I want to learn how the basic openses, opensesmp, and openseesp are different and where to code them differently.
Where can I find reference materials and explanations?
And before that, I want to know how each interpretation program is different from each other, what are the advantages and disadvantages. I don't know the difference between sp and mp just by explaining that it use multi-core.
opensees sp vs opensees mp
-
marafini.f
- Posts: 363
- Joined: Fri Nov 13, 2020 1:52 pm
Re: opensees sp vs opensees mp
Hi Gookki,
You can run OpenSees in three different ways:
- sequential - opensees
- parallel (single processor) - openseessp
- parallel (multiple processor) - openseesmp
If you run opensees sequential, your model will be run on a single processor of your computer in one process thread.
If you want to run parallel computing, in the first option, single processor, your model will be partitioned automatically and run on multiple processors, but your script itself will only run on the first processor, you will not access to the others to interrupt, request results or operate exchange of information between processors.
This is particularly useful when you have large models and analyses that require a high computational time to perform. With openseessp you can run large models and let opensees itself take care of the partitioning.
The other parallel computing option is openseesmp, multiple processors, with which you have access to all processors, but you will need to partition your model manually. The access to all processors allows you to run different analysis on the same model or vary parameters to run parametric analysis. With openseessp you can just run a large model in parts and exploit the computational power of multiple processors, with openseesmp you can run multiple models at the same time and operate multiple analysis.
Openseesmp is not used much to run large models because the partitioning operations are complex and tricky to do by hand. In STKO you have the partition tool, which does the partition automatically based on the number of processors you select. This is why with STKO with mainly use opensees sequential and openseesmp for both large models and parametric analysis.
I advise you to check out our introduction to OpenSees webinar:
https://asdeasoft.net/?get-started-webinars
the presentation is in the support file and at the end there are many resources link you can explore.
I suggest you give a look also to:
webinar n#10 https://asdeasoft.net/?webinar for more on sequential and parallel computing
and webinar n#21 https://asdeasoft.net/?webinar for more on parametric analysis
Enjoy your modeling
Francesca
You can run OpenSees in three different ways:
- sequential - opensees
- parallel (single processor) - openseessp
- parallel (multiple processor) - openseesmp
If you run opensees sequential, your model will be run on a single processor of your computer in one process thread.
If you want to run parallel computing, in the first option, single processor, your model will be partitioned automatically and run on multiple processors, but your script itself will only run on the first processor, you will not access to the others to interrupt, request results or operate exchange of information between processors.
This is particularly useful when you have large models and analyses that require a high computational time to perform. With openseessp you can run large models and let opensees itself take care of the partitioning.
The other parallel computing option is openseesmp, multiple processors, with which you have access to all processors, but you will need to partition your model manually. The access to all processors allows you to run different analysis on the same model or vary parameters to run parametric analysis. With openseessp you can just run a large model in parts and exploit the computational power of multiple processors, with openseesmp you can run multiple models at the same time and operate multiple analysis.
Openseesmp is not used much to run large models because the partitioning operations are complex and tricky to do by hand. In STKO you have the partition tool, which does the partition automatically based on the number of processors you select. This is why with STKO with mainly use opensees sequential and openseesmp for both large models and parametric analysis.
I advise you to check out our introduction to OpenSees webinar:
https://asdeasoft.net/?get-started-webinars
the presentation is in the support file and at the end there are many resources link you can explore.
I suggest you give a look also to:
webinar n#10 https://asdeasoft.net/?webinar for more on sequential and parallel computing
and webinar n#21 https://asdeasoft.net/?webinar for more on parametric analysis
Enjoy your modeling
Francesca
Re: opensees sp vs opensees mp
Hi STKO team,
I am doing parametric study, and running many 2D numerical cases. I tried to run more than 10 cases simultaneously using Opensees, but the CPU usage is around 15%. Is there any method to employ fully the CPU capacity for running many cases simultaneously?
Thank you for your support,
Buu
I am doing parametric study, and running many 2D numerical cases. I tried to run more than 10 cases simultaneously using Opensees, but the CPU usage is around 15%. Is there any method to employ fully the CPU capacity for running many cases simultaneously?
Thank you for your support,
Buu
Re: opensees sp vs opensees mp
If you run few examples (less than the number of your cores) it's normal