These are chat archives for pixijs/pixi.js

2nd
Feb 2016
Micheal Winger
@mordof
Feb 02 2016 04:58
@ivanpopelyshev just recently got finished moving. haven't had the chance to come in for a few days.
Tamas Hortobagyi
@hortom
Feb 02 2016 09:17
Anybody tested pixi.js, that which android phone/tablet/SOC can run it in fullHD at 40-60 fps?
Dave Moore
@themoonrat
Feb 02 2016 10:03
If you have something specific to test, then upload it somewhere and I can test for you on a variety of devices. I work in html5 games so we've got a cupboard full to test on!
Tamas Hortobagyi
@hortom
Feb 02 2016 11:09
@themoonrat Thank you! I prepare some test case, after work, and upload to somewhere.
Tamas Hortobagyi
@hortom
Feb 02 2016 12:48
@themoonrat Sorry, I can't wait for it, so here is a small Pixi test, with some touch/mouse events: http://pixitest.droppages.com/ It can adopt to screen size/rotation.
Lloyd Evans
@lloydevans90
Feb 02 2016 12:57
@hortom Samsung Tab 2 with Android 4.4.2:
Native browser - runs at a wobbly 30fps with dimensions: 980x441.
Chrome - solid 60fps with dimensions 1024x479.
Tamas Hortobagyi
@hortom
Feb 02 2016 13:16
@lloydevans90 Thank you! Can you please rotate the tablet for a quick test? The performance is not the same at portrait and landscape orientation.
Lloyd Evans
@lloydevans90
Feb 02 2016 13:18
In portrait the fps of native browser dropped to average 20, and Chrome dropped to 30. Looks like the increased scale causes that.
@hortom that was after a page refresh in portrait
Tamas Hortobagyi
@hortom
Feb 02 2016 13:35
Thanks!
Tamas Hortobagyi
@hortom
Feb 02 2016 13:43
Here the pixel count is not the same in portrait and landscape, too. There is about +65% pixels in portrait.
1280x631 vs 980x1361 or 807680px vs 1333780px => 1.0 vs 1.6514. So, I'm not sure, that it is scale related. I hope this is pixel count related, but I make a version without scaling.
Tamas Hortobagyi
@hortom
Feb 02 2016 13:59
I updated to a new version, without any scaling.
Lloyd Evans
@lloydevans90
Feb 02 2016 14:50
I’d expect it from the canvas renderer, but I'm a bit surpried the webGL renderer suffers so much in Chrome (it is a crappy samsung device though). I think there are four main things to consider - canvas element css scale, canvas element width + height, PIXI renderer dimensions, and the PIXI renderer resolution.
Chad Engler
@englercj
Feb 02 2016 16:07
WebGL perf on that test may be bad because it is creating new gl buffers every frame.
take a look at gl inspector
Lloyd Evans
@lloydevans90
Feb 02 2016 16:51
Yes, the PIXI examples get solid 60fps with WebGL on this device.
I’m at over 9000 bunnies here
Tamas Hortobagyi
@hortom
Feb 02 2016 18:37
Ok. How can I avoid this ever frame creating? I'm using draw for the circles.
Ivan Popelyshev
@ivanpopelyshev
Feb 02 2016 18:55
use sprites instead.
or draw it in canvas2d and use Texture.fromCanvas(canvas) and then use as a sprite :)
Tamas Hortobagyi
@hortom
Feb 02 2016 22:33
And if I want to change the circle texture in every frame? It's will be fast enough?
Ivan Popelyshev
@ivanpopelyshev
Feb 02 2016 22:34
you can create a number of textures and then re-assign them to circles each frame
the best option is to use atlas for that
that way it will be very cheap