Refactor tracking of bound vbufs and ibufs and vertex attrib arrays.

Refactor tracking of bound vbufs and ibufs and vertex attrib arrays.
Committed: https://code.google.com/p/skia/source/detail?r=7779

[bsalomon, 2013-02-19 19:09:14]

This adds a class that tracks bound VBOs and attrib array state. It's more
generic than the old code and it emits fewer unnecessary VBO binds and attrib
array calls. It also sets the stage for GrGL*Buffers that are backed by a
CPU-side array rather than a VBO.

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGLIndexBuffer.h
File src/gpu/gl/GrGLIndexBuffer.h (right):

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGLIndexBuffer.h...
src/gpu/gl/GrGLIndexBuffer.h:31: GrGLvoid* baseOffset() const { return 0; }
This function is added so that in a forthcoming CL it can return the CPU address
when we aren't really using a VBO.

[robertphillips, 2013-02-19 19:54:42]

LGTM + nit & question

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGpuGL.cpp
File src/gpu/gl/GrGpuGL.cpp (right):

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGpuGL.cpp#newco...
src/gpu/gl/GrGpuGL.cpp:2332: GrAssert(!vbuf->isLocked());
Could baseOffset return a size_t directly?

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGpuGL.h
File src/gpu/gl/GrGpuGL.h (right):

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGpuGL.h#newcode124
src/gpu/gl/GrGpuGL.h:124: // an into the index buffer. It does not account for
drawInfo.startIndex() but rather the start
retuned -> returned

[bsalomon, 2013-02-19 19:59:08]

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGpuGL.cpp
File src/gpu/gl/GrGpuGL.cpp (right):

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGpuGL.cpp#newco...
src/gpu/gl/GrGpuGL.cpp:2332: GrAssert(!vbuf->isLocked());
On 2013/02/19 19:54:42, robertphillips wrote:
> Could baseOffset return a size_t directly?

yeah... not a bad idea. will do.

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGpuGL.h
File src/gpu/gl/GrGpuGL.h (right):

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGpuGL.h#newcode124
src/gpu/gl/GrGpuGL.h:124: // an into the index buffer. It does not account for
drawInfo.startIndex() but rather the start
On 2013/02/19 19:54:42, robertphillips wrote:
> retuned -> returned

will do.

[JimVV, 2013-02-19 20:15:06]

LGTM + an FYI

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGLProgram.h
File src/gpu/gl/GrGLProgram.h (right):

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGLProgram.h#new...
src/gpu/gl/GrGLProgram.h:84: enum {
My plan is to replace these by the index of the corresponding attribute in the
currently active vertex attribute array. So they may become functions again.

[bsalomon, 2013-02-19 20:17:28]

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGLProgram.h
File src/gpu/gl/GrGLProgram.h (right):

https://codereview.appspot.com/7359045/diff/2001/src/gpu/gl/GrGLProgram.h#new...
src/gpu/gl/GrGLProgram.h:84: enum {
On 2013/02/19 20:15:06, JimVV wrote:
> My plan is to replace these by the index of the corresponding attribute in the
> currently active vertex attribute array. So they may become functions again.

Ok. Hopefully all but kPosition and kTexCoord (or their function equivalents)
can eventually be eliminated.