In general, to study the performance of an algorithm you should fist implement the algorithm and than implement a function that
- queries a timer
- call the algorithm giving it the required testing data
- queries the timer (again)
The difference between the values of the two queries is the elapsed.
About timers, you can fund some useful articles here
or by google "windows high resolution timer".
To come to a graph, you have to repeat that process a number of times, passing different type of data (or calling different algotithm implementations) to came to a "series of numbers".
This can be done:
- manually, calling the testing program many times after changing the data, then pasting the number into a spreadsheet (like excel) and produce a graph,
- semiautomatically by making the testing program able to do the repeated calls and save the results in a text file (that you will later import into a spreadsheet)
- Automatically by writing a testing program that also displays the data in the form of a graph (more complex, if you are unaware of how graphics work on windows, and in general not suitable as a "portable test", but certainly more "scenographic")