[R] ?summaryRprof running at 100% cpu for one hour ...

Mike Marchywka marchywka at hotmail.com
Mon Nov 22 16:13:01 CET 2010

> Date: Mon, 22 Nov 2010 12:03:54 -0300
> From: kjetilbrinchmannhalvorsen at gmail.com
> To: ligges at statistik.tu-dortmund.de
> CC: r-help at r-project.org
> Subject: Re: [R] ?summaryRprof running at 100% cpu for one hour ...
> see below.
> 2010/11/21 Uwe Ligges :
> >
> >
> > On 21.11.2010 18:13, Kjetil Halvorsen wrote:
> >>> ?save.image
> >>
> >> And at this point it has been running with one cpu at 100% for over an
> >> hour!
> >
> >
> > It's OK to take an hour (due to memory <-> disc IO) if it uses swap space
> > heavily. Factor of 60 is not much given memory is faster than harddiscs by
> > orders of magnitude.
> >
> > Uwe
> It takes much more than an hour! I started anew a process with the
> problem yesterday aroun 18.00, had to kill it this morning around
> 09.00. That's more than 1|5 hours.

Again, see if you can run it under gdb or at least look at
tools you have to determine page faults. My brain has been corrupted
with 'dohs but in task manager CPU usage drops when page faults start
or lock startvation etc. A blocking thread should yield IIRC. Waiting
for it to die a natural death may not be practical. 

I just posted something on this after following another's suggestion but
it should be easy for you to get developer tools, execute gdb,
point it at R and then break a few times. Debuggers don't speed anything
up but presumably it gets into its limit cycle ( infinite futile loop )
within a short time. Also sometimes you get these loops due to memory corruption
with native code etc etc so confusing results may take a few different approaches
to figure out. 

Turning on profiling will at best destry any memory coherence and worse
ad to VM thrashing. At least try to determine if you are faulting all over.

> Kjetil
> >
> ______________________________________________
> R-help at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
> and provide commented, minimal, self-contained, reproducible code. 		 	   		  

More information about the R-help mailing list