Menu

Discussion Forum

New Discussion
Filtrer

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

About Integration in Comsol

Posted 28 juin 2010, 09:12 UTC−4 3 Replies

chen wei
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Hi, guys!

I met a problem on Integration when I use Comsol to calculate Lennard-Jones Potential. I wanna know: in 3D, how to deal with "int_v1(dV1)int_v2(f(r)dV2)", dV1=dx1dy1dz1,dV2= dx2dy2dz2, v1 is subdomain1, v2 is subdomain2 and f(r) is the function of distance of dV1 and dV2, as shown EQ2 in attachment. Should I use PDE fomulation or integration coupling variables?

Thanks a lot!!!!!
Best Regards

Come on buddies!

Attachments:

3 Replies Last Post 30 juin 2010, 16:24 UTC−4
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 28 juin 2010, 10:37 UTC−4
Hi

have you checked how to use the "dest()" operator on the different source and destination domains ?

I believe this is jut writing out plainvise your integration equation, no ?

Have fun Comsoling
Ivar
Hi have you checked how to use the "dest()" operator on the different source and destination domains ? I believe this is jut writing out plainvise your integration equation, no ? Have fun Comsoling Ivar
chen wei
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 30 juin 2010, 12:38 UTC−4

Hi

have you checked how to use the "dest()" operator on the different source and destination domains ?

I believe this is jut writing out plainvise your integration equation, no ?

Have fun Comsoling
Ivar


Thank you! "dest" seems to be great!

But, one more question, when I use "integration coupling variable" in postprocessing, it appears only one CPU do work, actually I have 8 CPUs and use "-np 8 -64 ", however solving will, as expected, use all 8 CPUs. Is postprocessing can only use one CPU? Is there any way to work around it?

How about use PDE or ODE to tackle my problem?
[QUOTE] Hi have you checked how to use the "dest()" operator on the different source and destination domains ? I believe this is jut writing out plainvise your integration equation, no ? Have fun Comsoling Ivar [/QUOTE] Thank you! "dest" seems to be great! But, one more question, when I use "integration coupling variable" in postprocessing, it appears only one CPU do work, actually I have 8 CPUs and use "-np 8 -64 ", however solving will, as expected, use all 8 CPUs. Is postprocessing can only use one CPU? Is there any way to work around it? How about use PDE or ODE to tackle my problem?
Ivar KJELBERG COMSOL Multiphysics(r) fan, retired, former "Senior Expert" at CSEM SA (CH)
Send Private Message Flag post as spam

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 30 juin 2010, 16:24 UTC−4
Hi

well COMSOL set up the problems to multiple CPU when it can, but it cannot always, if you look carfully at a solving sequence (typially a non-linear one) you will notice that often you switch between 1 and N cpus, it all depends what COMSOL is doing, not all solvers nor all operations are (or can be easily) parallelised.

This is often wery frustrating when you are postprocessng a very large model with very mansy DOF (as need for solving correctly i.e. in RF or optics) but then you want to postprocess and often the desired values you are looking at do not have any need for the high mesh resolution, you still have to wait for ages to run through the hughe solver DB, and that for each plot.

Ideally, you should be able to define a sub-solution by decimation or 1/100 ... and then get your plots quicker. This will typically become an issue when you group to gether data from a large cluster, you would ned to have your graphics (post)processor run also hooked up on the supercluster/farm as the RAM required will not allow you to transfer it all, and then if only 1 CPU is running you will still need ages to dump a single plot.

So powerfull multiCPU PC's is not a full "quick" solution for large EM ...
Have fun Comsoling
Ivar
Hi well COMSOL set up the problems to multiple CPU when it can, but it cannot always, if you look carfully at a solving sequence (typially a non-linear one) you will notice that often you switch between 1 and N cpus, it all depends what COMSOL is doing, not all solvers nor all operations are (or can be easily) parallelised. This is often wery frustrating when you are postprocessng a very large model with very mansy DOF (as need for solving correctly i.e. in RF or optics) but then you want to postprocess and often the desired values you are looking at do not have any need for the high mesh resolution, you still have to wait for ages to run through the hughe solver DB, and that for each plot. Ideally, you should be able to define a sub-solution by decimation or 1/100 ... and then get your plots quicker. This will typically become an issue when you group to gether data from a large cluster, you would ned to have your graphics (post)processor run also hooked up on the supercluster/farm as the RAM required will not allow you to transfer it all, and then if only 1 CPU is running you will still need ages to dump a single plot. So powerfull multiCPU PC's is not a full "quick" solution for large EM ... Have fun Comsoling Ivar

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.

Suggested Content