Comments on: SquirrelFish Extreme: JIT comes to SquirrelFish with extreme results

By: ck2

ck2 — Thu, 26 Feb 2009 02:24:16 +0000

By the way, Safari 4 for Windows is the first browser to run my little benchmark smoothly (fullscreen). Everything else, Firefox 3.1 and even Chrome is not smooth enough.

Looks like the last pastebin went dead, so here’s another copy if anyone else wants to see what I mean:

http://pastebin.com/f4cd4165d

By: Jeria

Jeria — Mon, 22 Sep 2008 07:41:18 +0000

Where is the “fastest browser on earth”-Opera?

By: ck2

ck2 — Mon, 22 Sep 2008 05:24:18 +0000

@eyelidlessness, forgive the quicky/poorly made code but this will prove my point – open IE6 and Firefox full screen and run this code:

http://pastebin.com/m2fce0d31

Firefox will be half the speed of IE. Then try removing the background images and Firefox is fine. I suspect this is related to the Gecko smooth scrolling performance bugs which were finally dealt with for FF3 but still slow down overall performance. (Note that window size is critical, if you make the window small, you won’t see the slow down – but screen sizes have never been larger).

There is probably a better way to demonstrate this issue but was easiest way I could whip up quickly.

By: eyelidlessness

eyelidlessness — Sun, 21 Sep 2008 16:02:24 +0000

@ck2: Could you provide a benchmark demonstrating this?

By: ck2

ck2 — Sun, 21 Sep 2008 07:39:10 +0000

> I think that all benchmarks should use IE6 as their baseline, so
> that we can hammer home to IE6 users, how far today’s
> browsers have advanced. So how does this compare to IE6 performance?

IE6 JS sucks but it still has the fastest rendering performance when there are background images present during dhtml, compared to ALL other modern browsers that I have benchmarked (FF3.1 Safari3 Opera9.5)

So JS rendering has come a long way but rendering still has a way to go. I don’t understand with the advances in video cards why it’s still so slow except that maybe because it’s only 2D.

By: Nosredna

Nosredna — Sat, 20 Sep 2008 19:38:40 +0000

Bill Gates: What do you think about speed, Jerry?
.
Jerry Seinfeld: Woah, Bill. I’m already a little bit fidgety. I don’t think speed would suit me.
.
Bill: No Jerry, I was thinking about the fact that Internet Explorer beats Firefox when you’re concatenating strings using an array join.
.
Jerry: Are you a little hungry? I’m hungry.

By: Nosredna

Nosredna — Sat, 20 Sep 2008 19:35:11 +0000

>>well… IE6 beats the crap out of Firefox 2 and 3 in String concatting (if using Array + join )
.
Heh. Yeah. Maybe Microsoft should use that in their advertising.

By: Montago

Montago — Sat, 20 Sep 2008 18:24:21 +0000

well… IE6 beats the crap out of Firefox 2 and 3 in String concatting (if using Array + join )

By: pugmaster

pugmaster — Sat, 20 Sep 2008 16:12:29 +0000

I think that all benchmarks should use IE6 as their baseline, so that we can hammer home to IE6 users, how far today’s browsers have advanced. So how does this compare to IE6 performance?

By: Nosredna

Nosredna — Sat, 20 Sep 2008 16:05:55 +0000

>>Also, fast javascript is only useful to the users of that browser. For developers, you still have to deal with the slowest javascript of all your potential users, IE6 being the “modern” low end.
.
Each developer has to deal with the disparity of speed differences between browsers. You don’t want to limit what you can do just because IE6 exists.
.
What I do is write to modern browsers and then adjust for IE when running on IE.
.
For instance, I update my canvas charts only once every four frames in IE, and I turn off some other functionality.
.
I’m also planning on giving my IE6 users the ability to download a (naturally faster, since it uses WebKit) version as an Adobe AIR app.
.
To encourage IE6 users to at least grab Chrome, Opera, or Firefox (if they cannot throw out IE6 for some reason, such as needing IE6 to run corporate internet apps), IE6 users see the following message. I worked hard on it so as not to insult IE6 users, but I’m open to opinions on the wording.
.
“You are using Internet Explorer 6.0.
.
MYWEBSITENAME performs many calculations, so it works best with modern browsers such as IE7, IE8, Firefox, Opera, and Google Chrome.
.
If you have another browser, please switch to it now.
.
This version of MYWEBSITENAME does support IE6, but some operations will be very slow. Please be patient.”

By: TNO

TNO — Sat, 20 Sep 2008 13:35:12 +0000

@rizqi:
FYI more and more of the native C++ in FireFox is being moved over to JavaScript. Plus I guess you never heard of Thebes

By: rizqi

rizqi — Sat, 20 Sep 2008 06:25:11 +0000

I think the only way mozilla can catch up is to restart from scratch and rewrite all the codes. Firefox will be still slow if only the javascript engine gets updated, the rendering engine is still slower compared webkit.

By: ck2

ck2 — Sat, 20 Sep 2008 05:39:11 +0000

I wish they would list build numbers on their charts. Their results seem to contradict the info here: http://weblogs.mozillazine.org/roadmap/archives/2008/09/tracemonkey_update.html

50% faster javascript is nice but when they compare to Firefox, Firefox’s biggest problem is it’s rendering engine. Gecko crawls. Try just collapsing a 50 row table, row by row on a timer, in IE vs Safari vs Opera vs Chrome vs Firefox. Firefox loses by far, even with JIT turned on in 3.1

Also, fast javascript is only useful to the users of that browser. For developers, you still have to deal with the slowest javascript of all your potential users, IE6 being the “modern” low end.

So IE6 is the new NS4, ugh. We’ll still be throwing in code to deal with the stragglers through 2010.