Re: [AMBER] random number seed - range

From: Robert Duke <rduke.email.unc.edu>
Date: Fri, 13 Aug 2010 17:40:35 -0400

Okay, I am confused by this. PMEMD 10, at least in the source I have, and
all the bugfixes for 10 I have seen, sets ig to a constant unless you
specify it. If you specify a -1, then that gets used as the seed for
setting initial velocities (non-restart runs) and -1 + 1 = 0 gets used in a
reseeding for things like the Langevin or Anderson thermostats (this is off
the top of my head, so I may be missing something else). So I don't know
what values you initially set, and what other values you may be looking at.
Is it perhaps really pmemd 11 you are using? I don't want to lull folks
into thinking this stuff works in pmemd 10 or earlier; AFAIK it doesn't, my
knowledge based on looking at the source code tree and bugfixes. And I see
know traps for catching a -1 and error exiting. The problem came about
because I was not aware of the change being made in sander; poor
communication on the part of the developers (sorry).
Regards - Bob Duke
----- Original Message -----
From: "Hopkins, Robert" <hopkins.uhcl.edu>
To: "AMBER Mailing List" <amber.ambermd.org>
Sent: Friday, August 13, 2010 5:03 PM
Subject: Re: [AMBER] random number seed - range


> After having just completed over 240 ns of MD calculations on duplex DNA
> molecules using pmemd 10, I felt a certain amount of panic when I
> recently
> read that for ig=-1, the clock-based generation of a random number seed
> had not been implemented in pmemd 10. So, for any others who are using
> pmemd 10 and are concerned about problems arising from a constant
> default
> random number seed in calculations involving DNA, I'll add my "two
> cents"
> here.
>
> When I looked at the actual number being used for the random number seed
> (ig)
> in a series of calculations representing 40 ns, I found the following
> values
> for runs done at selected intervals of 5 ns:
>
> ns ig
>
> 5 84988
> 10 829000
> 15 637723
> 20 384269
> 25 113068
> 30 313712
> 35 466334
> 40 546673
>
> A definite feeling of relief!
>
> Bob Hopkins
>
>
> -----Original Message-----
> From: Robert Duke [mailto:rduke.email.unc.edu]
> Sent: Tuesday, July 27, 2010 2:13 PM
> To: AMBER Mailing List
> Subject: [Possible Spam: 10%] Re: [AMBER] random number seed - range
>
> Yes, I did not know that this functionality was going into amber 10
> sander,
> so I did not put it into pmemd 10, unfortunately (because it is
> definitely a
> source of some confusion). I think Ross' group put it into pmemd 11.
> Regards - Bob Duke
> ----- Original Message -----
> From: "case" <case.biomaps.rutgers.edu>
> To: "AMBER Mailing List" <amber.ambermd.org>
> Sent: Tuesday, July 27, 2010 3:02 PM
> Subject: Re: [AMBER] random number seed - range
>
>
>> On Tue, Jul 27, 2010, Keith Van Nostrand wrote:
>>>
>>> Because of correlation in the data, I had started using ig = -1 as
>>> well. I am
>>> doing Langevin dynamics with Amber 10 pmemd, and my output shows ig =
> -1
>>> everywhere:
>>
>> As far as I can tell from a quick code check, the special handling of
>> ig=-1
>> in pmemd did not occur until Amber11. If anyone has more detailed
> info,
>> they
>> can post it here.
>>
>> ....dac
>>
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>>
>
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
> _______________________________________________
> AMBER mailing list
> AMBER.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>
>


_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Aug 13 2010 - 15:00:03 PDT
Custom Search