20012013
Ericsson AB. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
fprof - The File Trace Profiler
Raimo Niskanen
nobody
nobody
no
2001-08-14
PA1
fprof_chapter.xml
fprof is a profiling tool that can be used to get a picture of
how much processing time different functions consumes and in which
processes.
fprof uses tracing with timestamps to collect profiling
data. Therfore there is no need for special compilation of any
module to be profiled.
fprof 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 own time i.e the
time used by a function for its own execution, and
accumulated time i.e execution time including called
functions.
Profiling is essentially done in 3 steps:
1
- Tracing; to file, as mentioned in the previous paragraph.
2
- 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.
3
- Analysing; the raw profile data is sorted and dumped
in text format either to file or console.
Since fprof 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 /tmp is usually a good choice,
while any NFS mounted disk is a lousy choice.
Fprof can also skip the file step and trace to a tracer process
of its own that does the profiling in runtime.
The following sections show some examples of how to profile with
Fprof. See also the reference manual
fprof(3).
Profiling from the source code
If you can edit and recompile the source code, it is convenient
to insert fprof:trace(start) and
fprof:trace(stop) before and after the code to be
profiled. All spawned processes are also traced. If you want
some other filename than the default try
fprof:trace(start, "my_fprof.trace").
Then read the trace file and create the raw profile data with
fprof:profile(), or perhaps
fprof:profile(file, "my_fprof.trace") for non-default
filename.
Finally create an informative table dumped on the console with
fprof:analyse(), or on file with
fprof:analyse(dest, []), or perhaps even
fprof:analyse([{dest, "my_fprof.analysis"}, {cols, 120}])
for a wider listing on non-default filename.
See the fprof(3) manual page
for more options and arguments to the functions
trace,
profile
and
analyse.
Profiling a function
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
fprof:apply(Module, Function, Args) and related for the
tracing step.
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
fprof:apply(M, F, Args, [continue | OtherOpts]).
The tracing has to be stopped at a suitable later time using
fprof:trace(stop).
Immediate profiling
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.
Do something like this:
{ok, Tracer} = fprof:profile(start),
fprof:trace([start, {tracer, Tracer}]),
%% Code to profile
fprof:trace(stop);
This puts less load on the filesystem, but much more on the
Erlang runtime system.