Hello everyone,
I am having some problems with my SimVascular simulation. At the inlet I am imposing patient specific boundary conditions using a customized BCT file. In my model only one inlet is present and here I am imposing customized velocities through the BCT file. This one seems to be read in the correct way since the velocity profile showed in the results is correct. The problem is in the computation of the flow. The flow profile and values are different from what I expect - I calculate the flow profile and values using a software and the same velocities written in the bct file. I have tried to reduce the time step, but the results are always the same, does anyone have any idea of what could be the problem?
I have also an issue with running the simulation job at cluster. One of my colleagues and I are using the same node of the cluster, which has 40 processors. Each one of us uses 13 processors but when we run two different simulation jobs at the same time the %CPU used lowers from 100% to 50%, is there a way to solve this problem?
Thank you for your help,
Valentina
Issues with flow calculation and with running two simulations at cluster at the same time
-
Valentina Scarponi
- Posts: 8
- Joined: Sat Mar 07, 2020 4:58 am
-
David Parker
- Posts: 1747
- Joined: Tue Aug 23, 2005 2:43 pm
Re: Issues with flow calculation and with running two simulations at cluster at the same time
Hi Valentina,
How are the flow results different than what you expect, too small, too large, wrong direction(!) ? Attach you .bct file and I'll have a look, make sure the normals are in the right direction.
Running a job on a cluster using mpiexec -np N should only use N processors. When you execute a solver job to you see a directory named N-procs_case created? Maybe this has something to do with hyperthreading on your cluster.
Cheers,
Dave
How are the flow results different than what you expect, too small, too large, wrong direction(!) ? Attach you .bct file and I'll have a look, make sure the normals are in the right direction.
Running a job on a cluster using mpiexec -np N should only use N processors. When you execute a solver job to you see a directory named N-procs_case created? Maybe this has something to do with hyperthreading on your cluster.
Cheers,
Dave
-
Valentina Scarponi
- Posts: 8
- Joined: Sat Mar 07, 2020 4:58 am
Re: Issues with flow calculation and with running two simulations at cluster at the same time
Thank you for your replay!
With my velocities the flow should always be entering the model and should show two peaks, the first one higher than the second one, while from my results the flow exits the model in the first 0.5s -the total simulation time is 0.94s-, then it reaches the first peak at the correct time step but with a value which is lower than I expected. The second peak is again at the right time step but with an higher value. The bct file is attached at the end of the message, I've included only a portion of it because the whole file was too large.
As far as the cluster is concerned, the directory is correctly created, so I'll check the management of hyperthreading.
Many thanks,
Valentina
With my velocities the flow should always be entering the model and should show two peaks, the first one higher than the second one, while from my results the flow exits the model in the first 0.5s -the total simulation time is 0.94s-, then it reaches the first peak at the correct time step but with a value which is lower than I expected. The second peak is again at the right time step but with an higher value. The bct file is attached at the end of the message, I've included only a portion of it because the whole file was too large.
As far as the cluster is concerned, the directory is correctly created, so I'll check the management of hyperthreading.
Many thanks,
Valentina
- Attachments
-
- bct1.7z
- (707.48 KiB) Downloaded 10 times
-
Weiguang Yang
- Posts: 110
- Joined: Mon Apr 07, 2008 2:17 pm
Re: Issues with flow calculation and with running two simulations at cluster at the same time
How about taking a step back? Use your own software and procedure to generate a time-dependent bct.dat with a parabolic profile and make sure it is identical to the bct.dat file generated by SV. Then make a steady inflow file with a nonidealized profile (customized nodal velocities). Do they work?