Homework 10: Sorting Algorithms
Write a program that compares the real (wall-clock elapsed) running times of the sorting algorithms that we have discussed in class. You will implement three O(N^2) sorts (selection, bubble, and insertion) and two O(N log N) sorts (mergesort and quicksort).
Your program should read an arbitrary sequence of integers into an array from a file, sort the array using a particular sorting algorithm, and then print your timing results. Your executable will prompt the user for two arguments before sorting: first the name of the input file, which will be printed back to the user. Second, the sort to use (i, s, b, m, or q). The input file will contain a count of the number of integers that follow on the first line, followed by one integer on each subsequent line. Once the data has been sorted, the program will prompt the user to optionally print the resultant sorted data set. Valid answers are `Y' or 'y' for yes (to print results), and anything else is interpreted as no. Sample files are available here:
Data Files Code
Note that if you want to put your program in a loop, you need to reread the file on each iteration (or else save the unsorted data in a "backup" array, because the timing would be skewed if you were to sort an already sorted array.File I/O is very straightforward in C++. There are object types named ifstream and ofstream (for input file stream and output file stream respectively) defined in the fstream library (necessitating a #include of fstream. To perform file input, you simply create an ifstream object with a constructor whose sole parameter is a string specifying the filename. Once this is done, you can use stream extraction to retrieve values from the stream, just as you would for cin. The same model of usage for output is exemplified in the makeints.cpp file above. You can also see www.cpluplus.com for examples of ifstream and ofstream usage.
To time your sorting algorithms, use the function gettimeofday() (recall that we used this function early in the semester). Please note that your timings should not include the time printing the sorted array to the console, nor should it include time retrieving the inputs from the user and the data from the file; all of these would add a significant amount of time and skew the results, especially for small inputs. To check your algorithm, print the results outside your sort functions, after timing the sort.
You can earn 3% extra credit for each sorting algorithm (up to 4) that performs better than your classmates' for a data set and size that I select. If time permits, we will hold this sort-off in class. This will be verified by my executions on a single platform.
The gettimeofday() function looks like this:
int gettimeofday(struct timeval *tv, struct timezone *tz);
The timeval structure has the following form:
struct timeval
{
long tv_sec; // seconds
long tv_usec; // microseconds
};
To time each sort, do something like this:
#include <sys/time.h>
...
timeval timeBefore, timeAfter;
long diffSeconds, diffUSeconds;
gettimeofday(&timeBefore, NULL);
sort_algorithm(A);
gettimeofday(&timeAfter, NULL);
diffSeconds = timeAfter.tv_sec - timeBefore.tv_sec;
diffUSeconds = timeAfter.tv_usec - timeBefore.tv_usec;
cout << diffSeconds + diffUSeconds/1000000.0 << " seconds" << endl;
Record and plot your results (input size vs. running time) for various size inputs ranging from 100 integers to at least 500,000. You should generate additional data points beyond the limited set I have given you above. Also, include a brief discussion of the results. Hand in a printed copy of your results and discussion in class, and email me a copy of your program.