ParaView Server with multiple Cores/GPus using mvapchi

ParaView Support
1

Hi,
For instance, I am using paraview 5.6.0 + mvapchi.
I’m confused on how running pvserver in parallel works. If I want to use 4 GPUs on a single machine,using MPI I will have to run 4 pvserver jobs using the parameters -display from 0.0 to 0.3 according to the given documentation

So far, this is what I got in the memory inspector:

pvserver_parallel1

However, when I execute the suggested Programmable Source to verify the processes on each display, I only get one ouput:

pvserver_parallel3
pvserver_parallel31442×551 39.9 KB

Which confuses me because I thought the first image confirmed the 4 jobs are working. My first guess on what happens is that I am running 4 cores and all of them are sharing the same GPU.
Also, If I connect to 2 machines with 4 GPUs each one:
pvserver_parallel4
I got the same output from the script

Do you guys have any thoughts? Is the given source code correct ? Have you guys test only with OpenMPI ?

Thank you for the help in advance

2

You mentioned by MP that you reproduce the issue with OpenMPI, could you elaborate?

3

From @kmilo9999 :

I completely switch to OpenMPI and now everything works as the documentation claims. The output I was expecting to see was being printed in the terminal, not in paraview console as I thought.