same model & compiler + different CPUs = differ. solutio

General scientific issues regarding ROMS

Moderators: arango, robertson

Post Reply
Message
Author
User avatar
donguiyo
Posts: 29
Joined: Thu Mar 11, 2004 5:17 pm
Location: BOEM

same model & compiler + different CPUs = differ. solutio

#1 Unread post by donguiyo »

I ran the same code with identical initial conditions, grid and forcing etc (compiled both with 32bit PG compiler).

The model's kinetic energy is not the same. In the dual Xeon the model's kinetic energy/volume is 0.000001 units smaller than in a quad Opteron, i.e., only difference is the type of processors. It seems like a small number but if you multiply by the total volume of the ocean it is not. Also, the flow between some islands seems wrong in the Quad server with the Opteron processors. any clues ? thx. Guillermo
ps: yes, both codes were compiled in the computer were I ran them.

User avatar
donguiyo
Posts: 29
Joined: Thu Mar 11, 2004 5:17 pm
Location: BOEM

Re

#2 Unread post by donguiyo »

got it.

If one uses the PG linux86 compiler (on the opteron server) you get similar results but not quite what it should be.

So, things worked when I compiled with the PG linux86-64 compiler (opteron is 32-64bit).

Guillermo.

Post Reply