Simon Fraser University

• CMPT 295

• Objective: Continuing on from our Lab 5 …  The purpose of this lab is to introduce us to the idea of microbenchmarking, i.e., testing the performance of very small units of computation. To do so, we shall use more precise tools than getrusage() (used in Lab 5). Submission done via Assignment 9  Include your data sheet and graph (last 2 pages of this lab) in your Assignment_9.pdf document. See Question 1 of our Assignment 9 for more details. Part 1: Benchmarking a Loop  Login to a CSIL machine in the Linux environment, download Lab6-files.zip and unpack it into sfuhome/CMPT295. Change directory into Lab6 where we will find the usual makefile and a collection of source files.  When we open main.c, we see something similar to the benchmarking code in Lab 5: an array of N integers randomly initialized, and a function sumPlus() which is benchmarked 20 times using getrusage().  The function sumPlus() is written in assembly. It computes the sum of all of the positive integers within an array of integers A[n]. Have a look at the code.  Build the executable, run it a few times and have a close look at the results. Woa! What is happening? Some of these runs conclude: “Average of 0 clock cycles”. Our testing does not seem to be working. The reason for these strange results is that we are measuring something very small: the time (number of clock cycles) it takes to compute the sum of the elements contained in a very small array, which is in the order of nanoseconds, with a tool (getrusage()) that is meant to measure something much larger, i.e., computation ranging from microseconds to seconds.