Mozilla developer Brian Hackett worked on a new feature called type inference for over a year. You may know that you need to define types for variables in some programming languages, while it is not necessary to do that in others. JavaScript for instance does not require type definitions for variables. The downside here is that this practice can have an impact on performance.

Type inference now refers to an algorithm that is automatically deducting the type of an expression.

Type inference is boosting Firefox JavaScript performance big time. Mozilla’s David Mandelin noticed JavaScript performance improvements of up to 44%.

On my machine, TI takes our Kraken score from about 4900 to about 3400, a 1.44x speedup. And on V8-v6, our score goes from about 5000 to about 6600, a 1.3x speedup.

Sebastian Anthony over at Extreme Tech did some measuring of his own and came to the conclusion that the performance gains depend largely on the benchmark used. He noticed performance improvements between 15% to 30% in benchmarks on a Intel i7 930 processor with 6 Gigabytes of RAM and a fairly speedy Nvidia GTX 460 video card.

With Firefox 9, and without type inference, the Kraken benchmark takes 3895 milliseconds; with type inference enabled it takes just 2763 milliseconds. Firefox 9, without type inference, scores 6075 on the V8 JavaScript Benchmark; with type inference, the score jumps up to 6585. Even on banal tests like Microsoft’s HTML5 Sudoku, type inference improves the solving time for 10,000 grids from 2.6 down to 1.62 seconds — and yes, the FishBowl framerate, at 2,000 fish, is increased by 15-20% with type inference enabled.

Interested users can download the latest Mozilla Central release with type inference enabled from the official Mozilla ftp server. The next Firefox channel to get type inference is Firefox Aurora, which is going to be released on September 27. Firefox Beta will get the feature on November 8, and the stable channel on December 20.

The Opera team speaks of a 30% speed increase over the previous version of the rendering engine. A test of the JavaScript performance of Opera 10 revealed astonishing results. Opera 10 completed the
Web Browser Javascript Benchmark in 188ms which is very fast compared to Firefox 3.0.4 which needed 349ms, Safari 3.1.2 with 344ms or Opera 9.52 with 420ms. It’s obviously not fair to compare a development release with web browsers who have already been released but it should give anyone an understanding about the performance increase in this new Opera version.

What else is new in Opera 10? Opera 10 adds a long requested inline spellchecker and an interesting auto-update function which can update Opera automatically whenever a new version has been released.

Web developers will like the support of web fonts, SVG and opacity through RGBA and HSLA. The current release schedule aims for a beta in early 2009 and a final release of Opera 10 later that year.

Interested users can download the Opera 10 alpha release for Windows, Linux or Mac.

Cnet used the SunSpider JavaScript Benchmark to test browser performance of the browsers. There are other factors that have an impact on JavaScript performance like the processing power of the computer system.

The results were that the latest Firefox 3.1 beta with the new JavaScript engine TraceMonkey was the fastest with a score of 2,257, followed by Chrome with 2,904, Firefox 3 beta 1 without TraceMonkey with 4,233, Safari 3.1.2 with 6,351 and Internet Explorer 8 Beta 2 at the end with a score of 9,025.

Cnet failed to test any version of Opera and should probably have tested the latest nightly test of Safari instead. Here are the results that a Windows XP service pack 3 with a Intel 8400 Wolfdale processor and 4 Gigabytes of computer memory.

• Firefox 3.1 Beta 1 without TraceMonkey- 1800
• Firefox 3.1 Beta 1 with Tracemonkey 1817
• Firefox 3.03 – 2139
• Opera 9.6 – 2736

It’s a bit surprising that Firefox without TraceMonkey was faster than Firefox with it enabled. It could be due to the fact that the portable version of Firefox has been used to run the tests. But the interesting question is, how does your browser compare? You need to run at least two browsers to be able to tell how it compares to the other browser.

Keep in mind that JavaScript performance is not everything and that page loading time is dependent on additional factors.