[Wien] Error in compiling mpi-parallel version

[Gavin Abo]

First, the latest WIEN2k version (14.2) removed some bugs in 14.1 [ 
http://www.wien2k.at/reg_user/updates/ ]; so it is quite recommended to 
not use 14.1.

Second, the error you are getting might be because the fftw3-mpi.f03 in 
SRC_lapw0 is probably still for version 3.3.2 of fftw3.  There are two 
different options you can try as described at:

http://www.mail-archive.com/wien%40zeus.theochem.tuwien.ac.at/msg06147.html

I think the same error was reported back in December:

http://www.mail-archive.com/wien%40zeus.theochem.tuwien.ac.at/msg11661.html

I don't think it was reported if they solved the problem or not. If they 
did, it might also be because they recompiled fftw and mpi (as mentioned 
in the December post) with the same compilers that they used for 
WIEN2k.  Mixing packages compiled with different compilers like gfortan 
and ifort, gcc and icc, might also lead to a library function 
interfacing error like you seem to have.

On 1/8/2015 11:02 PM, lung Fermin wrote:
> Dear all,
>
> Recently, I am trying to do a calculation with a supercell of about 
> 100 atoms. Previously I have tried to do it with k-point 
> parallelizatoin but it failed due to insufficient virtual memory. So 
> instead I am moving to the mpi parallelization.
>
> I tried to compile the lapw0 program first at a test. The following is 
> the settings in the Makefile:
> ----------------------------------------------------------------------------------------------------------------
> .SUFFIXES:        .F
> .SUFFIXES:        .F90
> SHELL = /bin/sh
> FC = ifort
> MPF = mpif90
> CC = cc
> FOPT =  -FR -mp1 -w -prec_div -pc80 -pad -ip -DINTEL_VML -traceback -r8
> FPOPT = -ffree-form -O2 -m64 -ffree-line-length-none -DFFTW3 
> -I/usr/local/fftw-3.3.3/mpi -I/usr/local/include
> DParallel = '-DParallel'
> FGEN = $(PARALLEL)
> LDFLAGS = $(FOPT) -L/opt/intel/Compiler/11.1/046/mkl/lib/em64t -pthread
> R_LIBS     = -L/usr/local/lib -lmkl_lapack -lmkl_intel_lp64 
> -lmkl_intel_thread -lmkl_core -openmp -lpthread -lguide
> RP_LIBS = -lfftw3_mpi -lmkl_scalapack_lp64 -lmkl_solver_lp64 
> -lmkl_blacs_lp64 $(R_LIBS)
> -------------------------------------------------------------------------------------------------------------------
> The following error is reported:
>
> touch .parallel
> make PARALLEL='-DParallel' ./lapw0_mpi \
>           FORT=mpif90 FFLAGS=' -ffree-form -O2 -m64 
> -ffree-line-length-none -DFFTW3 -I/usr/local/fftw-3.3.3/mpi 
> -I/usr/local/include '-DParallel''
> make[1]: Entering directory `/home/stretch/WIEN2k/SRC_lapw0'
> cc -c cputim.c
> mpif90 -ffree-form -O2 -m64 -ffree-line-length-none -DFFTW3 
> -I/usr/local/fftw-3.3.3/mpi -I/usr/local/include -DParallel -c modules.F
> mpif90 -ffree-form -O2 -m64 -ffree-line-length-none -DFFTW3 
> -I/usr/local/fftw-3.3.3/mpi -I/usr/local/include -DParallel -c 
> fft_modules.F
> fftw3-mpi.f03:33.14:
>     Included at fft_modules.F:69:
>
>       integer(), value :: comm
>               1
> Error: Expected initialization expression at (1)
> fftw3-mpi.f03:47.14:
>     Included at fft_modules.F:69:
>
>       integer(), value :: comm
>               1
> Error: Expected initialization expression at (1)
> fftw3-mpi.f03:57.14:
>     Included at fft_modules.F:69:
>
>       integer(), value :: comm
>               1
> -----------------------------------------------------------------------------------------------------------------
> What is the cause of those errors? Any suggestions in solving the 
> issue would be appreciated.
>
> ----------------------------------------------------------------------------------------------
> P.S. Some details about the system:
> * 45-nodes cluster formed by DELL R720/R620 servers (16 cores per node)
> * OS  : Rocks 6.1 (CentOS)
> * MPI : mpif90 for MVAPICH2 2.0a
> * FFTW ver 3.3.3
> * MKL ver 11.1.0.080
> * Scalapack ver 2.0.2
> * WIEN2k ver 14.1
> and the setting in the parallel_option file is:
> setenv TASKSET "no"
> setenv USE_REMOTE 1
> setenv MPI_REMOTE 1
> setenv WIEN_GRANULARITY 1
>
> -------------------------------------------------------
> Thanks and Regards,
> Fermin