[mpich-discuss] beginner for remote mem access

[Yi Gu]

Hi, sufeng:

I think you may first initialize array first, since you print out b without
initialization,
it could print out anything.

Yi


On Wed, Jul 3, 2013 at 1:12 PM, Sufeng Niu <sniu at hawk.iit.edu> wrote:

> Hi,
>
> I am a beginner and just try to use remote memory access, and I wrote a
> simple program to test it:
>
> #include "mpi.h"
> #include <stdio.h>
> #define SIZE 8
>
> int main(int argc, char *argv[])
> {
>         int numtasks, rank, source=0, dest, tag=1, i;
>         float a[64] =
>         {
>                 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0,
> 12.0, 13.0, 14.0, 15.0, 16.0,
>                 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0,
> 12.0, 13.0, 14.0, 15.0, 16.0,
>                 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0,
> 12.0, 13.0, 14.0, 15.0, 16.0,
>                 1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0, 11.0,
> 12.0, 13.0, 14.0, 15.0, 16.0,
>         };
>         float b[SIZE];
>
>         MPI_Status stat;
>
>         MPI_Init(&argc,&argv);
>         MPI_Comm_rank(MPI_COMM_WORLD, &rank);
>         MPI_Comm_size(MPI_COMM_WORLD, &numtasks);
>
>         MPI_Win win;
>
>         // check processor rank
>         char processor_name[MPI_MAX_PROCESSOR_NAME];
>         int name_len;
>         MPI_Get_processor_name(processor_name, &name_len);
>         printf("-- processor %s, rank %d out of %d processors\n",
> processor_name, rank, numtasks);
>
>         MPI_Barrier(MPI_COMM_WORLD);
>
>         if (numtasks == 4) {
>                 if (rank == 0) {
>                         printf("create window \n");
>                         MPI_Win_create(a, 8*sizeof(float), sizeof(float),
> MPI_INFO_NULL, MPI_COMM_WORLD, &win);
>
>                 }
>                 else {
>                         MPI_Win_create(MPI_BOTTOM, 0, sizeof(float),
> MPI_INFO_NULL, MPI_COMM_WORLD, &win);
>                 }
>
>                 MPI_Win_fence(0, win);
>
>                 if (rank == 1){
>                         MPI_Get(b, SIZE, MPI_FLOAT, 0, 8, SIZE, MPI_FLOAT,
> win);
>
>                         MPI_Win_fence(0, win);
>                 }
>
>                 printf("rank= %d  b= %3.1f %3.1f %3.1f %3.1f %3.1f %3.1f
> %3.1f %3.1f\n", rank,b[0],b[1],b[2],b[3],b[4],b[5],b[6],b[7]);
>         }
>         else
>                 printf("Must specify %d processors. Terminating.\n",SIZE);
>
>         MPI_Win_free(&win);
>         MPI_Finalize();
> }
>
> However the terminal gives some odd results:
> rank= 0  b= 0.0 0.0 0.0 0.0 0.0 0.0 -71847793475452928.0 0.0
> rank= 2  b= 0.0 0.0 0.0 0.0 0.0 0.0 222086852849451401216.0 0.0
> rank= 3  b= 0.0 0.0 0.0 0.0 0.0 0.0 -74882.4 0.0
> rank= 1  b= 9.0 10.0 11.0 12.0 13.0 14.0 15.0 16.0
> the rank 1 is correct results, but others should be all zero.
> terminal also give some comments: "Assertion failed in file
> src/mpid/ch3/src/ch3u_rma_sync.c at line 5061: win_ptr->my_counter >= 0
> internal ABORT - process 0"
>
> another question is if I use remote memory access. all process which does
> not create window for share must add additional line:
> MPI_Win_create(MPI_BOTTOM, 0, data_type, MPI_INFO_NULL, MPI_COMM_WORLD,
> &win); correct?
>
> Thanks a lot!
>
> --
> Best Regards,
> Sufeng Niu
> ECASP lab, ECE department, Illinois Institute of Technology
> Tel: 312-731-7219
>
> _______________________________________________
> discuss mailing list     discuss at mpich.org
> To manage subscription options or unsubscribe:
> https://lists.mpich.org/mailman/listinfo/discuss
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.mpich.org/pipermail/discuss/attachments/20130703/4414c5cd/attachment.html>