Re: [AMBER] amber14 CUDA GPU for workstation?

From: Jason Swails <jason.swails.gmail.com>
Date: Tue, 30 Dec 2014 17:55:33 -0500

On Tue, Dec 30, 2014 at 5:03 PM, Ross Walker <ross.rosswalker.co.uk> wrote:

> Hi Michael,
>
> This error means that the files required for AMBER 14 are not in your
> AMBERHOME directory. I suspect that only the AmberTools14.tar.bz2 file was
> extracted prior to commencing the installation. For simplicity I would
> start again from the beginning with:
>
> cd $AMBERHOME
> cd ../
> rm -rf amber14
> tar xvjf AmberTools14.tar.bz2
> tar xvjf Amber14.tar.bz2 (be sure to do both of these untar commands)
> cd amber14
> ./configure gnu (say yes to the updates)
> make install
> make test (optional)
>
> make clean
> ./configure -mpi gnu
> make install
> export DO_PARALLEL='mpirun -np 2' (or whatever syntax your mpi uses)
> make test (optional)
>
> make clean
> unset $DO_PARALLEL
>

​This should just be "unset DO_PARALLEL" (no $)



> ./configure -cuda gnu
> make install
> make test (optional)
> make clean
> ./configure -cuda -mpi gnu (only needed if you want to run across 2 GPUs
> in a single job at the same time).
> make install
> export DO_PARALLEL='mpirun -np 2' (or whatever syntax your mpi uses)
> make test (optional)
> make clean
> unset $DO_PARALLEL
>

T
​his should also just be "unset DO_PARALLEL"

I have nothing to add to the general advice here -- it is the typical
workflow for installing Amber in serial, parallel, and with CUDA. The wiki
page linked earlier goes into a little more detail, which will take longer
to work through but should hopefully provide more insight about what each
step is doing.

All the best,
Jason

-- 
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Dec 30 2014 - 15:00:02 PST
Custom Search