[Wien] segmentation fault in LAPW0 (MPI)

[Oleg Rubel]

Dear Wien2k community,

I compiled Wien2k v14.2 with Intel Composer XE 2013, mvapich2 v1.9 and FFTW v3.3.0
Both serial and MPI versions compiled error free, however the MPI code crushed in LAPW0 with segfault when executed on 8 cores.

It crashed when calling an FFTW subroutine:
call fftw_mpi_execute_dft(forward_plan, local_C, local_C)
in SRC_lapw0/fft_modules.F

DDT debugger provided a bit more information:

Processes 0-7:
Process stopped in __intel_ssse3_rep_memcpy with signal SIGSEGV (Segmentation fault).
Reason/Origin: address not mapped to object (attempt to access invalid address)
Your program will probably be terminated if you continue.
You can use the stack controls to see what the process was doing at the time.

Interestingly, LAPW0 did not crush in the case of 2-core MPI job; 8-, 12 and 16-core jobs failed.

I have no issues on another machine with much older Intel compilers and MKL, the same mvapich2 and FFTW 3.3.3


Any suggestions will be appreciated.

Thank you in advance
Oleg



P.S. Compilation options

current:FOPT:-FR -mp1 -w -prec_div -pc80 -pad -ip -DINTEL_VML -traceback -assume buffered_io -I/software/compilers/Intel/2013-sp1-14.0/composer_xe_2013_sp1.4.211/mkl/include
current:FPOPT:-FR -mp1 -w -prec_div -pc80 -pad -ip -DINTEL_VML -traceback -assume buffered_io -I/software/compilers/Intel/2013-sp1-14.0/composer_xe_2013_sp1.4.211/mkl/include
current:FFTW_OPT:-DFFTW3 -I/software/CentOS-6/libraries/FFTW-3.3/mvapich2-intel/include
current:FFTW_LIBS:-lfftw3_mpi -lfftw3 -L/software/CentOS-6/libraries/FFTW-3.3/mvapich2-intel/lib
current:LDFLAGS:$(FOPT) -L$(MKLROOT)/lib/intel64 -pthread -i-static
current:DPARALLEL:'-DParallel'
current:R_LIBS:-lmkl_intel_lp64 -lmkl_core -lmkl_sequential -lpthread -lm
current:RP_LIBS:-lfftw3_mpi -lmkl_scalapack_lp64 -lmkl_blacs_intelmpi_lp64 $(R_LIBS)
current:MPIRUN:mpirun -np _NP_ -machinefile _HOSTS_ _EXEC_
current:MKL_TARGET_ARCH:intel64