>>> SunSpider JS-benchark
>>> [...]
>>> - OWB: 4.1s

>> TenFourFox 9.0b1: 2.1s

> TenFourFox 9.0 (MethodJIT enabled): 1.9s

TenFourFox 10.0.2pre: 1.8s

> TenFourFox 10.0b1:
> 1104 points (rendering: 487, social networking: 722, complex
> graphics: 1764, data: 2364, DOM: 1160, text parsing: 1704)

TenFourFox 10.0.2pre:
1111 points (rendering: 465, social networking: 686, complex graphics: 1751, data: 2111, DOM: 1071, text parsing: 2349)

Just for the sake of having a complete set of results:

Powerbook 5,6 1.67GHz, running in OSX 'maximised' view (on 1280x854 screen):

Firefox 3.6.25 (standard PPC build)
Sunspider 0.91 - 6925.6ms +/- 1.3%
Peacekeeper - 196 points (rendering 9.82, HTML5 3/7, HTML5 canvas 1.71, data 3093.16, DOM operations 1037.25, Text parsing 5334.29)
Tests failed for WebGL, H264, WebM video, contrast 01 & 02.

Namoroka (Firefox 3.6.25 compiled for 7450 cpu),
Sunspider 0.91 - 7322.2ms +/- 0.8%
Peacekeeper - 191 points, (Rendering 10.08, HTML5 3/7, HTML5 canvas 1.69, Data 3010.77, DOM operations 1010.06, Text parsing 4888.30)
Tests failed for WebGL, H264, WebM video, contrast 01 & 02.

TenFourFox7450 (9.0)
Sunspider (methodJIT disabled) 0.91 - 2102.3ms +/- 1.8%
Sunspider (methodJIT enabled) 0.91 - 1930.3ms +/- 6.8%
Peacekeeper (methodJIT disabled FWIW) - 296 points, (Rendering 7.28, HTML5 5/7, HTML5 canvas 2.73, Data 4442.62, DOM operations 1952.92, Text parsing 13093.94)
Tests failed for WebGL, H264.
'Unresponsive script' warning thrown up during contrast 01 test.

Tenfourfox7450 (10.02pre)
Sunspider (methodJIT disabled) 0.91 - 8233.5ms +/- 0.6%
Sunspider (methodJIT enabled) 0.91 - 2008.3ms +/- 2.1%
Peacekeeper (methodJIT disabled FWIW) - 315 points, (rendering 7.44, HTML5 5/7, HTML canvas 3.15, Data 3262.39, DOM operations 1905.67, Text parsing 21201.80)
Tests failed for WebGL, H264.
'Unresponsive script' warning thrown up during contrast 01 & contrast 02 tests.

Conclusions:
Namoroka(7450) offers negligible difference to the standard 3.6.25 OSX/PPC build.
Tenfourfox7450 excels in most tests, and offers greater codec support.
The non-methodJIT fallback for 10.02 sucks. Bigtime. What have they (or I) done differently compared to 9.0?

[ Edited by boot_wb 15.02.2012 - 14:17 ]

> Powerbook 5,6 1.67GHz
> [...]
> TenFourFox7450 (9.0)
> Sunspider 0.91 - 2102.3ms

I wonder why you get only 2.1s on your 1.67 GHz machine when I get down to 1.9s on my 1.5 GHz machine. Do you have MethodJIT enabled?

@Andreas_Wolf

I've just checked 9.0, and these were set to false, so will update the sunspider scores accordingly in the post above.

I've also checked 10.02pre, and it seems that both
javascript.options.methodjit.chrome and javascript.options.methodjit.content
are set to 'True' by default.

I'll re-run the test with methodJIT disabled in 10.02 to get a comparison, and update the above post accordingly.


Rich

Hmm...

wrt to TenFourFox 10.02pre:
After setting the two methodJIT options to False, and getting such poor results, I thought I'd try enabling the analogous traceJIT options (to match the default config in 9.0).

No significant changes compared to just disabling methodJIT.

I then reset all four options back to default (methodJIT enabled, traceJIT disabled) and re-ran the Sunspider benchmark.

The result this time was 1747.2ms +/- 4.4%

This would be a better fit compared to your ~1900ms at 1.5GHz (being around 10% faster) assuming JS speed is largely dependent on cpu speed rather than memory access.

Whether toggling the options has had some effect, or whether something_random was taking up CPU cycles the first time around remains a mystery. No other options were changed, and there is nothing (of which I'm aware) running in the background which would have affected the results.

> After setting the two methodJIT options to False, and getting
> such poor results, I thought I'd try enabling the analogous
> traceJIT options (to match the default config in 9.0). No
> significant changes compared to just disabling methodJIT.

I think they removed traceJIT with TFF10, so it falls back to non-JIT with methodJIT disabled.

> I then reset all four options back to default (methodJIT enabled,
> traceJIT disabled) and re-ran the Sunspider benchmark. The result
> this time was 1747.2ms +/- 4.4% This would be a better fit compared
> to your ~1900ms at 1.5GHz

Yes, absolutely. But as I wrote it's 1.8s with TFF10. 1.9s was with TFF9 :-)

Ähm... now this may seem to be a totally daft question, but how does all this (the TenFourFox benchmarks and stuff) affect my surfing with OWB under MorphOS?

@Megander

Not sure, but it is on topic at least.

From the OP:
Quote:

And there's no way of bringing this PPC Altivec JIT Java thingy to MOS to help let's say improve OWB's speed?

> how does all this (the TenFourFox benchmarks and stuff) affect my surfing
> with OWB under MorphOS?

Apart from the fact that according to the thread opener, benchmarks of other browsers in comparison to OWB's performance is exactly what this thread is about (as boot_wb has said), the TenFourFox benchmarks also show what is possible on the very hardware MorphOS currently runs on and thus could be possible on MorphOS as well if someone ported the required implementations of the technologies which speed up TenFourFox the way they do.

> there's no way of bringing this PPC Altivec JIT Java thingy to MOS to help
> let's say improve OWB's speed?

You're mixing different things here. Basically, TenFourFox incorporates two approaches at leveraging PPC CPUs in a proper way. One approach is speeding up JavaScript code execution by means of a JIT compiler that directly compiles JavaScript code into PPC machine code. The second one is making as much use of AltiVec code as possible (but not on G3 obviously) and thus accelerating all kinds of operations where SIMD technology provides for an advantage. Those two approaches are independent of each other.
Bringing TenFourFox' AltiVec code to OWB shouldn't be too hard, at least in my layman's opinion. As for its JavaScript JIT, things look differently. Some comments on this:

https://morph.zone/modules/newbb_plus/viewtopic.php?forum=9&topic_id=7524&start=65
https://morph.zone/modules/newbb_plus/viewtopic.php?forum=32&topic_id=7103&start=10

I downloaded and installed TenFourFox on my 1.67GHz G4 PowerBook running MacOSX Leopard updated to the last 10.5.8 version, but have not noticed any difference in speed or performance between it and Firefox3.6.26.

Do you have any tips on setting up, or configuring TenFourFox that will show it's advantages over Firefox to me more clearly Andreas? I have also downloaded MacTubes, but have not installed it yet, since I have opted "IN" to YouTube's html5 optional program, I was not sure MacTubes was needed, or useful anymore.

> Do you have any tips on setting up, or configuring TenFourFox that
> will show it's advantages over Firefox to me more clearly Andreas?

Unfortunately not. Make sure to download the right and most current version, i.e. TenFourFox7450-10.0.2pre. On JavaScript-heavy pages you should notice an increased performance over Firefox. Also with decoding of WebM videos for instance.

> TenFourFox 10.0.2pre:
> 1111 points (rendering: 465, social networking: 686, complex
> graphics: 1751, data: 2111, DOM: 1071, text parsing: 2349)

TenFourFox 10.0.2:
1158 points (rendering: 479, social networking: 724, complex graphics: 1780, data: 2172, DOM: 1136, text parsing: 2441)

>>>> SunSpider JS-benchark
>>>> [...]
>>>> - OWB: 4.1s

>>> TenFourFox 9.0b1: 2.1s

>> TenFourFox 9.0 (MethodJIT enabled): 1.9s

> TenFourFox 10.0.2pre: 1.8s

TenFourFox 10.0.3pre: 1.7s

> TenFourFox 10.0.2:
> 1158 points (rendering: 479, social networking: 724, complex
> graphics: 1780, data: 2172, DOM: 1136, text parsing: 2441)

TenFourFox 10.0.3pre:
1206 points (rendering: 481, social networking: 740, complex graphics: 1778, data: 2446, DOM: 1169, text parsing: 2513)

For those who want better OWB performance:

HARDWARE ACCELERATED CAIRO BOUNTY

@magnetic

Quote:

As long as it doesn't end up as ARexx bounty I would contribute.

Try to find someone interested in doing that and I mean doing not just saying that will try and we on morphos.pl will be more than happy to setup that bounty.

>>>>> SunSpider JS-benchark
>>>>> [...]
>>>>> - OWB: 4.1s

>>>> TenFourFox 9.0b1: 2.1s

>>> TenFourFox 9.0 (MethodJIT enabled): 1.9s

>> TenFourFox 10.0.2pre: 1.8s

> TenFourFox 10.0.3pre: 1.7s

TenFourFox 12.0: 1.6s

> http://peacekeeper.futuremark.com
> [...]
> TenFourFox 8.0b1:
> 374 points (HTML5 capabilities: 4/6, rendering: 6.5,
> HTML5 canvas: 2.7, data: 13306, DOM operations: 1777,
> text parsing: 17741)

AuroraFox/10.5Fx 14.0a2:
391 points (HTML5 capabilities: 4/7, rendering: 7.9, HTML5 canvas: 4.4, data: 6053, DOM operations: 2229, text parsing: 19335)

>>> http://peacekeeper.futuremark.com

>> OWB 1.15:
>> 241 points (HTML5 capabilities: 4/6, rendering: 9.8, HTML5 canvas: 1.6,
>> data: 3316, DOM operations: 2067, text parsing: 7407)

> OWB 1.16:
> 219 points (rendering: 8.6, HTML5 capabilities: 5/7, HTML5 canvas: 1.5,
> data: 3167, DOM operations: 1718, text parsing: 7370)

OWB 1.17:
285 points (rendering: 10.2, HTML5 capabilities: 5/7, HTML5 canvas: 3.1, data: 3901, DOM operations: 1808, text parsing: 8562)

Nice speedup :-)

>>>>>> SunSpider JS-benchark
>>>>>> [...]
>>>>>> - OWB: 4.1s

>>>>> TenFourFox 9.0b1: 2.1s

>>>> TenFourFox 9.0 (MethodJIT enabled): 1.9s

>>> TenFourFox 10.0.2pre: 1.8s

>> TenFourFox 10.0.3pre: 1.7s

> TenFourFox 12.0: 1.6s

AuroraFox/10.5Fx 16.0a2: 1.4s

>>>>>>> SunSpider JS-benchark
>>>>>>> [...]
>>>>>>> - OWB: 4.1s

>>>>>> TenFourFox 9.0b1: 2.1s

>>>>> TenFourFox 9.0 (MethodJIT enabled): 1.9s

>>>> TenFourFox 10.0.2pre: 1.8s

>>> TenFourFox 10.0.3pre: 1.7s

>> TenFourFox 12.0: 1.6s

> AuroraFox/10.5Fx 16.0a2: 1.4s

AuroraFox/10.5Fx 18.0a2: 1.3s