[mpich-discuss] Question reg usage with MPI_THREAD_MULTIPLE
Ramesh Vinayagam
rvinayagam.85 at gmail.com
Mon Jun 25 22:31:08 CDT 2012
It hangs at the pthread join of the second process, which did not make
sense because its the same way I use for the first process, also I use
different communicators for communication.
Thanks
On Mon, Jun 25, 2012 at 5:50 PM, Darius Buntinas <buntinas at mcs.anl.gov>wrote:
> I didn't test your program, but I didn't see any obvious problems. MPICH
> does support using threads in the way you described. What is the output
> you get? Where does it hang?
>
> -d
>
>
> On Jun 25, 2012, at 5:58 PM, Ramesh Vinayagam wrote:
>
> > Hello,
> >
> >
> > I having a question regarding the usage of MPI_THREAD_MULTIPLE.
> >
> > I was wondering whether this scenario works with MPI
> >
> > Process 0
> > Thread 1 -> probes and receives
> > Thread 2 -> Sends to process 0
> >
> > Process 1
> >
> > Thread 1 -> probes and receives
> > Thread 2 -> Sends to process 1
> >
> > I wrote a simple code, and it seems to hang with mpich2. I am not sure
> whether or not this feature is supported with MPI, can anyone let me know
> whether what I do here is correct or wrong?
> >
> >
> > Here is the code.. This code hangs when I use it with two processes. It
> works with one process.
> >
> > Thanks
> > Ramesh
> >
> ____________________________________________________________________________________________________________________________________________________
> > #include <stdio.h>
> > #include <stdlib.h>
> > #include <mpi.h>
> > #include <string.h>
> > #include <assert.h>
> >
> > typedef struct
> > {
> > int thread;
> > MPI_Comm comm;
> > int rank;
> > } thdata;
> >
> > void probe_recv (void *ptr){
> > thdata *data;
> > data = (thdata *) ptr;
> > int count, flag = 0;
> > MPI_Status status;
> > char buf[256];
> >
> > printf("%d :Thread :%d Prbes and receives \n", data->rank,
> data->thread);
> > while (!flag){
> > MPI_Iprobe(data->rank, 0, data->comm, &flag, &status);
> > }
> > MPI_Get_count( &status, MPI_CHAR, &count);
> > printf("%d :Thread :%d count : %d \n", data->rank, data->thread,
> count);
> >
> > MPI_Recv( &buf,
> > count,
> > MPI_CHAR,
> > data->rank,
> > status.MPI_TAG,
> > data->comm,
> > MPI_STATUS_IGNORE);
> >
> > printf("%d Message : %s \n", data->rank,buf);
> > }
> >
> > void send(void *ptr){
> > thdata *data;
> > char buffer[256];
> > data = (thdata *) ptr;
> > printf("%d: Thread :%d Sends \n", data->rank, data->thread);
> > strcpy(buffer,"Hello World");
> > MPI_Send(&buffer,256, MPI_CHAR, data->rank, 0, data->comm);
> > printf("%d: Thread :%d send completed \n",
> > data->rank,
> > data->thread);
> > }
> >
> >
> > int main(int argc, char *argv[]){
> >
> > int provided, i[2], rank, size;
> > MPI_Status status;
> > pthread_t thread_a, thread_b;
> > MPI_Comm communicator[2];
> > thdata data1, data2;
> >
> > MPI_Init_thread(&argc, &argv, MPI_THREAD_MULTIPLE, &provided);
> >
> > if (provided != MPI_THREAD_MULTIPLE){
> > printf("Error\n");
> > MPI_Abort(911, MPI_COMM_WORLD);
> > }
> > MPI_Comm_rank(MPI_COMM_WORLD, &rank);
> > MPI_Comm_size(MPI_COMM_WORLD, &size);
> > assert (size == 2);
> > MPI_Comm_dup(MPI_COMM_WORLD,&communicator[rank]);
> >
> > data1.comm = communicator[rank];
> > data1.thread = 1;
> > data1.rank = rank;
> > pthread_create (&thread_a, NULL, (void *) &probe_recv, (void *)&data1);
> >
> > data2.comm = communicator[rank];
> > data2.thread = 2;
> > data2.rank = rank;
> > pthread_create (&thread_b, NULL, (void *) &send, (void *)&data2);
> >
> > pthread_join(thread_a, NULL);
> > pthread_join(thread_b, NULL);
> > MPI_Finalize();
> > }
> >
> >
> >
> > _______________________________________________
> > mpich-discuss mailing list mpich-discuss at mcs.anl.gov
> > To manage subscription options or unsubscribe:
> > https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>
> _______________________________________________
> mpich-discuss mailing list mpich-discuss at mcs.anl.gov
> To manage subscription options or unsubscribe:
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mcs.anl.gov/pipermail/mpich-discuss/attachments/20120625/bd889235/attachment.html>
More information about the mpich-discuss
mailing list