[AMBER] job in multiple GPUs

From: Vijay Manickam Achari <vjrajamany.yahoo.com>
Date: Wed, 15 Aug 2012 05:41:40 +0100 (BST)

Dear Amber users,

Here I would like to get some help on running a job in multiple GPUs. I tried running a job in multiple GPUs but it "hang" after running about 100ps. If I run the same job in single GPU, it runs without any problem.
I could compile pmemd.cuda.MPI and all the tests run fine.  
GPU type:  Tesla C2075
number if GPUs: 4 units
AMBER version : AMBER-12
MPI type: mpich2-1.4.1p1

note: Sorry for posting this question again. I had previously posted in this mailing list but my problem still unresolved.

Appreciate any help in advance.


Vijay Manickam Achari
(Phd Student)
Chemistry Department,
University of Malaya,
AMBER mailing list
Received on Tue Aug 14 2012 - 22:00:04 PDT
Custom Search