Hi Roy (and Jed),

I figured out the cause of the problem: -DNDEBUG in CXXFLAGS. With this 
flag on, the program returned bad convergence results both on my desktop 
and on the cluster. Without this flag, both machines yielded good 
results. So this problem is not to the intel compiler on the cluster.

Adding "-O2 -fp-model precise" and turning on "-snes_linesearchmonitor", 
as you suggested, changed the convergence behavior slightly but not 
fundamentally.

The intel compiler version I am using is "Compiler/11.0/083/". It is a 
quite old version.

Many thanks to both of you for your helpful advice.

Cheers,
Dafang

On 01/08/2014 03:57 PM, Roy Stogner wrote:
>
> On Wed, 8 Jan 2014, Dafang Wang wrote:
>
>> Thanks for your prompt reply. Did you mean turning -O2 "on" or 
>> "off"?  I got bad results when turning it on.
>
> I mean using "-O2 -fp-model precise".  IIRC that often gets icpc to
> turn off its inexact-in-floating-point optimizations without forcing
> it to turn off any optimizations which don't change the final code
> output.
>
> No guarantees, though.  I dug into the compiler configuration files
>
> https://github.com/manufactured-solutions/MASA/blob/master/m4/common/ax_cxx_minopt.m4
>  
>
>
> from the manufactured solutions code where we encountered these
> problems, to see which options we'd ended up needing, and it looks
> like we eventually gave up on any optimizations at all when we
> demanded maximum accuracy there.
> ---
> Roy

-- 
Dafang Wang, Ph.D
Postdoctoral Fellow
Institute of Computational Medicine
Department of Biomedical Engineering
Johns Hopkins University
Hackerman Hall Room 218
Baltimore, MD, 21218

------------------------------------------------------------------------------
Rapidly troubleshoot problems before they affect your business. Most IT 
organizations don't have a clear picture of how application performance 
affects their revenue. With AppDynamics, you get 100% visibility into your 
Java,.NET, & PHP application. Start your 15-day FREE TRIAL of AppDynamics Pro!
http://pubads.g.doubleclick.net/gampad/clk?id=84349831&iu=/4140/ostg.clktrk
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users

Reply via email to