Add static_use to shader variable info query.

Add static_use to shader variable info query.

BUG=249018
ANGLEBUG=465
R=kbr@chromium.org

Committed: https://code.google.com/p/angleproject/source/detail?r=4115932

[Zhenyao Mo, 2013-08-21 23:11:23]

Alok, Ken, please take a look.

Shannon, could you merge this to ES3Proto branch once it lands?

Others: FYI (welcome to provide feedback if you want to)

[Zhenyao Mo, 2013-08-21 23:12:32]

On 2013/08/21 23:11:23, Zhenyao Mo wrote:
> Alok, Ken, please take a look.
> 
> Shannon, could you merge this to ES3Proto branch once it lands?
> 
> Others: FYI (welcome to provide feedback if you want to)

Chrome side implementation is also done.  The varying matching related webgl
conformance tests are passing with this and chrome-side CL.

There are no regressions in webgl conformance tests.

Also tested with a bunch of WebGL contents.

[kbr1, 2013-08-22 00:47:45]

LGTM

https://codereview.appspot.com/13158043/diff/1/include/GLSLANG/ShaderLang.h
File include/GLSLANG/ShaderLang.h (right):

https://codereview.appspot.com/13158043/diff/1/include/GLSLANG/ShaderLang.h#n...
include/GLSLANG/ShaderLang.h:382: //            after pre-processing, no matter
whether the statment might
typo: statment -> statement

https://codereview.appspot.com/13158043/diff/1/include/GLSLANG/ShaderLang.h#n...
include/GLSLANG/ShaderLang.h:383: //            be executed or not.
Consider quoting the GLSL ES 3.0 specification more closely here:

"Returns 1 if the variable is accessed in a statement after pre-processing,
whether or not run-time flow of control will cause that statement to be
executed."

https://codereview.appspot.com/13158043/diff/1/src/compiler/VariableInfo.cpp
File src/compiler/VariableInfo.cpp (right):

https://codereview.appspot.com/13158043/diff/1/src/compiler/VariableInfo.cpp#...
src/compiler/VariableInfo.cpp:302: // TInterBinary node in that case. Since
attributes, unifroms,
preexisting typo: unifroms -> uniforms

[Zhenyao Mo, 2013-08-22 00:54:59]

https://codereview.appspot.com/13158043/diff/1/include/GLSLANG/ShaderLang.h
File include/GLSLANG/ShaderLang.h (right):

https://codereview.appspot.com/13158043/diff/1/include/GLSLANG/ShaderLang.h#n...
include/GLSLANG/ShaderLang.h:383: //            be executed or not.
On 2013/08/22 00:47:45, kbr1 wrote:
> Consider quoting the GLSL ES 3.0 specification more closely here:
> 
> "Returns 1 if the variable is accessed in a statement after pre-processing,
> whether or not run-time flow of control will cause that statement to be
> executed."

Done.

https://codereview.appspot.com/13158043/diff/1/src/compiler/VariableInfo.cpp
File src/compiler/VariableInfo.cpp (right):

https://codereview.appspot.com/13158043/diff/1/src/compiler/VariableInfo.cpp#...
src/compiler/VariableInfo.cpp:302: // TInterBinary node in that case. Since
attributes, unifroms,
On 2013/08/22 00:47:45, kbr1 wrote:
> preexisting typo: unifroms -> uniforms

Done.

[Zhenyao Mo, 2013-08-22 17:56:58]

Alok?

[Alok Priyadarshi, 2013-08-22 20:40:21]

https://codereview.appspot.com/13158043/diff/9001/include/GLSLANG/ShaderLang.h
File include/GLSLANG/ShaderLang.h (right):

https://codereview.appspot.com/13158043/diff/9001/include/GLSLANG/ShaderLang....
include/GLSLANG/ShaderLang.h:396: COMPILER_EXPORT void ShGetVariableInfo(const
ShHandle handle,
nit: The list of parameters is getting too long. At some point, we might want to
use a struct for the out parameter.

https://codereview.appspot.com/13158043/diff/9001/src/compiler/VariableInfo.cpp
File src/compiler/VariableInfo.cpp (right):

https://codereview.appspot.com/13158043/diff/9001/src/compiler/VariableInfo.c...
src/compiler/VariableInfo.cpp:170: static_use(false)
ANGLE style convention is to not use underscores in variable name.

https://codereview.appspot.com/13158043/diff/9001/src/compiler/VariableInfo.c...
src/compiler/VariableInfo.cpp:268: return true;
no need to define these functions anymore because TIntermTraverser::visitFoo
always returns true.

https://codereview.appspot.com/13158043/diff/9001/src/compiler/VariableInfo.h
File src/compiler/VariableInfo.h (right):

https://codereview.appspot.com/13158043/diff/9001/src/compiler/VariableInfo.h...
src/compiler/VariableInfo.h:46: TVariableInfoList& mAttribs;
It seems we should now store these variables in a map for faster searching
instead of a list.

[Zhenyao Mo, 2013-08-22 21:57:47]

Revised, please take another look.

https://codereview.appspot.com/13158043/diff/9001/include/GLSLANG/ShaderLang.h
File include/GLSLANG/ShaderLang.h (right):

https://codereview.appspot.com/13158043/diff/9001/include/GLSLANG/ShaderLang....
include/GLSLANG/ShaderLang.h:396: COMPILER_EXPORT void ShGetVariableInfo(const
ShHandle handle,
On 2013/08/22 20:40:22, Alok Priyadarshi wrote:
> nit: The list of parameters is getting too long. At some point, we might want
to
> use a struct for the out parameter.

We will switch to use Jammie's new API ShGetInfoPointer once it's implemented,
so this one will expire at that time.

https://codereview.appspot.com/13158043/diff/9001/src/compiler/VariableInfo.cpp
File src/compiler/VariableInfo.cpp (right):

https://codereview.appspot.com/13158043/diff/9001/src/compiler/VariableInfo.c...
src/compiler/VariableInfo.cpp:170: static_use(false)
On 2013/08/22 20:40:22, Alok Priyadarshi wrote:
> ANGLE style convention is to not use underscores in variable name.

Done.

https://codereview.appspot.com/13158043/diff/9001/src/compiler/VariableInfo.c...
src/compiler/VariableInfo.cpp:268: return true;
On 2013/08/22 20:40:22, Alok Priyadarshi wrote:
> no need to define these functions anymore because TIntermTraverser::visitFoo
> always returns true.

Done.

https://codereview.appspot.com/13158043/diff/9001/src/compiler/VariableInfo.h
File src/compiler/VariableInfo.h (right):

https://codereview.appspot.com/13158043/diff/9001/src/compiler/VariableInfo.h...
src/compiler/VariableInfo.h:46: TVariableInfoList& mAttribs;
On 2013/08/22 20:40:22, Alok Priyadarshi wrote:
> It seems we should now store these variables in a map for faster searching
> instead of a list.

We provide an API to iterate through all the uniforms/varyings/attributes in an
index based fashion, so if we change to map, that operation will be slow.  Per
our offline discussion, we will leave it at vector for now.

[Zhenyao Mo, 2013-08-22 23:18:06]

Committed patchset #3 manually as r4115932 (presubmit successful).