"Extreme JavaScript Performance" (from JSConf.eu, November 7)
Take a look at the slides from Extreme JavaScript Performance, a presentation by Thomas Fuchs, creator of the script.aculo.us JavaScript library, collaborator on the book Agile Web Development with Rails and one of the people behind FailCamp. He gave the presentation last week at JSConf.eu in Berlin.
In the presentation, Thomas looked at six simple things you can do to boost the performance of your JavaScript:
- Use inline functions instead of function calls.
- Embrace the language – using the language’s conventions yields unexpectedly faster code. Instantiate arrays using
var myArray = []
instead ofvar a = new Array
, and instantiate objects usingvar myObject = {}
instead ofvar o = new Object
. - Unroll your loops! (A trick so old that we covered it when I was in school!)
- Cache globals. If you’re going to access a global object, store a local reference and use that instead.
- Tune your boolean expressions: in logical “AND” (
&&
) operations, make the operand most likely to be false the first one. - Watch out for slow constructs such as with blocks, try/catch and features that JIT compilers don’t support well.
The presentation includes benchmarks for the four most common JavaScript engines:
- SpiderMonkey (Firefox 3.5)
- JavaScript Core (Safari 4)
- JScript (Internet Explorer 8)
- V8 (Google Chrome)
And yes, he does warn you – at least twice – of the dangers of premature optimization.
JavaScript Performance Rocks
If you liked Extreme JavaScript Performance, you’ll love the ebook Thomas co-authored with Amy Hoy, JavaScript Performance Rocks! Actually, it’s more than just a book – it’s four books and a profiling tool:
- Book 1: Understanding and Measuring Performance (or: “Dude, Where’s My Performance?”) – “In which our brave hero or heroine (that’s you!) apprentices to the cryptic-but-charming Master (that’s us) and learns how to get into the enemy’s head—the better to eat it alive, my dear.”
- Book 2: Loadtime (or: “The Land of Unicorn Tears”) – “Loadtime is a sad time, a time of of enormous, slow-loading assets; of maxed-out request queues; of bloated, waddling DOMs. Of limp white screens. Most of the world’s worst web performance woes? They live and breed in Loadtime. That’s why it’s the Land of Unicorn Tears, because unicorns hate slow web apps just as much as the rest of us. And they have magical horns. So there.“
- Book 3: Runtime (or: “’Cuz Tuning Loops is Hardcore”) – “The vast majority of the problems that the vast majority of apps will have can be solved with loadtime fixes of various stripes. But just in case you’re unique, and special, and have particularly intractable issues—or are just a glutton for punishment—we have written a third booklet, all about speeding up code when it runs.”
- Book 4: Interface Coping Strategies (or: “If You Can’t Fix It, Fake It”) – “Come across a performance problem that you really can’t fix? Long-running calculations? Slow server you can’t tune up? If you can’t make it, fake it. You can make your app feel faster to your customers, even if you can’t fix the underlying problem. And, let’s face it, that’s what your customers care about.”
- The DOM Monster profiling tool.
JavaScript Performance Rocks! isn’t your ordinary ebook, with material laid out for a dead-tree book simply cast in PDF form. It was designed from the ground up for onscreen reading, written in an entertaining way to keep you amused and your mind ready to learn, and written in a fun, irreverent way so that you don’t zone out. I know Thomas and Amy personally and have seen them teach; trust me – you want to learn from them.
JavaScript Performance Rocks! usually sells for US$49, but there’s a special deal right now – the first 500 books are selling for 10 dollars less — US$39. You’ve got two good reason to buy immediately: first, there’s this $10 discount, and second, the US/Canada exchange rate’s pretty good right now (as I write this, the PayPal exchange rate is CAD$1 = US$0.92).
I’ve already ordered my copy.
2 replies on “Taking JavaScript Performance to the Extreme with Thomas Fuchs”
Unless there is a huge performance difference between the tests, in which case you should put the expensive ones behind less expensive ones.
Mark Jaquith: Excellent caveat! Maybe I need to throw in an “all things being equal” qualifier somewhere in there.