Program #5 -- CS 491
Implement the problem described in Section 3.4 in the book
(pages 73 - 77)
Please use the following "inputs"
START NEW
Let the rod be 1 meter in length
Let "m" be 1000 (we are going to run each program for 1000 time steps)
Let "k" be one second (each step will be one second in length).
Number of segments (n) X Number of Processors
100 4
1000 8
16
You are to calculate "h"
You are to calculate "r"
This means your program should take 2 parameters, number of points and number
of processors (or read a control file that does this).
You must run each one of your pairs on a different set of processors (
you cannot use the default allocation of processors, but rather you have
to specify different processors).
Receivables include
1) A print out of your code WITH COMMENTS that make sense. Make it clear how
you figure out how the agglomeration and how you allocate the residual.
2) plots of the estimated points for each 200 time stamp stride (go 200 time
steps, then plot) Each plot should have 5 lines.
3) proof that your 6 runs were run on different, non-default, processor
groups.
4) A write up describing how you collected the data to create the plots
Explain how your agglomeration addresses the "reduced communication",
"load balance" issues. Explain your communication methods between
nodes