Thanks @Cumming ***@***.***> for the precious reviews

+[](){#ref-software-scivis} +# Scientific Visualization + +CSCS provides and supports a selection of scientific visualization applications on the computing systems: we usually build community codes that are adopted by several users on our systems. + +CSCS staff can also help users with performance tuning and ParaView Python code to optimise their workflow in production. + +* [ParaView][ref-uenv-paraview] + +## Other applications + +CSCS provides [tools and environments][ref-building] for installing applications that are not on the list of supported applications. + +!!! info ⬇️ Suggested change -!!! info +!!! under-construction

+ #SBATCH -A <account> + #SBATCH --uenv=paraview/5.13.2:v2 --view=paraview + #SBATCH --hint=nomultithread + + export MPICH_GPU_SUPPORT_ENABLED=0 + + srun --cpus-per-task=72 --cpu_bind=sockets /user-environment/ParaView-5.13/gpu_wrapper.sh /user-environment/ParaView-5.13/bin/pvbatch ParaViewPythonScript.py + ``` + Current observation is that best performance is achieved using [one MPI rank per GPU][ref-slurm-gh200-single-rank-per-gpu]. How to run multiple ranks per GPU is described [here][ref-slurm-gh200-multi-rank-per-gpu]. + +=== "Eiger" + + ```bash + #SBATCH -N 1 + #SBATCH --ntasks-per-node=144 there are 128 cores (2 * 64 core CPU socket) per node on Eiger. This will create 144 ranks on one node, so it is probably a mistake?

+ + srun --cpus-per-task=72 --cpu_bind=sockets /user-environment/ParaView-5.13/gpu_wrapper.sh /user-environment/ParaView-5.13/bin/pvbatch ParaViewPythonScript.py + ``` + Current observation is that best performance is achieved using [one MPI rank per GPU][ref-slurm-gh200-single-rank-per-gpu]. How to run multiple ranks per GPU is described [here][ref-slurm-gh200-multi-rank-per-gpu]. + +=== "Eiger" + + ```bash + #SBATCH -N 1 + #SBATCH --ntasks-per-node=144 + #SBATCH -A <account> + #SBATCH --uenv=paraview/5.13.2:v2 --view=paraview + #SBATCH --hint=nomultithread + + srun --cpus-per-task=144 /user-environment/ParaView-5.13/bin/pvbatch ParaViewPythonScript.py See the previous comment. Furthermore, is it 128 cpus-per-task, or 128 tasks-per-node?

+ #SBATCH --uenv=paraview/5.13.2:v2 --view=paraview + #SBATCH --hint=nomultithread + + srun --cpus-per-task=144 /user-environment/ParaView-5.13/bin/pvbatch ParaViewPythonScript.py + ``` + + +## Using ParaView in client-server mode + +A ParaView server can connect to a remote ParaView client installed on your desktop. Make sure to use the same version on both sides. Your local ParaView GUI client needs to create a SLURM job with appropriate parameters. We recommend that you make a copy of the file `/user-environment/ParaView-5.13/rc-submit-pvserver.sh` to your $HOME, such that you can further fine-tune it. + +You will need to add the corresponding XML code to your local ParaView installation, such that the Connect menu entry recognizes the ALPS cluster. The following code would be added to your **local** `$HOME/.config/ParaView/servers.pvsc` file + +!!! Example "XML code to add to your local ParaView settings" +```console This whole code block needs to be indented to be placed inside the example box (at the moment it is floating below an empty example box). Furthermore, instead of console you can mark it as xml to get xml highlighting. — Reply to this email directly, view it on GitHub<#234 (review)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/ABV3FECT2JJS5FFOGBCYF233OXGFTAVCNFSM6AAAAACEOXVKVCVHI2DSMVQWIX3LMV43YUDVNRWFEZLROVSXG5CSMV3GSZLXHMZTCNBQHAYDQMJRG4>. You are receiving this because you authored the thread.Message ID: ***@***.***>