O'Reilly Open Source Convention - August 1-5, 2005

Tutorial
Making Programs Faster
Mark-Jason Dominus, Chief Programmer, Plover Systems Co.

Track: Perl
Date: Tuesday, August 2nd, 2005
Time: 1:30pm - 5:00pm
Location: Portland 255
Almost every application must be made to run faster; some sooner, some later. Performance tuning of applications has long been a dark art, understood by few and riddled with terrible pitfalls. Stories abound of optimization projects that took weeks but yielded a pathetic 2% decrease in total run time. Don't let this happen to you.
The class begins with a brief introduction to the basic concepts of performance tuning. We'll then take an extensive look at modules for benchmarking and profiling, including reasons why you should avoid the standard Perl 'Benchmark' module, instructions for using the new 'Benchmark::Accurate' module, and common errors that even experts commit when benchmarking. We'll finish with a discussion of a few of the most important optimizations.
Throughout, the class emphasizes a rational, cost-benefit approach to performance tuning, discussing how to decide when to do performance tuning, alternative approaches, and presenting many amusing anecdotes and cautions about philosophical errors (such as the '1+1=0 principle' of optimizations) that can cost you time and money.
Outline:
I. Preliminaries: Different kinds of time; bogus optimizations; I/O-, CPU-, and memory-bound programs; measurement tools; Benchmark.pm.
II. Profiling examples: Mail folder analyzer; speeding up the perldoc program; the 90-10 rule; writing your own profiling tools.
III. Common blunders: Pseudo-hashes; the missing 15%; premature optimization; numerical calculation; Jackson's rules.