[amin_u50]

Hello,

I am going to launch CONVERGE V.2.4. with multiple cores. So, first I define exe. file of MPI through edit=>preference. And then I run it. Is this the correct way? If yes, it gives me an error. I would be happy if someone helps me.

Amin

[tmburton]

Amin,

Please give us some more details about the system you will be running on. Have you looked at https://convergecfd.com/support/getting-started-guide ?
It provides a lot of information about installing and launching CONVERGE.

Best regards,

Tristan

[amin_u50]

Hi,

I'm using windows 7, 64 bit and CONVERGE version is 2.4. I checked the link you attached but I think in version 2.4 running MPI is different than the older versions. I installed pmpi 2.4.13 win64. When I double click on it, there is an error says: the program cant start because libpcmpi64.dll is missing from your computer. Try reinstalling the program to fix this problem.

[tmburton]

Amin,

Have you tried installing the Microsoft MPI implementation? I have used it successfully with the MSMPI CONVERGE build in both Windows 7 and Windows 8.

Best regards,

Tristan

[amin_u50]

I just installed Microsoft MSPI. Now when I double click on MSMPI there is no error. So, how should I define the number of cores?

[tmburton]

Amin,

Here's the command I execute in a Command Prompt window to run the CONVERGE solver in parallel: mpiexec.exe -n 3 converge

Where I'm running on 3 cores, and my path is set so that the system can find mpiexec.exe (usually at C:\Program Files\Microsoft MPI\Bin) and the converge executable (name depends on the build, usually installed at C:\Program Files\CONVERGE)

Make sure you launch the run from the directory that contains all your CONVERGE input files.

CONVERGE will print text output to the screen so you'll probably want to redirect it to a file (otherwise it just flashes by on the screen and you lose it).

Best regards,

Tristan

[amin_u50]

I think I got confused. I copied/past the MSMPI to the folder where I have the converge input files. I double click on MSPI now and it automatically starts to run. It does not allow me to write any command.

[MFGT]

This is how i used to do it in 2012 as well

So, you open the cmd window in the desired path, then type the command (e.g. is from my system):
mpiexec -n 4 converge-2.4.12-msmpi-win64.exe >log.txt

[amin_u50]

Could you please give more details? I installed MSMPI. When I double click on it, it automatically starts running only by 1 core. How can I write command? How can I define cores?

[tmburton]

Amin,

As was noted by myself and another post, CONVERGE is run in parallel from a Command Prompt window (or a terminal window if you utilize cygwin), there is no clicking on anything required. Please re-read the previous posts for the detailed instructions and let me know if you have more specific questions about the process.

Best regards,

Tristan

[amin_u50]

Is it possible to connect to my computer and show me how to run multiple cores?

[MFGT]

1. Click Start.
2. Type cmd and press Enter.

navigate to your folder with: cd "Foldername"


Its also possible to open the cmd window directly with right click-> open command promt here (works in windows 7).


As next step, you should defined CONVERGE within the environment variables (https://en.wikipedia.org/wiki/Environment_variable), same as the MSMPI


Then you should be able to type the command e.g.:
mpiexec -n 4 converge-2.4.12-msmpi-win64.exe >log.txt


The >log.txt writes the screen output into a log file.

[amin_u50]

Hello,

I still cannot run with multi-cores. I really appreciate if anyone from support team connects to my PC and show me how to run with multi-cores? Thanks
Amin

[MFGT]

Where is the problem?

[amin_u50]

Originally Posted by MFGT

Where is the problem?

I fixed it. The license I've got from CONVERGE was for serial. Now it's good.

I found something strange. I run CONVERGE with 7 cores and all of them are working during the compression (CPU=85%). However, during the injection, only one core is working (CPU=13%). I really don't know why?

[tmburton]

Amin,

Check the load balancing in cell_count_ranks.out. Your cells may be concentrated on one core so the other cores are waiting for information from that core.

Best regards,

Tristan

[amin_u50]

Originally Posted by tmburton

Amin,

Check the load balancing in cell_count_ranks.out. Your cells may be concentrated on one core so the other cores are waiting for information from that core.

Best regards,

Tristan

Thank you. That's correct. I have a question. How can I restart the CONVERGE calculation when running multi-cores? for example from 100 CAD. I entered restart flag=1 and restart number=100. But it does not work.

[MFGT]

You simply restart the case like you would start a normal case.
Just change restart_flag to 1 and define a new unique number (restart number, you can do multiple restarts within the same folder)
Output files will have this number attached to differentiate them from original files.

[amin_u50]

Originally Posted by MFGT

You simply restart the case like you would start a normal case.
Just change restart_flag to 1 and define a new unique number (restart number, you can do multiple restarts within the same folder)
Output files will have this number attached to differentiate them from original files.

I exactly did that, but when I enter restart number=100, CONVERGE calculates the last step and then stops. I'm simulating closed cycle IVC=-169.7 to EVO=145.3.

[MFGT]

So your restart file is from +100°CA?

and in inputs.in you have

start_time=-169.7
end_time=145.3
restart_flag=1
restart_number=100

and the CONVERGE calculated only one time step and stops?
Can you share the log file/screen output?

you can open the restart file with a text editor to see the corresponding crank angle, it should be in the frist line.