Posts with the tag "profile".

The comparison of algorithms is an active area of work. When we start learning algorithms, or more advanced programming, we learn of different ways of doing the same complex task. The most usual first example is sorting, which introduces a series of different ways to sort a single array, such as selection sort, insertion sort, quick sort, merge sort, etc. When comparing these algorithms, we take into account a few things: how fast it is, how much memory it needs, what are the best/worst/average-case complexities, and so on.

One of the things I worked with was a Performance Profile tool, in the sense of Dolan and Moré. This tool, perprof-py, written in Python, reads input in the form of a table of problem with time and exitflag, and outputs a image/pdf with the performance profile. This tool is very important for the nonlinear optimization community, since it is one of the main comparison tools for the algorithms. The package is open source, available for changes, improvements and suggestions.