ParaView

ParaView is a scientific parallel visualizer.

This version supports CPU-only rendering without X context using the OSMesa library, it does not support GPU rendering.

Use the GALLIUM_DRIVER environment variable to choose a software renderer, it is recommended to use

GALLIUM_DRIVER=swr

for best performance.

Ray tracing using the OSPRay library is also supported.

Available Modules¶

module load ParaView/5.6.0-gimpi-2018b

Warning

The ParaView server loaded must be the same version as the client you have installed locally.

Setting up Client-Server Mode¶

If you want to use ParaView in client-server mode, use the following setup:

Load one of the ParaView Server modules listed above and launch the server in your interactive visualisation session on the HPC using;
```
module load ParaView
```
To start the ParaView server run;
```
pvserver
```

You should see;

Waiting for client...
Connection URL: cs://mahuika02:11111
Accepting connection(s): mahuika02:11111

Create an SSH tunnel for port "11111" from your local machine to the cluster. e.g.
```
ssh mahuika -L 11111:mahuika02:11111
```
Make sure the host name and socket match those given by the server earlier!
Launch the ParaView GUI on your local machine and go to "File > Connect" or click the button.
Click on "Add Server", choose server type "Client / Server", host "localhost" (as we will be using the SSH tunnel), and port "11111", then click on "Configure" .
Select the new server and click on "Connect"

Parallelisation¶

The CPU based versions of ParaView use the OpenSWR rasteriser as well as the OSPRay ray tracer for rendering graphics. These support parallel operation for better performance, but are configured to only use a single core by default. Run the following commands before launching ParaView GUI or ParaView Server if you want to use more cores (depending on the number of cores available in your session):

export KNOB_MAX_WORKER_THREADS=<number of cores>
export OSPRAY_THREADS=<number of cores>

ParaView Server also supports parallel execution using MPI.