ripc(1)ripc(1)NAMEripc - collect performance counter information from an application
SYNOPSISripc args [ app params | -p pid ]
DESCRIPTION
The ripc command produces a report of profile counter information
either over the entire run of an application, or over a sample of the
runtime of an existing pid.
OPTIONS
The following options are supported:
-c Output a list of the top four counters, paired up so that they
can be used as arguments for collect.
-f filename
Name of file to be used for graph. The default name is coun‐
ters.ps
-g If gnuplot is on the path, produce a graph of the performance
counter events, and other collected data, over time.
-h Print help information.
-o resultfile
File where the results are to be placed, by default the results
will go to stdout.
-p pid Process ID to track. This option cannot be used at the same time
as specifying commandline of an application.
-r Print raw counter data in addition to the derived stall info.
By default ripc will collect data about performance counters,
and compute information which is most likely to be interesting
to users (the derived stall information). If -r is specified,
ripc will also print the raw data (in addition to the derived
stall information).
-s Run the program multiple times collecting different pairs of
counters on each run. This option should be used when the pro‐
gram only runs for a short duration. Since the default is to
rotate through the performance counters, short running programs
may not accumulate sufficient data otherwise.
-t seconds
Specify the duration in seconds over which data should be col‐
lected. The default is 300 seconds (5 minutes).
-V Output version information.
-v Generate debug output
-x Collect extended counter information.
OUTPUT
The output of ripc comprises the following sections:
Stall time data
Estimates of the runtime lost due to the various processor stall
events.
Events per instruction
When extended data is collected, the number of events per
instruction is tracked.
Floating point traps
A count of the system-wide number of incomplete floating point
operations encountered during the run of the application.
Memory usage and time
Amount of memory that the process uses, and the amount of system
vs user time the process has accumulated.
SEE ALSOanalyzer(1), collect(1), er_archive(1), er_cp(1), er_export(1),
er_mv(1), er_print(1), er_rm(1), er_src(1), and the Performance Ana‐
lyzer manual.
January 2007 ripc(1)