David Mandelin's blog

I haven’t blogged in a while because I’ve been in heads-down mode working mostly on getting closures and the arguments keyword to trace in TraceMonkey and investigating topcrashes. But we’ve just solved the most common JS-engine topcrash, so I’m resurfacing with the story, and hopefully some useful tricks and analysis for anyone else working on [...]

I wrote an article on TraceMonkey on hacks.mozilla.org. It’s aimed at web developers and anyone else who wants to understand how TraceMonkey works and what makes it different from other JavaScript engines. The article also talks about what kinds of JavaScript run especially fast or slow in TraceMonkey and some introductory information on how to [...]

I’m back from PLDI in Dublin. The weather was nice and the Guinness was excellent. The research program was also very good. It seemed like this year there was a lot of variety and a different topic selection from previous years. I’m going to blog some notes on an arbitrary (slanted toward things with more [...]

Next week we’re taking the TraceMonkey show on the road to PLDI 2009 in Dublin, Ireland. Andreas Gal will of course be presenting our paper Thursday afternoon, and David Anderson (who is back at Moz for the summer) is also coming. Fortunately, the Mozilla 1.9.1 (FF 3.5) blocker list for JS is empty, and hopefully, [...]

Last fall, we submitted a paper on TraceMonkey to PLDI (officially: ACM SIGPLAN Conference on Programming Language Design and Implementation), one of the top conferences for programming language research. Our paper was accepted, and Andreas Gal will be presenting the paper on June 18 in Dublin, Ireland.
We’re hosting a PDF copy on the blog. [...]

In the evening at the OSQ retreat, we had some informal discussions about new scripting language designs and languages for mobile devices. The starting points for these discussions were (a) that scripting language programmers will soon want to use parallelism, and (b) that mobile devices will have uniprocessor performance at least 12x slower than laptops [...]

Random bits from the second half of yesterday’s OSQ events:
Bill McCloskey gave a quick talk on his experience doing some optimizations to Python. (From a VM optimizer’s point of view, Python and JS are very similar.) He had previously used whole-program type inference to drive a translation to C, and in this way got performance [...]

I’m at the Berkeley OSQ (Open Source Quality project) Retreat in Santa Cruz right now representing Mozilla. It’s an annual event where professors and grad students present their latest research results and ideas.
There’s a lot of good stuff here, so I’m just going to blog about a few things that seem particularly relevant to [...]

A little while ago, Rodrigo Sol at the Federal University of Minas Gerais submitted a Google Summer of Code proposal to create a new register allocator for TraceMonkey. I’m pleased to announce his proposal was accepted!
Rodrigo’s project idea is to implement Register Allocation By Puzzle Solving, a paper in PLDI 2008 by his advisor, Fernando [...]

I’ve been working on a visualization of TraceMonkey performance, with the goal of revealing what the JS VM is doing, and why it runs certain programs fast or slow, so we can figure out how to make the slow ones fast too. In this post, I want to show off the results and explain how [...]