Re: [AMBER] CUDA and Amber 18

From: David A Case <>
Date: Mon, 16 Jul 2018 16:00:11 -0400

On Mon, Jul 16, 2018, Beale, John wrote:

Just to expand on Darrin's post:

> I have noticed that when
> I use the pmemd.cuda.MPI file, the simulation runs on Amber 16.

You need to be clear (at least to yourself) about exactly what you
mean when you say "I use the pmemd.cuda.MPI file".

You can always give the full path to an executable: .e.g.

/path/to/amber18/bin/pmemd.cuda OR

If you just type "pmemd.cuda", what you get will depend on your PATH
variable. You can type "which pmemd.cuda" to see the path that gets
chosen, and "echo $PATH" to see what directories are searched.

You can also look in the /path/to/amber18/bin and /path/to/amber16/bin
directories to see what executables are there. It's certainly possible
that somehow you installed the Amber16 pmemd.cuda.MPI, but failed to
install the Amber18 version.

A note: I deliberately changed your references from "pmemd.cuda.MPI" to
"pmemd.cuda". Be sure you really need the MPI version for cuda: Only
those with specialized needs, and specialized hardware, get much benefit
from running pmemd.cuda in parallel.

...good luck...dac

p.s.: I hate to be picky, but Darrin's advice is not quite right, and
these items can be confusing, especially to those relatively new to the
command line. Darrin said this:

To run AMBER16
    csh ~/home/user/amber16/amber.csh
---and then run your simulations

The second line starts a new shell, sets up environment variables there,
then exits that shell, effectively doing nothing in the shell you are
running in. Replace the "csh" on the second line with "source". Also
note that you probably don't want the "~" in the second line.

Here is a more detailed explanation:

[Apologies to all, including Darrin, who already understand all this...]

AMBER mailing list
Received on Mon Jul 16 2018 - 13:30:03 PDT
Custom Search