AMBER Archive (2009)

Subject: Re: [AMBER] check MAXPR in locmem.f error on minimization

From: Robert Duke (rduke_at_email.unc.edu)
Date: Fri Nov 27 2009 - 08:44:36 CST

Hmmm, I just noted the complaint here about a "fill_bonded:max14 exceeded!"
error in pmemd. This is an error in some setup code written by Darden, but
then ported to pmemd by me. Could you give me pmemd version and patch level
info, and perhaps let me have a copy of the system so I can do some testing?
I know of one patch that may be relevant, but would need to check it out.
Thanks - Bob Duke
----- Original Message -----
From: "Jodi Hadden" <jodih_at_uga.edu>
To: "AMBER Mailing List" <amber_at_ambermd.org>
Sent: Wednesday, November 25, 2009 1:29 PM
Subject: Re: [AMBER] check MAXPR in locmem.f error on minimization

> Ah, I tried reducing the 2.5 to a smaller number before... I must have
> forgotten to recompile then. Sorry for my mistake. With a proper
> recompile, this remedies the problem. Thanks!
>
> The reason I was not using pmemd was because it crashed with the error:
> fill_bonded: max14 exceeded!!
> And the patch that was suggested to fix this error previously was for
> an earlier version of sander. Can a sander patch be applied to pmemd?
> I am not at all familiar with the code and was not sure of what I
> might break by trying something like this... The sander patch was
> suggested here: http://dev-archive.ambermd.org/200811/0021.html
>
> On 11/25/09, case <case_at_biomaps.rutgers.edu> wrote:
>> On Tue, Nov 24, 2009, Jodi Hadden wrote:
>>>
>>> I am trying to run a minimization of a solid glucose cluster with
>>> extra points (lone pairs on the oxygens) in Amber 11.
>>>
>>> On 2 cpus, the minimization runs for 1500 steps and then dies with the
>>> following error:
>>>
>>>
>>> * NB pairs 248 161448 exceeds capacity ( 161500) 0
>>> SIZE OF NONBOND LIST = 161500
>>> SANDER BOMB in subroutine nonbond_list
>>> Non bond list overflow!
>>> check MAXPR in locmem.f
>>>
>>> I have tried the suggestion by Ross Walker to change the locmem.f line
>>> maxpr_float = natom * (cutoffnb + skinnb)**3 / 3.0d0
>>> to
>>> maxpr_float = max(natom * (cutoffnb + skinnb)**3 / 3.0d0,500000)
>>> and recompile, but I still get the same error.
>>
>> That is because that line of code is only executed when numextra == 0,
>> which
>> is not true for your case. Try modifying the assignment to maxpr_float
>> two
>> lines below that.
>>
>>>
>>> The suggestion by Dave Case to change
>>> maxpr_float = natom * (cutoffnb + skinnb)**3 / 3.0d0
>>> to
>>> maxpr_float = natom * (cutoffnb + skinnb)**3 / 2.5d0
>>> was already implemented in the current version of Amber and so was
>>> unnecessary.
>>
>> Make the 2.5 value even smaller; sander is trying to estimate the number
>> of
>> nonbonded interactions, but the estimates may not be valid for solid
>> glucose with lots of extra points.
>>
>> Is there a reason you are not using pmemd?
>>
>> ....dac
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER_at_ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>
> _______________________________________________
> AMBER mailing list
> AMBER_at_ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
>

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber