"One example is to use the browser as a very simple Photoshop," said Shaver. "[Editing an image requires] things that, for each step, take the better part of a second. That's not a great user experience. But [with TraceMonkey], now you have something that comes close to interactive performance."
In his blog on Friday, Shreopfer posted a video that showed side-by-side comparisons between Firefox 3.0 and Firefox 3.1 with TraceMonkey. Users can also run the simple application themselves using Firefox 3.0 or the latest version of 3.1.
Mozilla has tentatively set the ship date of a finished Firefox 3.1 for late this year or early 2009.
TraceMonkey is based on a technique developed at University of California, Irvine called "trace trees," and it builds on code and ideas shared with the open-source Tamarin Tracing project. Shaver credited Eich; Andreas Gal, a project scientist at UC Irvine; David Anderson, a summer intern; and others for their work on the fast-track project.
"We really started work on this 51 days ago," Shaver said Friday. Gal, who took a summer leave of absence from UC Irvine, received particular kudos. "He was like an intern on steroids," Shaver said.
Shaver said he was sure that others would dive into comparative tests. That was an easy prediction, because also on Friday, Mason Chang, a graduate student at UC Irvine who worked with Gal, put the interpreters head-to-head. "If you do the aggregate speedup, which is the speedup for each test divided by the number of tests, it is about 2.4x faster," said Chang, who posted results from WebKit's own SunSpider benchmark test suite.
The Firefox 3.1 builds with TraceMonkey included but disabled can be downloaded from Mozilla's developer site.