And for pmemd I really want very precise control of optimizations and
libraries, and I don't want to take the time to write a real autoconfigure
myself, and I don't want to trust that someone else got it all right, or
have to deal with changing it as I change the way the things are built.
PMEMD is a part of amber where in my opinion, it is not the end of the world
for someone to have to know a little about the hardware, system software,
and performance issues. Getting things done both fast and correctly are the
overriding concerns. That is my 2 cents worth.
Best Regards - Bob
----- Original Message -----
From: "Ross Walker" <ross.rosswalker.co.uk>
To: <amber.scripps.edu>
Sent: Thursday, June 28, 2007 1:06 PM
Subject: RE: AMBER: configure question
> Hi Steve,
>
>> my issues I couldn't help but ask myself why isn't there a ./configure
>> script for this just like most of all the other open source
>> applications
>
> This is more a philosophical point than anything else. The issue is one of
> the fact that the Amber developers are scientists first and programmers
> second. We do not get any funding from NSF or NIH specifically to develop
> the software, and it is very hard to get such development included in
> grant
> requests. The science always has to come first. Thus the issue is one of
> maintenance. If you know how to build an autoconfigure script that will
> work
> on all platforms imaginable and all MPI implementations imaginable, and I
> mean from this everything from cygwin under windows to every flavour of
> linux, to AIX, to IRIX then by all means feel free to create it and send
> it
> to the developers we will be very grateful although it will have to be
> simple enough that the average graduate student contributing to the code
> can
> understand and maintain the script.
>
> The issue at heart is that Amber is significantly more complicated than
> most
> packages that use auto configure scripts and significant flexibility is
> required with regards to what code gets compiled on what platforms, which
> math libraries get linked in etc. If it was just a serial program running
> on
> Linux platforms then this might be somewhat possible but when you throw in
> both serial and MPI compilation + multiple platforms + multiple languages
> (F77, F95, c, c++, perl etc) + multiple packages with multiple
> dependencies
> (sander, sander.LES, sander.MPI, sander.PIMD, sander.PIMD.MPI, pmemd,
> ptraj,
> ambpdb, xleap, tleap etc etc...) such an effort to produce an all
> encompasing auto configure script gets out of hand very very quickly.
> Trust
> me I have tried it myself and after a day or so I gave up since I didn't
> have any more free time to work on it.
>
> An alternative would be to provide binaries for certain platforms but
> again
> this requires a significant amount of maintenance and testing that we
> simply
> don't have the manpower or funding to do.
>
> So, while I fully support what you are saying and agree that yes it really
> would help things there really isn't sufficient money, people and
> expertise
> to do this, maintain it, test it and teach everybody else how it works and
> how to add to it. I wish it was different but this is largely a function
> of
> the current funding climate in the US with regards to scientific software.
>
> Just my 2c...
>
> All the best
> Ross
>
> /\
> \/
> |\oss Walker
>
> | HPC Consultant and Staff Scientist |
> | San Diego Supercomputer Center |
> | Tel: +1 858 822 0854 | EMail:- ross.rosswalker.co.uk |
> | http://www.rosswalker.co.uk | PGP Key available on request |
>
> Note: Electronic Mail is not secure, has no guarantee of delivery, may not
> be read every day, and should not be used for urgent or sensitive issues.
>
>
> -----------------------------------------------------------------------
> 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 Sun Jul 01 2007 - 06:07:26 PDT
