Hi,

I am using IPDG methods with distributed computing framework. After reading
the paper on distributed computing, I gather that I need to make a point to
point communication for assembling the matrix. From this I understand, that
when I need to add the flux terms for the ghost cells, I need to send the
information to the processors that has this ghost cell as a local cell. I
have a couple of questions on this(probably very amateurish):

1. How do I send the matrix entry locations and the local contribution for
all the ghost cells to the corresponding processor?
   If I use MPI_Send and MPI_Recv , I need to mention the processor of the
current cell and the ghost cell processor and I don't know the function to
get that.
2. How do I ensure that the matrix entry locations and the local
contributions are not jumbled up when they are received by the neighboring
processor. Do I have to wait until each matrix entry location and local
contribution is added in the neighboring process before sending the next
one?
3. Is there a more elegant way of doing this?

Thanks
_______________________________________________
dealii mailing list http://poisson.dealii.org/mailman/listinfo/dealii

Reply via email to