Added support for EXT_frag_depth

Added support for EXT_frag_depth

This change also required that support be added for associating built-in
variables with an extension, similar to how functions could be associated with
extensions previously.

R=alokp@chromium.org

Committed: https://code.google.com/p/angleproject/source/detail?r=2248

[Shannon Woods, 2013-05-28 18:31:08]

Adding Nicolas as a reviewer.

[nicolas%transgaming.com, 2013-05-28 19:16:32]

Please also note that soon things will switch over to the dx11proto branch (but
using only the DX9 path for a while).

Anyway, looks pretty good!

[nicolas%transgaming.com, 2013-05-28 19:17:13]

Meant to publish these comments at the same time...

https://codereview.appspot.com/9827044/diff/1/src/compiler/Initialize.cpp
File src/compiler/Initialize.cpp (right):

https://codereview.appspot.com/9827044/diff/1/src/compiler/Initialize.cpp#new...
src/compiler/Initialize.cpp:543: symbolTable.insert(*new
TVariable(NewPoolTString("gl_FragDepthEXT"),               TType(EbtFloat,
EbpMedium, EvqFragDepth,   1)));
This should be EbpHight if resources.FragmentPrecisionHigh is set.

https://codereview.appspot.com/9827044/diff/1/src/compiler/glslang_tab.cpp
File src/compiler/glslang_tab.cpp (right):

https://codereview.appspot.com/9827044/diff/1/src/compiler/glslang_tab.cpp#ne...
src/compiler/glslang_tab.cpp:1: /* A Bison parser, made by GNU Bison 2.5.  */
Any chance you could regenerate these with Bison 2.7? Cygwin currently uses that
version.

[Alok Priyadarshi, 2013-05-28 19:25:35]

https://codereview.appspot.com/9827044/diff/1/src/compiler/OutputGLSL.cpp
File src/compiler/OutputGLSL.cpp (right):

https://codereview.appspot.com/9827044/diff/1/src/compiler/OutputGLSL.cpp#new...
src/compiler/OutputGLSL.cpp:26: TString name = node->getSymbol();
there is no need for this local variable "name".

https://codereview.appspot.com/9827044/diff/1/src/compiler/ParseHelper.cpp
File src/compiler/ParseHelper.cpp (right):

https://codereview.appspot.com/9827044/diff/1/src/compiler/ParseHelper.cpp#ne...
src/compiler/ParseHelper.cpp:934: if (symbol == 0) {
style nit: We do not use scopes { } for single statements.

[Brandon Jones, 2013-05-28 19:30:41]

I'm already working on ensuring these changes work with the dx11 branch. I
assume I can't include that branches changes in this CL?

https://codereview.appspot.com/9827044/diff/1/src/compiler/Initialize.cpp
File src/compiler/Initialize.cpp (right):

https://codereview.appspot.com/9827044/diff/1/src/compiler/Initialize.cpp#new...
src/compiler/Initialize.cpp:543: symbolTable.insert(*new
TVariable(NewPoolTString("gl_FragDepthEXT"),               TType(EbtFloat,
EbpMedium, EvqFragDepth,   1)));
On 2013/05/28 19:17:14, nicolas%transgaming.com wrote:
> This should be EbpHight if resources.FragmentPrecisionHigh is set.

I'm curious if this is something that should apply to the other outputs as well?
It seems like the gl_FragColor, gl_FragData, etc. should be treated the same
way.

[Shannon Woods, 2013-05-28 19:50:37]

On 2013/05/28 19:25:35, Alok Priyadarshi wrote:

>
https://codereview.appspot.com/9827044/diff/1/src/compiler/ParseHelper.cpp#ne...
> src/compiler/ParseHelper.cpp:934: if (symbol == 0) {
> style nit: We do not use scopes { } for single statements.

On style: please note that the ANGLE coding standard diverges in places from
Chrome's, and does not require single-line conditionals be unbraced (although it
inherits that allowance from the Google C++ Style Guide).

[nicolas%transgaming.com, 2013-05-28 20:50:38]

On 2013/05/28 19:30:41, Brandon Jones wrote:
> On 2013/05/28 19:17:14, http://nicolas%25transgaming.com wrote:
> > This should be EbpHight if resources.FragmentPrecisionHigh is set.
> 
> I'm curious if this is something that should apply to the other outputs as
well?
> It seems like the gl_FragColor, gl_FragData, etc. should be treated the same
> way.

Possibly, but I couldn't find any spec text mentioning it. Note that mediump
color suffices to render to 16-bit floating-point buffers (when
EXT_color_buffer_half_float is available), but mediump depth is very limited
since there are only 10 mantissa bits hence the strong need to define it as
highp when supported.

[Brandon Jones, 2013-05-28 20:54:34]

Regenerated glslang_tab.cpp with Bison 2.7.

I think I'll leave the braces on the if statement simply because that's
consistent with the rest of the file.

[Brandon Jones, 2013-05-28 21:01:57]

And I've now addressed the Medium/High precision issue as well

[Alok Priyadarshi, 2013-05-28 21:05:36]

https://codereview.appspot.com/9827044/diff/12001/src/compiler/OutputGLSL.cpp
File src/compiler/OutputGLSL.cpp (right):

https://codereview.appspot.com/9827044/diff/12001/src/compiler/OutputGLSL.cpp...
src/compiler/OutputGLSL.cpp:26: TString name = node->getSymbol();
Please delete this local variable "name".

[Brandon Jones, 2013-05-28 21:38:21]

On 2013/05/28 21:05:36, Alok Priyadarshi wrote:
> https://codereview.appspot.com/9827044/diff/12001/src/compiler/OutputGLSL.cpp
> File src/compiler/OutputGLSL.cpp (right):
> 
>
https://codereview.appspot.com/9827044/diff/12001/src/compiler/OutputGLSL.cpp...
> src/compiler/OutputGLSL.cpp:26: TString name = node->getSymbol();
> Please delete this local variable "name".

Done

[Alok Priyadarshi, 2013-05-28 21:43:37]

lgtm

Is there a WebGL conformance test for this?

[Brandon Jones, 2013-05-28 21:53:11]

On 2013/05/28 21:43:37, Alok Priyadarshi wrote:
> lgtm
> 
> Is there a WebGL conformance test for this?

https://www.khronos.org/registry/webgl/sdk/tests/conformance/extensions/ext-f...

[Brandon Jones, 2013-05-28 22:16:28]

Committed patchset #4 manually as r2248 (presubmit successful).