Hello all,

I have a simple model with a lot of triangles.  To reduce bandwidth, I'd li=
ke to use a VBO with the unique vertex coordinates, and have the vertex sha=
der generate the UV values (they're always {0, 1} {0, 0}, {1, 0} for each t=
riangle) for the fragment shader.

Is this possible?  I don't think my current setup allows for any other shad=
er stages...

Thanks!

On Wed, 23 Sep 2015 18:11:51 -0700, Edward Brekelbaum wrote:

> I have a simple model with a lot of triangles.  To reduce bandwidth, I'd
> like to use a VBO with the unique vertex coordinates, and have the vertex
> shader generate the UV values (they're always {0, 1} {0, 0}, {1, 0} for
> each triangle) for the fragment shader.
> 
> Is this possible?  I don't think my current setup allows for any other
> shader stages...

How does the vertex shader know which vertex should have which UV coordinates?

If you're rendering distinct triangles (rather than strips or fans), you
could use gl_VertexID % 3. Otherwise, you need some other way for the
vertex shader to know which of a triangle's vertices it's dealing with.

In the worst case, you can compress the UV coordinates down to one byte
per vertex, i.e. have a an extra attribute whose value is 0, 1 or 2 (or
-1, 0 or 1) which is then expanded to a (u,v) pair.

On Sunday, September 27, 2015 at 6:55:03 PM UTC-5, Nobody wrote:
> On Wed, 23 Sep 2015 18:11:51 -0700, Edward Brekelbaum wrote:
> 
> > I have a simple model with a lot of triangles.  To reduce bandwidth, I'd
> > like to use a VBO with the unique vertex coordinates, and have the vertex
> > shader generate the UV values (they're always {0, 1} {0, 0}, {1, 0} for
> > each triangle) for the fragment shader.
> > 
> > Is this possible?  I don't think my current setup allows for any other
> > shader stages...
> 
> How does the vertex shader know which vertex should have which UV coordinates?
> 
> If you're rendering distinct triangles (rather than strips or fans), you
> could use gl_VertexID % 3. Otherwise, you need some other way for the
> vertex shader to know which of a triangle's vertices it's dealing with.
> 
> In the worst case, you can compress the UV coordinates down to one byte
> per vertex, i.e. have a an extra attribute whose value is 0, 1 or 2 (or
> -1, 0 or 1) which is then expanded to a (u,v) pair.

Thanks.  I think I will go with sending an extra byte per vertex.