On Sat, Jul 5, 2014 at 11:20 PM, Roland Scheidegger <srol...@vmware.com> wrote: > Am 05.07.2014 06:07, schrieb Ilia Mirkin: >> The new location field can be either center, centroid, or sample, which >> indicates the location that the shader should interpolate at. > Looks good though maybe it should be mentioned in the interpolator > section in tgsi.rst (yeah centroid wasn't neither). > In any case, > Reviewed-by: Roland Scheidegger <srol...@vmware.com> > >> >> Signed-off-by: Ilia Mirkin <imir...@alum.mit.edu> >> --- >> >> I tried to make sure I hit all the uses, but I guess a bunch of drivers don't >> look at the Centroid field at all? > Well drivers not doing msaa or just fake one like llvmpipe aren't > interested in it. > And, this wasn't exposed via d3d9, though r300 reportedly can do it (the > docs even mention it for r5xx), I suspect using some of the crazy > backdoor mechanisms of d3d9. Some very quick look at the docs though > don't indicate how this could be done (if it's even possible per > interpolator) for these chips, which is I guess why centroid isn't handled.
d3d9 does support centroid. I used to use it in HLSL using this syntax: float4 v : TEXCOORD0_CENTROID: The reason centroid isn't supported by r300g is that the hw documentation doesn't say how to enable it. Centroid is also required by GL2. Marek _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
