[Public WebGL] ANGLE Dynamic Indexing of Arrays of Samplers

Tim Johansson [email protected]
Mon May 30 04:39:39 PDT 2011


On 2011-05-27 18:12, Gregg Tavares (wrk) wrote:
>
>
> On Thu, May 26, 2011 at 11:58 PM, Tim Johansson <[email protected] 
> <mailto:[email protected]>> wrote:
>
>     On 2011-05-26 19:19, Chris Marrin wrote:
>
>
>         I'm adding the following statement to the differences section:
>
>             The GLSL ES spec<a href="#refsGLES20GLSL">[GLES20GLSL]</a>
>          allows but does not mandate support for dynamic
>             indexing of arrays of samplers. WebGL disallows this support.
>
>         This is in support of tightening the language about array
>         bounds checking. Does ANGLE currently make this check and
>         disallow such indexing?
>
>     I would be very surprised if it did as it could not possible pass
>     the conformance tests if it did.
>
>     The test gl-min-textures.html has a shader which looks like:
>
>     "uniform sampler2D uni[8];
>     void main()
>     {
>        vec4 c = vec4(0,0,0,0);
>        for (int ii = 0; ii < NUM_TEXTURES; ++ii) {
>          c += texture2D(uni[ii], vec2(0.5, 0.5));
>        }
>        gl_FragColor = c;
>     }"
>
>
>
> Is that dynamic indexing? I thought since NUM_TEXTURES is replaced by 
> a constant it's static. I assumed dynamic meant that the indices can 
> change depending on the input. For example
>
You're right of cause, it is not dynamic so it must be a bug in the that 
specific OpenGL ES implementation. I'll just try to add a workaround for 
it instead.

//Tim

> uniform sampler2D ab[2];
> uniform int aOrB;
>
> void main () {
>   gl_FragColor = texture2D(ab[aOrB], vec2(0.5, 0.5));
> }
>
> is dynamic indexing since it can not be unrolled.
>
>
>     That is preventing us from passing the conformance tests on some
>     GLES implementations.
>
>     There is also an unrolled version of the test which does not use
>     dynamic indexing of sampler arrays - gl-min-textures-unroll.html.
>     Do the persons who wrote those two tests (svn indicates Gregg and
>     Mo) know if there are any known problems with the unrolled version
>     or if we can just switch to using that?
>
>     //Tim
>
>
>         -----
>         ~Chris
>         [email protected] <mailto:[email protected]>
>
>
>
>
>         -----------------------------------------------------------
>         You are currently subscribed to [email protected]
>         <mailto:[email protected]>.
>         To unsubscribe, send an email to [email protected]
>         <mailto:[email protected]> with
>         the following command in the body of your email:
>         unsubscribe public_webgl
>         -----------------------------------------------------------
>
>
>

-----------------------------------------------------------
You are currently subscribed to [email protected]
To unsubscribe, send an email to [email protected] with
the following command in the body of your email:
unsubscribe public_webgl
-----------------------------------------------------------




More information about the public_webgl mailing list