</strong></span>)</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#opgprof-details">5.1. Usage of <span class="command"><strong>opgprof</strong></span></a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#oparchive">6. Archiving measurements (<span class="command"><strong>oparchive</strong></span>)</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#oparchive-details">6.1. Usage of <span class="command"><strong>oparchive</strong></span></a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#opimport">7. Converting sample database files (<span class="command"><strong>opimport</strong></span>)</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#opimport-details">7.1. Usage of <span class="command"><strong>opimport</strong></span></a>
</span>
</dt>
</dl>
</dd>
</dl>
</dd>
<dt>
<span class="chapter">
<a href="#interpreting">5. Interpreting profiling results</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect1">
<a href="#irq-latency">1. Profiling interrupt latency</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#kernel-profiling">2. Kernel profiling</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#irq-masking">2.1. Interrupt masking</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#idle">2.2. Idle time</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#kernel-modules">2.3. Profiling kernel modules</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#interpreting-callgraph">3. Interpreting call-graph profiles</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#debug-info">4. Inaccuracies in annotated source</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#effect-of-optimizations">4.1. Side effects of optimizations</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#prologues">4.2. Prologues and epilogues</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#inlined-function">4.3. Inlined functions</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#wrong-linenr-info">4.4. Inaccuracy in line number information</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#symbol-without-debug-info">5. Assembly functions</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#overlapping-symbols">6. Overlapping symbols in JITed code</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#hidden-cost">7. Other discrepancies</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="chapter">
<a href="#ack">6. Acknowledgments</a>
</span>
</dt>
</dl>
</div>
<div class="chapter" title="Chapter 1. Introduction">
<div class="titlepage">
<div>
<div>
<h2 class="title"><a id="introduction"></a>Chapter 1. Introduction</h2>
</div>
</div>
</div>
<div class="toc">
<p>
<b>Table of Contents</b>
</p>
<dl>
<dt>
<span class="sect1">
<a href="#applications">1. Applications of OProfile</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#jitsupport">1.1. Support for dynamically compiled (JIT) code</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#requirements">2. System requirements</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#resources">3. Internet resources</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#install">4. Installation</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#uninstall">5. Uninstalling OProfile</a>
</span>
</dt>
</dl>
</div>
<p>
This manual applies to OProfile version 0.9.7-rc2.
OProfile is a profiling system for Linux 2.2/2.4/2.6 systems on a number of architectures. It is capable of profiling
all parts of a running system, from the kernel (including modules and interrupt handlers) to shared libraries
to binaries. It runs transparently in the background collecting information at a low overhead. These
features make it ideal for profiling entire systems to determine bottle necks in real-world systems.
</p>
<p>
Many CPUs provide "performance counters", hardware registers that can count "events"; for example,
cache misses, or CPU cycles. OProfile provides profiles of code based on the number of these occurring events:
repeatedly, every time a certain (configurable) number of events has occurred, the PC value is recorded.
This information is aggregated into profiles for each binary image.</p>
<p>
Some hardware setups do not allow OProfile to use performance counters: in these cases, no
events are available, and OProfile operates in timer/RTC mode, as described in later chapters.
</p>
<div class="sect1" title="1. Applications of OProfile">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="applications"></a>1. Applications of OProfile</h2>
</div>
</div>
</div>
<p>
OProfile is useful in a number of situations. You might want to use OProfile when you :
</p>
<div class="itemizedlist">
<ul class="itemizedlist" type="disc">
<li class="listitem">
<p>need low overhead</p>
</li>
<li class="listitem">
<p>cannot use highly intrusive profiling methods</p>
</li>
<li class="listitem">
<p>need to profile interrupt handlers</p>
</li>
<li class="listitem">
<p>need to profile an application and its shared libraries</p>
</li>
<li class="listitem">
<p>need to profile dynamically compiled code of supported virtual machines (see <a class="xref" href="#jitsupport" title="1.1. Support for dynamically compiled (JIT) code">Section 1.1, &#8220;Support for dynamically compiled (JIT) code&#8221;</a>)</p>
</li>
<li class="listitem">
<p>need to capture the performance behaviour of entire system</p>
</li>
<li class="listitem">
<p>want to examine hardware effects such as cache misses</p>
</li>
<li class="listitem">
<p>want detailed source annotation</p>
</li>
<li class="listitem">
<p>want instruction-level profiles</p>
</li>
<li class="listitem">
<p>want call-graph profiles</p>
</li>
</ul>
</div>
<p>
OProfile is not a panacea. OProfile might not be a complete solution when you :
</p>
<div class="itemizedlist">
<ul class="itemizedlist" type="disc">
<li class="listitem">
<p>require call graph profiles on platforms other than 2.6/x86</p>
</li>
<li class="listitem">
<p>don't have root permissions</p>
</li>
<li class="listitem">
<p>require 100% instruction-accurate profiles</p>
</li>
<li class="listitem">
<p>need function call counts or an interstitial profiling API</p>
</li>
<li class="listitem">
<p>cannot tolerate any disturbance to the system whatsoever</p>
</li>
<li class="listitem">
<p>need to profile interpreted or dynamically compiled code of non-supported virtual machines</p>
</li>
</ul>
</div>
<div class="sect2" title="1.1. Support for dynamically compiled (JIT) code">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="jitsupport"></a>1.1. Support for dynamically compiled (JIT) code</h3>
</div>
</div>
</div>
<p>
Older versions of OProfile were not capable of attributing samples to symbols from dynamically
compiled code, i.e. "just-in-time (JIT) code". Typical JIT compilers load the JIT code into
anonymous memory regions. OProfile reported the samples from such code, but the attribution
provided was simply:
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">"anon: &lt;tgid&gt;&lt;address range&gt;" </pre>
</td>
</tr>
</table>
<p>
Due to this limitation, it wasn't possible to profile applications executed by virtual machines (VMs)
like the Java Virtual Machine. OProfile now contains an infrastructure to support JITed code.
A development library is provided to allow developers
to add support for any VM that produces dynamically compiled code (see the <span class="emphasis"><em>OProfile JIT agent
developer guide</em></span>).
In addition, built-in support is included for the following:</p>
<div class="itemizedlist">
<ul class="itemizedlist" type="disc">
<li class="listitem">JVMTI agent library for Java (1.5 and higher)</li>
<li class="listitem">JVMPI agent library for Java (1.5 and lower)</li>
</ul>
</div>
<p>
For information on how to use OProfile's JIT support, see <a class="xref" href="#setup-jit" title="2. Setting up the JIT profiling feature">Section 2, &#8220;Setting up the JIT profiling feature&#8221;</a>.
</p>
</div>
</div>
<div class="sect1" title="2. System requirements">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="requirements"></a>2. System requirements</h2>
</div>
</div>
</div>
<div class="variablelist">
<dl>
<dt>
<span class="term">Linux kernel 2.2/2.4/2.6</span>
</dt>
<dd>
<p>
OProfile uses a kernel module that can be compiled for
2.2.11 or later and 2.4. 2.4.10 or above is required if you use the
boot-time kernel option <code class="option">nosmp</code>. 2.6 kernels are supported with the in-kernel
OProfile driver. Note that only 32-bit x86 and IA64 are supported on 2.2/2.4 kernels.
</p>
<p>
2.6 kernels are strongly recommended. Under 2.4, OProfile may cause system crashes if power
management is used, or the BIOS does not correctly deal with local APICs.
</p>
<p>
To use OProfile's JIT support, a kernel version 2.6.13 or later is required.
In earlier kernel versions, the anonymous memory regions are not reported to OProfile and results
in profiling reports without any samples in these regions.
</p>
<p>
PPC64 processors (Power4/Power5/PPC970, etc.) require a recent (&gt; 2.6.5) kernel with the line
<code class="constant">#define PV_970</code> present in <code class="filename">include/asm-ppc64/processor.h</code>.
</p>
<p>
Profiling the Cell Broadband Engine PowerPC Processing Element (PPE) requires a kernel version
of 2.6.18 or more recent.
Profiling the Cell Broadband Engine Synergistic Processing Element (SPE) requires a kernel version
of 2.6.22 or more recent. Additionally, full support of SPE profiling requires a BFD library
from binutils code dated January 2007 or later. To ensure the proper BFD support exists, run
the <code class="code">configure</code> utility with <code class="code">--with-target=cell-be</code>.
Profiling the Cell Broadband Engine using SPU events requires a kernel version of 2.6.29-rc1
or more recent.
</p>
<div class="note" title="Note" style="margin-left: 0.5in; margin-right: 0.5in;"><h3 class="title">Note</h3>Attempting to profile SPEs with kernel versions older than 2.6.22 may cause the
system to crash.</div>
<p>
</p>
<p>
Instruction-Based Sampling (IBS) profile on AMD family10h processors requires
kernel version 2.6.28-rc2 or later.
</p>
</dd>
<dt>
<span class="term">modutils 2.4.6 or above</span>
</dt>
<dd>
<p>
You should have installed modutils 2.4.6 or higher (in fact earlier versions work well in almost all
cases).
</p>
</dd>
<dt>
<span class="term">Supported architecture</span>
</dt>
<dd>
<p>
For Intel IA32, a CPU with either a P6 generation or Pentium 4 core is
required. In marketing terms this translates to anything
between an Intel Pentium Pro (not Pentium Classics) and
a Pentium 4 / Xeon, including all Celerons. The AMD
Athlon, Opteron, Phenom, and Turion CPUs are also supported. Other IA32
CPU types only support the RTC mode of OProfile; please
see later in this manual for details. Hyper-threaded Pentium IVs
are not supported in 2.4. For 2.4 kernels, the Intel
IA-64 CPUs are also supported. For 2.6 kernels, there is additionally
support for Alpha processors, MIPS, ARM, x86-64, sparc64, ppc64, AVR32, and,
in timer mode, PA-RISC and s390.
</p>
</dd>
<dt>
<span class="term">Uniprocessor or SMP</span>
</dt>
<dd>
<p>
SMP machines are fully supported.
</p>
</dd>
<dt>
<span class="term">Required libraries</span>
</dt>
<dd>
<p>
These libraries are required : <code class="filename">popt</code>, <code class="filename">bfd</code>,
<code class="filename">liberty</code> (debian users: libiberty is provided in binutils-dev package), <code class="filename">dl</code>,
plus the standard C++ libraries.
</p>
</dd>
<dt>
<span class="term">Required user account</span>
</dt>
<dd>
<p>
For secure processing of sample data from JIT virtual machines (e.g., Java),
the special user account "oprofile" must exist on the system. The 'configure'
and 'make install' operations will print warning messages if this
account is not found. If you intend to profile JITed code, you must create
a group account named 'oprofile' and then create the 'oprofile' user account,
setting the default group to 'oprofile'. A runtime error message is printed to
the oprofile daemon log when processing JIT samples if this special user
account cannot be found.
</p>
</dd>
<dt>
<span class="term">OProfile GUI</span>
</dt>
<dd>
<p>
The use of the GUI to start the profiler requires the <code class="filename">Qt</code> library.
Either <code class="filename">Qt 3</code> or <code class="filename">Qt 4</code> should work.
</p>
</dd>
<dt>
<span class="term">
<acronym class="acronym">ELF</acronym>
</span>
</dt>
<dd>
<p>
Probably not too strenuous a requirement, but older <acronym class="acronym">A.OUT</acronym> binaries/libraries are not supported.
</p>
</dd>
<dt>
<span class="term">K&amp;R coding style</span>
</dt>
<dd>
<p>
OK, so it's not really a requirement, but I wish it was...
</p>
</dd>
</dl>
</div>
</div>
<div class="sect1" title="3. Internet resources">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="resources"></a>3. Internet resources</h2>
</div>
</div>
</div>
<div class="variablelist">
<dl>
<dt>
<span class="term">Web page</span>
</dt>
<dd>
<p>
There is a web page (which you may be reading now) at
<a class="ulink" href="http://oprofile.sf.net/">http://oprofile.sf.net/</a>.
</p>
</dd>
<dt>
<span class="term">Download</span>
</dt>
<dd>
<p>
You can download a source tarball or check out code from
the code repository at the sourceforge page,
<a class="ulink" href="http://sf.net/projects/oprofile/">http://sf.net/projects/oprofile/</a>.
</p>
</dd>
<dt>
<span class="term">Mailing list</span>
</dt>
<dd>
<p>
There is a low-traffic OProfile-specific mailing list, details at
<a class="ulink" href="http://sf.net/mail/?group_id=16191">http://sf.net/mail/?group_id=16191</a>.
</p>
</dd>
<dt>
<span class="term">Bug tracker</span>
</dt>
<dd>
<p>
There is a bug tracker for OProfile at SourceForge,
<a class="ulink" href="http://sf.net/tracker/?group_id=16191&amp;atid=116191">http://sf.net/tracker/?group_id=16191&amp;atid=116191</a>.
</p>
</dd>
<dt>
<span class="term">IRC channel</span>
</dt>
<dd>
<p>
Several OProfile developers and users sometimes hang out on channel <span class="command"><strong>#oprofile</strong></span>
on the <a class="ulink" href="http://oftc.net">OFTC</a> network.
</p>
</dd>
</dl>
</div>
</div>
<div class="sect1" title="4. Installation">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="install"></a>4. Installation</h2>
</div>
</div>
</div>
<p>
First you need to build OProfile and install it. <span class="command"><strong>./configure</strong></span>, <span class="command"><strong>make</strong></span>, <span class="command"><strong>make install</strong></span>
is often all you need, but note these arguments to <span class="command"><strong>./configure</strong></span> :
</p>
<div class="variablelist">
<dl>
<dt>
<span class="term">
<code class="option">--with-linux</code>
</span>
</dt>
<dd>
<p>
Use this option to specify the location of the kernel source tree you wish
to compile against. The kernel module is built against this source and
will only work with a running kernel built from the same source with
exact same options, so it is important you specify this option if you need
to.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--with-java</code>
</span>
</dt>
<dd>
<p>
Use this option if you need to profile Java applications. Also, see
<a class="xref" href="#requirements" title="2. System requirements">Section 2, &#8220;System requirements&#8221;</a>, "Required user account". This option
is used to specify the location of the Java Development Kit (JDK)
source tree you wish to use. This is necessary to get the interface description
of the JVMPI (or JVMTI) interface to compile the JIT support code successfully.
</p>
<div class="note" title="Note" style="margin-left: 0.5in; margin-right: 0.5in;">
<h3 class="title">Note</h3>
<p>
The Java Runtime Environment (JRE) does not include the development
files that are required to compile the JIT support code, so the full
JDK must be installed in order to use this option.
</p>
</div>
<p>
By default, the Oprofile JIT support libraries will be installed in
<code class="filename">&lt;oprof_install_dir&gt;/lib/oprofile</code>. To build
and install OProfile and the JIT support libraries as 64-bit, you can
do something like the following:
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">
# CFLAGS="-m64" CXXFLAGS="-m64" ./configure \
--with-kernel-support --with-java={my_jdk_installdir} \
--libdir=/usr/local/lib64
</pre>
</td>
</tr>
</table>
<p>
</p>
<div class="note" title="Note" style="margin-left: 0.5in; margin-right: 0.5in;">
<h3 class="title">Note</h3>
<p>
If you encounter errors building 64-bit, you should
install libtool 1.5.26 or later since that release of
libtool fixes known problems for certain platforms.
If you install libtool into a non-standard location,
you'll need to edit the invocation of 'aclocal' in
OProfile's autogen.sh as follows (assume an install
location of /usr/local):
</p>
<p>
<code class="code">aclocal -I m4 -I /usr/local/share/aclocal</code>
</p>
</div>
</dd>
<dt>
<span class="term">
<code class="option">--with-kernel-support</code>
</span>
</dt>
<dd>
<p>
Use this option with 2.6 and above kernels to indicate the
kernel provides the OProfile device driver.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--with-qt-dir/includes/libraries</code>
</span>
</dt>
<dd>
<p>
Specify the location of Qt headers and libraries. It defaults to searching in
<code class="constant">$QTDIR</code> if these are not specified.
</p>
</dd>
<dt>
<a id="disable-werror"></a>
<span class="term">
<code class="option">--disable-werror</code>
</span>
</dt>
<dd>
<p>
Development versions of OProfile build by
default with <code class="option">-Werror</code>. This option turns
<code class="option">-Werror</code> off.
</p>
</dd>
<dt>
<a id="disable-optimization"></a>
<span class="term">
<code class="option">--disable-optimization</code>
</span>
</dt>
<dd>
<p>
Disable the <code class="option">-O2</code> compiler flag
(useful if you discover an OProfile bug and want to give a useful
back-trace etc.)
</p>
</dd>
</dl>
</div>
<p>
You'll need to have a configured kernel source for the current kernel
to build the module for 2.4 kernels. Since all distributions provide different kernels it's unlikely the running kernel match the configured source
you installed. The safest way is to recompile your own kernel, run it and compile oprofile. It is also recommended that if you have a
uniprocessor machine, you enable the local APIC / IO_APIC support for
your kernel (this is automatically enabled for SMP kernels). With many BIOS, kernel &gt;= 2.6.9 and UP kernel it's not sufficient to enable the local APIC you must also turn it on explicitly at boot time by providing "lapic" option to the kernel. On
machines with power management, such as laptops, the power management
must be turned off when using OProfile with 2.4 kernels. The power management software
in the BIOS cannot handle the non-maskable interrupts (NMIs) used by
OProfile for data collection. If you use the NMI watchdog, be aware that
the watchdog is disabled when profiling starts, and not re-enabled until the
OProfile module is removed (or, in 2.6, when OProfile is not running). If you compile OProfile for
a 2.2 kernel you must be root to compile the module. If you are using
2.6 kernels or higher, you do not need kernel source, as long as the
OProfile driver is enabled; additionally, you should not need to disable
power management.
</p>
<p>
Please note that you must save or have available the <code class="filename">vmlinux</code> file
generated during a kernel compile, as OProfile needs it (you can use
<code class="option">--no-vmlinux</code>, but this will prevent kernel profiling).
</p>
</div>
<div class="sect1" title="5. Uninstalling OProfile">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="uninstall"></a>5. Uninstalling OProfile</h2>
</div>
</div>
</div>
<p>
You must have the source tree available to uninstall OProfile; a <span class="command"><strong>make uninstall</strong></span> will
remove all installed files except your configuration file in the directory <code class="filename">~/.oprofile</code>.
</p>
</div>
</div>
<div class="chapter" title="Chapter 2. Overview">
<div class="titlepage">
<div>
<div>
<h2 class="title"><a id="overview"></a>Chapter 2. Overview</h2>
</div>
</div>
</div>
<div class="toc">
<p>
<b>Table of Contents</b>
</p>
<dl>
<dt>
<span class="sect1">
<a href="#getting-started">1. Getting started</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#tools-overview">2. Tools summary</a>
</span>
</dt>
</dl>
</div>
<div class="sect1" title="1. Getting started">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="getting-started"></a>1. Getting started</h2>
</div>
</div>
</div>
<p>
Before you can use OProfile, you must set it up. The minimum setup required for this
is to tell OProfile where the <code class="filename">vmlinux</code> file corresponding to the
running kernel is, for example :
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">opcontrol --vmlinux=/boot/vmlinux-`uname -r`</pre>
</td>
</tr>
</table>
<p>
If you don't want to profile the kernel itself,
you can tell OProfile you don't have a <code class="filename">vmlinux</code> file :
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">opcontrol --no-vmlinux</pre>
</td>
</tr>
</table>
<p>
Now we are ready to start the daemon (<span class="command"><strong>oprofiled</strong></span>) which collects
the profile data :
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">opcontrol --start</pre>
</td>
</tr>
</table>
<p>
When I want to stop profiling, I can do so with :
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">opcontrol --shutdown</pre>
</td>
</tr>
</table>
<p>
Note that unlike <span class="command"><strong>gprof</strong></span>, no instrumentation (<code class="option">-pg</code>
and <code class="option">-a</code> options to <span class="command"><strong>gcc</strong></span>)
is necessary.
</p>
<p>
Periodically (or on <span class="command"><strong>opcontrol --shutdown</strong></span> or <span class="command"><strong>opcontrol --dump</strong></span>)
the profile data is written out into the $SESSION_DIR/samples directory (by default at <code class="filename">/var/lib/oprofile/samples</code>).
These profile files cover shared libraries, applications, the kernel (vmlinux), and kernel modules.
You can clear the profile data (at any time) with <span class="command"><strong>opcontrol --reset</strong></span>.
</p>
<p>
To place these sample database files in a specific directory instead of the default location (<code class="filename">/var/lib/oprofile</code>) use the <code class="option">--session-dir=dir</code> option. You must also specify the <code class="option">--session-dir</code> to tell the tools to continue using this directory. (In the future, we should allow this to be specified in an environment variable.) :
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">opcontrol --no-vmlinux --session-dir=/home/me/tmpsession</pre>
</td>
</tr>
</table>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">opcontrol --start --session-dir=/home/me/tmpsession</pre>
</td>
</tr>
</table>
<p>
You can get summaries of this data in a number of ways at any time. To get a summary of
data across the entire system for all of these profiles, you can do :
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">opreport [--session-dir=dir]</pre>
</td>
</tr>
</table>
<p>
Or to get a more detailed summary, for a particular image, you can do something like :
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">opreport -l /boot/vmlinux-`uname -r`</pre>
</td>
</tr>
</table>
<p>
There are also a number of other ways of presenting the data, as described later in this manual.
Note that OProfile will choose a default profiling setup for you. However, there are a number
of options you can pass to <span class="command"><strong>opcontrol</strong></span> if you need to change something,
also detailed later.
</p>
</div>
<div class="sect1" title="2. Tools summary">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="tools-overview"></a>2. Tools summary</h2>
</div>
</div>
</div>
<p>
This section gives a brief description of the available OProfile utilities and their purpose.
</p>
<div class="variablelist">
<dl>
<dt>
<span class="term">
<code class="filename">ophelp</code>
</span>
</dt>
<dd>
<p>
This utility lists the available events and short descriptions.
</p>
</dd>
<dt>
<span class="term">
<code class="filename">opcontrol</code>
</span>
</dt>
<dd>
<p>
Used for controlling the OProfile data collection, discussed in <a class="xref" href="#controlling" title="Chapter 3. Controlling the profiler">Chapter 3, <i>Controlling the profiler</i></a>.
</p>
</dd>
<dt>
<span class="term">
<code class="filename">agent libraries</code>
</span>
</dt>
<dd>
<p>
Used by virtual machines (like the Java VM) to record information about JITed code being profiled. See <a class="xref" href="#setup-jit" title="2. Setting up the JIT profiling feature">Section 2, &#8220;Setting up the JIT profiling feature&#8221;</a>.
</p>
</dd>
<dt>
<span class="term">
<code class="filename">opreport</code>
</span>
</dt>
<dd>
<p>
This is the main tool for retrieving useful profile data, described in
<a class="xref" href="#opreport" title="2. Image summaries and symbol summaries (opreport)">Section 2, &#8220;Image summaries and symbol summaries (<span class="command"><strong>opreport</strong></span>)&#8221;</a>.
</p>
</dd>
<dt>
<span class="term">
<code class="filename">opannotate</code>
</span>
</dt>
<dd>
<p>
This utility can be used to produce annotated source, assembly or mixed source/assembly.
Source level annotation is available only if the application was compiled with
debugging symbols. See <a class="xref" href="#opannotate" title="3. Outputting annotated source (opannotate)">Section 3, &#8220;Outputting annotated source (<span class="command"><strong>opannotate</strong></span>)&#8221;</a>.
</p>
</dd>
<dt>
<span class="term">
<code class="filename">opgprof</code>
</span>
</dt>
<dd>
<p>
This utility can output gprof-style data files for a binary, for use with
<span class="command"><strong>gprof -p</strong></span>. See <a class="xref" href="#opgprof" title="5. gprof-compatible output (opgprof)">Section 5, &#8220;<span class="command"><strong>gprof</strong></span>-compatible output (<span class="command"><strong>opgprof</strong></span>)&#8221;</a>.
</p>
</dd>
<dt>
<span class="term">
<code class="filename">oparchive</code>
</span>
</dt>
<dd>
<p>
This utility can be used to collect executables, debuginfo,
and sample files and copy the files into an archive.
The archive is self-contained and can be moved to another
machine for further analysis.
See <a class="xref" href="#oparchive" title="6. Archiving measurements (oparchive)">Section 6, &#8220;Archiving measurements (<span class="command"><strong>oparchive</strong></span>)&#8221;</a>.
</p>
</dd>
<dt>
<span class="term">
<code class="filename">opimport</code>
</span>
</dt>
<dd>
<p>
This utility converts sample database files from a foreign binary format (abi) to
the native format. This is useful only when moving sample files between hosts,
for analysis on platforms other than the one used for collection.
See <a class="xref" href="#opimport" title="7. Converting sample database files (opimport)">Section 7, &#8220;Converting sample database files (<span class="command"><strong>opimport</strong></span>)&#8221;</a>.
</p>
</dd>
</dl>
</div>
</div>
</div>
<div class="chapter" title="Chapter 3. Controlling the profiler">
<div class="titlepage">
<div>
<div>
<h2 class="title"><a id="controlling"></a>Chapter 3. Controlling the profiler</h2>
</div>
</div>
</div>
<div class="toc">
<p>
<b>Table of Contents</b>
</p>
<dl>
<dt>
<span class="sect1">
<a href="#controlling-daemon">1. Using <span class="command"><strong>opcontrol</strong></span></a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#opcontrolexamples">1.1. Examples</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#eventspec">1.2. Specifying performance counter events</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#setup-jit">2. Setting up the JIT profiling feature</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#setup-jit-jvm">2.1. JVM instrumentation</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#oprofile-gui">3. Using <span class="command"><strong>oprof_start</strong></span></a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#detailed-parameters">4. Configuration details</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#hardware-counters">4.1. Hardware performance counters</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#rtc">4.2. OProfile in RTC mode</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#timer">4.3. OProfile in timer interrupt mode</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#p4">4.4. Pentium 4 support</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#ia64">4.5. Intel Itanium 2 support</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#ppc64">4.6. PowerPC64 support</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#cell-be">4.7. Cell Broadband Engine support</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#amd-ibs-support">4.8. AMD64 (x86_64) Instruction-Based Sampling (IBS) support</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#misuse">4.9. Dangerous counter settings</a>
</span>
</dt>
</dl>
</dd>
</dl>
</div>
<div class="sect1" title="1. Using opcontrol">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="controlling-daemon"></a>1. Using <span class="command"><strong>opcontrol</strong></span></h2>
</div>
</div>
</div>
<p>
In this section we describe the configuration and control of the profiling system
with opcontrol in more depth.
The <span class="command"><strong>opcontrol</strong></span> script has a default setup, but you
can alter this with the options given below. In particular,
if your hardware supports performance counters, you can configure them.
There are a number of counters (for example, counter 0 and counter 1
on the Pentium III). Each of these counters can be programmed with
an event to count, such as cache misses or MMX operations. The event
chosen for each counter is reflected in the profile data collected
by OProfile: functions and binaries at the top of the profiles reflect
that most of the chosen events happened within that code.
</p>
<p>
Additionally, each counter has a "count" value: this corresponds to how
detailed the profile is. The lower the value, the more frequently profile
samples are taken. A counter can choose to sample only kernel code, user-space code,
or both (both is the default). Finally, some events have a "unit mask"
- this is a value that further restricts the types of event that are counted.
The event types and unit masks for your CPU are listed by <span class="command"><strong>opcontrol
--list-events</strong></span>.
</p>
<p>
The <span class="command"><strong>opcontrol</strong></span> script provides the following actions :
</p>
<div class="variablelist">
<dl>
<dt>
<span class="term">
<code class="option">--init</code>
</span>
</dt>
<dd>
<p>
Loads the OProfile module if required and makes the OProfile driver
interface available.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--setup</code>
</span>
</dt>
<dd>
<p>
Followed by list arguments for profiling set up. List of arguments
saved in <code class="filename">/root/.oprofile/daemonrc</code>.
Giving this option is not necessary; you can just directly pass one
of the setup options, e.g. <span class="command"><strong>opcontrol --no-vmlinux</strong></span>.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--status</code>
</span>
</dt>
<dd>
<p>
Show configuration information.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--start-daemon</code>
</span>
</dt>
<dd>
<p>
Start the oprofile daemon without starting actual profiling. The profiling
can then be started using <code class="option">--start</code>. This is useful for avoiding
measuring the cost of daemon startup, as <code class="option">--start</code> is a simple
write to a file in oprofilefs. Not available in 2.2/2.4 kernels.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--start</code>
</span>
</dt>
<dd>
<p>
Start data collection with either arguments provided by <code class="option">--setup</code>
or information saved in <code class="filename">/root/.oprofile/daemonrc</code>. Specifying
the addition <code class="option">--verbose</code> makes the daemon generate lots of debug data
whilst it is running.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--dump</code>
</span>
</dt>
<dd>
<p>
Force a flush of the collected profiling data to the daemon.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--stop</code>
</span>
</dt>
<dd>
<p>
Stop data collection (this separate step is not possible with 2.2 or 2.4 kernels).
</p>
</dd>
<dt>
<span class="term">
<code class="option">--shutdown</code>
</span>
</dt>
<dd>
<p>
Stop data collection and kill the daemon.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--reset</code>
</span>
</dt>
<dd>
<p>
Clears out data from current session, but leaves saved sessions.
</p>
</dd>
<dt>
<span class="term"><code class="option">--save=</code>session_name</span>
</dt>
<dd>
<p>
Save data from current session to session_name.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--deinit</code>
</span>
</dt>
<dd>
<p>
Shuts down daemon. Unload the OProfile module and oprofilefs.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--list-events</code>
</span>
</dt>
<dd>
<p>
List event types and unit masks.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--help</code>
</span>
</dt>
<dd>
<p>
Generate usage messages.
</p>
</dd>
</dl>
</div>
<p>
There are a number of possible settings, of which, only
<code class="option">--vmlinux</code> (or <code class="option">--no-vmlinux</code>)
is required. These settings are stored in <code class="filename">~/.oprofile/daemonrc</code>.
</p>
<div class="variablelist">
<dl>
<dt>
<span class="term"><code class="option">--buffer-size=</code>num</span>
</dt>
<dd>
<p>
Number of samples in kernel buffer. When using a 2.6 kernel
buffer watershed need to be tweaked when changing this value.
</p>
</dd>
<dt>
<span class="term"><code class="option">--buffer-watershed=</code>num</span>
</dt>
<dd>
<p>
Set kernel buffer watershed to num samples (2.6 only). When it'll remain only
buffer-size - buffer-watershed free entry in the kernel buffer data will be
flushed to daemon, most usefull value are in the range [0.25 - 0.5] * buffer-size.
</p>
</dd>
<dt>
<span class="term"><code class="option">--cpu-buffer-size=</code>num</span>
</dt>
<dd>
<p>
Number of samples in kernel per-cpu buffer (2.6 only). If you
profile at high rate it can help to increase this if the log
file show excessive count of sample lost cpu buffer overflow.
</p>
</dd>
<dt>
<span class="term"><code class="option">--event=</code>[eventspec]</span>
</dt>
<dd>
<p>
Use the given performance counter event to profile.
See <a class="xref" href="#eventspec" title="1.2. Specifying performance counter events">Section 1.2, &#8220;Specifying performance counter events&#8221;</a> below.
</p>
</dd>
<dt>
<span class="term"><code class="option">--session-dir=</code>dir_path</span>
</dt>
<dd>
<p>
Create/use sample database out of directory <code class="filename">dir_path</code> instead of
the default location (/var/lib/oprofile).
</p>
</dd>
<dt>
<span class="term"><code class="option">--separate=</code>[none,lib,kernel,thread,cpu,all]</span>
</dt>
<dd>
<p>
By default, every profile is stored in a single file. Thus, for example,
samples in the C library are all accredited to the <code class="filename">/lib/libc.o</code>
profile. However, you choose to create separate sample files by specifying
one of the below options.
</p>
<div class="informaltable">
<table border="1">
<colgroup>
<col />
<col />
</colgroup>
<tbody>
<tr>
<td>
<code class="option">none</code>
</td>
<td>No profile separation (default)</td>
</tr>
<tr>
<td>
<code class="option">lib</code>
</td>
<td>Create per-application profiles for libraries</td>
</tr>
<tr>
<td>
<code class="option">kernel</code>
</td>
<td>Create per-application profiles for the kernel and kernel modules</td>
</tr>
<tr>
<td>
<code class="option">thread</code>
</td>
<td>Create profiles for each thread and each task</td>
</tr>
<tr>
<td>
<code class="option">cpu</code>
</td>
<td>Create profiles for each CPU</td>
</tr>
<tr>
<td>
<code class="option">all</code>
</td>
<td>All of the above options</td>
</tr>
</tbody>
</table>
</div>
<p>
Note that <code class="option">--separate=kernel</code> also turns on <code class="option">--separate=lib</code>.
When using <code class="option">--separate=kernel</code>, samples in hardware interrupts, soft-irqs, or other
asynchronous kernel contexts are credited to the task currently running. This means you will see
seemingly nonsense profiles such as <code class="filename">/bin/bash</code> showing samples for the PPP modules,
etc.
</p>
<p>
On 2.2/2.4 only kernel threads already started when profiling begins are correctly profiled;
newly started kernel thread samples are credited to the vmlinux (kernel) profile.
</p>
<p>
Using <code class="option">--separate=thread</code> creates a lot
of sample files if you leave OProfile running for a while; it's most
useful when used for short sessions, or when using image filtering.
</p>
</dd>
<dt>
<span class="term"><code class="option">--callgraph=</code>#depth</span>
</dt>
<dd>
<p>
Enable call-graph sample collection with a maximum depth. Use 0 to disable
callgraph profiling. NOTE: Callgraph support is available on a limited
number of platforms at this time; for example:
</p>
<p>
</p>
<div class="itemizedlist">
<ul class="itemizedlist" type="disc">
<li class="listitem">
<p>x86 with recent 2.6 kernel</p>
</li>
<li class="listitem">
<p>ARM with recent 2.6 kernel</p>
</li>
<li class="listitem">
<p>PowerPC with 2.6.17 kernel</p>
</li>
</ul>
</div>
<p>
</p>
<p>
</p>
</dd>
<dt>
<span class="term"><code class="option">--image=</code>image,[images]|"all"</span>
</dt>
<dd>
<p>
Image filtering. If you specify one or more absolute
paths to binaries, OProfile will only produce profile results for those
binary images. This is useful for restricting the sometimes voluminous
output you may get otherwise, especially with
<code class="option">--separate=thread</code>. Note that if you are using
<code class="option">--separate=lib</code> or
<code class="option">--separate=kernel</code>, then if you specification an
application binary, the shared libraries and kernel code
<span class="emphasis"><em>are</em></span> included. Specify the value
"all" to profile everything (the default).
</p>
</dd>
<dt>
<span class="term"><code class="option">--vmlinux=</code>file</span>
</dt>
<dd>
<p>
vmlinux kernel image.
</p>
</dd>
<dt>
<span class="term">
<code class="option">--no-vmlinux</code>
</span>
</dt>
<dd>
<p>
Use this when you don't have a kernel vmlinux file, and you don't want
to profile the kernel. This still counts the total number of kernel samples,
but can't give symbol-based results for the kernel or any modules.
</p>
</dd>
</dl>
</div>
<div class="sect2" title="1.1. Examples">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="opcontrolexamples"></a>1.1. Examples</h3>
</div>
</div>
</div>
<div class="sect3" title="1.1.1. Intel performance counter setup">
<div class="titlepage">
<div>
<div>
<h4 class="title"><a id="examplesperfctr"></a>1.1.1. Intel performance counter setup</h4>
</div>
</div>
</div>
<p>
Here, we have a Pentium III running at 800MHz, and we want to look at where data memory
references are happening most, and also get results for CPU time.
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">
# opcontrol --event=CPU_CLK_UNHALTED:400000 --event=DATA_MEM_REFS:10000
# opcontrol --vmlinux=/boot/2.6.0/vmlinux
# opcontrol --start
</pre>
</td>
</tr>
</table>
</div>
<div class="sect3" title="1.1.2. RTC mode">
<div class="titlepage">
<div>
<div>
<h4 class="title"><a id="examplesrtc"></a>1.1.2. RTC mode</h4>
</div>
</div>
</div>
<p>
Here, we have an Intel laptop without support for performance counters, running on 2.4 kernels.
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">
# ophelp -r
CPU with RTC device
# opcontrol --vmlinux=/boot/2.4.13/vmlinux --event=RTC_INTERRUPTS:1024
# opcontrol --start
</pre>
</td>
</tr>
</table>
</div>
<div class="sect3" title="1.1.3. Starting the daemon separately">
<div class="titlepage">
<div>
<div>
<h4 class="title"><a id="examplesstartdaemon"></a>1.1.3. Starting the daemon separately</h4>
</div>
</div>
</div>
<p>
If we're running 2.6 kernels, we can use <code class="option">--start-daemon</code> to avoid
the profiler startup affecting results.
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">
# opcontrol --vmlinux=/boot/2.6.0/vmlinux
# opcontrol --start-daemon
# my_favourite_benchmark --init
# opcontrol --start ; my_favourite_benchmark --run ; opcontrol --stop
</pre>
</td>
</tr>
</table>
</div>
<div class="sect3" title="1.1.4. Separate profiles for libraries and the kernel">
<div class="titlepage">
<div>
<div>
<h4 class="title"><a id="exampleseparate"></a>1.1.4. Separate profiles for libraries and the kernel</h4>
</div>
</div>
</div>
<p>
Here, we want to see a profile of the OProfile daemon itself, including when
it was running inside the kernel driver, and its use of shared libraries.
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">
# opcontrol --separate=kernel --vmlinux=/boot/2.6.0/vmlinux
# opcontrol --start
# my_favourite_stress_test --run
# opreport -l -p /lib/modules/2.6.0/kernel /usr/local/bin/oprofiled
</pre>
</td>
</tr>
</table>
</div>
<div class="sect3" title="1.1.5. Profiling sessions">
<div class="titlepage">
<div>
<div>
<h4 class="title"><a id="examplessessions"></a>1.1.5. Profiling sessions</h4>
</div>
</div>
</div>
<p>
It can often be useful to split up profiling data into several different
time periods. For example, you may want to collect data on an application's
startup separately from the normal runtime data. You can use the simple
command <span class="command"><strong>opcontrol --save</strong></span> to do this. For example :
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">
# opcontrol --save=blah
</pre>
</td>
</tr>
</table>
<p>
will create a sub-directory in <code class="filename">$SESSION_DIR/samples</code> containing the samples
up to that point (the current session's sample files are moved into this
directory). You can then pass this session name as a parameter to the post-profiling
analysis tools, to only get data up to the point you named the
session. If you do not want to save a session, you can do
<span class="command"><strong>rm -rf $SESSION_DIR/samples/sessionname</strong></span> or, for the
current session, <span class="command"><strong>opcontrol --reset</strong></span>.
</p>
</div>
</div>
<div class="sect2" title="1.2. Specifying performance counter events">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="eventspec"></a>1.2. Specifying performance counter events</h3>
</div>
</div>
</div>
<p>
The <code class="option">--event</code> option to <span class="command"><strong>opcontrol</strong></span>
takes a specification that indicates how the details of each
hardware performance counter should be setup. If you want to
revert to OProfile's default setting (<code class="option">--event</code>
is strictly optional), use <code class="option">--event=default</code>. Use of this
option over-rides all previous event selections.
</p>
<p>
You can pass multiple event specifications. OProfile will allocate
hardware counters as necessary. Note that some combinations are not
allowed by the CPU; running <span class="command"><strong>opcontrol --list-events</strong></span> gives the details
of each event. The event specification is a colon-separated string
of the form <code class="option"><span class="emphasis"><em>name</em></span>:<span class="emphasis"><em>count</em></span>:<span class="emphasis"><em>unitmask</em></span>:<span class="emphasis"><em>kernel</em></span>:<span class="emphasis"><em>user</em></span></code> as described in this table:
</p>
<div class="informaltable">
<table border="1">
<colgroup>
<col />
<col />
</colgroup>
<tbody>
<tr>
<td>
<code class="option">name</code>
</td>
<td>The symbolic event name, e.g. <code class="constant">CPU_CLK_UNHALTED</code></td>
</tr>
<tr>
<td>
<code class="option">count</code>
</td>
<td>The counter reset value, e.g. 100000</td>
</tr>
<tr>
<td>
<code class="option">unitmask</code>
</td>
<td>The unit mask, as given in the events list: e.g. 0x0f; or a symbolic name as
given by the first word of the description (only valid for unit masks having an "extra:" parameter)</td>
</tr>
<tr>
<td>
<code class="option">kernel</code>
</td>
<td>Whether to profile kernel code</td>
</tr>
<tr>
<td>
<code class="option">user</code>
</td>
<td>Whether to profile userspace code</td>
</tr>
</tbody>
</table>
</div>
<p>
The last three values are optional, if you omit them (e.g. <code class="option">--event=DATA_MEM_REFS:30000</code>),
they will be set to the default values (a unit mask of 0, and profiling both kernel and
userspace code). Note that some events require a unit mask.
</p>
<div class="note" title="Note" style="margin-left: 0.5in; margin-right: 0.5in;">
<h3 class="title">Note</h3>
<p>
For the PowerPC platforms, all events specified must be in the same group; i.e., the group number
appended to the event name (e.g. <code class="constant">&lt;<span class="emphasis"><em>some-event-name</em></span>&gt;_GRP9</code>) must be the same.
</p>
</div>
<p>
If OProfile is using RTC mode, and you want to alter the default counter value,
you can use something like <code class="option">--event=RTC_INTERRUPTS:2048</code>. Note the last
three values here are ignored.
If OProfile is using timer-interrupt mode, there is no configuration possible.
</p>
<p>
The table below lists the events selected by default
(<code class="option">--event=default</code>) for the various computer architectures:
</p>
<div class="informaltable">
<table border="1">
<colgroup>
<col />
<col />
<col />
</colgroup>
<tbody>
<tr>
<td>Processor</td>
<td>cpu_type</td>
<td>Default event</td>
</tr>
<tr>
<td>Alpha EV4</td>
<td>alpha/ev4</td>
<td>CYCLES:100000:0:1:1</td>
</tr>
<tr>
<td>Alpha EV5</td>
<td>alpha/ev5</td>
<td>CYCLES:100000:0:1:1</td>
</tr>
<tr>
<td>Alpha PCA56</td>
<td>alpha/pca56</td>
<td>CYCLES:100000:0:1:1</td>
</tr>
<tr>
<td>Alpha EV6</td>
<td>alpha/ev6</td>
<td>CYCLES:100000:0:1:1</td>
</tr>
<tr>
<td>Alpha EV67</td>
<td>alpha/ev67</td>
<td>CYCLES:100000:0:1:1</td>
</tr>
<tr>
<td>ARM/XScale PMU1</td>
<td>arm/xscale1</td>
<td>CPU_CYCLES:100000:0:1:1</td>
</tr>
<tr>
<td>ARM/XScale PMU2</td>
<td>arm/xscale2</td>
<td>CPU_CYCLES:100000:0:1:1</td>
</tr>
<tr>
<td>ARM/MPCore</td>
<td>arm/mpcore</td>
<td>CPU_CYCLES:100000:0:1:1</td>
</tr>
<tr>
<td>AVR32</td>
<td>avr32</td>
<td>CPU_CYCLES:100000:0:1:1</td>
</tr>
<tr>
<td>Athlon</td>
<td>i386/athlon</td>
<td>CPU_CLK_UNHALTED:100000:0:1:1</td>
</tr>
<tr>
<td>Pentium Pro</td>
<td>i386/ppro</td>
<td>CPU_CLK_UNHALTED:100000:0:1:1</td>
</tr>
<tr>
<td>Pentium II</td>
<td>i386/pii</td>
<td>CPU_CLK_UNHALTED:100000:0:1:1</td>
</tr>
<tr>
<td>Pentium III</td>
<td>i386/piii</td>
<td>CPU_CLK_UNHALTED:100000:0:1:1</td>
</tr>
<tr>
<td>Pentium M (P6 core)</td>
<td>i386/p6_mobile</td>
<td>CPU_CLK_UNHALTED:100000:0:1:1</td>
</tr>
<tr>
<td>Pentium 4 (non-HT)</td>
<td>i386/p4</td>
<td>GLOBAL_POWER_EVENTS:100000:1:1:1</td>
</tr>
<tr>
<td>Pentium 4 (HT)</td>
<td>i386/p4-ht</td>
<td>GLOBAL_POWER_EVENTS:100000:1:1:1</td>
</tr>
<tr>
<td>Hammer</td>
<td>x86-64/hammer</td>
<td>CPU_CLK_UNHALTED:100000:0:1:1</td>
</tr>
<tr>
<td>Family10h</td>
<td>x86-64/family10</td>
<td>CPU_CLK_UNHALTED:100000:0:1:1</td>
</tr>
<tr>
<td>Family11h</td>
<td>x86-64/family11h</td>
<td>CPU_CLK_UNHALTED:100000:0:1:1</td>
</tr>
<tr>
<td>Itanium</td>
<td>ia64/itanium</td>
<td>CPU_CYCLES:100000:0:1:1</td>
</tr>
<tr>
<td>Itanium 2</td>
<td>ia64/itanium2</td>
<td>CPU_CYCLES:100000:0:1:1</td>
</tr>
<tr>
<td>TIMER_INT</td>
<td>timer</td>
<td>None selectable</td>
</tr>
<tr>
<td>IBM iseries</td>
<td>PowerPC 4/5/970</td>
<td>CYCLES:10000:0:1:1</td>
</tr>
<tr>
<td>IBM pseries</td>
<td>PowerPC 4/5/970/Cell</td>
<td>CYCLES:10000:0:1:1</td>
</tr>
<tr>
<td>IBM s390</td>
<td>timer</td>
<td>None selectable</td>
</tr>
<tr>
<td>IBM s390x</td>
<td>timer</td>
<td>None selectable</td>
</tr>
</tbody>
</table>
</div>
</div>
</div>
<div class="sect1" title="2. Setting up the JIT profiling feature">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="setup-jit"></a>2. Setting up the JIT profiling feature</h2>
</div>
</div>
</div>
<p>
To gather information about JITed code from a virtual machine,
it needs to be instrumented with an agent library. We use the
agent libraries for Java in the following example. To use the
Java profiling feature, you must build OProfile with the "--with-java" option
(<a class="xref" href="#install" title="4. Installation">Section 4, &#8220;Installation&#8221;</a>).
</p>
<div class="sect2" title="2.1. JVM instrumentation">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="setup-jit-jvm"></a>2.1. JVM instrumentation</h3>
</div>
</div>
</div>
<p>
Add this to the startup parameters of the JVM (for JVMTI):
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen"><code xmlns="http://www.w3.org/1999/xhtml" class="option">-agentpath:&lt;libdir&gt;/libjvmti_oprofile.so[=&lt;options&gt;]</code> </pre>
</td>
</tr>
</table>
<p>
or
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen"><code xmlns="http://www.w3.org/1999/xhtml" class="option">-agentlib:jvmti_oprofile[=&lt;options&gt;]</code> </pre>
</td>
</tr>
</table>
<p>
</p>
<p>
The JVMPI agent implementation is enabled with the command line option
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen"><code xmlns="http://www.w3.org/1999/xhtml" class="option">-Xrunjvmpi_oprofile[:&lt;options&gt;]</code> </pre>
</td>
</tr>
</table>
<p>
</p>
<p>
Currently, there is just one option available -- <code class="option">debug</code>. For JVMPI,
the convention for specifying an option is <code class="option">option_name=[yes|no]</code>.
For JVMTI, the option specification is simply the option name, implying
"yes"; no option specified implies "no".
</p>
<p>
The agent library (installed in <code class="filename">&lt;oprof_install_dir&gt;/lib/oprofile</code>)
needs to be in the library search path (e.g. add the library directory
to <code class="constant">LD_LIBRARY_PATH</code>). If the command line of
the JVM is not accessible, it may be buried within shell scripts or a
launcher program. It may also be possible to set an environment variable to add
the instrumentation.
For Sun JVMs this is <code class="constant">JAVA_TOOL_OPTIONS</code>. Please check
your JVM documentation for
further information on the agent startup options.
</p>
</div>
</div>
<div class="sect1" title="3. Using oprof_start">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="oprofile-gui"></a>3. Using <span class="command"><strong>oprof_start</strong></span></h2>
</div>
</div>
</div>
<p>
The <span class="command"><strong>oprof_start</strong></span> application provides a convenient way to start the profiler.
Note that <span class="command"><strong>oprof_start</strong></span> is just a wrapper around the <span class="command"><strong>opcontrol</strong></span> script,
so it does not provide more services than the script itself.
</p>
<p>
After <span class="command"><strong>oprof_start</strong></span> is started you can select the event type for each counter;
the sampling rate and other related parameters are explained in <a class="xref" href="#controlling-daemon" title="1. Using opcontrol">Section 1, &#8220;Using <span class="command"><strong>opcontrol</strong></span>&#8221;</a>.
The "Configuration" section allows you to set general parameters such as the buffer size, kernel filename
etc. The counter setup interface should be self-explanatory; <a class="xref" href="#hardware-counters" title="4.1. Hardware performance counters">Section 4.1, &#8220;Hardware performance counters&#8221;</a> and related
links contain information on using unit masks.
</p>
<p>
A status line shows the current status of the profiler: how long it has been running, and the average
number of interrupts received per second and the total, over all processors.
Note that quitting <span class="command"><strong>oprof_start</strong></span> does not stop the profiler.
</p>
<p>
Your configuration is saved in the same file as <span class="command"><strong>opcontrol</strong></span> uses; that is,
<code class="filename">~/.oprofile/daemonrc</code>.
</p>
</div>
<div class="sect1" title="4. Configuration details">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="detailed-parameters"></a>4. Configuration details</h2>
</div>
</div>
</div>
<div class="sect2" title="4.1. Hardware performance counters">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="hardware-counters"></a>4.1. Hardware performance counters</h3>
</div>
</div>
</div>
<div class="note" title="Note" style="margin-left: 0.5in; margin-right: 0.5in;">
<h3 class="title">Note</h3>
<p>
Your CPU type may not include the requisite support for hardware performance counters, in which case
you must use OProfile in RTC mode in 2.4 (see <a class="xref" href="#rtc" title="4.2. OProfile in RTC mode">Section 4.2, &#8220;OProfile in RTC mode&#8221;</a>), or timer mode in 2.6 (see <a class="xref" href="#timer" title="4.3. OProfile in timer interrupt mode">Section 4.3, &#8220;OProfile in timer interrupt mode&#8221;</a>).
You do not really need to read this section unless you are interested in using
events other than the default event chosen by OProfile.
</p>
</div>
<p>
The Intel hardware performance counters are detailed in the Intel IA-32 Architecture Manual, Volume 3, available
from <a class="ulink" href="http://developer.intel.com/">http://developer.intel.com/</a>.
The AMD Athlon/Opteron/Phenom/Turion implementation is detailed in <a class="ulink" href="http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/22007.pdf">
http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/22007.pdf</a>.
For PowerPC64 processors in IBM iSeries, pSeries, and blade server systems, processor documentation
is available at <a class="ulink" href="http://www-01.ibm.com/chips/techlib/techlib.nsf/productfamilies/PowerPC/">
http://www-01.ibm.com/chips/techlib/techlib.nsf/productfamilies/PowerPC</a>. (For example, the
specific publication containing information on the performance monitor unit for the PowerPC970 is
"IBM PowerPC 970FX RISC Microprocessor User's Manual.")
These processors are capable of delivering an interrupt when a counter overflows.
This is the basic mechanism on which OProfile is based. The delivery mode is <acronym class="acronym">NMI</acronym>,
so blocking interrupts in the kernel does not prevent profiling. When the interrupt handler is called,
the current <acronym class="acronym">PC</acronym> value and the current task are recorded into the profiling structure.
This allows the overflow event to be attached to a specific assembly instruction in a binary image.
The daemon receives this data from the kernel, and writes it to the sample files.
</p>
<p>
If we use an event such as <code class="constant">CPU_CLK_UNHALTED</code> or <code class="constant">INST_RETIRED</code>
(<code class="constant">GLOBAL_POWER_EVENTS</code> or <code class="constant">INSTR_RETIRED</code>, respectively, on the Pentium 4), we can
use the overflow counts as an estimate of actual time spent in each part of code. Alternatively we can profile interesting
data such as the cache behaviour of routines with the other available counters.
</p>
<p>
However there are several caveats. First, there are those issues listed in the Intel manual. There is a delay
between the counter overflow and the interrupt delivery that can skew results on a small scale - this means
you cannot rely on the profiles at the instruction level as being perfectly accurate.
If you are using an "event-mode" counter such as the cache counters, a count registered against it doesn't mean
that it is responsible for that event. However, it implies that the counter overflowed in the dynamic
vicinity of that instruction, to within a few instructions. Further details on this problem can be found in
<a class="xref" href="#interpreting" title="Chapter 5. Interpreting profiling results">Chapter 5, <i>Interpreting profiling results</i></a> and also in the Digital paper "ProfileMe: A Hardware Performance Counter".
</p>
<p>
Each counter has several configuration parameters.
First, there is the unit mask: this simply further specifies what to count.
Second, there is the counter value, discussed below. Third, there is a parameter whether to increment counts
whilst in kernel or user space. You can configure these separately for each counter.
</p>
<p>
After each overflow event, the counter will be re-initialized
such that another overflow will occur after this many events have been counted. Thus, higher
values mean less-detailed profiling, and lower values mean more detail, but higher overhead.
Picking a good value for this
parameter is, unfortunately, somewhat of a black art. It is of course dependent on the event
you have chosen.
Specifying too large a value will mean not enough interrupts are generated
to give a realistic profile (though this problem can be ameliorated by profiling for <span class="emphasis"><em>longer</em></span>).
Specifying too small a value can lead to higher performance overhead.
</p>
</div>
<div class="sect2" title="4.2. OProfile in RTC mode">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="rtc"></a>4.2. OProfile in RTC mode</h3>
</div>
</div>
</div>
<div class="note" title="Note" style="margin-left: 0.5in; margin-right: 0.5in;">
<h3 class="title">Note</h3>
<p>
This section applies to 2.2/2.4 kernels only.
</p>
</div>
<p>
Some CPU types do not provide the needed hardware support to use the hardware performance counters. This includes
some laptops, classic Pentiums, and other CPU types not yet supported by OProfile (such as Cyrix).
On these machines, OProfile falls
back to using the real-time clock interrupt to collect samples. This interrupt is also used by the <span class="command"><strong>rtc</strong></span>
module: you cannot have both the OProfile and rtc modules loaded nor the rtc support compiled in the kernel.
</p>
<p>
RTC mode is less capable than the hardware counters mode; in particular, it is unable to profile sections of
the kernel where interrupts are disabled. There is just one available event, "RTC interrupts", and its value
corresponds to the number of interrupts generated per second (that is, a higher number means a better profiling
resolution, and higher overhead). The current implementation of the real-time clock supports only power-of-two
sampling rates from 2 to 4096 per second. Other values within this range are rounded to the nearest power of
two.
</p>
<p>
You can force use of the RTC interrupt with the <code class="option">force_rtc=1</code> module parameter.
</p>
<p>
Setting the value from the GUI should be straightforward. On the command line, you need to specify the
event to <span class="command"><strong>opcontrol</strong></span>, e.g. :
</p>
<p>
<span class="command">
<strong>opcontrol --event=RTC_INTERRUPTS:256</strong>
</span>
</p>
</div>
<div class="sect2" title="4.3. OProfile in timer interrupt mode">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="timer"></a>4.3. OProfile in timer interrupt mode</h3>
</div>
</div>
</div>
<div class="note" title="Note" style="margin-left: 0.5in; margin-right: 0.5in;">
<h3 class="title">Note</h3>
<p>
This section applies to 2.6 kernels and above only.
</p>
</div>
<p>
In 2.6 kernels on CPUs without OProfile support for the hardware performance counters, the driver
falls back to using the timer interrupt for profiling. Like the RTC mode in 2.4 kernels, this is not able to
profile code that has interrupts disabled. Note that there are no configuration parameters for
setting this, unlike the RTC and hardware performance counter setup.
</p>
<p>
You can force use of the timer interrupt by using the <code class="option">timer=1</code> module
parameter (or <code class="option">oprofile.timer=1</code> on the boot command line if OProfile is
built-in).
</p>
</div>
<div class="sect2" title="4.4. Pentium 4 support">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="p4"></a>4.4. Pentium 4 support</h3>
</div>
</div>
</div>
<p>
The Pentium 4 / Xeon performance counters are organized around 3 types of model specific registers (MSRs): 45 event
selection control registers (ESCRs), 18 counter configuration control registers (CCCRs) and 18 counters. ESCRs describe a
particular set of events which are to be recorded, and CCCRs bind ESCRs to counters and configure their
operation. Unfortunately the relationship between these registers is quite complex; they cannot all be used with one
another at any time. There is, however, a subset of 8 counters, 8 ESCRs, and 8 CCCRs which can be used independently of
one another, so OProfile only accesses those registers, treating them as a bank of 8 "normal" counters, similar
to those in the P6 or Athlon/Opteron/Phenom/Turion families of CPU.
</p>
<p>
There is currently no support for Precision Event-Based Sampling (PEBS), nor any advanced uses of the Debug Store
(DS). Current support is limited to the conservative extension of OProfile's existing interrupt-based model described
above. Performance monitoring hardware on Pentium 4 / Xeon processors with Hyperthreading enabled (multiple logical
processors on a single die) is not supported in 2.4 kernels (you can use OProfile if you disable hyper-threading,
though).
</p>
</div>
<div class="sect2" title="4.5. Intel Itanium 2 support">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="ia64"></a>4.5. Intel Itanium 2 support</h3>
</div>
</div>
</div>
<p>
The Itanium 2 performance monitoring unit (PMU) organizes the counters as four
pairs of performance event monitoring registers. Each pair is composed of a
Performance Monitoring Configuration (PMC) register and Performance Monitoring
Data (PMD) register. The PMC selects the performance event being monitored and
the PMD determines the sampling interval. The IA64 Performance Monitoring Unit
(PMU) triggers sampling with maskable interrupts. Thus, samples will not occur
in sections of the IA64 kernel where interrupts are disabled.
</p>
<p>
None of the advance features of the Itanium 2 performance monitoring unit
such as opcode matching, address range matching, or precise event sampling are
supported by this version of OProfile. The Itanium 2 support only maps OProfile's
existing interrupt-based model to the PMU hardware.
</p>
</div>
<div class="sect2" title="4.6. PowerPC64 support">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="ppc64"></a>4.6. PowerPC64 support</h3>
</div>
</div>
</div>
<p>
The performance monitoring unit (PMU) for the IBM PowerPC 64-bit processors
consists of between 4 and 8 counters (depending on the model), plus three
special purpose registers used for programming the counters -- MMCR0, MMCR1,
and MMCRA. Advanced features such as instruction matching and thresholding are
not supported by this version of OProfile.
</p>
<div class="note" title="Note" style="margin-left: 0.5in; margin-right: 0.5in;"><h3 class="title">Note</h3>Later versions of the IBM POWER5+ processor (beginning with revision 3.0)
run the performance monitor unit in POWER6 mode, effectively removing OProfile's
access to counters 5 and 6. These two counters are dedicated to counting
instructions completed and cycles, respectively. In POWER6 mode, however, the
counters do not generate an interrupt on overflow and so are unusable by
OProfile. Kernel versions 2.6.23 and higher will recognize this mode
and export "ppc64/power5++" as the cpu_type to the oprofilefs pseudo filesystem.
OProfile userspace responds to this cpu_type by removing these counters from
the list of potential events to count. Without this kernel support, attempts
to profile using an event from one of these counters will yield incorrect
results -- typically, zero (or near zero) samples in the generated report.
</div>
<p>
</p>
</div>
<div class="sect2" title="4.7. Cell Broadband Engine support">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="cell-be"></a>4.7. Cell Broadband Engine support</h3>
</div>
</div>
</div>
<p>
The Cell Broadband Engine (CBE) processor core consists of a PowerPC Processing
Element (PPE) and 8 Synergistic Processing Elements (SPE). PPEs and SPEs each
consist of a processing unit (PPU and SPU, respectively) and other hardware
components, such as memory controllers.
</p>
<p>
A PPU has two hardware threads (aka "virtual CPUs"). The performance monitor
unit of the CBE collects event information on one hardware thread at a time.
Therefore, when profiling PPE events,
OProfile collects the profile based on the selected events by time slicing the
performance counter hardware between the two threads. The user must ensure the
collection interval is long enough so that the time spent collecting data for
each PPU is sufficient to obtain a good profile.
</p>
<p>
To profile an SPU application, the user should specify the SPU_CYCLES event.
When starting OProfile with SPU_CYCLES, the opcontrol script enforces certain
separation parameters (separate=cpu,lib) to ensure that sufficient information
is collected in the sample data in order to generate a complete report. The
--merge=cpu option can be used to obtain a more readable report if analyzing
the performance of each separate SPU is not necessary.
</p>
<p>
Profiling with an SPU event (events 4100 through 4163) is not compatible with any other
event. Further more, only one SPU event can be specified at a time. The hardware only
supports profiling on one SPU per node at a time. The OProfile kernel code time slices
between the eight SPUs to collect data on all SPUs.
</p>
<p>
SPU profile reports have some unique characteristics compared to reports for
standard architectures:
</p>
<div class="itemizedlist">
<ul class="itemizedlist" type="disc">
<li class="listitem">Typically no "app name" column. This is really standard OProfile behavior
when the report contains samples for just a single application, which is
commonly the case when profiling SPUs.</li>
<li class="listitem">"CPU" equates to "SPU"</li>
<li class="listitem">Specifying '--long-filenames' on the opreport command does not always result
in long filenames. This happens when the SPU application code is embedded in
the PPE executable or shared library. The embedded SPU ELF data contains only the
short filename (i.e., no path information) for the SPU binary file that was used as
the source for embedding. The reason that just the short filename is used is because
the original SPU binary file may not exist or be accessible at runtime. The performance
analyst must have sufficient knowledge of the application to be able to correlate the
SPU binary image names found in the report to the application's source files.
<div class="note" title="Note" style="margin-left: 0.5in; margin-right: 0.5in;"><h3 class="title">Note</h3>
Compile the application with -g and generate the OProfile report
with -g to facilitate finding the right source file(s) on which to focus.
</div></li>
</ul>
</div>
</div>
<div class="sect2" title="4.8. AMD64 (x86_64) Instruction-Based Sampling (IBS) support">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="amd-ibs-support"></a>4.8. AMD64 (x86_64) Instruction-Based Sampling (IBS) support</h3>
</div>
</div>
</div>
<p>
Instruction-Based Sampling (IBS) is a new performance measurement technique
available on AMD Family 10h processors. Traditional performance counter
sampling is not precise enough to isolate performance issues to individual
instructions. IBS, however, precisely identifies instructions which are not
making the best use of the processor pipeline and memory hierarchy.
For more information, please refer to the "Instruction-Based Sampling:
A New Performance Analysis Technique for AMD Family 10h Processors" (
<a class="ulink" href="http://developer.amd.com/assets/AMD_IBS_paper_EN.pdf">
http://developer.amd.com/assets/AMD_IBS_paper_EN.pdf</a>).
There are two types of IBS profile types, described in the following sections.
</p>
<div class="sect3" title="4.8.1. IBS Fetch">
<div class="titlepage">
<div>
<div>
<h4 class="title"><a id="ibs-fetch"></a>4.8.1. IBS Fetch</h4>
</div>
</div>
</div>
<p>
IBS fetch sampling is a statistical sampling method which counts completed
fetch operations. When the number of completed fetch operations reaches the
maximum fetch count (the sampling period), IBS tags the fetch operation and
monitors that operation until it either completes or aborts. When a tagged
fetch completes or aborts, a sampling interrupt is generated and an IBS fetch
sample is taken. An IBS fetch sample contains a timestamp, the identifier of
the interrupted process, the virtual fetch address, and several event flags
and values that describe what happened during the fetch operation.
</p>
</div>
<div class="sect3" title="4.8.2. IBS Op">
<div class="titlepage">
<div>
<div>
<h4 class="title"><a id="ibs-op"></a>4.8.2. IBS Op</h4>
</div>
</div>
</div>
<p>
IBS op sampling selects, tags, and monitors macro-ops as issued from AMD64
instructions. Two options are available for selecting ops for sampling:
</p>
<div class="itemizedlist">
<ul class="itemizedlist" type="disc">
<li class="listitem">
Cycles-based selection counts CPU clock cycles. The op is tagged and monitored
when the count reaches a threshold (the sampling period) and a valid op is
available.
</li>
<li class="listitem">
Dispatched op-based selection counts dispatched macro-ops.
When the count reaches a threshold, the next valid op is tagged and monitored.
</li>
</ul>
</div>
<p>
In both cases, an IBS sample is generated only if the tagged op retires.
Thus, IBS op event information does not measure speculative execution activity.
The execution stages of the pipeline monitor the tagged macro-op. When the
tagged macro-op retires, a sampling interrupt is generated and an IBS op
sample is taken. An IBS op sample contains a timestamp, the identifier of
the interrupted process, the virtual address of the AMD64 instruction from
which the op was issued, and several event flags and values that describe
what happened when the macro-op executed.
</p>
</div>
<p>
Enabling IBS profiling is done simply by specifying IBS performance events
through the "--event=" options. These events are listed in the
<code class="function">opcontrol --list-events</code>.
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">
opcontrol --event=IBS_FETCH_XXX:&lt;count&gt;:&lt;um&gt;:&lt;kernel&gt;:&lt;user&gt;
opcontrol --event=IBS_OP_XXX:&lt;count&gt;:&lt;um&gt;:&lt;kernel&gt;:&lt;user&gt;
Note: * All IBS fetch event must have the same event count and unitmask,
as do those for IBS op.
</pre>
</td>
</tr>
</table>
</div>
<div class="sect2" title="4.9. Dangerous counter settings">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="misuse"></a>4.9. Dangerous counter settings</h3>
</div>
</div>
</div>
<p>
OProfile is a low-level profiler which allow continuous profiling with a low-overhead cost.
If too low a count reset value is set for a counter, the system can become overloaded with counter
interrupts, and seem as if the system has frozen. Whilst some validation is done, it
is not foolproof.
</p>
<div class="note" title="Note" style="margin-left: 0.5in; margin-right: 0.5in;">
<h3 class="title">Note</h3>
<p>
This can happen as follows: When the profiler count
reaches zero an NMI handler is called which stores the sample values in an internal buffer, then resets the counter
to its original value. If the count is very low, a pending NMI can be sent before the NMI handler has
completed. Due to the priority of the NMI, the local APIC delivers the pending interrupt immediately after
completion of the previous interrupt handler, and control never returns to other parts of the system.
In this way the system seems to be frozen.
</p>
</div>
<p>If this happens, it will be impossible to bring the system back to a workable state.
There is no way to provide real security against this happening, other than making sure to use a reasonable value
for the counter reset. For example, setting <code class="constant">CPU_CLK_UNHALTED</code> event type with a ridiculously low reset count (e.g. 500)
is likely to freeze the system.
</p>
<p>
In short : <span class="command"><strong>Don't try a foolish sample count value</strong></span>. Unfortunately the definition of a foolish value
is really dependent on the event type - if ever in doubt, e-mail </p>
<div class="address">
<p><code class="email">&lt;<a class="email" href="mailto:oprofile-list@lists.sf.net">oprofile-list@lists.sf.net</a>&gt;</code>.</p>
</div>
</div>
</div>
</div>
<div class="chapter" title="Chapter 4. Obtaining results">
<div class="titlepage">
<div>
<div>
<h2 class="title"><a id="results"></a>Chapter 4. Obtaining results</h2>
</div>
</div>
</div>
<div class="toc">
<p>
<b>Table of Contents</b>
</p>
<dl>
<dt>
<span class="sect1">
<a href="#profile-spec">1. Profile specifications</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#profile-spec-examples">1.1. Examples</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#profile-spec-details">1.2. Profile specification parameters</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#locating-and-managing-binary-images">1.3. Locating and managing binary images</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#no-results">1.4. What to do when you don't get any results</a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#opreport">2. Image summaries and symbol summaries (<span class="command"><strong>opreport</strong></span>)</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#opreport-merging">2.1. Merging separate profiles</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opreport-comparison">2.2. Side-by-side multiple results</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opreport-callgraph">2.3. Callgraph output</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opreport-diff">2.4. Differential profiles with <span class="command"><strong>opreport</strong></span></a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opreport-anon">2.5. Anonymous executable mappings</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opreport-xml">2.6. XML formatted output</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opreport-options">2.7. Options for <span class="command"><strong>opreport</strong></span></a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#opannotate">3. Outputting annotated source (<span class="command"><strong>opannotate</strong></span>)</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#opannotate-finding-source">3.1. Locating source files</a>
</span>
</dt>
<dt>
<span class="sect2">
<a href="#opannotate-details">3.2. Usage of <span class="command"><strong>opannotate</strong></span></a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#getting-jit-reports">4. OProfile results with JIT samples</a>
</span>
</dt>
<dt>
<span class="sect1">
<a href="#opgprof">5. <span class="command"><strong>gprof</strong></span>-compatible output (<span class="command"><strong>opgprof</strong></span>)</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#opgprof-details">5.1. Usage of <span class="command"><strong>opgprof</strong></span></a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#oparchive">6. Archiving measurements (<span class="command"><strong>oparchive</strong></span>)</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#oparchive-details">6.1. Usage of <span class="command"><strong>oparchive</strong></span></a>
</span>
</dt>
</dl>
</dd>
<dt>
<span class="sect1">
<a href="#opimport">7. Converting sample database files (<span class="command"><strong>opimport</strong></span>)</a>
</span>
</dt>
<dd>
<dl>
<dt>
<span class="sect2">
<a href="#opimport-details">7.1. Usage of <span class="command"><strong>opimport</strong></span></a>
</span>
</dt>
</dl>
</dd>
</dl>
</div>
<p>
OK, so the profiler has been running, but it's not much use unless we can get some data out. Fairly often,
OProfile does a little <span class="emphasis"><em>too</em></span> good a job of keeping overhead low, and no data reaches
the profiler. This can happen on lightly-loaded machines. Remember you can force a dump at any time with :
</p>
<p>
<span class="command">
<strong>opcontrol --dump</strong>
</span>
</p>
<p>Remember to do this before complaining there is no profiling data !
Now that we've got some data, it has to be processed. That's the job of <span class="command"><strong>opreport</strong></span>,
<span class="command"><strong>opannotate</strong></span>, or <span class="command"><strong>opgprof</strong></span>.
</p>
<div class="sect1" title="1. Profile specifications">
<div class="titlepage">
<div>
<div>
<h2 class="title" style="clear: both"><a id="profile-spec"></a>1. Profile specifications</h2>
</div>
</div>
</div>
<p>
All of the analysis tools take a <span class="emphasis"><em>profile specification</em></span>.
This is a set of definitions that describe which actual profiles should be
examined. The simplest profile specification is empty: this will match all
the available profile files for the current session (this is what happens
when you do <span class="command"><strong>opreport</strong></span>).
</p>
<p>
Specification parameters are of the form <code class="option">name:value[,value]</code>.
For example, if I wanted to get a combined symbol summary for
<code class="filename">/bin/myprog</code> and <code class="filename">/bin/myprog2</code>,
I could do <span class="command"><strong>opreport -l image:/bin/myprog,/bin/myprog2</strong></span>.
As a special case, you don't actually need to specify the <code class="option">image:</code>
part here: anything left on the command line is assumed to be an
<code class="option">image:</code> name. Similarly, if no <code class="option">session:</code>
is specified, then <code class="option">session:current</code> is assumed ("current"
is a special name of the current / last profiling session).
</p>
<p>
In addition to the comma-separated list shown above, some of the
specification parameters can take <span class="command"><strong>glob</strong></span>-style
values. For example, if I want to see image summaries for all
binaries profiled in <code class="filename">/usr/bin/</code>, I could do
<span class="command"><strong>opreport image:/usr/bin/\*</strong></span>. Note the necessity
to escape the special character from the shell.
</p>
<p>
For <span class="command"><strong>opreport</strong></span>, profile specifications can be used to
define two profiles, giving differential output. This is done by
enclosing each of the two specifications within curly braces, as shown
in the examples below. Any specifications outside of curly braces are
shared across both.
</p>
<div class="sect2" title="1.1. Examples">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="profile-spec-examples"></a>1.1. Examples</h3>
</div>
</div>
</div>
<p>
Image summaries for all profiles with <code class="constant">DATA_MEM_REFS</code>
samples in the saved session called "stresstest" :
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">
# opreport session:stresstest event:DATA_MEM_REFS
</pre>
</td>
</tr>
</table>
<p>
Symbol summary for the application called "test_sym53c8xx,9xx". Note the
escaping is necessary as <code class="option">image:</code> takes a comma-separated list.
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">
# opreport -l ./test/test_sym53c8xx\,9xx
</pre>
</td>
</tr>
</table>
<p>
Image summaries for all binaries in the <code class="filename">test</code> directory,
excepting <code class="filename">boring-test</code> :
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">
# opreport image:./test/\* image-exclude:./test/boring-test
</pre>
</td>
</tr>
</table>
<p>
Differential profile of a binary stored in two archives :
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">
# opreport -l /bin/bash { archive:./orig } { archive:./new }
</pre>
</td>
</tr>
</table>
<p>
Differential profile of an archived binary with the current session :
</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">
# opreport -l /bin/bash { archive:./orig } { }
</pre>
</td>
</tr>
</table>
</div>
<div class="sect2" title="1.2. Profile specification parameters">
<div class="titlepage">
<div>
<div>
<h3 class="title"><a id="profile-spec-details"></a>1.2. Profile specification parameters</h3>
</div>
</div>
</div>
<div class="variablelist">
<dl>
<dt>
<span class="term">
<code class="option">archive:</code>
<span class="emphasis">
<em>archivepath</em>
</span>
</span>
</dt>
<dd>
<p>
A path to an archive made with <span class="command"><strong>oparchive</strong></span>.
Absence of this tag, unlike others, means "the current system",
equivalent to specifying "archive:".
</p>
</dd>
<dt>
<span class="term">
<code class="option">session:</code>
<span class="emphasis">
<em>sessionlist</em>
</span>
</span>
</dt>
<dd>
<p>
A comma-separated list of session names to resolve in. Absence of this
tag, unlike others, means "the current session", equivalent to
specifying "session:current".
</p>
</dd>
<dt>
<span class="term">
<code class="option">session-exclude:</code>
<span class="emphasis">
<em>sessionlist</em>
</span>
</span>
</dt>
<dd>
<p>
A comma-separated list of sessions to exclude.
</p>
</dd>
<dt>
<span class="term">
<code class="option">image:</code>
<span class="emphasis">
<em>imagelist</em>
</span>
</span>
</dt>
<dd>
<p>
A comma-separated list of image names to resolve. Each entry may be relative
path, <span class="command"><strong>glob</strong></span>-style name, or full path, e.g.</p>
<table xmlns="" border="0" style="background: #E0E0E0;" width="90%">
<tr>
<td>
<pre class="screen">opreport 'image:/usr/bin/oprofiled,*op*,./opreport'</pre>
</td>
</tr>
</table>
</dd>
<dt>
<span class="term">
<code class="option">image-exclude:</code>
<span class="emphasis">
<em>imagelist</em>
</span>
</span>
</dt>
<dd>
<p>
Same as <code class="option">image:</code>, but the matching images are excluded.
</p>
</dd>
<dt>
<span class="term">
<code class="option">lib-image:</code>
<span class="emphasis">
<em>imagelist</em>
</span>
</span>
</dt>
<dd>
<p>
Same as <code class="option">image:</code>, but only for images that are for
a particular primary binary image (namely, an application). This only
makes sense to use if you're using <code class="option">--separate</code>.
This includes kernel modules and the kernel when using
<code class="option">--separate=kernel</code>.
</p>
</dd>
<dt>
<span class="term">
<code class="option">lib-image-exclude:</code>
<span class="emphasis">
<em>imagelist</em>
</span>
</span>
</dt>
<dd>
<p>
Same as <code class="option">lib-image:</code>, but the matching images
are excluded.
</p>
</dd>
<dt>
<span class="term">
<code class="option">event:</code>
<span class="emphasis">
<em>eventlist</em>
</span>
</span>
</dt>
<dd>
<p>
The symbolic event name to match on, e.g. <code class="option">event:DATA_MEM_REFS</code>.
You can pass a list of events for side-by-side comparison with <span class="command"><strong>opreport</strong></span>.
When using the timer interrupt, the event is always "TIMER".
</p>
</dd>
<dt>
<span class="term">
<code class="option">count:</code>
<span class="emphasis">
<em>eventcountlist</em>
</span>
</span>
</dt>
<dd>
<p>
The event count to match on, e.g. <code class="option">event:DATA_MEM_REFS count:30000</code>.
Note that this value refers to the setting used for <span class="command"><strong>opcontrol</strong></span>
only, and has nothing to do with the sample counts in the profile data
itself.
You can pass a list of events for side-by-side comparison with <span class="command"><strong>opreport</strong></span>.
When using the timer interrupt, the count is always 0 (indicating it cannot be set).
</p>
</dd>
<dt>
<span class="term">
<code class="option">unit-mask:</code>
<span class="emphasis">
<em>masklist</em>
</span>
</span>
</dt>
<dd>
<p>
The unit mask value of the event to match on