Bug 61722 (CamanJS_Perf)

Summary:	CamanJS Benchmark : WebKit is very slow compared to firefox
Product:	WebKit	Reporter:	rahmanih <haithem.rahmani>
Component:	JavaScriptCore	Assignee:	Nobody <webkit-unassigned>
Status:	RESOLVED WORKSFORME
Severity:	Normal	CC:	abecsi, barraclough, cmarcelo, dglazkov, dimich, ggaren, jamesr, jchaffraix, loki, mdelaney7, ojan, oliver, ryuan.choi, simon.fraser
Priority:	P2	Keywords:	InRadar
Version:	528+ (Nightly build)
Hardware:	All
OS:	All

rahmanih

Reported 2011-05-30 01:13:01 PDT

Hi all, I've found this blog: http://blog.meltingice.net/programming/camanjs-benchmark-firefox-4-beta-kicking-chromes-ass I run the benchmark http://dev.meltingice.net/canvas/test/benchmark.html and found the following results on a Core2 Duo 2.66 GHZ /Fedora 15 all webkit ports are really slow compared to firefox4 :( firefox4 Arora0.11.0(webkit 533.3) chrome (11.0.696.71) midori(webkitgtk-1.4.0) -------------------------------------------------------------------------------------------------------------- brightness 320ms 12570ms 3239ms 7613ms -------------------------------------------------------------------------------------------------------------- clip 352ms 23830ms 2723ms 7048ms -------------------------------------------------------------------------------------------------------------- colorize 313ms 34424ms 3461ms 6474ms -------------------------------------------------------------------------------------------------------------- contrast 937ms 29330ms 4758ms 7593ms -------------------------------------------------------------------------------------------------------------- gamma 376ms 49212ms 3920ms 6283ms -------------------------------------------------------------------------------------------------------------- greyscale 277ms 28935ms 3796ms 6150ms -------------------------------------------------------------------------------------------------------------- hue 853ms 33353ms 3675ms 6660ms -------------------------------------------------------------------------------------------------------------- invert 228ms 37749ms 2684ms 6118ms -------------------------------------------------------------------------------------------------------------- noise 247ms 24039ms 3325ms 6195ms -------------------------------------------------------------------------------------------------------------- saturation 488ms 17916ms 2943ms 6845ms -------------------------------------------------------------------------------------------------------------- sepia 244ms 31918ms 3420ms 6239ms -------------------------------------------------------------------------------------------------------------- regards. Haithem.

Attachments
Add attachment proposed patch, testcase, etc.

Gabor Loki

Comment 1 2011-05-30 06:11:24 PDT

I am not sure it is a canvas related issue. It would be nice if all canvas manipulations and accesses are disabled to make sure about the bottleneck. Maybe disabling the rendering part (putImageDate) is also enough.

Simon Fraser (smfr)

Comment 2 2011-05-30 08:20:10 PDT

<rdar://problem/9522963>

rahmanih

Comment 3 2012-04-01 02:57:28 PDT

On Windows Xp, Chrome "18.0.1025.142 m" is unable to run the benchmark at all.

Gavin Barraclough

Comment 4 2012-09-07 11:36:58 PDT

Benchmarking on my mac, we're now much faster than firefox. Chrome is slightly ahead, but results are mixed and I'm not sure we're going to want to track every single benchmark out there. I don't think we're going to take further action specifically targeting this particular benchmark, unless we were to significantly regress.

rahmanih

Comment 5 2012-09-08 04:16:27 PDT

Hi, I tests On chrome windows (20.0.1180.89) and results are much better: Filter Average brightness 60ms clip 64ms colorize 74ms contrast 876ms gamma 259ms greyscale 176ms hue 261ms invert 61ms noise 106ms saturation 292ms sepia 141ms are those improvements thanks to the DFG? regards Haithem.

Note You need to log in before you can comment on or make changes to this bug.