|
|
Hi Andrei,
thank you for the questions!
Yes of course we will assist you to get started.
For comparisons, it might be best to do those on a cluster, even if you
just use one node in an interactive session.
You might also want to take the upcoming 0.3.0 release for that, since
we improved many aspects of the code again. Feel free to contact us
performance wise, e.g. setting the correct GPU compute architecture
during `pic-configure` will be important to get optimal performance.
We also operate on Slurm clusters regularly and have a new section on it
in our manual under:
http://picongpu.readthedocs.io/en/dev/usage/tbg.html#batch-system-examples
You might just want to clone the tbg examples for slurm from the TU
Dresden cluster "Taurus", located here:
https://github.com/ComputationalRadiationPhysics/picongpu/tree/dev/src/picongpu/submit/taurus-tud
instead of modifying the quite different PBS submit scripts.
If you want to try PIConGPU on a laptop, I would recommend trying
setting it up via nivida-docker since I saw several people struggling in
the last weeks to get their individual software environments on Desktops
in a usable state (it's also easier to support you since it
automatically creates reproducible environments that you can share with
us via Dockerfiles).
Axel
On 24.05.2017 10:19, Andrei Berceanu wrote:
> Dear Axel,
>
> Thank you for the warm welcome and your detailed reply! While I am quite new to PIC codes, I have done scientific computing in the past and I now work on laser-matter interaction at ELI-NP [1], where we are currently evaluating the possibility of using PIConGPU instead of EPOCH. In fact, my previous question about OS/software stack was because I want to install a development version of PIConGPU on my personal laptop [2] for some benchmarks against EPOCH and for playing around with the code.
>
> On a separate note, one of my colleagues managed to install PIConGPU on a cluster, but he told me all the examples he could find in the docs refer to the PBS batch system, while the batch system on this particular cluster is Slurm. Are you aware of any examples of using PIConGPU with Slurm?
> Thanks,
>
> Andrei
>
> [1] http://www.eli-np.ro
> [2] https://www.asus.com/ROG-Republic-Of-Gamers/ROG-GL552JX/specifications/
>
> On Sat, 20 May 2017 17:10:34 +0200, Axel Huebl <a.huebl@hzdr.de> wrote:
>
>> Hi Andrei,
>>
>> Welcome to our list! Most of us devel on a Debian derivative, e.g. Debian testing, latest release of Ubuntu or Mint. Most of our runtime testing we do in interactive queues on various clusters.
>>
>> We don't have a vagrant setup yet but in case you like docker, we have a nvidia-docker file under:
>> https://github.com/ComputationalRadiationPhysics/picongpu/issues/829
>>
>> If you are an expert in vagrant, please don't hesitate to share you setup. I just did not find the time to check how well GPU support works with vagrant and if people are using it regularly. I know it exists but have not seen other scientists working with it yet.
>>
>> Package wise we usually need a rather recent version of cuda, cmake and boost with a working mpi installation, see our install requirements for exact versions :
>> https://picongpu.readthedocs.io/en/dev/install/dependencies.html
>>
>> An attractive and brand new alternative is also spack, which has working build receipts for all our dependencies :
>> https://github.com/LLNL/spack
>> https://spack.io
>>
>> Are you just generally interested or developing with a specific lab/university? If you need any help to get started or have a specific development you are aiming for, feel free to open threads here or issues on github.
>>
>> Again, welcome! :)
>>
>> Cheers,
>> Axel
>
>
>
> #############################################################
> This message is sent to you because you are subscribed to
> the mailing list <picongpu-users@hzdr.de>.
> To unsubscribe, E-mail to: <picongpu-users-off@hzdr.de>
> To switch to the DIGEST mode, E-mail to <picongpu-users-digest@hzdr.de>
> To switch to the INDEX mode, E-mail to <picongpu-users-index@hzdr.de>
> Send administrative queries to <picongpu-users-request@hzdr.de>
>
--
Axel Huebl
Phone +49 351 260 3582
https://www.hzdr.de/crp
Computational Radiation Physics
Laser Particle Acceleration Division
Helmholtz-Zentrum Dresden - Rossendorf e.V.
Bautzner Landstrasse 400, 01328 Dresden
POB 510119, D-01314 Dresden
Vorstand: Prof. Dr.Dr.h.c. R. Sauerbrey
Prof. Dr.Dr.h.c. P. Joehnk
VR 1693 beim Amtsgericht Dresden
|
|