[SOLVED] Compiling 6.12.3 on Blue Gene/P

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
pwhitfor
Posts: 2
Joined: Tue May 29, 2012 8:54 pm

[SOLVED] Compiling 6.12.3 on Blue Gene/P

Post by pwhitfor » Sun Jun 10, 2012 10:10 pm

I am trying to compile version 6.12.3 on a Blue Gene/P machine and am running into an issue with m_xmpi.F90

I have updated the xlf.cfg file so that it does not give errors with comments in the header files. I made it past the configure step, and have attached the bgp-fn.ac file that contains all values used (only uncommented lines shown).

enable_optim="yes"
CPP="/usr/bin/cpp"
CC="/bgsys/drivers/ppcfloor/comm/fast/bin/mpixlc_r"
CFLAGS="-qlanglvl=extc99 -O3 -qspill=64000 -qarch=450d -qtune=450 -qcache=auto -qstrict "
CXX="/bgsys/drivers/ppcfloor/comm/fast/bin/mpixlcxx_r"
FC="/bgsys/drivers/ppcfloor/comm/fast/bin/mpixlf90_r -F$HOME/xlf.cfg"
FCFLAGS="-O3 -qstrict -qarch=450d -qtune=450 -qmaxmem=-1 -qcache=auto "
enable_fc_wrapper="no"
AR="ar"
enable_mpi="yes"
enable_mpi_io="yes"
MPI_RUNNER="/bgsys/drivers/ppcfloor/bin/mpirun"
with_trio_flavor="none"
with_netcdf_incs="-I/usr/local/netcdf/include"
with_netcdf_libs="-L/usr/local/netcdf/lib -lnetcdf"
with_fft_flavor="fftw3"
with_fft_incs="-I$HOME/BIN/fftw-3.3.2_d/include/"
with_fft_libs="-L$HOME/BIN/fftw-3.3.2_d/lib -lfftw3"
with_linalg_flavor="essl"
with_linalg_incs="-I/opt/ibmmath/essl/4.4/include"
with_linalg_libs="-L/opt/apps/lapack/3.4.1/lib/ -L/opt/ibmmath/essl/4.4/lib -llapack -lblas -lesslsmpbg"
with_math_incs="-I/opt/ibmcmp/xlmass/bg/4.4/include"
with_math_libs="-L/opt/ibmcmp/xlmass/bg/4.4/bglib -lmass"
with_dft_flavor="none"
enable_clib="yes"
enable_gw_dpc="yes"
enable_maintainer_checks="no"
enable_test_timeout="yes"
with_fc_vendor="ibm"
with_fc_version="12.1"
enable_gui_build="no"


When I issue make, I get the following error. It appears that earlier version has worked on BGP, so perhaps something changed in 6.12.3. I haven't been able to find any discussion about this error in the forums


make[3]: Leaving directory `/home/p/abinit-6.12.3/src/11_qespresso_ext'
Making all in 12_hide_mpi
make[3]: Entering directory `/home/p/abinit-6.12.3/src/12_hide_mpi'
/bgsys/drivers/ppcfloor/comm/fast/bin/mpixlf2003_r -F/home/p/xlf.cfg -DHAVE_CONFIG_H -I. -I../.. -I../../src/incs -I../../src/incs -I/home/p/abinit-6.12.3/fallbacks/exports/include -WF,-DHAVE_CONFIG_H -qsuffix=cpp=F90:f=f90 -qfree=f90 -qmoddir=/home/p/abinit-6.12.3/src/mods -I/home/p/abinit-6.12.3/src/mods -O3 -qstrict -qarch=450d -qtune=450 -qmaxmem=-1 -qcache=auto -c -o m_xmpi.o m_xmpi.F90
"/tmp/F831084AhVh8X", line 11408.29: 1513-061 (S) Actual argument attributes do not match those specified by an accessible explicit interface.
"/tmp/F831084AhVh8X", line 11487.29: 1513-061 (S) Actual argument attributes do not match those specified by an accessible explicit interface.
"/tmp/F831084AhVh8X", line 11489.29: 1513-061 (S) Actual argument attributes do not match those specified by an accessible explicit interface.
"/tmp/F831084AhVh8X", line 11569.29: 1513-061 (S) Actual argument attributes do not match those specified by an accessible explicit interface.
"/tmp/F831084AhVh8X", line 11571.29: 1513-061 (S) Actual argument attributes do not match those specified by an accessible explicit interface.
"/tmp/F831084AhVh8X", line 11573.29: 1513-061 (S) Actual argument attributes do not match those specified by an accessible explicit interface.
"/tmp/F831084AhVh8X", 1520-031 (W) Option DLINES is ignored within Fortran 90 free form and IBM free form.
** m_xmpi === End of Compilation 1 ===
1501-511 Compilation failed for file m_xmpi.F90.
make[3]: *** [m_xmpi.o] Error 1
make[3]: Leaving directory `/home/p/abinit-6.12.3/src/12_hide_mpi'
make[2]: *** [all-recursive] Error 1
make[2]: Leaving directory `/home/p/abinit-6.12.3/src'
make[1]: *** [all-recursive] Error 1
make[1]: Leaving directory `/home/p/abinit-6.12.3'
make: *** [all] Error 2


I changed options in .ac file, so I could see the preprocessed files. The error messages persist

"m_xmpi_cpp.f90", line 10572.29: 1513-061 (S) Actual argument attributes do not match those specified by an accessible explicit interface.
"m_xmpi_cpp.f90", line 10612.29: 1513-061 (S) Actual argument attributes do not match those specified by an accessible explicit interface.
"m_xmpi_cpp.f90", line 10614.29: 1513-061 (S) Actual argument attributes do not match those specified by an accessible explicit interface.
"m_xmpi_cpp.f90", line 10657.29: 1513-061 (S) Actual argument attributes do not match those specified by an accessible explicit interface.
"m_xmpi_cpp.f90", line 10659.29: 1513-061 (S) Actual argument attributes do not match those specified by an accessible explicit interface.
"m_xmpi_cpp.f90", line 10661.29: 1513-061 (S) Actual argument attributes do not match those specified by an accessible explicit interface.


Here are the corresponding lines in m_xmpi_cpp.f90 file
call MPI_Type_hvector(ny,1,stride_x,column_type,new_type,mpi_err)
call MPI_Type_hvector(ny,1,stride_x,column_type,plane_type,mpi_err)
call MPI_Type_hvector(nz,1,ldy*stride_x,plane_type,new_type,mpi_err)
call MPI_Type_hvector(ny,1,stride_x,column_type,plane_type,mpi_err)
call MPI_Type_hvector(nz,1,ldy*stride_x,plane_type,vol_type,mpi_err)
call MPI_Type_hvector(na,1,ldz*ldy*stride_x,vol_type,new_type,mpi_err)

So, it appears to be the second argument that is causing trouble.

I'm guessing I am missing a trivial flag somewhere... Any help would be greatly appreciated. thanks

Paul Whitford
Rice University

kaneod
Posts: 42
Joined: Wed Mar 10, 2010 11:47 am

Re: Compiling 6.12.3 on Blue Gene/P

Post by kaneod » Tue Jun 12, 2012 7:39 am

Hi,

This could be similar to the MPICH2 bug with the hvector type definition (although you're not using MPICH). Try setting:

with_mpi_level="1"

and reconfigure. Unlike the MPICH case you probably can't just patch your MPI libraries to fix the fundamental issue here so the with_mpi_level change might be a workaround. Search the forums for the hvector discussion to see how the problem is fixed in MPICH2 with a patch.

Kane
Dr Kane O'Donnell
Postdoctoral Research Fellow
Australian Synchrotron

pwhitfor
Posts: 2
Joined: Tue May 29, 2012 8:54 pm

Re: Compiling 6.12.3 on Blue Gene/P

Post by pwhitfor » Tue Jun 26, 2012 9:02 pm

that worked. Thanks

Locked