[lammps-users] lammps on tesla

Hello all,
The default makefile for tesla uses SGI libraries. Can lammps be compiled on x86_64 with cuda ? An example makefile/flags would help.
many thanks

Ritwk

Hello all,

hello ritwik,

The default makefile for tesla uses SGI libraries. Can lammps be

please note this is for an (very old) SGI machine _called_ tesla,
not the nvidia tesla hardware (as crazy as it may seem).

compiled on x86_64 with cuda ? An example makefile/flags would help.

the short answer is: no

the long answer is:
to be compiled with CUDA, modules have to be programmed with
CUDA. i know of a number of efforts towards this, but none
is yet integrated into the official source. if you can live
with more limited functionality, i might interest you in the
HOOMD code, an MD code that is written to run _entirely_ on
GPUs and thus can make much better use of the GPU at the expense
of not being able to run in parallel (unless you script a parallel
replica method). its home page is at: http://www.ameslab.gov/hoomd/

cheers,
   axel