[gmx-users] Inconsistent results in different clusters and cores

[Mark Abraham]

On Fri, May 17, 2013 at 1:23 PM, Justin Lemkul <jalemkul at vt.edu> wrote:

>
>
> On 5/17/13 6:20 AM, tarak karmakar wrote:
>
>> I have read the 'implementation of charmm in gromacs' by bjelkmar,
>> JCTC. There they have used following cut-offs
>> coulombtype=PME
>> rcoulomb=1.2
>> vdwtype=switch
>> rvdw=1.2
>> rvdw-switch=1.0
>>
>> I am not sure about rlist.
>>
>>
> It's really unnecessary to spam the list with a dozen requests that ask
> the same question.
>
> By definition, with PME, rlist = rcoulomb.  For CHARMM, use rlistlong =
> 1.4; the message that is printed below is an error that has been fixed in
> newer versions.


Are you sure? Looking at the logic in src/mdlib/read_ir.c, I could not see
how Tarak's input could have led (correctly) to that note from grompp. I am
thinking that
https://github.com/gromacs/gromacs/blob/release-4-5-patches/src/kernel/readir.c#L170
is
setting rlistlong = rlist, and so
https://github.com/gromacs/gromacs/blob/release-4-5-patches/src/kernel/readir.c#L586
gets
double-crossed.

(Note that http://redmine.gromacs.org/issues/1164 pertained to a slightly
different issue.)

Mark