Add gpu/ files from the public skia project into the chrome build. This is in...

Add gpu/ files from the public skia project into the chrome build. This is in
preparation for experimenting with them to gpu-accelerate various aspects of
chrome's drawing. In this initial patch, the files will be built, but not
called.

Now with dependency on gpu lib for sharedlib build on linux

TEST=be sure that page rendering, and <canvas> rendering, look the same as before

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=71998

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=72103

[Nico, 2011-01-19 21:07:26]

LG with nits / questions.

Want me to send this to the tryservers?

http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp
File skia/skia.gyp (right):

http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode12
skia/skia.gyp:12: '../third_party/skia/gpu/src/GrAllocPool.cpp',
indent is too big

http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode651
skia/skia.gyp:651: '../gpu',
Why does skia need to depend on chrome's gpu directory?

http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode764
skia/skia.gyp:764: '../third_party/skia/gpu/include',
indent too big

[reed1, 2011-01-19 21:12:28]

Sorry about the indents -- VS tabs I think

We need the ../gpu in the search path so I can see the "correct" versions of
the gl headers. I'll investigate to see if that can be avoid through some
other trick.

Yes, tryserver would be great, thanks. -- is there a way I can fire one of
those?

On Wed, Jan 19, 2011 at 4:07 PM, <thakis@chromium.org> wrote:

> LG with nits / questions.
>
> Want me to send this to the tryservers?
>
>
> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp
> File skia/skia.gyp (right):
>
> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode12
> skia/skia.gyp:12: '../third_party/skia/gpu/src/GrAllocPool.cpp',
> indent is too big
>
> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode651
> skia/skia.gyp:651: '../gpu',
> Why does skia need to depend on chrome's gpu directory?
>
> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode764
> skia/skia.gyp:764: '../third_party/skia/gpu/include',
> indent too big
>
>
> http://codereview.chromium.org/6357006/
>

[Nico, 2011-01-19 21:24:28]

On Wed, Jan 19, 2011 at 1:12 PM, Mike Reed <reed@google.com> wrote:
> Sorry about the indents -- VS tabs I think
> We need the ../gpu in the search path so I can see the "correct" versions of
> the gl headers. I'll investigate to see if that can be avoid through some
> other trick.

Will this mostly live in the renderer and use the GPU IPC bridge to
execute GL commands? What if someone in the browser tries to use these
skia apis? Did you talk to beng, who's trying to get
hardware-accelerated drawing in the browser process up?

> Yes, tryserver would be great, thanks. -- is there a way I can fire one of
> those?

If you are a chromium committer and have a chrome checkout, run `gcl
try mychangelistname` in the src directory. If it asks, your
subversion password is at http://go/svn-pw

>
> On Wed, Jan 19, 2011 at 4:07 PM, <thakis@chromium.org> wrote:
>>
>> LG with nits / questions.
>>
>> Want me to send this to the tryservers?
>>
>>
>> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp
>> File skia/skia.gyp (right):
>>
>> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode12
>> skia/skia.gyp:12: '../third_party/skia/gpu/src/GrAllocPool.cpp',
>> indent is too big
>>
>> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode651
>> skia/skia.gyp:651: '../gpu',
>> Why does skia need to depend on chrome's gpu directory?
>>
>> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode764
>> skia/skia.gyp:764: '../third_party/skia/gpu/include',
>> indent too big
>>
>> http://codereview.chromium.org/6357006/
>
>

[reed1, 2011-01-20 01:54:22]

On 2011/01/19 21:24:28, Nico wrote:
> On Wed, Jan 19, 2011 at 1:12 PM, Mike Reed <mailto:reed@google.com> wrote:
> > Sorry about the indents -- VS tabs I think
> > We need the ../gpu in the search path so I can see the "correct" versions of
> > the gl headers. I'll investigate to see if that can be avoid through some
> > other trick.
> 
> Will this mostly live in the renderer and use the GPU IPC bridge to
> execute GL commands? What if someone in the browser tries to use these
> skia apis? Did you talk to beng, who's trying to get
> hardware-accelerated drawing in the browser process up?

The library can run in any process that has created an opengl context. Our
current experiment runs in the same thread as webcore renderer.

> 
> > Yes, tryserver would be great, thanks. -- is there a way I can fire one of
> > those?
> 
> If you are a chromium committer and have a chrome checkout, run `gcl
> try mychangelistname` in the src directory. If it asks, your
> subversion password is at http://go/svn-pw
> 
> >
> > On Wed, Jan 19, 2011 at 4:07 PM, <mailto:thakis@chromium.org> wrote:
> >>
> >> LG with nits / questions.
> >>
> >> Want me to send this to the tryservers?
> >>
> >>
> >> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp
> >> File skia/skia.gyp (right):
> >>
> >> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode12
> >> skia/skia.gyp:12: '../third_party/skia/gpu/src/GrAllocPool.cpp',
> >> indent is too big
> >>
> >> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode651
> >> skia/skia.gyp:651: '../gpu',
> >> Why does skia need to depend on chrome's gpu directory?
> >>
> >> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode764
> >> skia/skia.gyp:764: '../third_party/skia/gpu/include',
> >> indent too big
> >>
> >> http://codereview.chromium.org/6357006/
> >
> >

[Nico, 2011-01-20 02:02:21]

LG with green tryserver runs.

On Wed, Jan 19, 2011 at 5:54 PM,  <reed@google.com> wrote:
> On 2011/01/19 21:24:28, Nico wrote:
>>
>> On Wed, Jan 19, 2011 at 1:12 PM, Mike Reed <mailto:reed@google.com> wrote:
>> > Sorry about the indents -- VS tabs I think
>> > We need the ../gpu in the search path so I can see the "correct"
>> > versions of
>> > the gl headers. I'll investigate to see if that can be avoid through
>> > some
>> > other trick.
>
>> Will this mostly live in the renderer and use the GPU IPC bridge to
>> execute GL commands? What if someone in the browser tries to use these
>> skia apis? Did you talk to beng, who's trying to get
>> hardware-accelerated drawing in the browser process up?
>
> The library can run in any process that has created an opengl context. Our
> current experiment runs in the same thread as webcore renderer.

Hm, I think we no longer have a GL context in the renderer side, it
all goes through webkit's GraphicsContext3D. Maybe the GLES20
functions are still around, but they're no longer really used by
anything. Well, maybe they still work.

For the browser side, beng created a canvas interface that has skia
and D3D implementations. We can't use OpenGL directly on windows
because many boxes don't have opengl drivers installed iirc.

>
>
>> > Yes, tryserver would be great, thanks. -- is there a way I can fire one
>> > of
>> > those?
>
>> If you are a chromium committer and have a chrome checkout, run `gcl
>> try mychangelistname` in the src directory. If it asks, your
>> subversion password is at http://go/svn-pw
>
>> >
>> > On Wed, Jan 19, 2011 at 4:07 PM, <mailto:thakis@chromium.org> wrote:
>> >>
>> >> LG with nits / questions.
>> >>
>> >> Want me to send this to the tryservers?
>> >>
>> >>
>> >> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp
>> >> File skia/skia.gyp (right):
>> >>
>> >> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode12
>> >> skia/skia.gyp:12: '../third_party/skia/gpu/src/GrAllocPool.cpp',
>> >> indent is too big
>> >>
>> >> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode651
>> >> skia/skia.gyp:651: '../gpu',
>> >> Why does skia need to depend on chrome's gpu directory?
>> >>
>> >> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode764
>> >> skia/skia.gyp:764: '../third_party/skia/gpu/include',
>> >> indent too big
>> >>
>> >> http://codereview.chromium.org/6357006/
>> >
>> >
>
>
>
> http://codereview.chromium.org/6357006/
>

[reed1, 2011-01-20 02:14:46]

The renderer has a synthetic context, directing all of the opengl calls into
the command-buffer. It is working today (with another patch) for the
<canvas> element.

On Wed, Jan 19, 2011 at 9:02 PM, Nico Weber <thakis@chromium.org> wrote:

> LG with green tryserver runs.
>
> On Wed, Jan 19, 2011 at 5:54 PM,  <reed@google.com> wrote:
> > On 2011/01/19 21:24:28, Nico wrote:
> >>
> >> On Wed, Jan 19, 2011 at 1:12 PM, Mike Reed <mailto:reed@google.com>
> wrote:
> >> > Sorry about the indents -- VS tabs I think
> >> > We need the ../gpu in the search path so I can see the "correct"
> >> > versions of
> >> > the gl headers. I'll investigate to see if that can be avoid through
> >> > some
> >> > other trick.
> >
> >> Will this mostly live in the renderer and use the GPU IPC bridge to
> >> execute GL commands? What if someone in the browser tries to use these
> >> skia apis? Did you talk to beng, who's trying to get
> >> hardware-accelerated drawing in the browser process up?
> >
> > The library can run in any process that has created an opengl context.
> Our
> > current experiment runs in the same thread as webcore renderer.
>
> Hm, I think we no longer have a GL context in the renderer side, it
> all goes through webkit's GraphicsContext3D. Maybe the GLES20
> functions are still around, but they're no longer really used by
> anything. Well, maybe they still work.
>
> For the browser side, beng created a canvas interface that has skia
> and D3D implementations. We can't use OpenGL directly on windows
> because many boxes don't have opengl drivers installed iirc.
>
> >
> >
> >> > Yes, tryserver would be great, thanks. -- is there a way I can fire
> one
> >> > of
> >> > those?
> >
> >> If you are a chromium committer and have a chrome checkout, run `gcl
> >> try mychangelistname` in the src directory. If it asks, your
> >> subversion password is at http://go/svn-pw
> >
> >> >
> >> > On Wed, Jan 19, 2011 at 4:07 PM, <mailto:thakis@chromium.org> wrote:
> >> >>
> >> >> LG with nits / questions.
> >> >>
> >> >> Want me to send this to the tryservers?
> >> >>
> >> >>
> >> >> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp
> >> >> File skia/skia.gyp (right):
> >> >>
> >> >>
> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode12
> >> >> skia/skia.gyp:12: '../third_party/skia/gpu/src/GrAllocPool.cpp',
> >> >> indent is too big
> >> >>
> >> >>
> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode651
> >> >> skia/skia.gyp:651: '../gpu',
> >> >> Why does skia need to depend on chrome's gpu directory?
> >> >>
> >> >>
> http://codereview.chromium.org/6357006/diff/1/skia/skia.gyp#newcode764
> >> >> skia/skia.gyp:764: '../third_party/skia/gpu/include',
> >> >> indent too big
> >> >>
> >> >> http://codereview.chromium.org/6357006/
> >> >
> >> >
> >
> >
> >
> > http://codereview.chromium.org/6357006/
> >
>

[reed1, 2011-01-20 02:23:06]

patch 4 updates the skia version to 716, which has a runtime fix found testing
on windows.

[Nico, 2011-01-20 03:02:48]

SLG

[commit-bot: I haz the power, 2011-01-20 18:57:56]

No LGTM from reviewers yet.

[reed1, 2011-01-20 18:58:34]

layout failures do not appear to be related to this CL

[Stephen White, 2011-01-20 19:05:08]

LGTM

[Nico, 2011-01-20 19:12:26]

LGTM

[commit-bot: I haz the power, 2011-01-20 20:01:00]

Commit queue had an internal error

[M-A Ruel, 2011-01-20 20:03:17]

On 2011/01/20 20:01:00, commit-bot wrote:
> Commit queue had an internal error

Dang, small error on my side, reseting the bit.

[Stephen White, 2011-01-20 20:05:55]

On 2011/01/20 20:03:17, Marc-Antoine Ruel wrote:
> On 2011/01/20 20:01:00, commit-bot wrote:
> > Commit queue had an internal error
> 
> Dang, small error on my side, reseting the bit.

I committed this one by hand, and it broke the shlib build (so I had  to revert
it).  Please don't commit it for now.

[commit-bot: I haz the power, 2011-01-20 20:56:57]

Change committed as 71998

[M-A Ruel, 2011-01-20 20:57:55]

On 2011/01/20 20:56:57, commit-bot wrote:
> Change committed as 71998

OMG