The ABINIT Users Mailing List ( CLOSED )

[zahra hooshmand <z.hooshmand@gmail.com>]

Dear Alian,
thanks, I will do it and I will give you results. but just another question:
does gcc work too? does other kinds of CPU work, like AMD or others? which kind of compiler is suitable then?
and does abinit have any restart option? for continuing the job from stopped level nad to help not beginning from start. For example in GW calculations, the screening level is so long, some times it occured to me thst due to some problems it stopped after 14 days and I had to start it from beginnig. what should I do to avoid such a problems?

Again Thank You!
Best Regards
Zahra

On Tue, Dec 1, 2009 at 4:19 PM, Alain Jacques <ajs.jacques@gmail.com> wrote:

Dear Zahra,

Let's try it from scratch ...

Locate you mpif90 compiler wrapper ; make sure that it is in your path -
let say that mpif90 is in /root_of_mpich2/bin/mpif90, export
PATH=/root_of_mpich2/bin:$PATH if it is not done already.
Check with a "mpif90 -show" that you are using a suitable backend
compiler - any reasonably recent ifort should work, gfortran >4.1 should
be OK
 Try to compile and run the following test program:

PROGRAM HelloWorld
include 'mpif.h'

call mpi_init(ierr)
call mpi_comm_size(MPI_COMM_WORLD,npes, ierr)
call mpi_comm_rank(MPI_COMM_WORLD,irank,ierr)

print*,'Hello World! I am ',irank,' of ',npes

call mpi_finalize(ierr)

END PROGRAM

Name the code mpi_hello.f90 Compile it with a "mpif90  mpi_hello.f90"
... should produce an a.out file. Run 4 instances of it with a "mpiexec
-np 4 ./a.out". You should receive a ...

 Hello World! I am            0  of            4
 Hello World! I am            2  of            4
 Hello World! I am            1  of            4
 Hello World! I am            3  of            4

If it works, go on with abinit; if it doesn't, try to fix it - there is
no logic to try to compile half a million lines of code if your
toolchain is broken.

Back to abinit ... untar the 5.8.4p version, go to abinit-5.8.4 source
tree directory, create a new build subdirectory - let's say tmp-build
and change to it.
From that build directory, invoke "../configure --disable-config-file
--enable-mpi --with-mpi-prefix=/root_dir_of_mpich2"
Check that configure goes up to the end, look at config.log. Do a
"make".  When the compilation is over, do a "ldd src/98_main/abinip" and
check that there are no unresolved libraries calls.

When you have the binary, send a message about how to test it.

Kind regards,

Alain

zahra hooshmand wrote:
> Dear Alian
> Itry it on two cluster with 2 different Linux(scientific Linux,
> Ubuntu) in both of them, i had errors as:
> error[1], error[2]
> finally I find following commands and it make me mentioned Abinip file:
> ./configure --enable mpi=yes
> sudo make all install
>
> As I said this didn't work! the MPI on both systems are: MPICH2, the
> compilers are:
> ifort in one case, and Gcc in another.
> By the way, doaes abinit have any restart option just from where the
> running was stopped. for example in GW calculations, in screening file
> calculation?
>
> Best Regards
> Zahra

-- 
Regards
Zahra Hooshmand