Managing system performance

Formatting the collected data (prfpr)

The prfpr command formats the contents of file (data that was collected by prfdc or prfsnap). The prfpr command has the following format:

   /usr/sbin/prfpr [-P processor_id[, . . . ] | ALL] [-t] file [cutoff]
Each text address is converted to a system function name, and the percentage of time used by that function is printed if the activity percentage is greater than the cutoff number you specify. The range of cutoff is 0 percent to 99.99 percent; that is the value of cutoff can be any decimal value greater than or equal to zero and less than 100. When the value is 0, all non-zero contents print. The default cutoff is 1 percent.

The following screen display shows the output of the prfpr command.

   # /usr/sbin/prfpr temp 0.05

09/19/95 14:27:48 09/19/95 14:28:48

sfs_readi 0.06 sfs_getpage 0.06 struct_zero 0.13 bcopy 0.86 segmap_getmap 0.08 segmap_release 0.05 idle 95.35 systrap 0.06 lock_nodbg 0.28 unlock_nodbg 1.03 rw_unlock 0.07 v3read 0.06 user 0.28

The function calls in the kernel are listed. For more information about the function calls, refer to the manual pages, or contact an experienced user.

In normal use, the profiler characterizes a workload by its kernel resource usage, in which case percentage time per function is a useful statistic. When you want to know the absolute amount of time spent executing each function, use prfpr -t to report the accumulated counter difference in clock ticks instead of percentages.

On multiprocessing systems, prfpr accepts the -P option to report per-processor statistics, or ALL can be specified to report on all processors. The default behavior on multiprocessing systems is to report the system-wide mean.

Next topic: Displaying and changing parameters
Previous topic: Collecting profiling data (prfdc, prfsnap)

© 2007 The SCO Group, Inc. All rights reserved.
SCO OpenServer Release 6.0.0 -- 05 June 2007