problem.txt

Offline assignment on Java Threading (Section A)
Submission Rules

1.Submission Steps:

a.In your local machine, create a new folder; the name of the folder should be your 7 digit roll number.

b.You can code your solution in multiple files, if you want. Your main function must be in a file called Main.java. Put all the source code files in the folder created in (a)

c.Finally, compress the folder created in (a) to produce a .zip file. The name of the .zip file should be your 7 digit roll number.

d.Submit the .zip file in Moodle.

2.Do not copy code. You will get negative score otherwise.

3.On the evaluation day, you must be able to explain your code properly. You must bring your code on that day too.

 

Problem #1

Matrix Multiplication

In this task you need to tackle the problem of multiplying 2 matrices in a multi-threaded fashion. Both the matrixes (A, B) are given in row major order. The multiplication is guaranteed to be possible. Parallelize the task as follows:

·Create a set of worker threads. How many threads should be created - this will be a parameter given by user to your program. Let's call that parameter n and call this set of threads W. (Therefore, n = |W|). It is guaranteed that row(A) * column(B) is a multiple of |W|.

·Write a class called WorkItem. Each WorkItem object holds ith row of A (Ai) and jth column of B (Bj) for all 1 <= i <= row(A) and 1 <= j <= column(B). A WorkItem also stores the indices i and j. So there will be row(A) * column(B) work items. All these work items are put in a queue. This can be done by your main thread.

·A worker thread from W extracts a work item from the queue and performs the summation of multiplication between the Ai and Bj. Each worker performs exactly row(A) * column(B) / |W| number of work. The worker writes the result to the appropriate cell of the resultant matrix.

·Main thread writes the resultant matrix to the console.

Be sure to consider the following:

·Be prudent in using synchronization. Do not use it unnecessarily. Also, use enough synchronization to avoid data corruption.

·Perform wait, notify and joining, wherever you find it appropriate.

·Analyze your code to detect bugs. Ensure good parallelism and integrity of data

·Double check your results with a regular (single threaded) matrix multiplication routine.