AMBER Archive (2005)

Subject: Re: AMBER: AMBER8 Build on Linux x64 running Opterons

From: Robert Duke (rduke_at_email.unc.edu)
Date: Wed May 11 2005 - 10:14:50 CDT


You may want to look at the 'March 2005 configuration update for PMEMD'.
The config files for pmemd are different than the ones used by all the rest
of amber, mostly in that more aggressive optimization is sometimes used.
However, I think that for pathscale compilers, chances are good that the
compilers options/libraries listed here will get you started (I think that
pathscale is easy in that you don't need to reference the compiler libraries
explicitly); you should of course leave out the pmemd optimization defines
(-D...), though I doubt that including them is harmful. I found pathscale
to be very easy to use and solid.
Regards - Bob Duke

----- Original Message -----
From: "Mark Rance" <rance_at_rabi.med.uc.edu>
To: <amber_at_scripps.edu>
Sent: Wednesday, May 11, 2005 11:02 AM
Subject: Re: AMBER: AMBER8 Build on Linux x64 running Opterons

>I encountered the same problem Thomas has reported, with xleap compiling
> ok
> but giving a segmentation fault on start-up (after it prints out four
> lines of
> text stating it was adding entries to the search path). This was on an
> opteron system
> runniing SuSE Enterprise Server 9 and the pathscale compilers. The only
> change
> I had made in the supplied software was to change the entry in the
> makefile
> for leap for the XALEAP_LIB environment variable from -L$(XHOME)/lib
> to -L$(XHOME)/lib64 so that it could find libraries like Xt. For the
> config.h
> file, since there is no option in configure for the pathscale compilers, I
> made some
> best guesses as to compiler flags, but if anyone has an optimized config.h
> file
> for an opteron/pathscale system, I'd be interested to see it.
>
> The only other problem I had with the amber8 install was that the
> fantasian program
> (in nmr_aux) would not compile, due to not finding a subroutine/function
> named rand.
>
> Mark
>
>
> Thomas Patko wrote:
>
>> AMBER Users:
>>
>> For those of you that have had trouble getting AMBER8 to
>> compile correctly on a 64-bit Opteron Linux system (I have
>> RHEL 4.0 AS) with the old PGI 5.2 compiler (as I certainly
>> did), there is a new instruction set of how to build
>> AMBER8 on the PGI website for 64-bit Opteron and Intel
>> processors.
>>
>> http://www.pgroup.com/resources/amber/amber8_pgi60.htm
>>
>> I upgraded my PGI compiler to 6.0-2 and followed the PGI
>> directions after applying the latest bugfixall. Everything
>> compiled correctly and passed the tests (with some minor
>> numerical differences that can be expected). xleap
>> compiles and links correctly (or so it would seem) but
>> gives a segmentation fault upon start, so I an still using
>> the 32-bit binary from my previous ifc 8.1 32-bit build
>> (no problem there). PGI support said that they were
>> unaware of the issue and are looking into it.
>>
>> Cheers,
>>
>> Thomas
>>
>> -- Graduate Student
>> Department of Chemistry & Biochemistry
>> California State University, Long Beach
>> Dr. Marco Lopez Research Group
>> -----------------------------------------------------------------------
>> The AMBER Mail Reflector
>> To post, send mail to amber_at_scripps.edu
>> To unsubscribe, send "unsubscribe amber" to majordomo_at_scripps.edu
>
> --
>
> **********************************************************
>
> Dr. Mark Rance
> Associate Professor
> Dept. of Molecular Genetics, Biochemistry and Microbiology
> University of Cincinnati College of Medicine
> 231 Albert Sabin Way
> Cincinnati, Ohio 45267-0524
> phone: 513-558-0066
> fax: 513-558-8474
> e-mail: mark.rance_at_UC.Edu
>
> **********************************************************
>
>
>
> -----------------------------------------------------------------------
> 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