Why does the simulation run very slowly?

[孟珂 孙]

I have set two different size before I run the simulation:
aorta:2.457 coronary:0.5
aorta:4.85 coronary:0.75
And the mesh info is as follow:
Number of Nodes:165851, Number of Elems:958338, Number of Edges:83292, Number of Faces:55528
Number of Nodes:47215, Number of Elems:256393, Number of Edges:36315, Number of Faces:24210
The path, segment and model are exactly the same.
But why is the speed of running in almost two cases almost the same?

[David Parker]

You will need to provide detailed information about how you created your meshes and how you are running your simulations. For example:

1) What version of SV are you using
2) What OS are you running on
3) What solver are you using, svSolver or svFSI
4) Are you using more than one Meshes Tool
5) Are you using more than one Simulation Tool
6) How are you executing the simulation, from the GUI or command line
7) Are you using MPI

My guess is that you are not running the simulation on the mesh you think you are. You can check the echo.dat file under your project's simulation directory to see how many elements are in the mesh.

And be sure to save your project after meshing.

[孟珂 孙]

Thank you very much. This problem has been solved. I just forgot save my project after meshing.

[David Parker]

This is actually a fairly common problem for new users, not to save their project before closing SV. I am planning on adding the display of a popup window asking if the user wants to save the project data when closing the project window.

Cheers,
Dave

[gang fang]

hi, David,
how are you now? hope you are healthy!
I am considering how to speed up the SV solver? Is there some tricks for speeding up?

I ran a case about 0.5 miilion elements and 16000steps for interval is 0.0001 .
OS : win 10 x64
CPU: i7 8750H @2.2GHz physics 6core logic 12cores
RAM: 32G

Best,
Gang

[gang fang]

hi, David,
how are you now? hope you are healthy!
I am considering how to speed up the SV solver? Is there some tricks for speeding up?

I ran a case about 0.5 miilion elements and 16000steps for interval is 0.0001 .
OS : win 10 x64
CPU: i7 8750H @2.2GHz physics 6core logic 12cores
RAM: 32G

Best,
Gang

[gang fang]

sorry forget to solver time, I run 12cores and about 7Hours finished.

[David Parker]

Hi Gang,

I am healthy, thanks, hope you are too!

There are no tricks to speed up the solver. Adjusting solver solution parameters is tricky, could produce incorrect results and may not speed up the solution by that much.

Here are a few things you can try

1) Try using the Intel Fortan compiler. I have not done a comparison between gcc and ifort so I don't know what the speed up might be.

2) Perform a convergence study to determine the number of elements needed for a good solution. This might allow you to reduce the number of elements.

3) Reduce the time step. The time step does not need to strictly enforce the CFL condition. Try setting the time step for a CFL condition around 1.0.

Of course you could run the problems on a cluster!

Cheers,
Dave

[Charlie Bright]

Hi Gang,

Incase you're still struggling with this, I read that CFD solvers tend not to take advantage of virtual cores/hyper-threading on your CPU.
Try running the simulation using 6 cores on the slider in place of 12.
I tried this on a 4 core 8 thread i7-3770 and a 6 core 12 thread i7-8750H and choosing the physical number of cores saved around 30% processing time for me.
I tried disabling the extra threads in the BIOS as well but there was no difference in time from just selecting the number of cores you have, so you may as well keep hyper-threading on and take advantage of it while you do other work.

I hope this helps,
Charlie