There will likely be a slight perf hit (a percent or two). I know that some
people tend to worry about memory consumption by apps like this. But from
my viewpoint, wasting available resources is a much bigger crime. Memory is
cheap (The difference between a 1.5 GB GTX 580 and a 3 GB GTX 580 is about
$60 on Amazon today). Time isn't.
That said, I'm adding code to detect cheap GPUs and do this maneuver
automagically.
On Tue, Aug 23, 2011 at 6:51 AM, Dmitry Mukha <dvmukha.gmail.com> wrote:
> Thank you, Scott, your solution works. How does it affect the simulation?
>
> What I did was
> sed -i 's[gpu->sim.randomSteps =
> MAX_RANDOM_STEPS;[gpu->sim.randomSteps = MAX_RANDOM_STEPS /
> 4;[g'
> $AMBERHOME/src/pmemd/src/cuda/gpu.cpp
>
> 2011/8/22 Scott Le Grand <varelse2005.gmail.com>
>
> > Yes, try dividing by 4 on a 1 GB card. It ought to get just enough
> room...
> >
>
> --
> Sincerely,
> Dmitry Mukha
> Institute of Bioorganic Chemistry, NAS, Minsk, Belarus
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Aug 23 2011 - 09:00:03 PDT
