to your home directory (or subdirectory) on H:, e.g.

For the following exercises, compile the programs on a login node. Create a batch file and submit the job(s) to the batch nodes.

Exercise

Input file:

Fortran lab files:

Fortran solution files:

C lab files:

C solution files:

Output file (to compare C/Fortran results):

Program Description

Wave_mw implements the concurrent wave equation described in Chapter 5 of Fox et al. (1988) utilizing a master-worker model. It calculates the amplitude of points along a vibrating string for a specified number of time intervals. The equation it solves is:

   newval[i] = (2.0 * values[i]) - oldval[i] 
	       + (sqtau * (values[i-1] - (2.0 * values[i]) + values[i+1]));

The decomposition can be viewed as:

Each task is assigned a contiguous block of points ("block" decomposition). Each task has all the data needed to update its interior points. To update its endpoints, the task must receive values for the points bordering the block (boundary values) from the tasks that "own" them. It must also send the values of its own endpoints to these tasks. Non-blocking sends (followed by a call to MPI_Wait) and blocking receives are used for this communication. Thus, the main computational loop exchanges endpoints and updates amplitudes.

Modifying the Code to Use Scatterv and Gatherv

In the present implementation of wave_mw, the master task reads the initial amplitudes from a file. Blocks of data must be distributed to workers. Similarly, at the end of the run, the updated data must be collected on the master for writing to a file. These steps are the focus of this lab.

The outline below shows the wave_mw program structure. The selection boxes identify changes that must be made to specific subroutines. These subroutines contain comments that provide information on how to proceed. In addition, you may want to refer to the code fragment presented in the MPI Collective Communication II talk.

begin program

learn number of tasks and rank (main program)

if master: get program parameters and initial data (get_data)

broadcast program parameters (get_data)

distribute initial data

distribute_dataFirst, calculate for each task:* number of points assigned* displacement in data array of first point assignedThen call scatterv

update values (update)

collect results

collect_dataCall gatherv

if master: store results (out_master)

end

Compile and run the modified program wave_mw_vv.exe on 2-8 processors. Remember, you must copy the input file wave_inp.data into your local directory space.

The code prints out several values for validation. The correct values are:

 1: .00   11:-.20   21:-.32   31:-.32   41:-.19   51: .01   
61: .21   71: .32   81: .31   91: .18  100: .00
 

The resulting output file, cwave_vv_out.data or fwave_vv_out.data, can be compared (using diff) to the generic output file (wave_vv_out.data) in the lab directory to give additional confirmation that the code is running correctly.