On Wed, Aug 14, 2019, Jose Javier Muņoz Criollo wrote:
>
>I'm trying to compile Amber16 with AmberTools19 ...
Note that this not something we recommend or really support. There have been
(more in the past than now) connections between a given version of Amber and
the corresponding version of AmberTools, so you can't (generally) just mix and
match the two.
What I recommend:
1. Download and install AmberTools19: this will be in a folder headed by
"amber18".
2. Extract your Amber16 tarfile and the most recent matching AmberTools (which
is AmberTools17). These will both be in a folder headed with "amber16".
3. Build pmemd and pmemd.cuda in the "amber16" folder.
4. Set AMBERHOME to point to the "amber18" folder. When you need to use
pmemd, use its full path (.../amber16/bin/pmemd), or put that in your PATH.
The pmemd program does not reference AMBERHOME.
Of course, someone on the list might know how to get you combination to work.
But be sure you understand that pmemd.cuda.MPI is only useful is specialized
situations; generally, parallel scaling of pmemd.cuda is only modest.
And, please understand that there is no reason to use Intel compilers for
pmemd.cuda: all the code that does the real work is compiled by the nvidia
compilers, so you should consider using gnu compilers for all cuda builds.
....good luck...dac
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Aug 14 2019 - 18:30:03 PDT
