[SOLVED]Abinit 6.8.1 + OpenMPI = FAIL

[trebor]

I have compiled many sources of abinit and their corresponding dependencies, but I have always obtained the same ERROR when trying to run a parallel job.

Finally I decided to make a clean linux install to ensure that there is no conflict within libraries. So I have installed abinit 6.8.1 and open mpi through "apt-get install openmpi-bin" on my ubuntu amd64 version.

I have executed:

mpiexec -np 4 abinit < t_kpt+spin.files >& log

And my log file gives the same error as allways:

" Your architecture is not able to handle 16, 8, 4 or 2-bytes FORTRAN file record markers!
You cannot use ABINIT and MPI/IO.
MPI_ERROR_STRING: Unknown error. Please file a bug report.
ABINIT

Give name for formatted input file:
At line 130 of file iofn1.F90 (unit = 5, file = 'stdin')
Fortran runtime error: End of file
application called MPI_Abort(MPI_COMM_WORLD, 13) - process 0
[unset]: aborting job:
application called MPI_Abort(MPI_COMM_WORLD, 13) - process 0 "

I have to say that despite this error abinit makes the run sequentially.

I would be very grateful if you could please give me a clue on how to solve it.

Thank you,

Trebor

[jbeuken]

Hello,

this is my recipe for a Core-i7 / ubuntu 10.04 (64bits) / abinit 6.8.2 or 6.10.1 ( tested under our testfarm ) :

from a fresh install :

1) installation of theses packages :

Code: Select all

apt-get -y install gcc gfortran
apt-get -y install patch
apt-get -y install libnetcdf4  libnetcdf-dev
apt-get -y install fftw3 libfftw3-dev

apt-get -y install libatlas-base-dev  libatlas3gf-base
apt-get -y install liblapack3gf liblapack-dev
apt-get -y install libblas3gf libblas-dev

apt-get -y install openmpi-bin openmpi-common libopenmpi-dev

2) download of abinit sources ( http://www.abinit.org )

Code: Select all

cd /tmp
wget http://ftp.abinit.org/abinit-6.8.2.tar.gz
tar xzf abinit-6.8.2.tar.gz
cd abinit-6.8.2

3) create a "ubuntu.ac" file :

Code: Select all

prefix="/usr/local"
enable_mpi="yes"
enable_mpi_io="yes"
with_mpi_prefix="/usr"
with_netcdf_incs="-I/usr/include"
with_netcdf_libs="-L/usr/lib64 -lnetcdf -lnetcdff"
with_fft_flavor="fftw3"
with_fft_incs="-I/usr/include/"
with_fft_libs="-L/usr/lib64 -lfftw3"
with_linalg_flavor="atlas"
with_linalg_libs="-L/usr/lib64 -llapack -lf77blas -lcblas -latlas"
with_dft_flavor="atompaw+bigdft+libxc+wannier90"
enable_gw_dpc="yes"
enable_maintainer_checks="no"
enable_test_timeout="yes"
enable_gui_build="no"

4) make abinit & co

Code: Select all

./configure --with-config-file="./ubuntu.ac"
make mj4
make install

abinit binary is in "/usr/local/bin" folder

hope it works for you...

regards

jmb

[trebor]

Dear jmb,

I have successfully installed and run parallel abinit 6.8.2 following your instructions. Thank you very much, I spend many time on this issue and I was being upset.

I noticed you use a similar processor as mine, it is an intel core i7 920 2.67 GHz, which has 4 cores and 8 cores simulated through hyper threading. In order to check the maximum speed I can reach in parallel, I made some tests with the t_kpt+spin tutorial files. I observed that the speed never reaches x4 times that with the sequential run, I suppose it has to do with the architecture of the processor and I suppose that you may have the same handicap. I attach you a graph below.




Thank you very much for your time,

Trebor

[Zhang Xiaole]

Dear jmb,

I have successfully installed and run parallel abinit 6.8.2 following your instructions. Thank you very much, I spend many time on this issue and I was being upset.

I noticed you use a similar processor as mine, it is an intel core i7 920 2.67 GHz, which has 4 cores and 8 cores simulated through hyper threading. In order to check the maximum speed I can reach in parallel, I made some tests with the t_kpt+spin tutorial files. I observed that the speed never reaches x4 times that with the sequential run, I suppose it has to do with the architecture of the processor and I suppose that you may have the same handicap. I attach you a graph below.




Thank you very much for your time,

Trebor

Hi Trebor
i'm wondering have you compared the difference between your two compilings? Because after my compiling with gfortran+openmpi i also got the same erorr as you got when run the parallel abinint. But since i'm a fedora user , i really confused which part of my compiling was wrong.

[lamsalc]

Dear jmb, trebor and all
I exactly followed the steps as mentioned above but after running for sometime I got the following error message:
mpiexec noticed that process rank 0 with PID 2333 on node "PC_NAME" exited on signal 11 (Segmentation fault).
Could you please tell me what is going wrong? Thank you!

[lamsalc]

I repeated the process as suggested; Now it is working. Thank you jmb for your "recipe".

[maxim]

a recipe for installing abinit 7.0.5 on ubuntu-12.04 x64 (intel core 2 quad) :

1) just after installing ubuntu disable all updates in ubuntu first (click "Dash Home" shortcut in upper-right corner and type "update" and press enter, then select "Update Manager" and then press "Settings..." and disable updates there)
2) to invoke terminal do the same as in (1), except typing "terminal"
3) Unlock Root Account in Ubuntu - see http://blog.sudobits.com/2012/05/03/how ... 12-04-lts/
4) select "ubuntu software center" -> type "netcdf" and press enter and install about 6-7 most relevant packages (they appears first in the list)
5) see (past text to terminal line-by-line) viewtopic.php?f=3&t=1387&p=4311&hilit=mpiexec#p4311 and use "ubuntu.ac" file attached here:

ubuntu.ac:

Code: Select all

    prefix="/usr/local"
    enable_mpi="yes"
    enable_mpi_io="yes"
    with_mpi_prefix="/usr"
    with_netcdf_incs="-I/usr/include"
    with_netcdf_libs="-L/usr/lib64 -lnetcdf -lnetcdff"
    with_fft_flavor="fftw3"
    with_fft_incs="-I/usr/include/"
    with_fft_libs="-L/usr/lib64 -lfftw3"
    with-linalg-flavor="none"
    with_linalg_libs="-L/usr/lib64 -llapack -lf77blas -lcblas -latlas"
    with_dft_flavor="atompaw+libxc+wannier90"
    enable_gw_dpc="yes"
    enable_maintainer_checks="no"
    enable_test_timeout="yes"
    enable_gui_build="no"
    enable_gpu="no"


[srasoul]

hello
im a newworker in ubuntu & abinit
i dont understand it:
prefix="/usr/local"
enable_mpi="yes"
enable_mpi_io="yes"
with_mpi_prefix="/usr"
with_netcdf_incs="-I/usr/include"
with_netcdf_libs="-L/usr/lib64 -lnetcdf -lnetcdff"
with_fft_flavor="fftw3"
with_fft_incs="-I/usr/include/"
with_fft_libs="-L/usr/lib64 -lfftw3"
with_linalg_flavor="atlas"
with_linalg_libs="-L/usr/lib64 -llapack -lf77blas -lcblas -latlas"
with_dft_flavor="atompaw+bigdft+libxc+wannier90"
enable_gw_dpc="yes"
enable_maintainer_checks="no"
enable_test_timeout="yes"
enable_gui_build="no"

how i can get it?

thanks very much

[pouillon]

See ~abinit/doc/config/build-config.ac, as stated in the forum rules at the top of this page.

[srasoul]

i solved it
thank you
regards

[srasoul]

hello
I exactly followed the steps as mentioned above but after "make mj4" I got the following error message:

"call scrhdr_mpio_skip(mpi_fh,test_fform,offset)
1
Error: Type mismatch in argument 'offset' at (1); passed INTEGER(8) to INTEGER(4)
make[5]: *** [m_io_screening.o] Error 1
make[5]: *** Waiting for unfinished jobs....
make[5]: Leaving directory `/home/queen/Desktop/abinit-6.8.2/src/62_iowfdenpot'
make[4]: *** [all-recursive] Error 1
make[4]: Leaving directory `/home/queen/Desktop/abinit-6.8.2/src'
make[3]: *** [all-recursive] Error 1
make[3]: Leaving directory `/home/queen/Desktop/abinit-6.8.2'
make[2]: *** [all] Error 2
make[2]: Leaving directory `/home/queen/Desktop/abinit-6.8.2'
make[1]: *** [multi] Error 2
make[1]: Leaving directory `/home/queen/Desktop/abinit-6.8.2'
make: *** [mj4] Error 2"

Could you please tell me what is going wrong? Thank you!

[gmatteo]

=== modified file 'src/62_iowfdenpot/m_io_screening.F90'
--- src/62_iowfdenpot/m_io_screening.F90 2013-01-22 09:22:38 +0000
+++ src/62_iowfdenpot/m_io_screening.F90 2013-05-02 11:49:40 +0000
@@ -1516,7 +1516,7 @@
integer :: test_fform,mpi_err,ierr,sc_mode
integer :: bsize_frm,mpi_type_frm
integer :: mpi_fh,buf_dim !,mat_ggw,mat_ggwq
- integer(MPI_OFFSET_KIND) :: offset,displ_wq !,my_offpad
+ integer(XMPI_OFFSET_KIND) :: offset,displ_wq !,my_offpad
!complex(dpc) :: ctmp
#endif
character(len=500) :: msg

[srasoul]

Dear gmatteo
thanks for your post but i dont understand it and i cant fix my problem.
could you please help me?
regards

[Alain_Jacques]

Hi,

Copy Matteo's code

Code: Select all

=== modified file 'src/62_iowfdenpot/m_io_screening.F90'
--- src/62_iowfdenpot/m_io_screening.F90 2013-01-22 09:22:38 +0000
+++ src/62_iowfdenpot/m_io_screening.F90 2013-05-02 11:49:40 +0000
@@ -1516,7 +1516,7 @@
integer :: test_fform,mpi_err,ierr,sc_mode
integer :: bsize_frm,mpi_type_frm
integer :: mpi_fh,buf_dim !,mat_ggw,mat_ggwq
- integer(MPI_OFFSET_KIND) :: offset,displ_wq !,my_offpad
+ integer(XMPI_OFFSET_KIND) :: offset,displ_wq !,my_offpad
!complex(dpc) :: ctmp
#endif
character(len=500) :: msg

to a text file, save under the name matteo.txt then go to abinit root directory and apply the patch with

Code: Select all

patch -p0 < matteo.txt

Kind regards,

Alain