[Mesa-dev] [PATCH 22/22] meta/blit: Use GL_EXT_shader_samples_identical in MSAA-SS resolve blit

[Iago Toral]

On Fri, 2016-02-19 at 11:47 +0000, Neil Roberts wrote:
> Iago Toral <itoral at igalia.com> writes:
> 
> > I don't know much about this, but using shader_samples_identical
> > should only give a benefit if we actually get identical samples,
> > otherwise it means more work, right? I noticed that the test renders a
> > quad with random colors for each vertex that will be interpolated
> > across the region, so could it be that we are hitting few cases of
> > identical samples? Shouldn't the results be expected to be better if
> > we rendered a flat color instead?
> 
> It should all be identical regardless of the color chosen because it is
> doing per-pixel shading rather than per-sample which means the fragment
> shader is only run once per pixel and the same value is written to all
> of the samples. The only case where you get differing samples normally
> is at the edges of a polygon where the single color is only written to a
> few of the samples.
> 
> In this case it is even more identical because the random color is set
> on a uniform rather than a vertex so it is actually using the same color
> for the whole framebuffer. However, it is rendered as two triangles
> which might mean that along the diagonal of the framebuffer where the
> two triangles meet there will be a polygon edge which means the hardware
> might not recognise that these all the same color, but that should only
> be a few pixels and probably won't affect the results very much.

You're right, I read through the test code too fast :)

Iago