Putra -
Okay, it may be possible, but I think it would also be ill-advised. If I
understand correctly, mpich-g2 is "grid-enabled", meaning supporting the use
of different architecture machines, and perhaps also widely distributed. I
know that for pmemd there could be problems in mixing architectures, in that
there are a few calculations done on all nodes where identical results from
the computation are expected. Thus, you may be able to mix p3's and p4's,
but if you threw a completely different chip into the mix, there might be
problems. Secondly, mpich-g2 may need to do data translation between
different architectures, which would not be particularly efficient, and
mpich over gb ethernet is already stretched to handle the inter-node
communications load. Finally, I think a grid could be rather widely
distributed, with some rather low bandwidth segments in the net. Also not
so good for a tightly coupled application like sander or pmemd. If you ran
this stuff in such a way that you selected machines of the same architecture
connected by a high speed interconnect, it should basically work, though.
Regards - Bob Duke
----- Original Message -----
From: "putra abdullah" <putra77.rocketmail.com>
To: <amber.scripps.edu>
Sent: Sunday, March 06, 2005 9:55 PM
Subject: AMBER: amber with mpich-G2
> Hello,
>
> Just want to ask if amber can be use with mpich-G2 or not.
>
> __________________________________________________
> Do You Yahoo!?
> Tired of spam? Yahoo! Mail has the best spam protection around
> http://www.rosswalker.co.uk/adsense_alternatives/
> -----------------------------------------------------------------------
> The AMBER Mail Reflector
> To post, send mail to amber.scripps.edu
> To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
>
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Mon Mar 07 2005 - 04:53:01 PST
