[OMPI users] openmpi-1.7.3rc2r29276 doesn't honour --with-wrapper-libs
Hi, I installed openmpi-1.7.3rc2r29276 on my platforms (Solaris Sparc, Solaris x86_64, and Linux x86_64) with Sun C 5.12 and gcc-4.8.0 in 32- and 64-bit versions. On Solaris Sparc I configured with LIBS="-lgcc_s" \ --with-wrapper-cflags="-std=c11 -m64" \ --with-wrapper-libs="-lgcc_s" \ "-lgcc_s" is neccessary to build Open MPI and later compile MPI programs. tyr openmpi-1.7.3rc2r29276-SunOS.sparc.64_gcc 19 ../openmpi-1.7.3rc2r29276/configure \ --help | grep wrapper-libs --with-wrapper-libs Extra flags to add to LIBS when using wrapper tyr openmpi-1.7.3rc2r29276-SunOS.sparc.64_gcc 20 more config.log This file contains any messages produced by compilers while running configure, to aid debugging if configure makes a mistake. It was created by Open MPI configure 1.7.3rc2, which was generated by GNU Autoconf 2.69. Invocation command line was $ ../openmpi-1.7.3rc2r29276/configure --prefix=/usr/local/openmpi-1.7_64_gcc \ --libdir=/usr/local/openmpi-1.7_64_gcc/lib64 \ --with-jdk-bindir=/usr/local/jdk1.7.0_07/bin/sparcv9 \ --with-jdk-headers=/usr/local/jdk1.7.0_07/include \ JAVA_HOME=/usr/local/jdk1.7.0_07 LDFLAGS=-m64 \ -L/usr/local/gcc-4.8.0/lib/sparcv9 LIBS=-lgcc_s \ CC=gcc CXX=g++ FC=gfortran CFLAGS=-m64 CXXFLAGS=-m64 FCFLAGS=-m64 CPP=cpp \ CXXCPP=cpp CPPFLAGS= CXXCPPFLAGS= --enable-cxx-exceptions --enable-mpi-java \ --enable-heterogeneous --enable-opal-multi-threads --enable-mpi-thread-multiple \ --with-threads=posix --with-hwloc=internal --without-verbs --without-udapl \ --with-wrapper-cflags=-std=c11 -m64 --with-wrapper-libs=-lgcc_s --enable-debug ... tyr hello_1 114 which mpicc /usr/local/openmpi-1.7_64_gcc/bin/mpicc tyr hello_1 115 ls -l /usr/local/openmpi-1.7_64_gcc/bin/mpicc lrwxrwxrwx 1 root root 12 Sep 28 10:30 /usr/local/openmpi-1.7_64_gcc/bin/mpicc -> opal_wrapper tyr hello_1 116 mpicc -showme gcc -I/usr/local/openmpi-1.7_64_gcc/include -fexceptions -pthread -std=c11 -m64 \ -L/usr/local/openmpi-1.7_64_gcc/lib64 -lmpi Unfortunately the wrapper compiler doesn't contain "-lgcc_s" so that I cannot compile programs. tyr hello_1 117 mpicc hello_1_mpi.c /usr/local/openmpi-1.7_64_gcc/lib64/libmpi.so: undefined reference to `__muldc3@GCC_4.0.0' /usr/local/openmpi-1.7_64_gcc/lib64/libmpi.so: undefined reference to `__mulsc3@GCC_4.0.0' /usr/local/openmpi-1.7_64_gcc/lib64/libmpi.so: undefined reference to `__multc3@GCC_4.0.0' collect2: error: ld returned 1 exit status I would be grateful if somebody can fix the problem. Perhaps you can even add "-lgcc_s" as a default library to the build process and wrapper compiler. Thank you very much for your help in advance. Kind regards Siegmar
[OMPI users] mpi_barrier
Dear All, In my code I implement mpi_send/mpi_receive for an three dimensional real array, and process is as follows: all other processors send the array to rank 0 and then rank 0 receives the array and put these arrays into a complete array. Then mpi_bcast is called to send the complete array from rank 0 to all others. This is very basic usage of mpi_send and mpi_receive. In my fortran code I found that if I added call mpi_barrier(...) before the mpi_send and mpi_reive statements the wall time (60s) for this sending and receiving will be much lower than that if mpi_barrier is not called (2s). I used mpi_wtime to count the time. I think mpi_send and mpi_recv are blocking subroutines and thus no additional mpi_barrier is needed. Can anybody tell me what is the reason for this phenomena? Thank you very much. best regards, Huangwei
Re: [OMPI users] mpi_barrier
On Sep 29, 2013, at 01:19 , Huangwei wrote: > Dear All, > > In my code I implement mpi_send/mpi_receive for an three dimensional real > array, and process is as follows: > > all other processors send the array to rank 0 and then rank 0 receives the > array and put these arrays into a complete array. Then mpi_bcast is called to > send the complete array from rank 0 to all others. This pattern of communication reminds me of an MPI_Allgather (or the more flexible version MPI_Allgatherv). > This is very basic usage of mpi_send and mpi_receive. In my fortran code I > found that if I added call mpi_barrier(...) before the mpi_send and > mpi_receive statements the wall time (60s) for this sending and receiving > will be much lower than that if mpi_barrier is not called (2s). I used > mpi_wtime to count the time. In a parallel application each process is out of sync to the others. I have no idea how you measure your time in the original version but I guess that in the MPI_Barrier case you start your timer after the barrier. As the barrier put in sync all processes, you only measure the real time to exchange the data, which might seem shorter. > I think mpi_send and mpi_recv are blocking subroutines and thus no additional > mpi_barrier is needed. Can anybody tell me what is the reason for this > phenomena? Thank you very much. Yes, these operations are indeed blocking, which is why you see the slowdown. If one single process is late to send its contribution, the entire operation is be penalized (as the root , aka. process zero, is waiting for contributions in order). So you should either try to use the collective pattern I highlighted before, switch to using non-blocking point-to-point instead of blocking, or look into the potential benefit of using a non-blocking collective. George. > > best regards, > Huangwei > > > > > ___ > users mailing list > us...@open-mpi.org > http://www.open-mpi.org/mailman/listinfo.cgi/users
