Re: gfortran and a compilation

[Francesco Pietra <frapietra@alice.it>]

On Monday 26 June 2006 19:46, Stefano Simonucci wrote:
> On Mon, 2006-06-26 at 16:14 +0100, Jo Shields wrote:
> > We make extensive use of Portland, Intel and GNU compilers (both C and
> > Fortran) at work - all of them work fine, and there are certainly no
> > issues with G77 on AMD64. That said, our GAMESS installation was
> > compiled with Intel 9.0 (as GNU is almost always considerably slower
> > than PGI or Intel)
>
> I have used the Intel compiler to compile my programs and I have not
> seen improvement with respect to GNU fortran.
>
> Stefano

Stefano, accepting your experience I started preparing Makefile.config 
indicaring g77 for Fortran and gcc for C (I am at amd64 debian etch, parallel 
machine dual opteron) yesterday I went on obtaining successfully 
makefile.config and an invite to run make, although none of the mathem 
libraries had been found. The program specifically said "should I proceed 
without"? I said yes, but I did not run make.

To day
#apt-file update
#apt-file search 
returns
/usr/lib/libatlas.a
/usr/lib/libblas.a
which are just two of the mathematic files requested by ./configure

Making clear that I have
/usr/lib
/usr/lib64-> lib
[where /lib only contains 64bit libraries (I have purged from lib32 that base 
installation had provided)]
looking into /usr/lib
I only found
libblas.so.3 -> libblas.so.3.0
libblas.so.3.0
not exacly what ap-file search had indicated.
Is that correct?

Moreover (although this may be too much to ask because it may be specific of 
the software), I should manually (the software is not yet automatic from this 
viewpoint) indicate in Makefile.config where the libs have to be searched

nano examination of Makefile.config reveals empty

MPI_INCLUDE_DIR =
MPI_LIB_PATH =
MPI_LIB =

Does this make clear where the libs should be indicated?

Thanks a lot to all guys suggesting me how to go on
francesco