Skip site navigation (1)Skip section navigation (2)
Date:      Fri, 29 Jul 2005 12:03:37 +0200
From:      Suleiman Souhlal <ssouhlal@FreeBSD.org>
To:        Joseph Koshy <jkoshy@FreeBSD.org>
Cc:        Martin Cracauer <cracauer@cons.org>, freebsd-current@FreeBSD.org
Subject:   Re: PMC (performance counters) problems - kernel panic in 6-beta1 
Message-ID:  <63730E21-8766-4A87-8878-EACE167E64C5@FreeBSD.org>
In-Reply-To: <20050729031107.C2FD616A41F@hub.freebsd.org>
References:  <20050729031107.C2FD616A41F@hub.freebsd.org>

next in thread | previous in thread | raw e-mail | index | archive | help
Hi,

On Jul 29, 2005, at 5:11 AM, Joseph Koshy wrote:

>
>
>
>> Another question, what is the difference of -s versus -S and -p  
>> versus
>> -P, respectively?
>>
>
> -S and -P turn on sampling.  -S does 'whole system' sampling.  -P is
> process specific (i.e., the counters only count when the target  
> process(es)
> are running).
>
> So:
>
>  # pmcstat -S instructions -O /tmp/foo.out
>
> will collect samples.
>
>  # pmcstat -g -R /tmp/foo.out
>
> will convert the sample log to gprof(1) compatible 'gmon.out' files.
>
> Sampling on the AMD64 requires a patch to current [I'll commit it
> shortly].

Is sampling supposed to work on i386? Because it doesn't seem to  
collect anything. Counting works fine, on the other hand.

# pmcstat -P instructions -O sample.out -d fsck_ffs /usr > /dev/null &
[1]  + done       pmcstat -P instructions -O sample.out -d fsck_ffs / 
usr > /dev/null
# pmcstat -R sample.out -g
# ls k7-retired-instructions
#

--
Suleiman Souhlal     | ssouhlal@vt.edu
The FreeBSD Project  | ssouhlal@FreeBSD.org




Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?63730E21-8766-4A87-8878-EACE167E64C5>