AMBER Archive (2005)

Subject: Re: AMBER: REM difficulties

From: Carlos Simmerling (carlos_at_ilion.bio.sunysb.edu)
Date: Thu Nov 03 2005 - 13:41:05 CST


I think I know what's going on - in Amber8 remd, the first
exchange is actually discarded and no output is written from
the first sander call. If you don't do an exchange, none of the
outputs are written. Amber9 no longer does this. The workaround
here is to not use REMD, just use multisander. Try -rem 0.

Carlos

Peter Varnai wrote:

>Thanks for the response.
>
>
>
>>>* in my particular case using default i/o I have problems with mixed
>>>ext 000/001 (although in test it seemed to be OK):
>>>
>>>
>>>
>>are you writing replica outputs or temperature-based outputs?
>>
>>
>
>I used the default as in the test, that should be for a given
>temperature.
>
>
>
>>the code cannot handle the same temp0, this exchange is temperature-based
>>so you end up dividing by zero if they match. Amber9 will check the inputs for
>>this.
>>
>>
>
>sander seems to run; delta remains zero, but I did not see division
>with delta.
>
>
>
>>It's not clear excatly what you mean by the last statement.
>>Did you get a "run2.out.000" file? Did you get one called
>>"run2.out.001"?
>>I haven't seen any similar behavior myself (e.g. having multisander
>>not output files from each "group member").
>>
>>
>
>again, at the same temp0 (that I should not use) only run2.out and
>run2.out.000 are produced not run2.out.001. run2.out is supposed to be
>the results before the first exchange that is discarded?
>
>
>
>>What happens if you leave out the "-rem 1" flag?
>>
>>
>
>I could not get replica exchange using groupfile with or without the
>"-rem 1" flag.
>
>Peter
>
>-----------------------------------------------------------------------
>The AMBER Mail Reflector
>To post, send mail to amber_at_scripps.edu
>To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
>

-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber_at_scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu