You are on page 1of 1

Parallel Computing I

Homework 5: Performance
Thorsten Grahs, Andres Rodriguez
1. June 2015
This will be your fifth homework in the exercise parallel computing. Send your solution to
a.rodriguez-escobar@tu-braunschweig.de until June 10st 2015 23:55.
Prepare a pdf file for your written text and attach the source code of your program to the mail.

Task 1 (20 Points)


In this task you are going to experiment with different implementation of the inner product
computation:
a) Determine the peak performance (in GFLOPS) of the serial implementation of the inner
product from lecture 7 (Parallel performance). To do so, plot a graph for different n until
you reach the peak performance.
b) Implement a parallel version of the program, using the recursive doubling algorithm of
lecture 7. Calculate the speedup and the parallel efficiency for different numbers of processors P.
c) Implement a parallel version of the program, using the butterfly algorithm of lecture 7.
Calculate the speedup and the parallel efficiency for different numbers of processors P.
The result at the end of the calculation should be available on all processors. Plot the result of
your computation for different P (e.g 2, 4, 6, 8, . . . , 32). Give an explanation of your result.

You might also like