[resending: it does not seem to have reached the case log the first
time]
I am very pleased to sponsor the following case for Chad Mynhier. This
case enhances the ptime(1) utility to provide greater precision and
adds useful features.
The timer is set for 10/22/2007.
Note that this is an open case submitted by a community member; please
remember to CC Chad on all correspondence.
Thanks,
-dp
------- ------- ------- ------- ------- ------- ------- ------- -------
ptime(1) Improvements
Chad Mynhier <cmynhier at gmail.com>
SUMMARY
This fast-track enhances the ptime command to address two existing
RFEs[1,2] requesting higher resolution for reporting (microseconds
or nanoseconds), the option to see the full complement of microstate
statistics, and the ability to see statistics for a currently running
process.
The new options are committed interfaces; this case seeks minor
release binding.
INTERFACE TABLE
Interface Stability Binding
----------------------------------------------------------------
Output Uncommitted Minor
-m option Committed Minor
-p option Committed Minor
----------------------------------------------------------------
DETAILS
Overview
Currently, the ptime command reports three statistics for a process,
real (i.e., wall clock) time, user time and system time, where these
statistics are presented at a millisecond resolution. In addition,
ptime can only report these statistics for a process started as a
child of the command. It does not provide the ability to look at these
statistics for an existing process.
We plan to address these issues as follows:
1. We will change the default resolution for the statistics reported
by ptime to nanoseconds instead of milliseconds. (Note that this
may break scripts that match any number of digits to the right of
the decimal point but assume that the resolution is milliseconds.
But note also that the man page states that the human-readable
output is unstable.)
2. We will add a -m option that will generate a report containing the
full set of microstate statistics.
3. We will add a -p <pid> option to allow ptime to attach to a running
process and print a snapshot of the statistics for that process.
EXAMPLE
This example demonstrates nanosecond-resolution reporting:
# ./ptime /bin/ls > /dev/null
real 0.001734904
user 0.000433082
sys 0.001190076
#
This example demonstrates full microstate reporting:
# ./ptime -m /bin/ls > /dev/null
real 0.001680855
user 0.000430206
sys 0.001139441
trap 0.000000949
tflt 0.000000000
dflt 0.000000000
kflt 0.000000000
lock 0.000000000
slp 0.000000000
lat 0.000004242
stop 0.000007332
#
This example demonstrates attaching to an existing process:
# ./ptime -m -p `pgrep syslogd`
real 49:06:52.585250745
user 0.035726760
sys 0.034103073
trap 0.000154120
tflt 0.000000000
dflt 0.000000000
kflt 0.000000000
lock 343:48:03.815831359
slp 245:34:21.457591100
lat 0.065530109
stop 0.000043641
#
REFERENCES
[1] ptime should report microseconds or nanoseconds instead of just
milliseconds
(http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=6234106)
[2] ptime should report resource usage statistics from /proc for running
processes
(http://bugs.opensolaris.org/bugdatabase/view_bug.do?bug_id=4532599)
------- ------- ------- ------- ------- ------- ------- ------- -------
--
Daniel Price - Solaris Kernel Engineering - dp at eng.sun.com - blogs.sun.com/dp
