<?xml version="1.0" encoding="latin1" ?>
<!DOCTYPE chapter SYSTEM "chapter.dtd">
<chapter>
<header>
<copyright>
<year>2001</year><year>2009</year>
<holder>Ericsson AB. All Rights Reserved.</holder>
</copyright>
<legalnotice>
The contents of this file are subject to the Erlang Public License,
Version 1.1, (the "License"); you may not use this file except in
compliance with the License. You should have received a copy of the
Erlang Public License along with this software. If not, it can be
retrieved online at http://www.erlang.org/.
Software distributed under the License is distributed on an "AS IS"
basis, WITHOUT WARRANTY OF ANY KIND, either express or implied. See
the License for the specific language governing rights and limitations
under the License.
</legalnotice>
<title>fprof - The File Trace Profiler</title>
<prepared>Raimo Niskanen</prepared>
<responsible>nobody</responsible>
<docno></docno>
<approved>nobody</approved>
<checked>no</checked>
<date>2001-08-14</date>
<rev>PA1</rev>
<file>fprof_chapter.xml</file>
</header>
<p><c>fprof</c> is a profiling tool that can be used to get a picture of
how much processing time different functions consumes and in which
processes.
</p>
<p><c>fprof</c> uses tracing with timestamps to collect profiling
data. Therfore there is no need for special compilation of any
module to be profiled.
</p>
<p><c>fprof</c> presents wall clock times from the host machine OS,
with the assumption that OS scheduling will randomly load the
profiled functions in a fair way. Both <em>own time</em> i.e the
time used by a function for its own execution, and
<em>accumulated time</em> i.e execution time including called
functions.
</p>
<p>Profiling is essentially done in 3 steps:</p>
<taglist>
<tag><c>1</c></tag>
<item>Tracing; to file, as mentioned in the previous paragraph.</item>
<tag><c>2</c></tag>
<item>Profiling; the trace file is read and raw profile data is
collected into an internal RAM storage on the node. During
this step the trace data may be dumped in text format to file
or console.</item>
<tag><c>3</c></tag>
<item>Analysing; the raw profile data is sorted and dumped
in text format either to file or console.</item>
</taglist>
<p>Since <c>fprof</c> uses trace to file, the runtime performance
degradation is minimized, but still far from negligible,
especially not for programs that use the filesystem heavily
by themselves. Where you place the trace file is also important,
e.g on Solaris <c>/tmp</c> is usually a good choice,
while any NFS mounted disk is a lousy choice.
</p>
<p>Fprof can also skip the file step and trace to a tracer process
of its own that does the profiling in runtime.
</p>
<p>The following sections show some examples of how to profile with
Fprof. See also the reference manual
<seealso marker="fprof">fprof(3)</seealso>.
</p>
<section>
<title>Profiling from the source code</title>
<p>If you can edit and recompile the source code, it is convenient
to insert <c>fprof:trace(start)</c> and
<c>fprof:trace(stop)</c> before and after the code to be
profiled. All spawned processes are also traced. If you want
some other filename than the default try
<c>fprof:trace(start, "my_fprof.trace")</c>.
</p>
<p>Then read the trace file and create the raw profile data with
<c>fprof:profile()</c>, or perhaps
<c>fprof:profile(file, "my_fprof.trace")</c> for non-default
filename.
</p>
<p>Finally create an informative table dumped on the console with
<c>fprof:analyse()</c>, or on file with
<c>fprof:analyse(dest, [])</c>, or perhaps even
<c>fprof:analyse([{dest, "my_fprof.analysis"}, {cols, 120}])</c>
for a wider listing on non-default filename.
</p>
<p>See the <seealso marker="fprof">fprof(3)</seealso> manual page
for more options and arguments to the functions
<seealso marker="fprof#trace">trace</seealso>,
<seealso marker="fprof#profile">profile</seealso>
and
<seealso marker="fprof#analyse">analyse</seealso>.
</p>
</section>
<section>
<title>Profiling a function</title>
<p>If you have one function that does the task that you want to
profile, and the function returns when the profiling should
stop, it is convenient to use
<c>fprof:apply(Module, Function, Args)</c> and related for the
tracing step.
</p>
<p>If the tracing should continue after the function returns, for
example if it is a start function that spawns processes to be
profiled, you can use
<c>fprof:apply(M, F, Args, [continue | OtherOpts])</c>.
The tracing has to be stopped at a suitable later time using
<c>fprof:trace(stop)</c>.
</p>
</section>
<section>
<title>Immediate profiling</title>
<p>It is also possible to trace immediately into the profiling
process that creates the raw profile data, that is to short
circuit the tracing and profiling steps so that the filesystem
is not used.
</p>
<p>Do something like this:</p>
<pre>
{ok, Tracer} = fprof:profile(start),
fprof:trace([start, {tracer, Tracer}]),
%% Code to profile
fprof:trace(stop);</pre>
<p>This puts less load on the filesystem, but much more on the
Erlang runtime system.
</p>
</section>
</chapter>
