8 replies to this topic

[TheNut]

I just heard about the release of Firefox 18.0. The biggest feature announcement is the introduction of the IonMonkey JavaScript engine, which claims to boost performance by using better optimizations in their JIT compiler. There were no specific numbers that I could find, so I did my own performance benchmarking to compare browsers on my machine. The results are as follows.

Machine
Windows 7 64bit
AMD Thuban X6 2.8GHz

Browsers
1. Firefox 16.0.2
2. Firefox 18.0
3. Chrome 23.0.1271.97

Google V8 Benchmark

Firefox 16.0.2	 Firefox 18.0	Chrome 23.0.1271.97
4614		 6911		 10570

Kraken Benchmark

-- Firefox 16.0.2 -----------------------------		-- Firefox 18.0 ----------		-- Chrome 23.0.1271.97 ---		
Total:				4423.7ms +/- 0.9%	  Total:  3014.3ms +/- 1.8%		Total:  3397.1ms +/- 4.6%
-----------------------------------------------		--------------------------		--------------------------

  ai:				 198.3ms +/- 1.2%	  202.5ms +/- 1.4%			276.3ms +/- 3.6%
	astar:			 198.3ms +/- 1.2%	  202.5ms +/- 1.4%			276.3ms +/- 3.6%

  audio:			 1595.4ms +/- 1.7%	  1266.1ms +/- 3.4%			1240.8ms +/- 7.1%
	beat-detection:		  408.6ms +/- 2.9%	  320.0ms +/- 9.1%			276.4ms +/- 8.4%
	dft:			  532.6ms +/- 0.9%	  392.9ms +/- 9.5%			507.8ms +/- 5.0%
	fft:			  315.8ms +/- 3.4%	  212.0ms +/- 11.3%			208.9ms +/- 10.4%
	oscillator:		  338.4ms +/- 2.6%	  341.2ms +/- 11.8%			208.9ms +/- 10.4%

  imaging:			 1697.9ms +/- 0.4%	  658.0ms +/- 1.2%			695.1ms +/- 6.5%
	gaussian-blur:		  924.7ms +/- 0.6%	  280.6ms +/- 1.8%			266.3ms +/- 2.5%
	darkroom:		  460.5ms +/- 1.3%	  238.6ms +/- 1.2%			258.2ms +/- 15.6%
	desaturate:		   312.7ms +/- 0.9%	  138.8ms +/- 4.9%			170.6ms +/- 5.9%

  json:				 185.9ms +/- 1.5%	  188.7ms +/- 2.1%			272.5ms +/- 10.0%
	parse-financial:	 110.0ms +/- 2.1%	  113.8ms +/- 3.1%			135.5ms +/- 5.6%
	stringify-tinderbox:	 75.9ms +/- 2.9%	   74.9ms +/- 2.6%			137.0ms +/- 20.1%

  stanford:			746.2ms +/- 1.7%	  699.0ms +/- 3.9%			912.4ms +/- 7.7%
	crypto-aes:		170.7ms +/- 2.8%	  149.5ms +/- 3.1%			254.4ms +/- 10.2%
	crypto-ccm:		123.4ms +/- 5.4%	  127.9ms +/- 1.8%			221.4ms +/- 9.3%
	crypto-pbkdf2:		334.8ms +/- 1.7%	  312.5ms +/- 8.8%			282.7ms +/- 8.8%
	crypto-sha256-iterative:  117.3ms +/- 0.7%	  109.1ms +/- 2.9%			153.9ms +/- 10.0%

It's interesting to see that Firefox now outperforms Chrome on the Kraken tests, but most importantly I think is that Firefox improved their image processing, which has been very slow for quite some time. I haven't had much time to run it through my own tests, but according to these results it should now be on-par with Chrome to manipulate RGB data on an Image object, which is useful for games that want to take advantage of grayscale tinting or generating procedural textures.

Now, if Mozilla would implement web audio in Firefox I would be a happy camper.

[Dia]

Wow, that's impressive!
Looking at the release notes, it looks like they also have preliminary support for WebRTC. Doesn't that pave the way for web audio?

[touch_the_sky]

Interesting,

If FF has some decent web audio support, then I'll look into making Space Retards available for FF as well

Also, Nut, I wanted to visit your site but my Firewall / Antivir says it's scary?

[fireside]

I had about the same thing. Chrome won on v8, Firefox won on Kraken. I can't see a noticeable difference in speed between the two.

[TheNut]

Dia said:

Wow, that's impressive!
Looking at the release notes, it looks like they also have preliminary support for WebRTC. Doesn't that pave the way for web audio?

I haven't seen anything official. There's an issue to track progress and it looks like they're making steps. They have to deprecate their audio api, which I would guess is prolonging development.

touch_the_sky said:

If FF has some decent web audio support, then I'll look into making Space Retards available for FF as well

Is that what you're using for your game now? I didn't hear any fancy effects, so I thought you were just using the simpler audio object (the music volume was low BTW). I'm looking forward to web audio though. It will have more features than DSound and OpenAL

touch_the_sky said:

Also, Nut, I wanted to visit your site but my Firewall / Antivir says it's scary?

Not sure as my site is harmless. McAfee flags my site for some reason, but coming from a company whose owner is in trouble with the law doesn't impress me much.

[touch_the_sky]

Yes I am using Webkit Web Audio for my game at the moment. Not because I want to apply a compressor real-time etc., but because I tried the standard audio thing first and had problems with performance (it was laggy). And come on, there is some fancy sounds, just exported into files;)

The music volume is low cause I thought that to get 'the right vibe' I just need to expose the arcade-style noises more, but without any music in the background it just felt a bit empty. Maybe it should be louder though? But that's something I might adjust, as it's still a w.i.p.

So the game currently uses Canvas 2d, Webkit Web Audio and Typekit.

Back to the browser performance thing:
The reason why I am recommending latest Chrome for testing my game and not Safari (also webkit) is this: I tried it on an older iMac at work in both browsers. I don't have any stats to hand, but the difference between Safari and Chrome was basically unplayable vs perfect 60Hz.
The thing you posted about FF makes everything more interesting though.

Also, McAfee owner is in trouble with the law?

[Kenneth Gorking]

Chrome 24 is out today, so how about an updated comparison?

[Stainless]

touch_the_sky, on 10 January 2013 - 10:51 PM, said:

Also, McAfee owner is in trouble with the law?

I guess you missed it.

He's been on the run for ages, eventually got cornered in Guatamala and has been extradited.

http://www.bbc.co.uk...merica-20621271

[touch_the_sky]

Haha, I did miss it!

"The 67-year old revealed that in order to go unnoticed, he changed his appearance by dying his hair and beard, sticking chewed bubble gum to his upper gums to fatten his face and staining his teeth."

What a cunning plan!