John Resig on TraceMonkey

by sant0sk1 • 18 years ago

66 20 18 years ago

sh1mmer• 18 years ago

John does a really job of discussion what implications this had as well as how it does it. Great article.

Install the update and do the image manipulation test. For best results, run the test before and after. The difference is astounding.

babyshake• 18 years ago

I thought "astounding" would be an overstatement. Turns out you're right.

KirinDave• 18 years ago

Has anyone run a comparison between this new engine and SquirrelFish yet? It seems to me like two major ideas are competing:

1. Better JIT for interpreters. 2. Faster interpreters, period.

I'm really curious to see which has the faster ramp-up and which has the best end-game results.

jeresig• 18 years ago

One comparison was posted in the comments of the blog post: http://www.masonchang.com/2008/08/tracemonkey-vs-squirrelfis...

nostrademons• 18 years ago

It's funny, 2 months after I give up on GameClay (partially because of performance issues in the prototypes), this comes out. Maybe I should go back to it. Well, that's why I chose not to open-source the code and my cofounder's hanging on to the domain names.

Though any application to games needs to get over the fact that 90% of the casual game market still uses IE...

tokipin2• 18 years ago

check out Steve Yegge's video/transcript about dynamic language optimization in case you missed it:

http://steve-yegge.blogspot.com/2008/05/dynamic-languages-st...

andreyf• 18 years ago

Things have been moving in this direction for quite some time, I think. Here's a master's thesis from March 2006:

http://www.cs.wustl.edu/~plezbert/contcom/

MaysonL• 18 years ago

Michael Franz, the PI for the work at UCI, has been working on this stuff for almost 20 years. See his 1994 PHD thesis: Code-Generation On-the-Fly ftp://ftp.inf.ethz.ch/pub/publications/diss/th10497.ps [postscript]. Reading it was one of my "Holy Shit!" moments.

invisible4• 18 years ago

I think the key point of this article over the other related article (http://news.ycombinator.com/item?id=284192) is that John points out that tracing will be able to be applied to DOM manipulation/property access in the future.

  "Being able to trace through a DOM method would successfully speed up, not only, math and object-intensive applications (as it does now) but also regular DOM manipulation and property access."

As I said on the previous article, "I believe tracing very well could be the key to unlocking advances in that [DOM] area as well."

To me, this means that JavaScript will finally be integrated with the browser in a way that is required to advance the web. I've always wondered what would be the "big thing" to lead up to the move towards the web even more, and this is definitely the answer.

The only thing left now is making it so JavaScript can be compiled (the only way I can think of how to accomplish preventing stealing code).

gruseom• 18 years ago

speed up, not only, math and object-intensive applications (as it does now) but also regular DOM manipulation

Agreed. The sooner they get to that, the better!

Edit: As an aside, the web arguably wouldn't have happened without "stealing code". I see this as a strength not a weakness.

tokipin2• 18 years ago

true, but it's not as straightforward when the things you're stealing are whole applications

invisible4• 18 years ago

I'm not against open sourcing code, or giving away helpful small things, but at the moment anyone can copy every bit of useful functionality of a JS application and just create their own back-end. That is a very ridiculous concept, and there isn't much that can be done about it if that person isn't within the same country.

John Resig on TraceMonkey

Related Stories