Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

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.

creating plots is extremely slow

Please login with a confirmed email address before reporting spam

Dear Forum,

creating simple line plots (with global scope) of model results is extremely slow.
For quite large model files (2-3 GB of size) it sometimes takes several hours for one simple plot on a machine with 4 GB RAM and 8 cores. COMSOL needs more time for creating the graphical output of the results than for solving the model.

Did anyone experience the same? Is there a way to accelerate graphical output?

Thanks for hints on this issue,
Juergen

4 Replies Last Post 21 août 2013, 10:31 UTC−4
Sven Friedel COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 21 août 2013, 04:05 UTC−4
Hi Jürgen,

we would need to know a bit better what kind of expressions you want to display.
Note that if you have integrals to be evaluated and maybe integrals of integrals: such evaluations are done only during the postprocessing, and all over again. If you want to have such outputs quickly, you can define an ODE with a degree of freedom in whcih the reult of that expression is stored. Then it is part of the solution and can be displayed right away.

If this does not help, feel free to send us a workable example in support, i.e. a file that si not GB big and takes hours but sufficeintly small and fast but illustrates your problem. If necessary use our FTP site to submit the file.

Best regards,

Sven
Hi Jürgen, we would need to know a bit better what kind of expressions you want to display. Note that if you have integrals to be evaluated and maybe integrals of integrals: such evaluations are done only during the postprocessing, and all over again. If you want to have such outputs quickly, you can define an ODE with a degree of freedom in whcih the reult of that expression is stored. Then it is part of the solution and can be displayed right away. If this does not help, feel free to send us a workable example in support, i.e. a file that si not GB big and takes hours but sufficeintly small and fast but illustrates your problem. If necessary use our FTP site to submit the file. Best regards, Sven

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 21 août 2013, 08:59 UTC−4
Dear Sven,

thanks for your quick and helpful reply.
There are indeed some "intops" in the global plot expressions. I was not aware of the fact that integrals are evaluated "live" during postprocessing.
As you suggested I have put these integrals into an ODE and I am currently running the model to see if it improves the speed of graphical output.
If it does not help I would like to come back to your offer and send you a small example model that illustrates the problem.

Thanks and best regards,
Juergen
Dear Sven, thanks for your quick and helpful reply. There are indeed some "intops" in the global plot expressions. I was not aware of the fact that integrals are evaluated "live" during postprocessing. As you suggested I have put these integrals into an ODE and I am currently running the model to see if it improves the speed of graphical output. If it does not help I would like to come back to your offer and send you a small example model that illustrates the problem. Thanks and best regards, Juergen

Edgar J. Kaiser Certified Consultant

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 21 août 2013, 09:17 UTC−4
Hi Sven and Jürgen,

I am sometimes encountering this issue when doing far field plots in RF simulations. It is not that bad for line plots like polar antenna pattern diagrams, but it gets really time consuming in any 3D plots. Fortunately for the quantitative work the polar plots mostly do the job and 3D patterns are a nice to have. But sometimes you would like to have them, and then it easily takes longer than the solution time.

Could implementing the respective equations to make the far field data part of the solution be a good way here? Is COMSOL considering this in order to make far field plots easier and quicker? Or would we just extend the time for solving this way?

Cheers
Edgar

--
Edgar J. Kaiser
emPhys Physical Technology
Hi Sven and Jürgen, I am sometimes encountering this issue when doing far field plots in RF simulations. It is not that bad for line plots like polar antenna pattern diagrams, but it gets really time consuming in any 3D plots. Fortunately for the quantitative work the polar plots mostly do the job and 3D patterns are a nice to have. But sometimes you would like to have them, and then it easily takes longer than the solution time. Could implementing the respective equations to make the far field data part of the solution be a good way here? Is COMSOL considering this in order to make far field plots easier and quicker? Or would we just extend the time for solving this way? Cheers Edgar -- Edgar J. Kaiser emPhys Physical Technology

Magnus Olsson COMSOL Employee

Please login with a confirmed email address before reporting spam

Posted: 1 decade ago 21 août 2013, 10:31 UTC−4
Hi Edgar,

Are you making use of the 3D far field plot feature where you have explicit control over the phi and theta resolution? Compared to plotting the far field on for example a boundary, that should be a lot faster. When plotting the far field on radiation boundaries, the required mesh resolution (lambda/5) is usually unnecessarily fine for plotting far fields which makes it slow.

We could indeed compute the far field as an extra step during solution but it would not be any faster. I recommend to use the 3D far field plot feature with coarse phi and theta resolution and request a finer plot when needed.

--
Magnus
Hi Edgar, Are you making use of the 3D far field plot feature where you have explicit control over the phi and theta resolution? Compared to plotting the far field on for example a boundary, that should be a lot faster. When plotting the far field on radiation boundaries, the required mesh resolution (lambda/5) is usually unnecessarily fine for plotting far fields which makes it slow. We could indeed compute the far field as an extra step during solution but it would not be any faster. I recommend to use the 3D far field plot feature with coarse phi and theta resolution and request a finer plot when needed. -- Magnus

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.