0xbench_2d_hw: Forked from 0xbench for android HW accelerated 2d rendering benchmark

232 views
Skip to first unread message

Jim Huang

unread,
Mar 2, 2012, 3:51:50 PM3/2/12
to 0xlab-devel
Hello list,

Since Android 4.x supports hardware accelerated 2D rendering, it is
quite important to perform corresponding measurement.

rogeryi contributed a modified version of 0xbench:
https://github.com/rogeryi/0xbench_2d_hw

The work was forked from 0xbench.

Sincerely,
Jim Huang (jserv)
http://0xlab.org/

Roger.Yi

unread,
Mar 3, 2012, 2:04:04 AM3/3/12
to 0xlab-devel
Hello Everyone,

I am Roger, this branch is focus on HW accelerated 2d rendering tests
above honeycomb and ics, the same 2d graphics test suit from original
0xbench but run on four different modes.

1, 2D(SV), rendering happens on a SurfaceView, through a SW canvas in
an individual rendering thread.
2, 2D(HW), rendering happens on a normal View inside a HW accelerated
window, through a HW canvas in UI thread.
3, 2D(SW1), rendering happens on a normal View inside a None HW
accelerated window, through a SW canvas in UI thread.
4, 2D(SW2), rendering happens on a normal View inside a HW accelerated
window, but the View has a SW layer, through a SW canvas in UI thread.

more information you can found on my blog: http://rogeryi.wordpress.com/
and this slide (include two test results, one from a galaxy tab 10.1
android 3.2, other one from a Nexus S android 4.0.3):
http://www.slideshare.net/rogeryi/android-hardware-accelerated-2d-rendering
also this one: http://www.slideshare.net/rogeryi/beyond-android-views-windowsurfacespecial-viewsand-more

Julian Chu

unread,
Mar 4, 2012, 7:41:34 PM3/4/12
to 0xlab...@googlegroups.com
Hi Roger,

I am Julian, the original author of 0xbench. (yeah, Jim Huang asked Bizkit
removing my contribute information with unknown reason and never told me,
I don't know what does he thinking.)

Since nobody in 0xlab has time to maintain 0xbench yet, I am appreciate you
fork this project to reuse this source code.

Cheers,

-Ju1ian


--
You received this message because you are subscribed to the Google Groups "0xlab-devel" group.
To post to this group, send email to 0xlab...@googlegroups.com.
To unsubscribe from this group, send email to 0xlab-devel...@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/0xlab-devel?hl=en.




--
Julian Chu.  a.k.a. WalkingIce -- 會走路的冰塊
http://walkingice.twbbs.org  只放拖鞋的鞋櫃

Jim Huang

unread,
Mar 5, 2012, 12:23:08 AM3/5/12
to 0xlab...@googlegroups.com, Roger.Yi
2012/3/3 Roger.Yi <roge...@gmail.com>:

> Hello Everyone,
>
> I am Roger, this branch is focus on HW accelerated 2d rendering tests
> above honeycomb and ics, the same 2d graphics test suit from original
> 0xbench but run on four different modes.

hi Roger,

Thanks for your work.

> 1, 2D(SV), rendering happens on a SurfaceView, through a SW canvas in
> an individual rendering thread.
> 2, 2D(HW), rendering happens on a normal View inside a HW accelerated
> window, through a HW canvas in UI thread.
> 3, 2D(SW1), rendering happens on a normal View inside a None HW
> accelerated window, through a SW canvas in UI thread.
> 4, 2D(SW2), rendering happens on a normal View inside a HW accelerated
> window, but the View has a SW layer, through a SW canvas in UI thread.

These cases seem to be what WebKit in ICS behaves. It is great to quantify.

What is the typical usage of condition (4)?

> more information you can found on my blog: http://rogeryi.wordpress.com/
> and this slide (include two test results, one from a galaxy tab 10.1
> android 3.2, other one from a Nexus S android 4.0.3):
> http://www.slideshare.net/rogeryi/android-hardware-accelerated-2d-rendering
> also this one: http://www.slideshare.net/rogeryi/beyond-android-views-windowsurfacespecial-viewsand-more

Nice materials. Thanks again for sharing.

易旭昕

unread,
Mar 5, 2012, 12:54:37 AM3/5/12
to 0xlab...@googlegroups.com, Roger.Yi
Hi Jim,

ICS WebKit port can run on both mode 2, 3, and 4, i.e. the web page rendering code path support both HW accelerated rendering and the old SW rendering. Which code path will be selected depends on the Canvas pass to WebView's onDraw method.

Mode 4 is most used to slove compatible issues, because HW canvas do not support all methods as old SW canvas, or some drawing operations are too slow. And you can use mode 4 to make some Views rendered by CPU to a off screen Bitmap then copy to window surface, and other Views can rendered by GPU directly to the HW acclerated window surface throuhg OpenGL ES.


在 2012年3月5日星期一UTC+8下午1时23分08秒,Jim Huang写道:
Reply all
Reply to author
Forward
0 new messages