Apple

Google's latest version of Chrome has claimed the lead in my JavaScript speed tests, but Apple's new Safari 4 beta is the first browser to challenge it on Google's own performance benchmark.

JavaScript is a programming language that powers not just innumerable ordinary Web sites, but also many Web-based applications such as Google Docs. With the computing industry's major push to cloud computing, Web application performance is increasingly important, and there's a race on to see who's got the best JavaScript engine. JavaScript engines even have become a named feature, with Chrome's V8, Firefox's TraceMonkey, Opera's Futhark and upcoming Carakan, and now the Safari's newly branded Nitro, which is Apple's version of WebKit's Squirrelfish.

On the SunSpider test, the new Safari 4 beta scored third place.

(Credit: Stephen Shankland/CNET News)

Just about every browser out there now is trying to grab the crown for fastest performance for running JavaScript, the programming language that powers many increasingly sophisticated Web-based applications. The latest development is from the programmers behind Apple's Safari.

Mozilla bragged earlier this month about TraceMonkey, a new JavaScript engine due to ship in Firefox 3.1 near the end of 2008. Next came Google's Chrome, a leading feature of which is the performance of its V8 JavaScript engine. Now the WebKit programmers, whose open-source code is used in Apple's Safari browser and the Konqueror browser of the KDE interface software sometimes used on Linux systems, have a new version of their JavaScript technology.

It's called Squirrelfish Extreme, and the WebKit programmers said Thursday in a blog posting that it's more than twice as fast as the first-generation Squirrelfish announced in June and more than three times faster than the current WebKit 3.1 version. They based their conclusions on one benchmark, SunSpider.

"SquirrelFish Extreme uses more advanced techniques, including fast native code generation, to deliver even more JavaScript performance," the programmers said.

For details of Squirrelfish's techniques--bytecode optimization, a polymorphic inline cache, a context-threaded just-in-time compiler, and a regular expression just-in-time compiler--check the WebKit blog.

Charles Ying also performed SunSpider tests that showed Squirrelfish beating Google's V8 and Mozilla's Tracemonkey on a 2.4GHz iMac.

WebKit's SquirrelFish Extreme is faster than its three-month-old predecessor on the SunSpider JavaScript benchmark.

(Credit: WebKit)

The iPhone isn't a true mobile computer yet, but it's on the right track, according to a Mozilla executive.

Will there be two separate Firefox browsers for smartphones and PCs or one to rule them all?

(Credit: Mozilla)

"Getting a no-compromise web experience on devices requires significant memory (>=64MB) as well as significant CPU horsepower. High end devices today are just approaching these requirements and will be commonplace soon," wrote Mike Schroepfer, vice president of engineering at Mozilla, in a blog post Tuesday, implying that while the iPhone and its current competitors don't quite have what it takes under the hood to be full-fledged mobile computers, we're not all that far away.

It seems to me like there's a bit of a chicken-and-egg thing going on here. Are smartphones slower than people would like because the hardware is too rudimentary, or because truly useful software is too bloated for the limited memory and power requirements of smartphones? I don't think too many people bought an iPhone expecting it would be just as zippy as their PC, but just how much slower is it than a PC?

Schroepfer thinks, based on third-party tests, that the iPhone is about 10 to 100 times slower than a MacBook Pro on scripting benchmarks and about 3 to 5 times slower than a ThinkPad T40 laptop when operating on the same Wi-Fi network. "But rapid improvements in mobile processors will close this gap within a few years," he wrote.

He estimates that the iPhone is using about 128MB of system RAM, and a processor (known to be an ARM-based chip from Samsung) running at between 400MHz and 600MHz. Apple's iPhone application development policy means we're not going to see Firefox on the iPhone anytime soon, but that's information that Mozilla is using to work on future mobile browsers for devices like the iPhone that won't be able to run unmodified PC software for several years.

As Schroepfer notes, the nice thing about the chip industry is that we can be reasonably sure that there will be more performance to work with every couple of years. Both ARM and Intel have set aggressive performance and power consumption goals for chips due out over the next several years.

But Schoepfer seems to be operating under the assumption that it's the hardware that is holding back a true Internet experience on a smartphone. "Up until very recently, device limitations required writing new mobile browsers from the ground up," he wrote. I wonder if that was such a bad thing; I'm sure to save time and effort developers would rather port as much of their PC code as is feasible over to smartphones, but is it better to develop mobile software that's designed specifically for mobile devices or to investigate ways to move the multitude of software that's already out there for PCs to a new category of mobile devices?

Mozilla wants to work both sides of the fence, not wanting to throw away all the work they've done on PC development when mobile processors are bound to get more capable, but recognizing that mobile-computing requirements are different. "There is far from a dominant player in this marketplace and even the best mobile browsers today have compromises in user experience, performance, and compatibility. There is still *plenty* of room for innovation," Schroepfer wrote.

I'm no software developer, and I'd welcome feedback about this from those who are examining this problem. It seems pretty clear to me that true mobile computing is going to require new thinking about software development in addition to faster hardware, the same way multicore processors have shaken up the PC software development industry. And those concepts are even going to merge at some point: by 2010 ARM's partners will have multicore mobile processors on the market.

Does that mean personal-computing software development is headed down two different development paths or that smartphone developers and PC developers are converging at some point down the road? Let me know what you think.