Re: [AMBER] Error in Running heating in GPU enabled AMBER

From: David A Case <david.case.rutgers.edu>
Date: Fri, 17 Aug 2018 21:28:41 -0400

On Fri, Aug 17, 2018, kuldeep singh wrote:

> But when i run the heating for 500 ps,, I got the following error:-
>
> Error: an illegal memory access was encountered launching kernel
> kcleaForces cudaFree GpuBuffer::Deallocate failed an illegal memory access
> was encountered

Usual debugging: try to run with the cpu version of pmemd. Sometimes
equilibration needs to be done in full double precision, especially if
you starting configuration is not optimal.

If you still have problems with pmemd.cuda, try short runs with a small
value of ntpr: see if there is any indication of what the problem is.
Compare these short runs with the CPU version.

If you still have problems, you may need to post files that would allow
someone to try to reproduce the problem. When doing so, be sure to
indicate which version of Amber you are using.

Note: if you are using Amber16, be sure you have update.13 installed
(and have re-compiled with the modified code.) To check, cd to
AMBERHOME, and type "./update_amber --help". (Update.13 deals with
problems that can arise with smallish unit cells.)

...good luck....dac


_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Aug 17 2018 - 18:30:02 PDT
Custom Search