Hello,
Can you only run SimVasc on 4 processors?
Thanks,
Vitaly
Multiple Processors
- Adam Updegrove
- Posts: 236
- Joined: Fri Aug 30, 2013 4:11 pm
Re: Multiple Processors
You can run the SimVascular solver on as many processors as there are available. The solver scales well, so this is an advantage of installing the solver on a cluster.
Thanks!
SimVascular Development Team
Thanks!
SimVascular Development Team
- Vitaly Kheyfets
- Posts: 35
- Joined: Mon Jun 22, 2015 2:23 pm
Re: Multiple Processors
Perfect, thanks!
-Vitaly
-Vitaly
- Rodrigo Romarowski
- Posts: 16
- Joined: Sat Apr 15, 2017 10:06 am
Re: Multiple Processors
Hi!
Do you guys have any experience in large clusters? I would like to run some simulations in a cluster with around 2k cores. Did you test strong and weak scalability of the svSolver and can give me any hint? do you know the N of cores where i will lose linear scaling?
In this case... do you suggest using CG, NS or GMRES as a linear solver? I am performing some tests in my local computer but maybe you can share with me your experience.
Thanks a lot!
Rodrigo
Do you guys have any experience in large clusters? I would like to run some simulations in a cluster with around 2k cores. Did you test strong and weak scalability of the svSolver and can give me any hint? do you know the N of cores where i will lose linear scaling?
In this case... do you suggest using CG, NS or GMRES as a linear solver? I am performing some tests in my local computer but maybe you can share with me your experience.
Thanks a lot!
Rodrigo
- Jongmin Seo
- Posts: 1
- Joined: Fri May 13, 2016 2:44 pm
Re: Multiple Processors
Hi,
I have been testing performance of linear solvers in high performance clusters. Overall NS shows pretty good strong scalability (almost linear) up to 384 processors in a 8M mesh size problem. A weak scaling study shows that NS works fine up to 160 processors and the computing time increases after using 320 processors. I can share detailed test results, so please send me an email to jmseo.eng@gmail.com. if you wish.
Jongmin
I have been testing performance of linear solvers in high performance clusters. Overall NS shows pretty good strong scalability (almost linear) up to 384 processors in a 8M mesh size problem. A weak scaling study shows that NS works fine up to 160 processors and the computing time increases after using 320 processors. I can share detailed test results, so please send me an email to jmseo.eng@gmail.com. if you wish.
Jongmin
- Rodrigo Romarowski
- Posts: 16
- Joined: Sat Apr 15, 2017 10:06 am
Re: Multiple Processors
Thanks for the comment.
Do you have more details about the platform you make those tests? Was it Broadwell, KNL, Haswell?
In which computer? Did you try other linear solvers within the software?
Do you have more details about the platform you make those tests? Was it Broadwell, KNL, Haswell?
In which computer? Did you try other linear solvers within the software?