Re: AMBER: A question on AMBER8 installation

From: M. L. Dodson <bdodson.scms.utmb.edu>
Date: Thu, 14 Oct 2004 13:41:15 -0500

Pardon me for butting in, can you build leap all by itself?
Leap does not involve any fortran and works just fine with
gcc/g++. See below.

On Thursday 14 October 2004 12:57 pm, Hwankyu Lee wrote:
> I appreciate your answers..but, still have problems..
> I exactly followed Ross's procedures...I started all the procedures from the
> top.
> 1)tar xvf AMBER8.tar
> 2)cd /home/leehk/local/amber8
> setenv AMBERHOME /home/leehk/local/amber8
> 3)cd home/leehk/local/amber8/
> patch -p0 -N -r patch_rejects <bugfix.all (going well)
> 4)cd home/leehk/local/amber8/src
> setenv MKL_HOME /opt/intel/mkl70
> 5)cd /leehk/local/amber8/src
> ./configure -p4 -static -nopar ifort
> 6)Edit config.h and change gcc and g++ to icc
> 7)make clean
> 8)make or make serial
> 9)cd /leehk/local/amber8/test
> make test
>
> AFter 8), I have error like this at the end of execution.
> --------------
> bondtype.o(.gnu.linkonce.t._ZSt24__uninitialized_copy_auxIN9__gnu_cxx17__normal_iteratorIP5PointSt6vectorIS2_SaIS2_EEEES7_ET0_T_S9_S8_12__false_type+0x91):
> In function `destr_detour20.0.156':
> : undefined reference to `__cxa_end_catch'
> bondtype.o(.eh_frame+0x12): In function `.eh_frame_seg':
> : undefined reference to `__gxx_personality_v0'
> make[1]: *** [bondtype] Error 1
> make[1]: Leaving directory `/home/leehk/local/amber8/src/antechamber'
> make: *** [serial] Error 2
> -----------------
>
> After 9), I have this error.
> -----------------
> cd dmp; ./Run.dmp
> diffing mdout.dmp.save with mdout.dmp
> PASSED
> ==============================================================
> cd adenine; ./Run.adenine
> diffing mdout.adenine.save with mdout.adenine
> PASSED
> ==============================================================
> cd cytosine; ./Run.cytosine
> Badly placed ()'s.
> make: *** [test.sander.no_lmod] Error 1
> --------------------
>
> In terms of intrinsic.h file, I couldn't find the file. In /usr/X11R6/include,
> thers is no file there. I'm using fedora 2.
> Could you give me other hints for this problem, please?
> Thanks for you help.
>
> best,
> Hwankyu.
>
>

The file should be:
/usr/X11R6/include/X11/Intrinsic.h

Note the exact spelling and be aware that unix file system naming
is case sensitive.

If you really mean there is no file there (i.e., the directory is
empty), you probably need to install the X development system
(rpm? I don't normally do Linux.) If you just mean that intrinsic.h is
not there, then it is because you are ignoring the capitalization
of the Intrinsic.h file name.

If leap now builds correctly, pardon my intrusion. I have no
useful comments on the ins and outs of Intel compilers, I just use
ifc version 7 and live with the performance hit.

Bud Dodson

>
> Quoting Ross Walker <ross.rosswalker.co.uk>:
>
> > Dear Hwankyu,
> >
> > > ./configure -p4 -nopar -ifort
> >
> > You definitely did ./configure -p4 -nopar ifort here yes (I.e no - before
> > the ifort)?
> >
> > > icc -c -O2 -o makeDIST_RST.o makeDIST_RST.c
> > > makeDIST_RST.c(338): warning #266: function declared implicitly
> > > if ( !lookuprtyp(atma.rnum,atma.nrtyp) ||
> >
> > Don't worry about these. They are just warnings. The compiler thinks the
> > routines are declared implicitly because the call to the routine comes
> > earlier in the file than the actual routine and the function is not
> > prototyped at the beginning of the file. This occurs because icc is not
> > really smart enough to scan through the file twice before reporting
> > warnings. This sort of error is only a problem if it occurs at the link
> > stage (when all routines must have been declared in the object files or
> > libraries).
> >
> > > bondtype.o(.eh_frame+0x12): undefined reference to
> > > `__gxx_personality_v0'
> >
> > Are you certain you did:
> >
> > cd $AMBERHOME/src/
> > make clean
> >
> > Before you did the make?
> >
> > The reason I ask is because the __ in front of gxx_peronality_v0 looks like
> > the naming convention used by gcc. As far as I am aware icc only expects one
> > underscore. Hence it looks like all the object files were not cleaned from
> > the directory before you ran the make again.
> >
> > If you have not done it, I would suggest the following:
> >
> > Remove your existing amber8 directory tree.
> >
> > 1) Get hold of the original tar file from the amber8 cd (so you know it is
> > unmodified).
> > 2) Untar this to a folder on your machine.
> > 3) set AMBERHOME to this directory setenv AMBERHOME /path_to_inst/amber8
> > 3) Obtain the latest bugfixes from here:
> > http://amber.scripps.edu/bugfixes/8.0/bugfix.all - save it in the amber8
> > directory the untaring created.
> > 4) Patch your amber8 copy with these bugfixes:
> > cd $AMBERHOME
> > patch -p0 -N -r patch_rejects < bugfix.all
> > 5) cd into the src directory
> > cd $AMBERHOME/src
> > 6) Create the configure script
> > ./configure -p4 -static -nopar ifort
> > 7) Edit config.h and change all gcc and g++ to icc
> > 8) clean everything
> > make clean
> > 9) make
> > 10) cd $AMBEROME/test
> > 11) make
> >
> > See how you get on. The important thing is to make sure you are starting
> > from a clean un-aulterated copy of amber8.
> >
> > All the best
> > Ross
> >
> >
> > /\
> > \/
> > |\oss Walker
> >
> > | Department of Molecular Biology TPC15 |
> > | The Scripps Research Institute |
> > | Tel:- +1 858 784 8889 | EMail:- ross.rosswalker.co.uk |
> > | http://www.rosswalker.co.uk/ | PGP Key available on request |


-- 
M. L. Dodson                                bdodson.scms.utmb.edu
409-772-2178                                FAX: 409-747-8608
-----------------------------------------------------------------------
The AMBER Mail Reflector
To post, send mail to amber.scripps.edu
To unsubscribe, send "unsubscribe amber" to majordomo.scripps.edu
Received on Thu Oct 14 2004 - 19:53:00 PDT
Custom Search