[aspect-devel] ": internal error: ** The compiler has encountered an unexpected problem.

[John Naliboff]

Hi Lev,

There was indeed a speedup between the different optimization levels, 
but off the top of my head I can't recall how much. That code was 
written in Fortran, used a direct solver, not parallel, etc, so the 
observed speedups may be completely different than with ASPECT. It would 
be interesting to see exactly how much the code speedups between 
optimization levels O0, O1 and O2 for a small model using 1 thread.

Cheers,
John

*************************************************
John Naliboff
Assistant Project Scientist, CIG
Earth & Planetary Sciences Dept., UC Davis

On 07/06/2017 10:31 AM, Lev Karatun wrote:
> Ok, will keep it in mind, thanks John!
> By the way, did you notice any difference in performance when you were 
> switching between optimization levels?
>
> Best regards,
> Lev Karatun.
>
> 2017-07-06 11:37 GMT-04:00 John Naliboff <jbnaliboff at ucdavis.edu 
> <mailto:jbnaliboff at ucdavis.edu>>:
>
>     Hi Lev,
>
>     Glad to hear the compilation worked. An anecdotal word of caution
>     with regard to the intel optimization levels - with a different
>     code we found some of our benchmarks were not passing when we
>     switched from level 2 to level 3 optimization. That may not be the
>     case here, but perhaps something to watch out for.
>
>     Cheers,
>     John
>
>     *************************************************
>     John Naliboff
>     Assistant Project Scientist, CIG
>     Earth & Planetary Sciences Dept., UC Davis
>
>     On 07/06/2017 07:57 AM, Lev Karatun wrote:
>>     Hi Wolfgang,
>>
>>     yes, I was actually trying to do it yesterday -- to compile with
>>     level 2 optimization and on 1 core. It actually compiled
>>     successfully on 1 core for both level 2 and 3 optimizations.
>>     Thanks everyone for the help!
>>
>>     Best regards,
>>     Lev Karatun.
>>
>>     2017-07-06 9:39 GMT-04:00 Wolfgang Bangerth <bangerth at tamu.edu
>>     <mailto:bangerth at tamu.edu>>:
>>
>>         On 07/05/2017 10:45 AM, Lev Karatun wrote:
>>
>>             The cluster support says it's recommended to use intel
>>             compilers because they provide significant speed-ups on
>>             their machines, and all modules are installed using icc
>>             (including trilinos and p4est).
>>
>>
>>         From all of our experiments, the difference between icc and
>>         gcc is actually quite small (at best a few percent). I
>>         wouldn't worry about it.
>>
>>
>>             If I want to install Aspect using gcc, does it mean I
>>             have to reinstall deal.II, trilinos and p4est using gcc
>>             first?
>>
>>
>>         Yes, probably. (It may be that you don't have to reinstall
>>         p4est because it's just a C library, but it may be easiest to
>>         just do that as well.)
>>
>>         But I don't see why you need to switch to gcc. I believe that
>>         the compiler just ran out of memory. You may as well just try
>>         to compile everything again with, say, 'make -j1' and see
>>         whether that makes the compilation succeed.
>>
>>         Best
>>          W.
>>
>>
>>         -- 
>>         ------------------------------------------------------------------------
>>         Wolfgang Bangerth          email: bangerth at colostate.edu
>>         <mailto:bangerth at colostate.edu>
>>                                    www:
>>         http://www.math.colostate.edu/~bangerth/
>>         <http://www.math.colostate.edu/%7Ebangerth/>
>>
>>
>>
>>
>>     _______________________________________________
>>     Aspect-devel mailing list
>>     Aspect-devel at geodynamics.org <mailto:Aspect-devel at geodynamics.org>
>>     http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>>     <http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel>
>
>
>     _______________________________________________
>     Aspect-devel mailing list
>     Aspect-devel at geodynamics.org <mailto:Aspect-devel at geodynamics.org>
>     http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel
>     <http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel>
>
>
>
>
> _______________________________________________
> Aspect-devel mailing list
> Aspect-devel at geodynamics.org
> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/aspect-devel

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.geodynamics.org/pipermail/aspect-devel/attachments/20170706/69f11711/attachment.html>