[NOT FOR COMMIT] Hacks to get hardware canvas

[NOT FOR COMMIT] Hacks to get hardware canvas

These will break definitely break other platforms, and possibly break
software canvas.

BUG=239864

[changjun.yang, 2014-01-06 09:35:03]

Hi Bo,
I am also working on a CL to enable the Canvas2D hardware acceleration:
https://codereview.chromium.org/110583008
https://codereview.chromium.org/101163004
The method has something in common with your CL while it might resolve the
software canvas breaking issue.
Please help to review and comment.
Thanks!

[boliu, 2014-01-06 18:20:31]

On 2014/01/06 09:35:03, changjun.yang wrote:
> Hi Bo,
> I am also working on a CL to enable the Canvas2D hardware acceleration:
> https://codereview.chromium.org/110583008
> https://codereview.chromium.org/101163004
> The method has something in common with your CL while it might resolve the
> software canvas breaking issue.
> Please help to review and comment.
> Thanks!

Wow I didn't realize there are intel folks working on this too. Thanks for the
interest.

I left some comments in other patches, they are mostly doing the same thing as
here, which are just hacks not addressing the underlying issues. Also I expect
that you might run into subtle deadlock issues:
https://code.google.com/p/chromium/codesearch#chromium/src/android_webview/br...

There is another issue that is not addressed here. We don't want to run
canvas/webgl on the ui thread run by the gl functor, for security and
performance reasons. We hinted that in the bug but haven't done any work here
yet.

Sorry that you are still testing on m30. I'm poking internal people who are
working on supporting building chromium trunk in aosp, but for now it's still
not possible. I also realize webview doesn't have a lot of documentation, but
for now, if any of the things I said doesn't make sense, please just ask.

And lastly, I locked crbug.com/239864 for comments so random app developers
don't come with me too comments, but I think I can add you or any other intel
folks to the cc list so you can comment. Just give me a list.

[changjun.yang, 2014-01-08 09:24:27]

On 2014/01/06 18:20:31, boliu wrote:
> On 2014/01/06 09:35:03, changjun.yang wrote:
> > Hi Bo,
> > I am also working on a CL to enable the Canvas2D hardware acceleration:
> > https://codereview.chromium.org/110583008
> > https://codereview.chromium.org/101163004
> > The method has something in common with your CL while it might resolve the
> > software canvas breaking issue.
> > Please help to review and comment.
> > Thanks!
> 
> Wow I didn't realize there are intel folks working on this too. Thanks for the
> interest.
> 
> I left some comments in other patches, they are mostly doing the same thing as
> here, which are just hacks not addressing the underlying issues. Also I expect
> that you might run into subtle deadlock issues:
>
https://code.google.com/p/chromium/codesearch#chromium/src/android_webview/br...
> 
> There is another issue that is not addressed here. We don't want to run
> canvas/webgl on the ui thread run by the gl functor, for security and
> performance reasons. We hinted that in the bug but haven't done any work here
> yet.
> 

We also have interests in separating the ui thread, do you already have plans or
could we take some efforts for this?

> Sorry that you are still testing on m30. I'm poking internal people who are
> working on supporting building chromium trunk in aosp, but for now it's still
> not possible. I also realize webview doesn't have a lot of documentation, but
> for now, if any of the things I said doesn't make sense, please just ask.
> 

Do you have estimated schedule for this issue or is there any workaround for
testing before it is done?

> And lastly, I locked crbug.com/239864 for comments so random app developers
> don't come with me too comments, but I think I can add you or any other intel
> folks to the cc list so you can comment. Just give me a list.

Could you please add: changjun.yang#intel.com, jin.a.yang#intel.com,
kangyuan.shu#intel.com for the comment authority?
Thanks a lot!