Re: [AMBER] Problems with cutoff (?)

From: David Case <>
Date: Sat, 12 Jan 2019 21:45:47 +0000

On Fri, Jan 11, 2019, Margarita I. Bernal-Uruchurtu wrote:
>I'm having a very strange problem with the cutoff option.
>The system is a box with a water/air interface and a couple of organic
>molecules. The starting geometry was previously equilibrated with TIP3P and
>GAFF and I'll use SEBOMD and equilibrate with the lambda option.

Can you be a bit more explicit: are you using SEBOMD in the runs that
fail? Or is that just an option you plan to use in the future? That
is, is this likely to be a problem with the SEBOMD code, or with some
other part of Amber?

>With amber18 in a MacBookPro /Mojave it finds no problem. As soon as I
>switch to the cluster I have an error that appears using both, the amber18
>or amber16 versions.

We are probably going to need get enough files to try to reproduce your
error, espeically in a situation like this were the code code works on
one OS/compiler combination but not on another.

>It exits with the following message:
>Program received signal SIGSEGV: Segmentation fault - invalid memory

As you probably know, this error conveys no useful information other
than that you had a bad memory reference.

>and then I changed the size of the cutoff from 8.0 to 7.0 With the smaller
>cutoff it doesn't exit with an error but if I increase the size of the
>cutoff from 7.4 to 7.5 then I'll get again the same error than with a
>cutoff of 8.0

What is your box size? In each direction, it must be greater than twice
the cutoff plus the value of skinnb (which is 2 by default). Do all
values for the cutoff work on the Mac?

Along with files to reproduce, it would be great to get some information
about what you call "the cluster": what OS and compilers are used there?


AMBER mailing list
Received on Sat Jan 12 2019 - 14:00:02 PST
Custom Search