This is the mail archive of the
systemtap@sourceware.org
mailing list for the systemtap project.
Re: static user probe performance
- From: fche at redhat dot com (Frank Ch. Eigler)
- To: systemtap at sourceware dot org
- Date: Wed, 07 Apr 2010 10:43:12 -0400
- Subject: Re: static user probe performance
- References: <4BABACFE.60406@redhat.com> <4BB65994.5070906@redhat.com> <y0mvdc6exho.fsf@fche.csb>
>> run with stap
>> [...]
>> elapsed time
>> volatile semaphore 1858.71
>> no volatile no semaphore 1865.68
>> volatile no semaphore 1873.07
>> no volatile semaphore 1878.09
>
> So for this particular microbenchmark, systemtap probing imposes a 30x
> slowdown, yikes. How many probe hits does this represent?
>From scox's later numbers, that's 192 million hits, for about 9
us/hit, which seems rather high. (In this microbenchmark, it seems
that the instrumentation is taking an impressive 98% of the cpu time,
which ought to have triggered the overload protection. I assume there
was an -DSTP_NO_OVERLOAD in effect.)
- FChE