I managed to compile it with an old arch.make copied from Siesta 2.0.2, but
that's only because the configure script from FoX runs successful in this case.
I want to use a new MPI version (OpenMPI) instead, because the old one was
specific to InfiniPath, and I don't have that on the complete cluster. If I
specify the OpenMPI mpif90 instead, it fails.
Here is what I do (trying to compile it from scratch):
#tar --gzip -xf ~hf63/siesta-3.0-b.tgz
#cd siesta-3.0-b/Obj
#sh ../Src/obj_setup.sh
#cp ~/arch.make .
#make
This gives the following output:
Compilation architecture to be used: intel9-cmkl8-mpi
If this is not what you want, create the right
arch.make file using the models in Src/Sys
Hit ^C to abort...
(cd FoX; touch arch.make ; \
CONFIGURE="/usr/local/siesta-3.0-b/Src/FoX/configure"; \
$CONFIGURE VPATH="/usr/local/siesta-3.0-b/Src/FoX" \
FC="/opt/bin/mpif90" FCFLAGS="-O2" \
--enable-wcml || false )
checking build system type... x86_64-unknown-linux-gnu
checking host system type... x86_64-unknown-linux-gnu
checking for linker flag to name executables... -o
checking for Fortran compiler default output file name... conftest
checking whether the Fortran compiler works... configure: error: cannot run
Fortran compiled programs.
If you meant to cross compile, use `--host'.
See `config.log' for more details.
make: *** [FoX/.config] Error 1
I don't want it to run compiled programs! I want it to compile with he settings
I have in my arch.make. Is there an easy way to switch this off?
Thanks in advance,
Herbert
Quoting "Barraza-lopez, Salvador" <[email protected]>:
>
>
> Â Marcos is right. It seems like at the beginning something is being
> configured on the FoX directory, but as long as you first copy your own
> arch.make that used to work in previous versions you are fine (are you
> compiling SIESTA for the very first time?) .
>
>
>
> I compiled 3.0b with my arch.make from a previous version. The only problem
> that arose at some point during compilation was in trying to get the location
> of some mod files from the FoX directory. For that, I just passed the include
> directoy location (~/ siesta-3.0-b/Obj/FoX/objs/finclude in my case) on my
> fortran compiler line:
>
>
> FC=ifort -lmpi -I/u/ac/sbl3/siesta-3.0-b/Obj/FoX/objs/finclude
>
>
>
> Â That did it for me. With best regards,
>
> -Salvador.
>
>
> ----- Original Message -----
> From: "Marcos VerÃssimo Alves" <[email protected]>
> To: [email protected]
> Sent: Saturday, December 12, 2009 12:58:42 PM GMT -05:00 US/Canada Eastern
> Subject: Re: [SIESTA-L] parallel compilation of 3.0b
>
> No. The configure of the modules is "harmless" because it uses the same
> options as arch.make. If you have an arch.make that works, just type "make",
> sit back, relax and be happy afterwards ;)
>
>
> On Sat, Dec 12, 2009 at 6:53 PM, Herbert Fruchtl <
> [email protected] > wrote:
>
>
> The problem is that even if I supply my own arch.make in Obj, "make" will
> start
> the configure script from other directories (FoX, but I'm not sure if this is
> the only one). Do I have to copy my arch.make to all subdirectories of Src?
>
> Â Herbert
>
>
>
>
> Quoting Marcos VerÃssimo Alves < [email protected] >:
>
> > Yes, there is... Just skip the "configure" part, use an arch.make supplied
> > by you and type "make". It will run a few configure's for the modules of
> > siesta, which them selves have make files, and start building everything.
> If
> > you are too lost about arch.make's, there are plenty of them flying around
> > on the list archives and of course, on the Src/Sys directory, with many
> > flavors of compilers and math libraries - but you will most certainly have
> > to adjust:
> >
> > 1) the paths for the libraries (especially blacs and scalapack, which have
> > to be compiled individually)
> > 2) the paths for the mpi stuff, if your system has some particular/weird
> > configuration.
> >
> > Now I have realized one thing... are you trying to customize an arch.make
> > and *after* customizing it typing ../../Src/configure ? I've never done it
> > myself but I'd guess configure would overwrite whatever you put in the
> > arch.make with its own guesses, hence no matter what you put in configure
> > would generate complaints.
> >
> > Cheers,
> >
> > Marcos
> >
> > On Sat, Dec 12, 2009 at 1:31 AM, Herbert Fruchtl <
> > [email protected] > wrote:
> >
> > > Is there a way to compile 3.0b without it trying (and usually failing) to
> > > determine all the compiler flags by itself? No matter what I put into
> > > arch.make, it will complain that it cannot determine how to create an
> > > executable, or run compiled programs (presumably because they need
> > > mpirun...)
> > >
> > > Â Herbert
> > >
> > >
> > > ------------------------------------------------------------------
> > > University of St Andrews Webmail: https://webmail.st-andrews.ac.uk
> > >
> > >
> >
>
>
> --
> Herbert Fruchtl
> EaStCHEM Fellow
> School of Chemistry
> University of St Andrews
>
>
>
>
>
> ------------------------------------------------------------------
> University of St Andrews Webmail: https://webmail.st-andrews.ac.uk
>
>
>
>
> --
> Salvador Barraza-Lopez
> Postdoctoral Fellow
> School of Physics
> The Georgia Institute of Technology
>
> Office N205
> 837 State Street Atlanta, Georgia 30332-0430 U.S.A
> Tel: (404) 894-0892 Fax: (404) 894-9958
>
--
Herbert Fruchtl
EaStCHEM Fellow
School of Chemistry
University of St Andrews
------------------------------------------------------------------
University of St Andrews Webmail: https://webmail.st-andrews.ac.uk
