AMBER Archive (2005)

Subject: Re: AMBER: compiling amber with pgi 5.2.4

From: Robert Duke (rduke_at_email.unc.edu)
Date: Mon Oct 24 2005 - 07:44:42 CDT


Hannes -
For my edification, please compile pmemd separately and let me know how that
goes. As I presume you know, you do a separate configure and make in the
pmemd directory, and presumably nothing has been broken in the plain old
fortran 90 code. Your problems below reference problems with include files
of the standard template library for c++. Now, I am not currently active in
c++ development, but I know from recent experience trying to get some stuff
built for someone that the c++ language definition has been severely dinked
with, especially in the area of stl, recently. So if you have gcc
3.2-compliant code, there may be problems moving it forward to 3.3, 4.0,
etc. There may be problems with interaction between pgi and gcc, etc., and
who the heck knows what has been done with pgi in terms of following these
language definitions (ie., I have no idea what standards and gcc version
compatibility pgi is currently supporting). I would therefore expect that
the antechamber guys will have to look at their code seriously in regard to
this.
Best Regards - Bob Duke

----- Original Message -----
From: "Hannes Barsch" <hbarsch_at_physik.uni-bielefeld.de>
To: <amber_at_scripps.edu>
Sent: Monday, October 24, 2005 8:21 AM
Subject: AMBER: compiling amber with pgi 5.2.4

> Dear AMBER community!
>
> On a newly installed Opteron system I am not able to compile AMBER 8 using
> the Portland Group compilers. I use the same settings as in a former
> install - the only differences are the version of PGI (now 5.2.4, before
> 5.2.2) and the Rocks-OS version (now 4.0.0 (based on CentOS 4.0), before
> 3.2.0 (based on RHEL 3.0)). I also included Bob Duke's configuration
> updates
> for PMEMD now, but the compilation error occurs already earlier at the
> compilation of bondtype.C in the antechamber module:
>
> pgcc -O2 atomtype.o -lm -o atomtype
> pgCC -c -o bondtype.o bondtype.C
> "/usr/local/pgi/linux86-64/5.2/include/CC/stl/_bvector.h", line 86:
> error:
> identifier "__CHAR_BIT__" is undefined
> if (_M_offset++ == __WORD_BIT - 1) {
> ^
>
> "/usr/local/pgi/linux86-64/5.2/include/CC/stl/_bvector.h", line 94:
> error:
> identifier "__CHAR_BIT__" is undefined
> _M_offset = __WORD_BIT - 1;
> ^
>
> "/usr/local/pgi/linux86-64/5.2/include/CC/stl/_bvector.h", line 106:
> error:
> identifier "__CHAR_BIT__" is undefined
> _M_p += __n / __WORD_BIT;
> ^
>
> "/usr/local/pgi/linux86-64/5.2/include/CC/stl/_bvector.h", line 116:
> error:
> identifier "__CHAR_BIT__" is undefined
> return __WORD_BIT * (_M_p - __x._M_p) + _M_offset - __x._M_offset;
> ^
>
> 4 errors detected in the compilation of "bondtype.C".
> make[1]: *** [bondtype.o] Error 2
> make[1]: Leaving directory
> `/usr/src/redhat/BUILD/amber-8/src/antechamber'
> make: *** [serial] Error 2
> error: Bad exit status from /var/tmp/rpm-tmp.72495 (%build)
>
> The /usr/local/pgi/linux86-64/5.2/include/CC/stl/_bvector.h does not
> differ
> from the header file of the older PGI installation.
>
> Has anyone compiled amber with this combination of HW / OS / compilers and
> realized similar compile errors? Does anybody know why this error might
> happen? If so, please give me a hint!
> Thanks for your help! Regards,
>
> Hannes Barsch
>
>
> ____________________________________________________________________________
> _
>
> The RPM spec file looks like this:
>
> Summary: AMBER 8
> Name: amber
> Version: 8
> Release: 0
> Copyright: University of California
> Group: Sciences/Chemistry
> Source0: %{name}-%{version}.tar.gz # amber 8 source from CD
> Source1: %{name}-%{version}.pmemd.tar.gz # new configuration files from
> http://amber.scripps.edu/pmemd-mar05.html
> Source2: %{name}-%{version}.local.tar.gz # configuration files for
> opterons/PGI from http://www.pgroup.com/resources/amber/amber8_pgi52.htm
> Patch0: %{name}-%{version}.patch0 # amber bugfixes from
> http://amber.scripps.edu/bugfixes/8.0/bugfix.all
> Patch1: %{name}-%{version}.patch1 # PGI fixes adopted from
> http://www.pgroup.com/resources/amber/amber8_pgi52.htm
> URL: http://amber.scripps.edu
> Packager: Hannes Barsch, University of Bielefeld
> AutoReq: no
> Requires: /bin/csh, /bin/sh, /bin/sh, /bin/sh, /usr/bin/perl,
> libICE.so.6()(64bit), libSM.so.6()(64bit), libX11.so.6()(64bit),
> libXext.so.6()(64bit), libXt.so.6()(64bit), libc.so.6()(64bit),
> libc.so.6(GLIBC_2.2.5)(64bit), libc.so.6(GLIBC_2.3)(64bit),
> libgm.so.0()(64bit), libm.so.6()(64bit), libm.so.6(GLIBC_2.2.5)(64bit),
> libpthread.so.0()(64bit), libpthread.so.0(GLIBC_2.2.5)(64bit), perl(CPin),
> perl(Carp), perl(Data::Dumper), perl(Getopt::Long), perl(IO::File),
> perl(lib), perl(strict), rpmlib(CompressedFileNames)<= 3.0.4-1,
> rpmlib(PayloadFilesHavePrefix) <= 4.0-1
>
>
> %description
> Amber refers to two things: a set of molecular mechanical force fields
> for the simulation of biomolecules (which are in the public domain, and
> are
> used in a variety of simulation programs); and a package of molecular
> simulation programs which includes source code and demos.
>
> %prep
> %setup -q
> %setup -qDTb 1
> %setup -qDTb 2
> %patch -P 0
> %patch -P 1
>
> %build
> PATH=/usr/local/pgi/linux86-64/5.2/bin:$PATH
> export PGI=/usr/local/pgi
> export AMBERHOME=$RPM_BUILD_DIR/%{name}-%{version}
> export GM_LIBDIR=/opt/gm/lib
> cd $AMBERHOME/src
> ./configure -nopar -opteron -acml -lapack pgf90
> make -e YACC="/usr/bin/bison -y" X11LIBDIR="lib64" serial
> cd pmemd
> ./new_configure linux_opteron pgf90 nopar
> make install
> cd $AMBERHOME/exe
> mv pmemd pmemd.serial
> mv sander sander.serial
> mv sander.LES sander.LES.serial
>
> cd $AMBERHOME/src
> ./configure -mpich -opteron -acml -lapack pgf90
> make -e YACC="/usr/bin/bison -y" X11LIBDIR="lib64" parallel
> cd pmemd
> ./new_configure linux_opteron pgf90 mpich
> make install
>
> [...]
>
> -----------------------------------------------------------------------
> 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