Re: [OMPI users] comm_join and singleton init

On Tue, Mar 14, 2006 at 12:37:52PM -0600, Edgar Gabriel wrote: > I think I know what goes wrong. Since they are in different 'universes', > they will have exactly the same 'Open MPI name', and therefore the > algorithm in intercomm_merge can not determine which process should be > first and whic

Re: [OMPI users] Problem with MPI_Comm_split and intercomms

thanks for reporting the bug, I found the problem and fixed it on the trunk. Your program is correct, and produces now the results which you expected. Thanks Edgar Bert Wesarg wrote: Hello list, i get some erroneous results from calls to MPI_Comm_split with an intercommunicator as communicat

Re: [OMPI users] pnetcdf & Open MPI

On Tue, May 02, 2006 at 10:32:56PM +0200, Dries Kimpe wrote: > It looks as if the problem is not really due to Open MPI, but to the > included ROM-IO: > > All tests fail with the same error message: > > For example, test/test_double/test_write shows: > > Testing write ... Error: Unsupported data

Re: [OMPI users] fortran mpi io malloc error

Dear Brian and Jeff, I have downloaded the latest nightly snapshot (openmpi-1.0.3a1r9798) and the problem is gone. Thanks a lot, Bernard. On 5/3/06, Jeff Squyres (jsquyres) wrote: Thanks for reporting this. I have committed the fixes to the v1.0 and v1.1 branches; they will show up in all o

Re: [OMPI users] Problem with MPI_Comm_split and intercomms

I'm having a look into that ... Thanks Edgar Bert Wesarg wrote: Hello list, i get some erroneous results from calls to MPI_Comm_split with an intercommunicator as communicator. this is a brokendown testcase, the source is attached: test_split.c: mpi_comm_spwan(world, "test_split_client", 1,

Re: [OMPI users] MPI_Intercomm_merge broken

Michael, I think you are right, the logic is exactly switched in our code. I'll commit a fix on the trunk, and I assume that this fix will also be backported to v1.1. and maybe v1.0 Thanks Edgar Michael Kluskens wrote: MPI_Intercomm_merge is broken in OpenMPI 1.1a4r9788 (and likely all versi

[OMPI users] Problem with MPI_Comm_split and intercomms

Hello list, i get some erroneous results from calls to MPI_Comm_split with an intercommunicator as communicator. this is a brokendown testcase, the source is attached: test_split.c: mpi_comm_spwan(world, "test_split_client", 1, &inter) for i=1:np if rank < i mpi_comm_split(inter, 0, 0, &i

[OMPI users] MPI_Intercomm_merge broken

MPI_Intercomm_merge is broken in OpenMPI 1.1a4r9788 (and likely all versions) Details: the second argument, high, of MPI_Intercomm_merge is a logical in Fortran (pg 216 of Using MPI) and an int in C. This now correct with regards to the f90 interfaces in OpenMPI 1.1. The meaning of "hi