The ABINIT Users Mailing List ( CLOSED )

[Eric Roman <ERoman@lbl.gov>]

I've just built abinip 4.3.3 on an IBM SP.  It builds fine, and I have the
serial version (abinis) works.

I'm having trouble running parallel jobs (more than 1 processor with abinip).
The job runs, but I get the following error:

poe ../../abinip.mpi -nodes 1 -tasks_per_node 2 < abfiles.run > tA2.log
ERROR: 0032-117 User pack or receive buffer is too small  (0) in MPI_Recv, 
task 0
ERROR: 0031-250  task 0: Terminated
ERROR: 0031-250  task 1: Terminated

I'm using the Test_paral input files.  The output below is shown for test A.

I can make the problem go away if I set prtwf to 0 in the output file.
(Code executes, but obviously produces no WFK file.)

I'd appreciate any advice to correct this.

Thank you

Eric


The last few lines of tA2.log are:
  sigma(3 3)=  4.73018315E-03  sigma(2 1)=  0.00000000E+00

 scprqt:  WARNING -
  nstep=    2 was not enough SCF cycles to converge;
  maximum residual=  1.557E-06 exceeds tolwfr=  1.000E-20

-P-0000  leave_test : synchronization done...
================================================================================

 ----iterations are completed or convergence reached----

 outwf  : write wavefunction to file tA2.o_WFK
-P-0000  leave_test : synchronization done...

My makefile_macros file looks like this:
# For the sequential version 

# Machine type
MACHINE=ibm
# Fortran optimized compilation
FC=mpxlf90_r
# original: FFLAGS=-q64 -O3 -qnolm -qarch=pwr3 -qtune=pwr3 -qstrict
FFLAGS=-q64 -O3 -qarch=pwr3 -qtune=pwr3 -qstrict
MAIN_ROUTINE_FFLAGS=-qcharlen=16000 -q64
FFLAGS_LIBS=-O3 -q64 -qnolm -qarch=pwr3 -qtune=pwr3 -qstrict -qfixed
FLINK=-bloadmap:loadmap -q64 -bmaxdata:2048000000
AR_COMMAND= ar -X64

# C preprocessor, used to preprocess the fortran source.
# CPP=/usr/ccs/lib/cpp
CPP=cpp
CPP_FLAGS=-P

# C optimized compilation.
CC=mpcc_r
CFLAGS=-O -q64 -qstrict

# Location of perl . Used to generate the script fldiff, in ~ABINIT/Utilities 
.
PERL=/bin/perl
# Definition of libraries
LIBS= $(LAPACK) $(BLAS) $(FFTNEW) $(NUMRECIP)

####################################################################
# For the parallel version 
_BASE=/usr/lpp/ppe.poe

# Location of directory with header files *.h for parallel version
FFLAGS_PAR=$(FFLAGS) -I$(_BASE)/include/thread64
# Location of the MPI library
LIBS_PAR=$(LIBS) -L$(_BASE)/lib  -lmpi


-- 
Eric Roman                       Computational Research Division
510-486-6420                     Berkeley Lab