XPROF

Section: User Commands (1)
Updated: 7th July 1992
Index Return to Main Contents

NAME

xprof - X11 client performance profiler.

SYNOPSIS

xprof [ -options ]

DESCRIPTION

xprof is a performance profiler for client programs in the X Window System. It follows a protocol-level profiling methodology. Since the X Window System follows the client-server model of computing, accurate and complete profiles are difficult to arrive at by means of conventional execution profilers. A client program could invoke considerable computation at the display server, or X Server, and this computation is an important part of the execution profile. xprof generates meaningful profiles for X Window applications by estimating the time spent in servicing the request messages at the display server. The central idea is to analyze a protocol-level trace of the interaction between the application and the display server and thereby construct an execution profile from the trace and a set of supplied metrics about the target display server.

xprof is used along with two associated programs: the X Window protocol tracing tool xscope and the X11 Server performance measurement program xmeasure.

OPTIONS

xprof accepts the options listed below:

<tracefile>: The trace of the protocol level interaction between the client program and the X server. Typically, this trace is produced by running xscope, the protocol tracing tool, in its most verbose mode, i.e., xscope -v3 > tracefile
-pperffile: Specify the X Server performance metrics file. Typically this file is produced by running the program xmeasure on the target X Server, i.e., xmeasure > perffile
-SNetSpeed: Speed of the target network in bytes per second.
-LNetLatency: Network latency in milliseconds.
When these network parameters are specified, xprof estimates the effect of network transport on the execution profile. Otherwise, the profile reflects only the time spent in executing requests at the server.
-bbuckets: Size of histogram for the size distribution of requests (default 4096). Larger numbers will result in a finer grain of measurement, but will cause higher usage of heap memory at run-time.
-strapsignal: Should xprof trap run-time signals.
0 => do not trap any signals.
1 => trap the following signals: SIGHUP causes xprof to print out current statistics and continue processing the input. SIGINT, SIGQUIT, SIGFPE, SIGBUS, SIGSEGV, and SIGXCPU cause xprof to print out statistics and terminate.
-vverboselevel: How verbose should the output statistics be.
0 => Print overviews of statistics.
1 => Also, print the range, mode, median, mean, and standard deviations of the size and
interarrival time distributions of the requests.
2 => Also, print the raw data for the above distributions.

USAGE

In order to use xprof you need to collect protocol-level traces of your client execution and also obtain the parameters of interest for your target server.

The X Window program xscope, which is distributed with the source code from MIT, may be used to collect the protocol-level trace. In order to collect the trace you would typically run it as follows:

xscope -v3 -hhostname > tracefile
where hostname is the host machine of the X Window Server. Next, start the client program as follows: xclient -display hostname:1
This procedure will set up the client program to communicate with xscope, which will pass on all protocol-level interaction to the X Server at hostname while collecting the trace in tracefile. For furthur information consult the xscope documentation.

The server parameters may be collected by the xmeasure program, which is distributed as part of the xprof package. Typically it should suffice to collect the server parameters by running xmeasure as follows for a local connection on a freshly started server.

xmeasure > perffile
For furthur information consult the xmeasure documentation.

X DEFAULTS

There are no X defaults used by this program.

COPYRIGHT

AUTHOR

Aloke Gupta

This document was created by man2html, using the manual pages.
Time: 10:14:13 GMT, January 04, 2023