ABINIT Discussion Forums

The meeting place for ABINIT users and developers
https://forum.abinit.org/

Unable to compile in parallel using mvapich

https://forum.abinit.org/viewtopic.php?f=3&t=2697

Page 1 of 1

Unable to compile in parallel using mvapich

Posted: Thu Jun 12, 2014 11:01 am
by IvanHito
Greetings to all!

I'm quite new on this forum and not well enough in abinit code. So i ask for an excuse if there would be something wrong in my question.

My current problem is to compile abinit on a super computer with mvapich release of mpi.
There are configuration options i use:

Code: Select all

./configure --enable-debug=paranoid --enable-mpi="yes" --disable-mpi-inplace --disable-mpi-io --with-mpi-prefix="/common/mvapich-1.2rc1-gcc" --enable-64bit-flags="yes"


The configuration always goes well (as far as i can conclude :roll: ). The config log is in the attachment. The real problem arises on the make step. There is the error:

Code: Select all

/common/mvapich-1.2rc1-gcc/bin/mpif90 -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -I/gpfs/NETHOME/ipsm1/Lobzenko/abinit-7.6.4/fallbacks/exports/include   -free -module /gpfs/NETHOME/ipsm1/Lobzenko/abinit-7.6.4/src/mods  -O0 -g -debug all -check uninit -ftrapuv -traceback -warn all -fp-stack-check -extend-source -vec-report0 -noaltparam -nofpscomp   -c -o m_GreenHyb.o m_GreenHyb.F90
...
...
------------------------------^
m_GreenHyb.F90(866): error #6404: This name does not have a type and must have an excplicit type. [MPI_IN_PLACE]
      CALL MPI_ALLGATHERV(MPI_IN_PLACE, O, MPI_DATA_TYPE_NULL,&
------------------------------^
compilation aborted from m_GreenHyb.F90 (code 1)
make[3]: *** [m_GreenHyb.o] Error 1
make[2]: *** [all-recursive] Error 1
make[1]: *** [all-recursive] Error 1
make: *** [all] Error 2


Appreciate any hints on solving this. :)

Re: Unable to compile in parallel using mvapich

Posted: Thu Jun 12, 2014 12:15 pm
by pouillon
This is a bug in Abinit. The case where MPI_IN_PLACE is not supported has not been taken into account in the faulty module. I'll notify the developers of this section and we'll keep you in touch.

Thank you for your report.

Re: Unable to compile in parallel using mvapich  [SOLVED]

Posted: Thu Jun 12, 2014 6:05 pm
by Jordan
Sorry for the bug,

Here is a patch to apply that should work. Let me know if it does not.

Download and copy the attached file into your abinit directory

Code: Select all

$ pwd
/foo/bar/abinit-7.6.4/
$ patch -p0 < 62_ctqmc.patch.log

Then try to compile again.

Cheers,

Jordan

Re: Unable to compile in parallel using mvapich

Posted: Mon Jun 16, 2014 1:54 pm
by IvanHito
I didn't have access to the computer for last three days. Now I'm back to work again.

I did the patch you kindly prepared for me and restarted the make (with distclean and the same as above ./configure befor it). But it failed again... :( though, in the other part of the compilation. Now that's what it complains about:

Code: Select all

make[3]: Entering directory `/gpfs/NETHOME/ipsm1/Lobzenko/abinit-7.6.4/src/98_main'
/common/mvapich-1.2rc1-gcc/bin/mpif90 -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -I/gpfs/NETHOME/ipsm1/Lobzenko/abinit-7.6.4/fallbacks/exports/include     -free -module /gpfs/NETHOME/ipsm1/Lobzenko/abinit-7.6.4/src/mods  -O0 -g -debug all -check uninit -ftrapuv -traceback -warn all -fp-stack-check -extend-source -vec-report0 -noaltparam -nofpscomp    -g -debug all -check uninit -ftrapuv -traceback -warn all -fp-stack-check -extend-source -vec-report0 -noaltparam -nofpscomp   -c -o abinit-abinit.o `test -f 'abinit.F90' || echo './'`abinit.F90
make[3]: Warning: File `../../src/95_drive/lib95_drive.a' has modification time 4.1 s in the future
/common/mvapich-1.2rc1-gcc/bin/mpif90 -free -module /gpfs/NETHOME/ipsm1/Lobzenko/abinit-7.6.4/src/mods  -O0 -g -debug all -check uninit -ftrapuv -traceback -warn all -fp-stack-check -extend-source -vec-report0 -noaltparam -nofpscomp    -g -debug all -check uninit -ftrapuv -traceback -warn all -fp-stack-check -extend-source -vec-report0 -noaltparam -nofpscomp   -static-intel -static-libgcc  -static-intel -static-libgcc  -o abinit abinit-abinit.o -static-intel -static-libgcc  ../../src/95_drive/lib95_drive.a ../../src/79_seqpar_mpi/lib79_seqpar_mpi.a ../../src/77_ddb/lib77_ddb.a ../../src/77_suscep/lib77_suscep.a ../../src/72_response/lib72_response.a ../../src/71_bse/lib71_bse.a ../../src/70_gw/lib70_gw.a ../../src/69_wfdesc/lib69_wfdesc.a ../../src/68_dmft/lib68_dmft.a  ../../src/68_recursion/lib68_recursion.a ../../src/68_rsprc/lib68_rsprc.a ../../src/67_common/lib67_common.a ../../src/66_paw/lib66_paw.a ../../src/66_wfs/lib66_wfs.a ../../src/65_psp/lib65_psp.a ../../src/65_nonlocal/lib65_nonlocal.a  ../../src/64_atompaw/lib64_atompaw.a ../../src/62_occeig/lib62_occeig.a ../../src/62_iowfdenpot/lib62_iowfdenpot.a ../../src/62_wvl_wfs/lib62_wvl_wfs.a ../../src/62_poisson/lib62_poisson.a ../../src/62_cg_noabirule/lib62_cg_noabirule.a ../../src/62_ctqmc/lib62_ctqmc.a ../../src/61_ionetcdf/lib61_ionetcdf.a ../../src/57_iovars/lib57_iovars.a ../../src/57_iopsp_parser/lib57_iopsp_parser.a ../../src/56_recipspace/lib56_recipspace.a ../../src/56_xc/lib56_xc.a ../../src/56_mixing/lib56_mixing.a ../../src/56_io_mpi/lib56_io_mpi.a ../../src/53_abiutil/lib53_abiutil.a ../../src/53_spacepar/lib53_spacepar.a ../../src/53_ffts/lib53_ffts.a  ../../src/52_fft_mpi_noabirule/lib52_fft_mpi_noabirule.a ../../src/51_manage_mpi/lib51_manage_mpi.a ../../src/49_gw_toolbox_oop/lib49_gw_toolbox_oop.a ../../src/47_xml/lib47_xml.a ../../src/45_geomoptim/lib45_geomoptim.a ../../src/44_abitypes_defs/lib44_abitypes_defs.a ../../src/43_wvl_wrappers/lib43_wvl_wrappers.a ../../src/43_ptgroups/lib43_ptgroups.a ../../src/42_parser/lib42_parser.a ../../src/42_nlstrain/lib42_nlstrain.a ../../src/42_libpaw/lib42_libpaw.a ../../src/41_xc_lowlevel/lib41_xc_lowlevel.a ../../src/41_geometry/lib41_geometry.a ../../src/32_util/lib32_util.a ../../src/28_numeric_noabirule/lib28_numeric_noabirule.a ../../src/27_toolbox_oop/lib27_toolbox_oop.a ../../src/21_psiesta_noabirule/lib21_psiesta_noabirule.a ../../src/18_timing/lib18_timing.a ../../src/16_hideleave/lib16_hideleave.a  ../../src/14_hidewrite/lib14_hidewrite.a ../../src/12_hide_mpi/lib12_hide_mpi.a ../../src/11_qespresso_ext/lib11_qespresso_ext.a ../../src/11_memory_mpi/lib11_memory_mpi.a ../../src/10_defs/lib10_defs.a  ../../src/01_linalg_ext/lib01_linalg_ext.a ../../src/01_interfaces_ext/lib01_interfaces_ext.a  -L/gpfs/NETHOME/ipsm1/Lobzenko/abinit-7.6.4/fallbacks/exports/lib -llapack -lblas -lrt -L/usr/lib64 -L/common/mvapich-1.2rc1-gcc/lib/shared -L/common/mvapich-1.2rc1-gcc/lib -L/opt/intel/composerxe-2011.3.174/compiler/lib/intel64 -L/opt/intel/composerxe-2011.3.174/mkl/lib/intel64 -L/usr/lib/gcc/x86_64-redhat-linux/4.1.2 -L/usr/lib/gcc/x86_64-redhat-linux/4.1.2/../../../../lib64 -L/usr/lib/gcc/x86_64-redhat-linux/4.1.2/../../.. -L/lib64 -L/lib -L/usr/lib -lmpichf90nc -lmpichfarg -lmpich -libverbs -libumad -lpthread -lrt -lifport -lifcore -limf -lsvml -lm -lipgo -lirc -lirc_s -ldl -L/usr/lib64 -L/common/mvapich-1.2rc1-gcc/lib/shared -L/common/mvapich-1.2rc1-gcc/lib -L/opt/intel/composerxe-2011.3.174/compiler/lib/intel64 -L/opt/intel/composerxe-2011.3.174/mkl/lib/intel64 -L/usr/lib/gcc/x86_64-redhat-linux/4.1.2 -L/usr/lib/gcc/x86_64-redhat-linux/4.1.2/../../../../lib64 -L/usr/lib/gcc/x86_64-redhat-linux/4.1.2/../../.. -L/lib64 -L/lib -L/usr/lib -lmpichf90nc -lmpichfarg -lmpich -libverbs -libumad -lpthread -lrt -lifport -lifcore -limf -lsvml -lm -lipgo -lirc -lirc_s -ldl
../../src/12_hide_mpi/lib12_hide_mpi.a(m_xmpi.o): In function `xmpi_comm_set_errhandler':
/gpfs/NETHOME/ipsm1/Lobzenko/abinit-7.6.4/src/12_hide_mpi/m_xmpi.F90:1779: undefined reference to `mpi_comm_get_errhandler_'
/gpfs/NETHOME/ipsm1/Lobzenko/abinit-7.6.4/src/12_hide_mpi/m_xmpi.F90:1780: undefined reference to `mpi_comm_set_errhandler_'
make[3]: *** [abinit] Error 1
make[3]: Leaving directory `/gpfs/NETHOME/ipsm1/Lobzenko/abinit-7.6.4/src/98_main'
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory `/gpfs/NETHOME/ipsm1/Lobzenko/abinit-7.6.4/src'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/gpfs/NETHOME/ipsm1/Lobzenko/abinit-7.6.4'
make: *** [all] Error 2

Re: Unable to compile in parallel using mvapich

Posted: Mon Jun 16, 2014 8:08 pm
by gmatteo
Enable the calls to MPI_Errhandler_get and comment the
calls to MPI_comm_get_Errhandler in m_xmpi.F90
in subroutine xmpi_comm_set_errhandler


=== modified file 'src/12_hide_mpi/m_xmpi.F90'
--- src/12_hide_mpi/m_xmpi.F90 2014-05-12 09:22:46 +0000
+++ src/12_hide_mpi/m_xmpi.F90 2014-06-16 18:06:01 +0000
@@ -2179,14 +2179,14 @@

mpierr1=MPI_SUCCESS; mpierr2=MPI_SUCCESS

-#if defined HAVE_MPI1
+!#if defined HAVE_MPI1
call MPI_Errhandler_get(my_comm,old_err_handler,mpierr1)
call MPI_Errhandler_set(my_comm,new_err_handler,mpierr2)
-#endif
-#if defined HAVE_MPI2
- call MPI_comm_get_Errhandler(my_comm,old_err_handler,mpierr1)
- call MPI_comm_set_Errhandler(my_comm,new_err_handler,mpierr2)
-#endif
+!#endif
+!#if defined HAVE_MPI2
+! call MPI_comm_get_Errhandler(my_comm,old_err_handler,mpierr1)
+! call MPI_comm_set_Errhandler(my_comm,new_err_handler,mpierr2)
+!#endif

ierror=MPI_SUCCESS
if (mpierr1/=MPI_SUCCESS) then

Re: Unable to compile in parallel using mvapich

Posted: Tue Jun 17, 2014 1:49 pm
by IvanHito
Didn't get it from the first look, but after some attempts I understood, what you mean.

The compilation have been successfully finished :D

I used the following version of m_xmpi.F90:

Code: Select all

 mpierr1=MPI_SUCCESS; mpierr2=MPI_SUCCESS

#if defined HAVE_MPI1
   call MPI_Errhandler_get(my_comm,old_err_handler,mpierr1)
   call MPI_Errhandler_set(my_comm,new_err_handler,mpierr2)
#endif
#if defined HAVE_MPI2
!! call MPI_comm_get_Errhandler(my_comm,old_err_handler,mpierr1)
!! call MPI_comm_set_Errhandler(my_comm,new_err_handler,mpierr2)
   call MPI_Errhandler_get(my_comm,old_err_handler,mpierr1)
   call MPI_Errhandler_set(my_comm,new_err_handler,mpierr2)
#endif

 ierror=MPI_SUCCESS


So, basicaly i just commented two calls for MPI2 and copy/paste those for MPI1. If it wouldn't corrupt my calculations in future, I say that the problem is SOLVED. Is it?

Re: Unable to compile in parallel using mvapich

Posted: Tue Jun 17, 2014 10:32 pm
by gmatteo
[quote]
So, basicaly i just commented two calls for MPI2 and copy/paste those for MPI1. If it wouldn't corrupt my calculations in future, I say that the problem is SOLVED. Is it?
[/quote]

Yes, the compilation error due to the missing MPI functions is solved. Note however, that your MPI library is outdated.
Try to run the tests in ~abinit/tests with the runtests.py script to validate the build.
If you encounter problem with Abinit in parallel, try a more recent version of mvapich
All times are UTC+02:00
Page 1 of 1

Powered by phpBB® Forum Software © phpBB Limited