AMBER Archive (2002)

Subject: Re: sander (CAP) chokes on efc (Itanium2/Linux)

From: Tadashi Takemori (takemori_at_bk.tsukuba.ac.jp)
Date: Wed Dec 18 2002 - 03:38:25 CST

Thank you for the advice. Yes, I am using Amber7 with efc7.0 on Red Hat
Linux. I have tried modifying Machine.nec_tx for my hp workstation
zx2000 (Red Hat Linux/Itanium2/efc-ecc), but got the same result (chokes
on Run.cap as on a few others). I instead tried compiling with -O2
wherever the option was -O3 and it worked perfectly! I therefore still
suspect this to be a bug in the efc. Thank you for the help all the
same. (Perhaps this info should be put up somewhere for
Itanium2/Linux/efc users?)

Tadashi Takemori
Institute of Materials Science
University of Tsukuba,
Tsukuba Ibaraki 305-8573, Japan
takemori_at_bk.tsukuba.ac.jp
+81-298-53-5312

Karen Haskell wrote:

>Are you using Amber 7?
>
>See the Amber web page under "Amber-related links" to download
>files and instructions we (at NEC) used to build Amber 7 on
>Itanium. The files and instructions you want to look at are for
>the NEC_TX systems. This was initially done on Itanium but
>should be fine for Itanium2. If you have any further questions
>please contact us.
>
>--------------------------------------------------
> Karen Haskell
>--------------------------------------------------
> NEC Solutions (America), Inc.
> Advanced Technology Computing Center
> 4200 Research Forest Drive, Suite 400
> The Woodlands, TX 77381-4257
>
> E-mail: khaskell_at_atcc.necsys.com
> Phone: 281-465-1529
> Fax: 281-465-1599
>--------------------------------------------------
>
>
>-----Original Message-----
>From: Tadashi Takemori [mailto:takemori_at_bk.tsukuba.ac.jp]
>Sent: Tuesday, December 17, 2002 1:11 AM
>To: amber-request_at_heimdal.compchem.ucsf.edu;
>amber_at_heimdal.compchem.ucsf.edu
>Subject: sander (CAP) chokes on efc (Itanium2/Linux)
>
>
>I have compiled AMBER on an Itanium2/Linux machine using the
>Intel
>Fortran Compiler that comes by the name of 'efc' instead of 'ifc'
>by
>just making a symbolic link ifc -> efc and using Machine.ifc.
>The test
>runs passed perfectly on most items but failed dismally on a
>small
>number of items notably in Run.cap in nonper, reporting a series
>of
>question marks '?????????' instead of numbers (indicating the
>number
>blew up). Given that the same program works perfectly in most
>runs but
>blows up completely in others, I suspect this is a compiler
>issue, or
>some slight unintended/unannonced modification in the Fortran
>grammar
>from ifc to efc. Can you possibly provide me with some helpful
>information (e.g. a Machine file specifically tuned for use with
>efc/Red
>Hat Linux)?
>
>Thank you.
>
>Tadashi Takemori
>Institute of Materials Science,
>University of Tsukuba,
>Tsukuba Ibaraki 305-8573, Japan
>tel: +81-298-53-5312
>
>
>
>
>