[General] SquirrelFish

The latest WebKit got a bytecoded JavaScript interpreter - details at:

        http://webkit.org/blog/189/announcing-squirrelfish/

Guess that won't hurt livelyness :)

- Bert -

Bert Freudenberg <[hidden email]> wrote...

>The latest WebKit got a bytecoded JavaScript interpreter - details at:
>
> http://webkit.org/blog/189/announcing-squirrelfish/
>
>Guess that won't hurt livelyness :)

Indeed!

We've been playing with it a bit, and I can report the following results...

I see a clean 2x improvement over Safari 3.1 (and I recall this was nearly 2x over the earlier WebKit...

        Safari 3.1 1.7M sends/sec 71M ops/sec
        Firefox 3.1 3.2M sends/sec 130M ops/sec
        webkit 5/21 3.9M sends/sec 154M ops/sec

... as measured by http://weather-dimensions.com/Dan/JavaScriptBenchmark.html

They have a way to go to equal a good interpreter (Squeak does 11M sends, 525M ops on the same machine), but this is finally becoming a serious programming medium!  Isn't it nice to have all the browser makers working for us!

In fact most Lively apps run about the same speed because our current bottleneck is the DOM interface to SVG graphics and SVG itself.  One notable exception is text composition, where we do all the "heavy lifting" in JavaScript and, yes indeed, it's now twice as fast.

        - Dan

Goddard Jiri <[hidden email]> replied offline...

>I noticed that the UI is quite slow in comparison to another JS
>based UIs like the one@<http://www.meebo.com>www.meebo.com
>Will it get better?

Hi, Jiri -

I sure hope so.

Now is the time for the SVG folks to start doing some of the same
sort of engineering that is being done on JavaScript.  The Lively
Kernel could run *much* faster with just a bit of work.  If no one
does this, eventually we will, but for now we'll keep holding a
carrot out there to see what others can do.

Our current priorities are

>  Exploit the delightful uniformity of doing everything ourselves in JavaScript

>  Explore high-security variants (Caja etc)

>  Experiment with end-user programming and easy wiki-like web content creation

>  Explore possibilities of multi-user web pages (ie, collaboration)

>  Build the ultimate power-user web programming tool

None of this is about performance... yet.

At any point we (or some of you out there in Lurky-Land) could work
on a faster Canvas model (but beware text), do a lively plugin (we
did this once already), or even link to native widgets.  Right now,
though, our current strategy leaves the entire system much more
malleable, and therefore a better vehicle for these explorations.

        - Dan
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://livelykernel.sunlabs.com/pipermail/general/attachments/20080603/57c94dea/attachment.html