Getting Started
Download CS365_Lab04.zip to your cluster head node account. Unzip it. You will edit the code in column_transfer.c.
Your Task
As implemented, the program runs with two parallel processes, each of which allocates an array of integers which it treats as a 2D grid of NROWS rows and NCOLS columns. This should remind you of your Grid Datatype implementation. Process 0 initializes the rightmost “column” of values with predetermined values (multiples of 9), and sends them to process 1. Process 1 receives these values in its leftmost “column” of values, and prints them out. The output of the program should be:
Process 1: 9 18 27 36 45 54 63 72 81 90
Your task is to use an MPI derived datatype to allow the entire column of values to be sent using a single MPI_Send operation and received using a single MPI_Recv operation. You should modify the code marked by the TODO comments.
After modifying the program, it should produce the same output as before.
Testing
Use make to compile the program. Use the command
./runpar
to run the program. (It will run with exactly two processes.)
Hints
Use MPI_Type_vector to define the column datatype.
The “grid” allocated by each process is laid out row-by-row. So, cells in the same column are NCOLS number of elements apart.
The elements of the grid are ints, so use the MPI_INT data type as the element type.
