[SOLVED] MPI problem when running 'make'

option, parallelism,...

Moderators: fgoudreault, mcote

Forum rules
Please have a look at ~abinit/doc/config/build-config.ac in the source package for detailed and up-to-date information about the configuration of Abinit 8 builds.
For a video explanation on how to build Abinit 7.x for Linux, please go to: http://www.youtube.com/watch?v=DppLQ-KQA68.
IMPORTANT: when an answer solves your problem, please check the little green V-like button on its upper-right corner to accept it.
Locked
asorini
Posts: 13
Joined: Wed Apr 07, 2010 12:23 am

[SOLVED] MPI problem when running 'make'

Post by asorini » Wed Apr 07, 2010 12:37 am

Hi forum,

I'm trying to compile the parallel version of abinit-6.0.3 with OpenMPI on my group's computing cluster. I have previously been able to compile 5.8 and 5.7 versions of abinit, but 6.0 fails to compile when I run make.

For this failed attempt at compiling I first configured using:

$ ./configure CC=mpicc CXX=mpicxx FC=mpif90 --enable-mpi=yes --disable-all-plugins --enable-mpi-io

and configure runs without any problems (I've attached config.log).

Then I use:

$ make

which runs for a while and then fails in ./src/12_hide_mpi when it chokes on some MPI_RECV and MPI_SEND for 5d arrays in xexch_mpi.F90 (lines 507 and 510). Below I've pasted some of the output from make:

Making all in 12_hide_mpi
make[3]: Entering directory `/afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src/12_hide_mpi'
mpif90 -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -free -module /afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src/mods -O2 -g -extend-source -vec-report0 -c -o xallgather_mpi.o xallgather_mpi.F90
mpif90 -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -free -module /afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src/mods -O2 -g -extend-source -vec-report0 -c -o xallgatherv_mpi.o xallgatherv_mpi.F90
mpif90 -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -free -module /afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src/mods -O2 -g -extend-source -vec-report0 -c -o xalltoallv_mpi.o xalltoallv_mpi.F90
mpif90 -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -free -module /afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src/mods -O2 -g -extend-source -vec-report0 -c -o xalltoall_mpi.o xalltoall_mpi.F90
mpif90 -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -free -module /afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src/mods -O2 -g -extend-source -vec-report0 -c -o xbarrier_mpi.o xbarrier_mpi.F90
mpif90 -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -free -module /afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src/mods -O2 -g -extend-source -vec-report0 -c -o xcast_mpi.o xcast_mpi.F90
mpif90 -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -free -module /afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src/mods -O2 -g -extend-source -vec-report0 -c -o xcomm_rank.o xcomm_rank.F90
mpif90 -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -free -module /afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src/mods -O2 -g -extend-source -vec-report0 -c -o xcomm_size.o xcomm_size.F90
mpif90 -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -free -module /afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src/mods -O2 -g -extend-source -vec-report0 -c -o xerror_string.o xerror_string.F90
mpif90 -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -free -module /afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src/mods -O2 -g -extend-source -vec-report0 -c -o xexch_mpi.o xexch_mpi.F90
xexch_mpi.F90(507): error #6285: There is no matching specific subroutine for this generic subroutine call. [MPI_RECV]
call MPI_RECV(vrecv,n1*n2*n3*n4*n5,MPI_DOUBLE_PRECISION,sender,etiq,spaceComm,statux,ier)
-------^
xexch_mpi.F90(510): error #6285: There is no matching specific subroutine for this generic subroutine call. [MPI_SEND]
call MPI_SEND(vsend,n1*n2*n3*n4*n5,MPI_DOUBLE_PRECISION,recever,etiq,spaceComm,ier)
-------^
compilation aborted for xexch_mpi.F90 (code 1)
make[3]: *** [xexch_mpi.o] Error 1
make[3]: Leaving directory `/afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src/12_hide_mpi'
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory `/afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3/src'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/afs/slac.stanford.edu/package/ABINIT/abinit-6.0.3'
make: *** [all] Error 2


I would greatly appreciate any help/ideas from the Abinit gurus of the forum. Thanks in advance. Cheers,

Adam
Attachments
config.log
(89.14 KiB) Downloaded 402 times

User avatar
pouillon
Posts: 651
Joined: Wed Aug 19, 2009 10:08 am
Location: Spain
Contact:

Re: MPI problem when running 'make'

Post by pouillon » Wed Apr 07, 2010 4:45 pm

Could you try the following?

Code: Select all

./configure --with-mpi-prefix="/afs/slac.stanford.edu/package/OpenMPI" --enable-mpi=yes --disable-all-plugins --enable-mpi-io
Yann Pouillon
Simune Atomistics
Donostia-San Sebastián, Spain

nicheal
Posts: 3
Joined: Wed Apr 07, 2010 5:32 pm

Re: MPI problem when running 'make'

Post by nicheal » Thu Apr 08, 2010 6:15 am

I'm in the same predicament with OPENMPI1.33, Intel fortran 11.1.046

asorini
Posts: 13
Joined: Wed Apr 07, 2010 12:23 am

Re: MPI problem when running 'make'

Post by asorini » Thu Apr 08, 2010 9:40 am

Hi,

Thanks for getting back to me so quickly. I tried running configure with the flags you suggested, but I get the exact same error when I run make. Any other suggestions? Cheers,
Adam

User avatar
pouillon
Posts: 651
Joined: Wed Aug 19, 2009 10:08 am
Location: Spain
Contact:

Re: MPI problem when running 'make'

Post by pouillon » Thu Apr 08, 2010 12:59 pm

Then the problem comes from the MPI implementation you're using. It might be related to the way it has been compiled.

You should ask your system administrator.
Yann Pouillon
Simune Atomistics
Donostia-San Sebastián, Spain

asorini
Posts: 13
Joined: Wed Apr 07, 2010 12:23 am

Re: MPI problem when running 'make'

Post by asorini » Thu Apr 08, 2010 10:16 pm

Thanks very much for your help. I'll talk to the system admin. Cheers,
Adam

nicheal
Posts: 3
Joined: Wed Apr 07, 2010 5:32 pm

Re: MPI problem when running 'make'

Post by nicheal » Fri Apr 09, 2010 10:29 am

I uninstall the OPENMPI1.3.2, and update to openmpi 1.3.3. Then No error with Openmpi 1.3.3! You can try the new version!

asorini
Posts: 13
Joined: Wed Apr 07, 2010 12:23 am

Re: [SOLVED] MPI problem when running 'make'

Post by asorini » Mon Apr 12, 2010 6:57 pm

Thanks very much for the advice. Cheers, Adam

mverstra
Posts: 655
Joined: Wed Aug 19, 2009 12:01 pm

Re: [SOLVED] MPI problem when running 'make'

Post by mverstra » Tue Apr 20, 2010 12:19 am

Matthieu Verstraete
University of Liege, Belgium

Locked