Questions regarding the compilation of VASP on various platforms: hardware, compilers and libraries, etc.
Moderators: Global Moderator, Moderator
-
peterklaver
- Newbie
![Newbie Newbie](./images/ranks/star.gif)
- Posts: 31
- Joined: Thu Apr 21, 2005 9:28 am
- Location: Netherlands
-
Contact:
#1
Post
by peterklaver » Mon Feb 06, 2017 9:06 am
Hello people,
Our group has purchased three Dell C4130 nodes with 4 K80s in each, connected with infiniband. I'm quite eager to test the gpu verison of VASP on them. I've managed to compile an executale, but I can't seem to find documentation dealing with how to run the gpu executable. Specifically, is there a page detailing how to control the number of gpus versus cpu cores, how to tie certain gpus to specific cpu cores (which can make a performance difference on C4130 nodes), etc? Or does VASP sort out the optimal solution for such things itself automatically without user input?
I looked through the manual, but couldn't find a single instance of the word 'gpu' in there. And on the wiki, I only find the page about the gpu port where the section 'Running' is empty.
Is there a 'gpu-VASP for dummies' page anywhere?
greets,
Peter
-
admin
- Administrator
![Administrator Administrator](./images/ranks/staradmin.gif)
- Posts: 2921
- Joined: Tue Aug 03, 2004 8:18 am
- License Nr.: 458
#2
Post
by admin » Wed Mar 01, 2017 9:16 am
In the current version of the code there are no input parameters for GPU in INCAR.
Vasp automatically sort out the optimal solution without user input.
-
guiyang_huang1
- Newbie
![Newbie Newbie](./images/ranks/star.gif)
- Posts: 12
- Joined: Tue Nov 12, 2019 7:00 pm
#3
Post
by guiyang_huang1 » Wed Nov 13, 2019 3:39 pm
For gpu, NSIM=14 can be set if no test is done for the system I used. But for different system and computer, it should be different.
NSIM can be adjusted in order to have a faster speed.
LREAL=Auto should be set.