User Agent: Mozilla/5.0 (X11; Linux x86_64; rv:11.0a1) Gecko/20111117 Firefox/11.0a1 Build ID: 20111117030939 Steps to reproduce: Drawing a line with a lot of points is terribly slow. More over, time for drawing line between two points increase while increasing points in chart. Actual results: Example: http://jsperf.com/stroke-n-points/2 ops (from jsperf) — line redraws per second; t = (1/ops * 10e6)/dots — how long to draw line between interconnected points (in us); On my machine (Linux, 64bit): Mozilla/5.0 (X11; Linux x86_64; rv:11.0a1) Gecko/20111116 Firefox/11.0a1 250 dots: ops=141 t=28 us 500 dots: ops=82.42 t=24 us 750 dots: ops=49.57 t=27 us 1000 dots: ops=32.44 t=31 us 2000 dots: ops=10.58 t=47 us 3000 dots: ops=5.76 t=58 us Opera 11.51 250 dots: ops=204 t=20 us 500 dots: ops=131 t=15 us 750 dots: ops=104 t=13 us 1000 dots: ops=86.16 t=12 us 2000 dots: ops=39.63 t=13 us 3000 dots: ops=28.68 t=12 us On Windows machine a bit different results, but increasing too: 250 dots: t=80 us 500 dots: t=92 us 750 dots: t=101 us 1000 dots: t=103 us 2000 dots: t=106 us 3000 dots: t=105 us Expected results: The value `t' shouldn't grow while increase `dots'.
On Mac, I get these values: 250 dots: ops=865 t=4.6 us 500 dots: ops=521 t=3.8 us 750 dots: ops=395 t=3.4 us 1000 dots: ops=315 t=3.2 us 2000 dots: ops=174 t=2.9 us 3000 dots: ops=121 t=2.8 us So this looks like yet another cairo path issue...
Summary: stroke degradate on many lineTo()'s → Canvas stroke performance is super-linear in number of points in the path
Whiteboard: [cairo path stuff]
Ok, I add related bug in cairo bugtracker: https://bugs.freedesktop.org/show_bug.cgi?id=43064
Cairo quite stable on the range 1000..3000 points. Firefox on that range slowed down to 2 times.
#568526 looks like related bug.
You need to log in before you can comment on or make changes to this bug.