[Mesa-dev] Intel: Please help define the common rectangle primitive type

[Jason Ekstrand]

On Wed, Aug 24, 2016 at 2:42 PM, Marek Olšák <maraeo at gmail.com> wrote:

> On Wed, Aug 24, 2016 at 10:42 PM, Jason Ekstrand <jason at jlekstrand.net>
> wrote:
> > On Wed, Aug 24, 2016 at 11:46 AM, Marek Olšák <maraeo at gmail.com> wrote:
> >>
> >> On Fri, Aug 5, 2016 at 1:22 AM, Kenneth Graunke <kenneth at whitecape.org>
> >> wrote:
> >> > On Thursday, August 4, 2016 11:28:28 AM PDT Jason Ekstrand wrote:
> >> >> > > - No cut index (GS is supposed to be fine and Tess looks OK too)
> >> >> >
> >> >> > I'd be OK with dropping GS and Tess support for rectangles.
> >> >>
> >> >> I don't know that we need to but the docs say cut index doesn't work.
> >> >
> >> > Perhaps we should throw a GL error when using rectangles if primitive
> >> > restart (i.e. cut index) is enabled.
> >> >
> >> > I could potentially see GS being useful - maybe someone would want
> >> > to do layered rendering with rectangles, and generate rectangles
> >> > for each?  Tessellation on rectangles seems pretty nonsensical.
> >>
> >> AMD_vertex_shader_layer is better IMO. I don't think we can do GS +
> >> rectangles.
> >>
> >> The last remaining thing is whether the order of vertices matters for
> >> Intel.
> >
> >
> > I think I said this already, but I don't think it actually does.  I know
> > I've gotten the order wrong in the past with no ill consequences.  That
> > said, I haven't exhaustively tested it.
>
> Well, if you say "I don't think", it means "I don't know for sure".
> But OK, I'll assume the order doesn't matter.
>

Here's some data points:

 1) We've been doing it wrong in blorp since the dawn of time and have had
no problems.
 2) I just tried all 6 possible permutations on my Sky Lake and they all
seem to work.  The simulator gets a bit grumpy but the hardware doesn't
care.

We'll have to write a good piglit test and throw it at the CI system (all
the hardware!) to be 100% sure but, as far as I can tell, the hardware
doesn't actually care what order the points are in.  I think it's safe to
assume order doesn't matter at the moment.  If it turns out it does, we'll
cross that bridge when we come to it.  It'll be much easier to thoroughly
test it once we have it plumbed through and can write piglit tests.

--Jason
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20160824/c01c1cae/attachment.html>