522, CSc Assignment 1
Due date: February 6, 2009, at noon. Please place a hardcopy of your assignment in the box that I will provide. The box will be located in the mail room. No late assignments will be accepted. You are required to do the following: 1. From the text, problems 2.7, 2.10, 2.14, 2.17, 2.21, 2.23, 2.31. Some notes: For problem 2.7, you can write pseudocode. For problem 2.23, I strongly recommend that you rst write out a proof for the search algorithm code given in the example of the while statementexcept using repeat-until. This will help you generate an inference rule for repeat-until. In addition, proving equivalence of {P } S1 {Q} and {P } S2 {Q} means showing that {P } S1 {Q} {P } S2 {Q}. You will nd, assuming you get the correct repeat-until rule, that one direction is much easier than the other. For problem 2.31, the triples are from dierent arms of a co statement. In addition, ignore part (A) and just answer part (B) 2. Consider the shared-memory matrix multiplication algorithm given in class, in which we discussed that work could be assigned in a block fashion or a wrapped fashion. example, For if there are two threads and n rows of the result matrix, we can assign rows 0 through n/2 1 to thread 1 and rows n/2 through n to thread 2 (block), or we can assign rows 0, 2, . . ., n 2 to thread 1 and rows 1, 3, . . ., n 1 to thread 2 (wrapped). Assuming that the n is large, what is the dierence in terms of the hardware main memory cache, if any, between these two work assignments. 3. Write a parallel matrix multiplication program on Voltron using pthreads. Make the program congurable so that on the command line, any number of threads can be specied. Also write a sequential version so that you can measure speedup, which should always be rel...