The questions below pertain to the CONVERGE 3.x installation process. For a more general list of FAQs pertaining to internal combustion, physical modeling, and other topics, go to https://convergecfd.com/support/faqs.
If you need help resolving an issue, please email [email protected] (US), [email protected] (EU), or [email protected] (India).
Why is CONVERGE running slowly?
If you observe poor performance, we recommend checking your latency as a first step. CONVERGE prints the node-to-node latency to the screen and to mpi_latency.log at the start of a simulation. You can also run the check_system utility to check the latency without starting a simulation. Example commands:
CONVERGE 3.0: mpirun -np 16 converge-ompi -C
CONVERGE 3.1+: mpirun -np 16 converge-ompi -u check_system
CONVERGE does not check out a license when you run this utility.
For high-speed interconnects like InfiniBand and Omni-Path, we have seen latencies ranging from about 2 µs to 25 µs. These numbers are highly dependent on system configuration. If a high-speed interconnect is available for your system and your latency is higher than you expect, it is possible that the MPI library is not configured to use the high-speed interconnect. Work with your IT team to check your MPI setup.
Do I have to use one of the MPI packages that was installed with the CONVERGE 3.x solver?
No. We include MPI packages in the CONVERGE 3.x installer only for convenience. If your IT team has set up their own MPI package in a supported version, we recommend using that instead. To do so on Linux, load the module for your organization's MPI environment immediately after loading the CONVERGE module. To do so on Windows, prepend the path to your organization's MPI environment files to the Path environment variable.
Can I copy the CONVERGE executable from the installation package to a different directory (on Linux) without running the INSTALL script?
No. To install CONVERGE 3.x on Linux, you must run the INSTALL script.
In CONVERGE 3.x, you do not need to copy or link the CONVERGE executable to your Case Directory when you run a simulation. Instead, you set up your environment so that the executable can be run from any directory.
I've already installed CONVERGE 3.x.y. How do I upgrade to CONVERGE 3.x.z?
Download the installation package for the new version and run the INSTALL script (Linux) or the .msi installer (Windows). The directories for the new version (e.g., 3.1.4) are installed parallel to the directories for the old version (e.g., 3.1.3).
I just upgraded to a new minor version. How do I run an older version of CONVERGE 3.x?
Linux: Set up the CONVERGE environment for the older version that you want to run, and then run CONVERGE as you normally do.
Windows: Instead of invoking the *.ps1 script, provide the full path to the CONVERGE executable for the version you want to run. An example for running CONVERGE 3.1.7 with MS-MPI would be
"C:\Program Files\Microsoft MPI\Bin\mpiexec.exe" -n 4 "C:\Program Files\
Convergent_Science\CONVERGE\3.1.7\bin\msmpi\converge.exe"
The *.ps1 script will always run the version of CONVERGE that was most recently installed.
Can I use the same MPI version that I used for CONVERGE 2.4?
No. You must use one of the supported MPI versions for CONVERGE 3.x.