Re: [OMPI users] Using Open MPI as a communication library

You still need a rendezvous of some kind for comm_connect to work. If you want to try using the file-based method from early MPI days, you can try it - haven’t seen anyone test it in some time, and so I honestly don’t know if it will work. You’d still require some way of telling the application

Re: [OMPI users] Using Open MPI as a communication library

I guess this is client server. Can we do communications collective communications with this approach? Thanks, Supun.. On Fri, Jul 8, 2016 at 5:37 PM, Jeff Hammond wrote: > Why wouldn't https://www.open-mpi.org/doc/v1.8/man3/MPI_Comm_connect.3.php > and friends work after MPI_Init is called, reg

Re: [OMPI users] Using Open MPI as a communication library

Why wouldn't https://www.open-mpi.org/doc/v1.8/man3/MPI_Comm_connect.3.php and friends work after MPI_Init is called, regardless of how the process is spawned? Jeff On Fri, Jul 8, 2016 at 9:55 AM, Ralph Castain wrote: > You’d need to have some rendezvous mechanism. I suppose one option would >

[OMPI users] Ability to overlap communication and computation on Infiniband

Hi OMPI_Users and OMPI_Developers, I would like someone to verify if my understanding is correct concerning Open MPI ability to overlap communication and computations on Infiniband when using non-blocking MPI_Isend() and MPI_Irecv() functions (i.e. the computation is done between the non-blocki

Re: [OMPI users] Can OMPI 1.8.8 or later support LSF 9.1.3 or 10.1?

Not exactly, just want to confirm before we try them together. Thanks for the answer. On Fri, Jul 8, 2016 at 3:25 PM, Ralph Castain wrote: > There shouldn’t be any issue, nor anything required to make it work. Are > you hitting a problem? > > On Jul 8, 2016, at 12:22 PM, Gang Chen wrote: > > Ar

Re: [OMPI users] Can OMPI 1.8.8 or later support LSF 9.1.3 or 10.1?

There shouldn’t be any issue, nor anything required to make it work. Are you hitting a problem? > On Jul 8, 2016, at 12:22 PM, Gang Chen wrote: > > Are they supposed to work together? Is there any change done in OpenMPI in > order to support LSF 9.1.3 and 10.1? > > On Fri, Jul 8, 2016 at 2:31

Re: [OMPI users] Can OMPI 1.8.8 or later support LSF 9.1.3 or 10.1?

Are they supposed to work together? Is there any change done in OpenMPI in order to support LSF 9.1.3 and 10.1? On Fri, Jul 8, 2016 at 2:31 PM, Ralph Castain wrote: > I’m sure IBM is doing so, but nothing that has been publicly disclosed on > the mailing list. > > > On Jul 8, 2016, at 11:20 AM,

Re: [OMPI users] Can OMPI 1.8.8 or later support LSF 9.1.3 or 10.1?

I’m sure IBM is doing so, but nothing that has been publicly disclosed on the mailing list. > On Jul 8, 2016, at 11:20 AM, Gang Chen wrote: > > Hi, > > I am wondering if there's integration test conducted with v1.8.8 and IBM LSF > 9.1.3 or 10.1, especially the cpu affinity parts. Is there som

[OMPI users] Can OMPI 1.8.8 or later support LSF 9.1.3 or 10.1?

Hi, I am wondering if there's integration test conducted with v1.8.8 and IBM LSF 9.1.3 or 10.1, especially the cpu affinity parts. Is there somewhere I can find detail info? Thanks, Gordon

Re: [OMPI users] Using Open MPI as a communication library

Thanks Ralph. Will have a look in these. Supun.. On Fri, Jul 8, 2016 at 1:05 PM, Ralph Castain wrote: > I don’t know anything about ZooKeeper. You would need another plugin to > support it, probably down in the OPAL pmix framework I suppose. > > There is a list of info you’d have to get from it

Re: [OMPI users] Using Open MPI as a communication library

I don’t know anything about ZooKeeper. You would need another plugin to support it, probably down in the OPAL pmix framework I suppose. There is a list of info you’d have to get from it: https://github.com/pmix/master/wiki/2.8-Pmix-Server-Data-Requirements

Re: [OMPI users] Using Open MPI as a communication library

I would like to discover the processes using a ZooKeeper server. The purpose is to use MPI as a communication library for applications managed by a resource manager such as Mesos or Yarn. Thanks, Supun.. On Fri, Jul 8, 2016 at 12:55 PM, Ralph Castain wrote: > You’d need to have some rendezvous

Re: [OMPI users] Using Open MPI as a communication library

You’d need to have some rendezvous mechanism. I suppose one option would be to launch a set of PMIx servers on the nodes (and ensure they know about each other) to support these things, but that’s all mpirun really does anyway. What did you have in mind? > On Jul 8, 2016, at 9:49 AM, Supun Kam

Re: [OMPI users] Using Open MPI as a communication library

Thanks for the quick response. Is there a way for extending OpenMPI so that it can discover the processes using other means? Supun. On Fri, Jul 8, 2016 at 12:45 PM, Ralph Castain wrote: > If not spawned by mpirun, and not spawned by a resource manager, then it > won’t work. There is no way for

Re: [OMPI users] Using Open MPI as a communication library

If not spawned by mpirun, and not spawned by a resource manager, then it won’t work. There is no way for the procs to wireup. > On Jul 8, 2016, at 9:42 AM, Supun Kamburugamuve > wrote: > > Yes, the processes are not spawned by MPI and they are not spawned by > something like Slurm/PBS. > >

Re: [OMPI users] Using Open MPI as a communication library

Yes, the processes are not spawned by MPI and they are not spawned by something like Slurm/PBS. How does MPI get to know what processes running in what nodes in a general sense? Do we need to write some plugin so that it can figure out this information? I guess this must be the way it is supportin

Re: [OMPI users] Using Open MPI as a communication library

You mean you didn’t launch those procs via mpirun, yes? If you started them via some resource manager, then you might just be able to call MPI_Init and have them wireup. > On Jul 8, 2016, at 8:55 AM, Supun Kamburugamuve > wrote: > > Hi, > > I have a set of processes running and these are no

[OMPI users] Using Open MPI as a communication library

Hi, I have a set of processes running and these are not managed/spawned by Open MPI. Is it possible to use Open MPI as a pure communication library among these processes? Thanks, Supun..

Re: [OMPI users] mpi_reduce implementation

On Jul 8, 2016 4:04 PM, "Juan Francisco Martínez" < juan.francisco.marti...@est.fib.upc.edu> wrote: > > George, > > Thanks for your rapid answer. > > I just ask for "simple synchronized reduction implementation" because I > am using a simple (and therefore really rapid) mpi communications > simulat

Re: [OMPI users] Java-OpenMPI returns with SIGSEGV

Hi Gundram Could you configure without the disable dlopen option and retry? Howard Am Freitag, 8. Juli 2016 schrieb Gilles Gouaillardet : > the JVM sets its own signal handlers, and it is important openmpi dones > not override them. > this is what previously happened with PSM (infinipath) but t

Re: [OMPI users] mpi_reduce implementation

George, Thanks for your rapid answer. I just ask for "simple synchronized reduction implementation" because I am using a simple (and therefore really rapid) mpi communications simulator that models all collectives as synchronized collectives and I appreciate a huge differences between the real a

Re: [OMPI users] mpi_reduce implementation

Juan, if you need synchronization after MPI_Reduce, you can reimplement the function like this int MPI_Reduce(...) { PMPI_Reduce(...); return PMPI_Barrier(); } Cheers, Gilles On Friday, July 8, 2016, Juan Francisco Martínez < juan.francisco.marti...@est.fib.upc.edu> wrote: > Hi everybod

Re: [OMPI users] mpi_reduce implementation

On Jul 8, 2016 3:16 PM, "Juan Francisco Martínez" < juan.francisco.marti...@est.fib.upc.edu> wrote: > > Hi everybody! > > First of all I want to congratulate all of you because the quality of > the community, I have solved a lot of doubts just reading the mailing > list. > > However I have a questi

[OMPI users] mpi_reduce implementation

Hi everybody! First of all I want to congratulate all of you because the quality of the community, I have solved a lot of doubts just reading the mailing list. However I have a question that I can not solve... Until now I though that all the collective operations have an implicit sincronization,

Re: [OMPI users] Java-OpenMPI returns with SIGSEGV

the JVM sets its own signal handlers, and it is important openmpi dones not override them. this is what previously happened with PSM (infinipath) but this has been solved since. you might be linking with a third party library that hijacks signal handlers and cause the crash (which would explain why

Re: [OMPI users] Java-OpenMPI returns with SIGSEGV

You made the best of it... thanks a lot! Whithout MPI it runs. Just adding MPI.init() causes the crash! maybe I installed something wrong... install newest automake, autoconf, m4, libtoolize in right order and same prefix check out ompi, autogen configure with same prefix, pointing to the sam

Re: [OMPI users] Java-OpenMPI returns with SIGSEGV

I am running out of ideas ... what if you do not run within slurm ? what if you do not use '-cp executor.jar' or what if you configure without --disable-dlopen --disable-mca-dso ? if you mpirun -np 1 ... then MPI_Bcast and MPI_Barrier are basically no-op, so it is really weird your program is sti

Re: [OMPI users] Java-OpenMPI returns with SIGSEGV

In any cases the same error. this is my code: salloc -n 3 export IPATH_NO_BACKTRACE ulimit -s 10240 mpirun -np 3 java -cp executor.jar de.uros.citlab.executor.test.TestSendBigFiles2 also for 1 or two cores, the process crashes. On 07/08/2016 12:32 PM, Gilles Gouaillardet wrote: you can try

Re: [OMPI users] Java-OpenMPI returns with SIGSEGV

you can try export IPATH_NO_BACKTRACE before invoking mpirun (that should not be needed though) an other test is to ulimit -s 10240 before invoking mpirun. btw, do you use mpirun or srun ? can you reproduce the crash with 1 or 2 tasks ? Cheers, Gilles On Friday, July 8, 2016, Gundram Leifert

Re: [OMPI users] Java-OpenMPI returns with SIGSEGV

Hello, configure: ./configure --enable-mpi-java --with-jdk-dir=/home/gl069/bin/jdk1.7.0_25 --disable-dlopen --disable-mca-dso 1 node with 3 cores. I use SLURM to allocate one node. I changed --mem, but it has no effect. salloc -n 3 core file size (blocks, -c) 0 data seg size