Re: [OMPI users] Memory leak in my code

[Eugene Loh]

Mark Allan wrote: Dear all, With this simple code I find I am getting a memory leak when I run on 2 processors.  Can anyone advise why? I'm fairly new to MPI (have only done very simple things in the past).  I'm trying to do a non-blocking send/recv (from any proc to any proc) but the receiving processor doesn't know how much data it is going to be sent, hence the the blocking recv of the size in order to allocate the buffer.  Is there a better way of doing this? Thanks, Mark #include <mpi.h> MPI_Request nonBlockingSend(int *t, int size, const int tag, const int destinationRank) {     MPI_Request request1;     MPI_Isend(&size,1,MPI_INT,destinationRank,0,MPI_COMM_WORLD,&request1); After this non-blocking send, you need an MPI call to complete the operation.  So, you're launching a bunch of non-blocking sends without completing them. Another approach is to use MPI_Probe on the receive side.  You can then just send a message (without first sending its size).  On the receive side, you MPI_Probe the message first to get its size, then allocate the buffer, and then MPI_Recv that same message into the allocated buffer.     MPI_Request request;     MPI_Isend(t,size,MPI_INT,destinationRank,tag,MPI_COMM_WORLD,&request);     return request; } MPI_Request nonBlockingRecv(int *&t, int &size, const int tag, const int senderRank) {     MPI_Status s1;     MPI_Recv(&size,1,MPI_INT,senderRank,0,MPI_COMM_WORLD,&s1);     t = (int *) malloc(size*sizeof(int));     MPI_Request request;     MPI_Irecv(t,size,MPI_INT,senderRank,tag,MPI_COMM_WORLD,&request);     return request; } void communicationComplete(MPI_Request &r) {     MPI_Status status;     MPI_Wait(&r,&status); } void barrier() {     MPI_Barrier(MPI_COMM_WORLD); } int main(int argc, char *argv[]) {     MPI_Init(&argc,&argv);         int numProcs,rank;     MPI_Comm_size(MPI_COMM_WORLD,&numProcs);     MPI_Comm_rank(MPI_COMM_WORLD,&rank);     int numIts = 10000000;     int bufSize = 10;     // Setup send buffers     int *sendData = (int *) malloc(bufSize*sizeof(int));     for(int i=0;i<bufSize;i++)     sendData[i] = i;         // Perform send and recvs     for(int i=0;i<numIts;i++)     {     if(rank==0)     {         for(int proc = 1; proc<numProcs;proc++)         {         MPI_Request r = nonBlockingSend(sendData,bufSize,proc,proc);         communicationComplete(r);         }     }     else     {         int *recvData;         int size;         MPI_Request r = nonBlockingRecv(recvData,size,rank,0);         communicationComplete(r);         free(recvData);     }     barrier();     }     free(sendData);     MPI_Finalize();     return 1; } _______________________________________________ users mailing list us...@open-mpi.org http://www.open-mpi.org/mailman/listinfo.cgi/users