The objective of this assignment is for you to gain experience in the following parallel programming languages/models:
Complete the following programming questions. For each, design efficient well comment codes and evaluate them carefully.
1) Create an MPI-based program that implements sample sort as described in "Parallel Sorting by Regular Sampling" by Hanmao Shi and Jonathan Schaeffer. Journal of Parallel and Distributed Computing, vol. 14, no. 4, pp. 361-372, 1992. You should design your sort to work with records that contain a key and data field.. You do not need to fully balance the resulting sorted output over the machine. Your code may either read the input and generated the output to files distributed over your parallel machine or start and end with the data on PE 0, whichever you find easier. I have created an outline of a simple sample sort to aid you. You will need millions of data items per processor to see speedup.
2) The goal of this question is to give you some experience using OpenMP to speed-up an existing C++ code. Start by reading carefully the following links
Particle Swarm Optimization (PSO) is a derivitive free optimization method (like Genetic Algorithms or Simulated Alealing) that uses a population of "solutions" to home in on a final answer to a provided search problem. Proceed as follows:
Note: For each program you will be asked to evaluate its performance carefully. More details on the expected analysis and performance evaluation will follow in class. Please see parallel_trap.c as an example of how to time your application.