AMBER Archive (2009)

Subject: Re: [AMBER] Testing parallel amber10: no libmkl_lapack.so

From: Francesco Pietra (chiendarret_at_gmail.com)
Date: Wed Apr 08 2009 - 10:14:43 CDT


On Wed, Apr 8, 2009 at 1:20 PM, Atro Tossavainen
<atro.tossavainen+amber_at_helsinki.fi> wrote:
>> I understand this. However, everything about is sourced correctly from
>> my .bashrc from the intel *.sh.
>
> echo $LD_LIBRARY_PATH

/opt/intel/mkl/10.0.1.014/lib/em64t/:/opt/intel/cce/10.1.015/lib:/opt/intel/fce/10..1.015/lib:usr/local/lib

>
> please.
>
>> If I do not succeed to compile amber10 thereafter, I'll remove
>> everything about intels to try with gnu compilers and libraries.
>> This will be nothing for amber, though bad for the electronic
>> calculations. But I was unable to do better.
>
> Fixing the LD_LIBRARY_PATH is something you can do in a shell simply
> by setting the environment variable.
>
> Adding the -rpath is something you can do in the AMBER configuration
> files before compiling, or manually after it has built the executable
> (just copy/paste the line that builds the executable, add the -rpath
> THAT_DIRECTORY flag, press enter, you're done)
>
> The measures you suggest are on par with "Blah, I can't work out how
> to use this car key, I think I'll just walk." when the entire problem
> is on par with you trying to stick in the plastic end rather than the
> metal one.  But it's up to you.  Keep ignoring the answers you get,
> you're well on your way to becoming killfile fodder as far as I'm
> concerned.

You are wright. I was deeply discouraged about this installation. I
have reinstalled (and cheked positively with examples) the 10.1.015
intels and 10.0.1014 mkl.
However, there is another problem if I want to go fully with intels,
as I did plainly before in amd64 etch:

with ambertools 1.2 (fully patched)

./configure_at icc
make -f Makefile_at
returns (between ===):

=================
mkdir /usr/local/amber10/bin /usr/local/amber10/lib /usr/local/amber10/include
mkdir: cannot create directory `/usr/local/amber10/bin': File exists
mkdir: cannot create directory `/usr/local/amber10/lib': File exists
mkdir: cannot create directory `/usr/local/amber10/include': File exists
make: [install] Error 1 (ignored)
(cd ucpp-1.3 && make install )
make[1]: Entering directory `/usr/local/amber10/src/ucpp-1.3'
Makefile:35: warning: overriding commands for target `.c.o'
../config.h:99: warning: ignoring old commands for target `.c.o'
icc -c -wd117,177,266,880,1011 -DBINTRAJ -DSTAND_ALONE nhash.c
/usr/include/limits.h(125): catastrophic error: could not open source
file "limits.h"
  # include_next <limits.h> ^

compilation aborted for nhash.c (code 4)
make[1]: *** [nhash.o] Error 4
make[1]: Leaving directory `/usr/local/amber10/src/ucpp-1.3'
make: *** [install] Error 2
======================

which limits.h
limits: /usr/include/limits.h

>From "echo:$PATH"
/usr/include

I have lost trace of a web warning that this is an issue of this
version of intel compilers not finding the path of g++. True or not, I
planned to install intels 11. However, as before with version 10, the
program alien did a bad job. The deb package needs fixing, and I have
noticed that intel has changed once again wehre things are
located.That fixing would take long in my hands and i am not
comfortable with installing a rpm package on debian.

thanks a lot for your advice
francesco

I circumvented the problem by compiling AmberTools 1.2 with gcc, which
is fine for me. Then I dropped in the above problems with amber10.

>
> --
> Atro Tossavainen (Mr.)               / The Institute of Biotechnology at
> Systems Analyst, Techno-Amish &     / the University of Helsinki, Finland,
> +358-9-19158939  UNIX Dinosaur     / employs me, but my opinions are my own.
> < URL : http : / / www . helsinki . fi / %7E atossava / > NO FILE ATTACHMENTS
>
> _______________________________________________
> AMBER mailing list
> AMBER_at_ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber
>

_______________________________________________
AMBER mailing list
AMBER_at_ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber