- The important thing you should take away from this post is Mozilla has no plans to include Glitz.
- The stats in this post has been proven useless. Better benchmark results.
All hail Opera 9.51 our new browser overlords! I ran the latest version of Opera through a SVG benchmark and was just about floored. The two things that it measures is shape creations per second and shape moves per second.
To give you a comparison here are the average results the author got for the various browsers:
|
The joys of hardware accelerated graphics. [It looks like Opera 9.51 isn't using hardware acceleration. It is something they are working on for a future release] With the release of Firefox 3.0, firefox now has the capabilities to use hardware acceleration for their SVG rendering. Please don't read that sentence wrong, as there seems to be some confusion in the community about this. The libary firefox uses to render SVG called Cairo supports hardware acceleration when coupled a backend called Glitz. But Firefox 3.0 is NOT coupled with glitz and thus has no hardware acceleration. There is an option to enable the glitz backend for cairo during the compilation of Firerfox 3.0. Don't try to build it though because as a bugzilla entry reveals it errors out. No one in the blogosphre has noticed comment #9 from Vladimir Vukicevic who states
Chances are this will become WONTFIX -- glitz won't be the way forward for us
for OpenGL/accelerated support, and what little bits of code are in there right
now don't work very well (with or without build patches).
So, the mozilla team has no plans on incorporating Glitz in the future. Does an alternative to Gltiz even exist? Does mozilla plan to write their own backend? These are currently open questions but don't hold your breath to Firefox performing as well as Opera until they make up their minds.
Update:
While Opera is shining with its SVG rendering times, Firefox 3.0 out preforms Opera on every SunSpider JavaScript benchmark minus one. The battle continues!
TEST COMPARISON FROM TO DETAILS
=============================================================================
** TOTAL **: *1.78x as slow* 3158.8ms +/- 0.6% 5631.2ms +/- 1.6% significant
=============================================================================
3d: *1.56x as slow* 362.6ms +/- 1.9% 565.4ms +/- 5.7% significant
cube: *1.55x as slow* 133.2ms +/- 1.7% 206.2ms +/- 20.6% significant
morph: *1.82x as slow* 108.4ms +/- 0.6% 196.8ms +/- 8.9% significant
raytrace: *1.34x as slow* 121.0ms +/- 3.7% 162.4ms +/- 6.4% significant
access: *1.43x as slow* 518.0ms +/- 1.1% 741.0ms +/- 3.5% significant
binary-trees: - 47.6ms +/- 6.6% 44.0ms +/- 37.2%
fannkuch: *1.40x as slow* 268.2ms +/- 0.2% 375.2ms +/- 5.2% significant
nbody: *1.44x as slow* 130.0ms +/- 3.2% 187.6ms +/- 0.4% significant
nsieve: *1.86x as slow* 72.2ms +/- 1.9% 134.2ms +/- 7.8% significant
bitops: *1.71x as slow* 365.0ms +/- 0.9% 625.0ms +/- 3.2% significant
3bit-bits-in-byte: 1.15x as fast 71.8ms +/- 1.4% 62.4ms +/- 1.1% significant
bits-in-byte: *1.32x as slow* 102.0ms +/- 0.0% 134.2ms +/- 7.8% significant
bitwise-and: *3.44x as slow* 82.6ms +/- 3.1% 284.4ms +/- 3.1% significant
nsieve-bits: *1.33x as slow* 108.6ms +/- 1.0% 144.0ms +/- 6.3% significant
controlflow: *1.41x as slow* 44.4ms +/- 4.2% 62.4ms +/- 1.1% significant
recursive: *1.41x as slow* 44.4ms +/- 4.2% 62.4ms +/- 1.1% significant
crypto: *1.38x as slow* 212.8ms +/- 1.0% 293.4ms +/- 6.9% significant
aes: *1.96x as slow* 79.8ms +/- 1.3% 156.2ms +/- 8.7% significant
md5: *1.08x as slow* 63.8ms +/- 0.9% 68.6ms +/- 15.6% significant
sha1: - 69.2ms +/- 2.0% 68.6ms +/- 15.6%
date: *3.33x as slow* 284.6ms +/- 1.4% 947.0ms +/- 1.1% significant
format-tofte: *1.66x as slow* 182.6ms +/- 1.3% 303.4ms +/- 3.6% significant
format-xparb: *6.31x as slow* 102.0ms +/- 1.9% 643.6ms +/- 1.3% significant
math: *1.20x as slow* 356.8ms +/- 2.2% 428.0ms +/- 4.4% significant
cordic: *1.21x as slow* 165.4ms +/- 2.2% 200.0ms +/- 8.4% significant
partial-sums: *1.25x as slow* 120.0ms +/- 6.0% 149.8ms +/- 7.4% significant
spectral-norm: *1.10x as slow* 71.4ms +/- 1.0% 78.2ms +/- 0.7% significant
regexp: *2.17x as slow* 234.8ms +/- 2.7% 509.4ms +/- 4.4% significant
dna: *2.17x as slow* 234.8ms +/- 2.7% 509.4ms +/- 4.4% significant
string: *1.87x as slow* 779.8ms +/- 2.2% 1459.6ms +/- 2.4% significant
base64: *1.76x as slow* 90.6ms +/- 1.8% 159.6ms +/- 10.1% significant
fasta: *1.63x as slow* 178.8ms +/- 2.9% 290.6ms +/- 3.7% significant
tagcloud: *2.16x as slow* 137.2ms +/- 3.4% 296.8ms +/- 4.7% significant
unpack-code: *2.05x as slow* 268.2ms +/- 2.8% 549.8ms +/- 4.6% significant
validate-input: *1.55x as slow* 105.0ms +/- 2.5% 162.8ms +/- 6.4% significant
Opera 9.51 benchmark url:
http://webkit.org/perf/sunspider-0.9/sunspider-results.html?%7B%223d-cube%22:%5B188,266,187,187,203%5D,%223d-morph%22:%5B219,187,187,203,188%5D,%223d-raytrace%22:%5B156,156,171,157,172%5D,%22access-binary-trees%22:%5B32,47,31,63,47%5D,%22access-fannkuch%22:%5B359,375,391,391,360%5D,%22access-nbody%22:%5B187,187,188,188,188%5D,%22access-nsieve%22:%5B141,140,140,125,125%5D,%22bitops-3bit-bits-in-byte%22:%5B63,62,62,63,62%5D,%22bitops-bits-in-byte%22:%5B140,125,140,141,125%5D,%22bitops-bitwise-and%22:%5B282,281,297,281,281%5D,%22bitops-nsieve-bits%22:%5B141,140,141,157,141%5D,%22controlflow-recursive%22:%5B62,62,62,63,63%5D,%22crypto-aes%22:%5B141,172,156,156,156%5D,%22crypto-md5%22:%5B63,62,62,78,78%5D,%22crypto-sha1%22:%5B63,62,78,78,62%5D,%22date-format-tofte%22:%5B313,297,313,297,297%5D,%22date-format-xparb%22:%5B640,656,640,641,641%5D,%22math-cordic%22:%5B187,219,187,203,204%5D,%22math-partial-sums%22:%5B156,157,156,140,140%5D,%22math-spectral-norm%22:%5B78,79,78,78,78%5D,%22regexp-dna%22:%5B516,531,516,500,484%5D,%22string-base64%22:%5B141,156,157,172,172%5D,%22string-fasta%22:%5B297,297,281,297,281%5D,%22string-tagcloud%22:%5B281,297,296,297,313%5D,%22string-unpack-code%22:%5B531,531,562,578,547%5D,%22string-validate-input%22:%5B172,172,157,156,157%5D%7D
Firefox 3.0 benchmark url:
http://webkit.org/perf/sunspider-0.9/sunspider-results.html?%7B%223d-cube%22:%5B131,133,132,135,135%5D,%223d-morph%22:%5B108,108,108,109,109%5D,%223d-raytrace%22:%5B121,120,116,126,122%5D,%22access-binary-trees%22:%5B47,46,52,46,47%5D,%22access-fannkuch%22:%5B269,268,268,268,268%5D,%22access-nbody%22:%5B131,127,130,135,127%5D,%22access-nsieve%22:%5B71,72,72,74,72%5D,%22bitops-3bit-bits-in-byte%22:%5B71,72,73,72,71%5D,%22bitops-bits-in-byte%22:%5B102,102,102,102,102%5D,%22bitops-bitwise-and%22:%5B83,82,81,86,81%5D,%22bitops-nsieve-bits%22:%5B108,108,110,109,108%5D,%22controlflow-recursive%22:%5B44,47,43,44,44%5D,%22crypto-aes%22:%5B80,79,81,79,80%5D,%22crypto-md5%22:%5B64,63,64,64,64%5D,%22crypto-sha1%22:%5B69,69,69,68,71%5D,%22date-format-tofte%22:%5B181,184,184,180,184%5D,%22date-format-xparb%22:%5B102,104,103,100,101%5D,%22math-cordic%22:%5B162,170,164,165,166%5D,%22math-partial-sums%22:%5B128,124,117,117,114%5D,%22math-spectral-norm%22:%5B72,71,72,71,71%5D,%22regexp-dna%22:%5B243,236,233,231,231%5D,%22string-base64%22:%5B90,92,92,89,90%5D,%22string-fasta%22:%5B174,184,175,181,180%5D,%22string-tagcloud%22:%5B135,142,136,133,140%5D,%22string-unpack-code%22:%5B259,276,268,270,268%5D,%22string-validate-input%22:%5B105,108,105,105,102%5D%7D
Opera rocks. :)
ReplyDeleteI'd be interested to see the comparison from 9.50 to 9.51, though.
Hi, I am the author of the simple Dojo/SVG benchmark and have a critical remark:
ReplyDeleteIn your result graphic you compared YOUR Opera 9.51 results with MY Safari / Firefox / old Opera scores.
But we use different machines.
The benchmark itself is, well, only a benchmark. So I don't really know what real-world(tm) propositions you can derive from the results anyway.
However, in order to that at least the relative results are sane, you have to compare all of the browsers on your (and only your) machine, otherwise you measure the performance difference of Opera vs. the rest and the performance difference of your CPU/RAM/OS/... vs mine.
I also placed some new results on my site, for Firefox and Safari on Leopard, they are:
Safari 3.0.4 on Leopard: ~1600 CPS, ~4900 MPS.
Firefox 2.0.0.8 on Leopard: ~800 CPS, ~3500 MPS.
Unfortunately I am not allowed to publish the latest Safari 4 developer preview results, but I can tell you they are *fast*.
Best regards
Ralf