to overflow or not to overflow ... with different MohidWater

MOHID Water forum. Questions and discussion around MOHID Water
vvieira@ualg.pt
MOHID Beginner
Posts: 13
Joined: Thu Jan 01, 1970 00:00

to overflow or not to overflow ... with different MohidWater

Postby vvieira@ualg.pt » Mon Nov 07, 2011 23:02

Dear supporters,

I run a stabilization simulation with Mohid Water 4.9.2 (64bits) default exe and it overflows (floating overflow) with a dt=2 and a courrant arround 2.27. I run the same simulation with MOHID_Intel_Single_Traceback.exe and it goes just fine, the hdf5 outputted are fine ... no problems! ... ??? ... how come?
I saw a related post by anamartin but in this case it was a matter of different pcs with different processors. Now it is the same pc with equivalent MohidWater.exe that are supposed to have the same code lines for the calculus, right?
I guess I could simply keep using the MOHID_Intel_Single_Traceback.exe. However, I think this one is much slower, isn't it?

cheers,
Vasco

frank.maretec@ist.utl.pt
MOHID Guru
Posts: 151
Joined: Thu Jan 01, 1970 00:00
Location: Technical University of Lisbon
Contact:

Re: to overflow or not to overflow ... with different MohidW

Postby frank.maretec@ist.utl.pt » Tue Nov 08, 2011 21:58

Dear all,

this is not an easy question but must urgently be addressed. During the last course we saw that in some ocasions that a recent executable, compiled without specific optimizations, didn't run on AMD processor machines when activating discharges...

That all what I can say right now...

Happy MOHIDing,
Frank
In order to understand recursivity you must first understand recursivity.

vvieira@ualg.pt
MOHID Beginner
Posts: 13
Joined: Thu Jan 01, 1970 00:00

Re: to overflow or not to overflow ... with different MohidW

Postby vvieira@ualg.pt » Wed Nov 09, 2011 18:36

Dear all,

My processor is an Intel Core 2 Extreme, Quad Core. The machine runs in 64 bits.
Now that I updated the MOHID Studio (and therefore also the MohidWater within it) to the new version made available last week, the problem was solved! It all goes well! I did absolutely anything else! So, I guess this was one more issue that was solved with the new developments. Whatever it was that you changed, it solved this problem!
Moreover, using the:
older MohidWaterIntelTraceback.exe took 3h43m to do the calculus,
new MohidWater_single.exe took 1h20m to do the calculus,
new MohidWater_single_openmp.exe took 50m to do the calculus,
new MohidWater_double_openmp.exe took 1h20m to do the calculus.

I would like to take this chance to aknowledge the MOHID developers, Frank, Guillaume, Luís and whoever I may be forgeting, for their developments as I really am amazed with the performance improvements. As I remember, some years ago it took almost one day! But I may be mistaken as it seams too much of a difference!

Nevertheless, the MOHID Studio Map is still giving me trouble with large hdf5!

Best wishes


Return to “MOHID Water”

Who is online

Users browsing this forum: No registered users and 3 guests