the bash built-in time provides far less data than the /usr/bin/time does.
One issue to think about is if you are looking to see what's happening
with the process while it's running (in which case Andrew's program,
iotop, etc help), or if you are looking for the totals after the run is
completed (in which case anything looking in proc is going to miss some
info)
David Lang
On Tue, 19 Jun 2012, Andrew Hume wrote:
Date: Tue, 19 Jun 2012 11:02:15 -0700
From: Andrew Hume <[email protected]>
To: [email protected]
Cc: Christopher R Webber <[email protected]>, [email protected]
Subject: Re: [lopsa-tech] Job/Process Profiling
because of the variability of whose "time" you get and so forth, i gave up and
wrote a linux-specific
function to do this. there are of course pros and cons, but take this if you
want:
#include "basic.h"
#include "libng.h"
#include "libw.h"
#include <asm-i386/param.h>
void
resource_summary(FILE *fp)
{
char buf[4096];
FILE *mp;
int r_tps, r_rss, r_sp, r_text, r_lib, r_data, r_dt, r_pid;
unsigned long r_utime, r_stime, r_vsize;
long r_nthread;
sprintf(buf, "/proc/%d/statm", getpid());
if((mp = fopen(buf, "r")) == NULL){
perror(buf);
return;
}
if(fscanf(mp, "%d %d %d %d %d %d %d", &r_tps, &r_rss, &r_sp, &r_text, &r_lib,
&r_data, &r_dt) != 7){
fprintf(stderr, "bad format statm\n");
fclose(mp);
return;
}
fclose(mp);
sprintf(buf, "/proc/%d/stat", getpid());
if((mp = fopen(buf, "r")) == NULL){
perror(buf);
return;
}
if(fscanf(mp, "%d %*s %*c %*d %*d %*d %*d %*d %*u %*u %*u %*u %*u %lu \
%lu %*d %*d %*d %*d %ld %*u %*u %lu %*u %*u %*u %*u %*u
\
%*u %*u %*u %*u %*u %*u %*u %*u %*d %*d %*u %*u %*u %*u %*d
",
&r_pid, &r_utime, &r_stime, &r_nthread, &r_vsize) != 5){
fprintf(stderr, "bad format stat\n");
fclose(mp);
return;
}
fclose(mp);
fprintf(fp, "tps=%d rss=%d sp=%d text=%d lib=%d data=%d pid=%d utime=%.1fs
stime=%.1fs nthr=%ld vs=%.1fMB\n",
r_tps, r_rss, r_sp, r_text, r_lib, r_data, r_pid,
r_utime/(double)HZ, r_stime/(double)HZ, r_nthread, r_vsize/(1024*1024.0));
}
of course, its only doing what i was interested in but it may be a useful start.
On Jun 19, 2012, at 10:27 AM, [email protected] wrote:
/usr/bin/time will provide a lot of stats, take a look and see how close it
comes to what you want.
David Lang
On Tue, 19 Jun 2012, Christopher R Webber wrote:
Date: Tue, 19 Jun 2012 05:34:14 -0700
From: Christopher R Webber <[email protected]>
To: [email protected]
Subject: [lopsa-tech] Job/Process Profiling
All,
I have students and postdocs that write software as part of their research and
occasionally they want to answer questions like:
- How much memory does this take?
- How much time does this take?
- What does CPU utilization look like?
To add to the list above, I would love to see I/O statistics as well. Any
suggestions on a command line profiling tool for linux that will do this? While
most of my users are savy enough to use the command line, I would like to see
something where you don't have to understand a whole lot of details to get some
basic info.
Ideas?
Thanks.
-- cwebber
Christopher Webber - Systems Administrator
Bioinformatics - University of California, Riverside
Twitter: @cwebber
Tel: 951.867.7108
http://cwebber.ucr.edu
_______________________________________________
Tech mailing list
[email protected]
https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech
This list provided by the League of Professional System Administrators
http://lopsa.org/
_______________________________________________
Tech mailing list
[email protected]
https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech
This list provided by the League of Professional System Administrators
http://lopsa.org/
------------------
Andrew Hume (best -> Telework) +1 623-551-2845
[email protected] (Work) +1 973-236-2014
AT&T Labs - Research; member of USENIX and LOPSA
_______________________________________________
Tech mailing list
[email protected]
https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech
This list provided by the League of Professional System Administrators
http://lopsa.org/
