Closed lopez-c closed 5 years ago
Hi, we would be more interested in a live system to test our code and to play around with.
Ok, we are getting a new server with a P100 (very similar to the one installed @ SwissFEL). As soon as we have it installed in our lab we will give you access to it.
Hi,
we had the idea that we could install our version of the software on your system with all the needed libraries and write down a step by step guide for you. That way you could test our new versions and you would be able to reinstall it (in case this should be necessary) in a fairly short time. For this, we would need access to current system. What do you think?
Cheers, Jonas
Hi Jonas,
With your external user account you should be able to ssh our server (mpc2198). You need to ssh hop.psi.ch first, and then ssh our server:
The server is very similar to the one installed at SwissFEL and includes a Tesla P100. We have installed CUDA-10 drivers on it.
Let me know if there is any software you need to run your application that it is not installed yet.
Cheers, Carlos
Hello Carlos,
unfortunately I am not able to log into your server (even after going over hop.psi.ch), the connection is immediately closed by the host. Are you sure that I have access rights to this machine?
Cheers, Jonas
Hi Jonas, I'm sorry, could you please try again. It should be fine now Cheers, Carlos
Hi Carlos, thanks for the quick reply! It works now but unfortunately I don't have a home directory on this server and also lack the permission to create one. Could you create one for me? Cheers, Jonas
Hi Jonas, I've created a new user on that machine that you can use to access it: "l_hzdr" The password is: "photoncounter", you can change it
Remember to access hop.psi.ch with your external user account first, and then acces mpc2198 with the new local user account
Cheers, Carlos
Hello Carlos, the access to the server doesn't seem to work currently. The connection is immediately closed by the host. Could you give me access to this machine again? Cheers, Jonas
Hi, Could you try again? Cheers, Carlos
Hi, thanks, it works now! Cheers, Jonas
Hello,
I've managed to get our code working on your system. For this, I installed gcc, boost and cmake locally using the spack package manager. Specifically boost >= 1.65.1, cmake >= 3.11 and either gcc 6.4, gcc 7.2 or gcc 7.3 are required. Additionally, alpaka is needed, which can be installed with the spack package manager and the HZDR spack repository for alpaka (see spack documentation on how to add repositories to spack). Alternatively it is also possible to just clone the alpaka repository and export the path to it using export alpaka_DIR=/path/to/alpaka
.
After these installations it should be possible to get the program running by following the instructions in the readme (our current working branch is clustering
).
On a different note: I'm having trouble login into your system again with the same error message as the previous times. Would it be possible to get some kind of a permanent fix for this issue? If so, I could test the program after major changes and see if it still works on your hardware.
Cheers, Jonas
Hi, One of your requests was to have access to the current online system, but after the meeting we were not 100% sure of your request, if you wanted to have access access to the live system at SwissFEL to check how it looks, or to another server with a Jungfrau connected to it where you could play with it.