[Piglit] [PATCH 1/2] Add a test of tessellation of transform feedback primitives.

[Paul Berry]

On 9 November 2011 17:38, Eric Anholt <eric at anholt.net> wrote:

> On Tue,  8 Nov 2011 16:03:39 -0800, Paul Berry <stereotype441 at gmail.com>
> wrote:
> > According to the OpenGL 3.0 spec (section 2.15: Transform Feedback):
> >
> >     "When quads and polygons are provided to transform feedback with a
> >     primitive mode of TRIANGLES, they will be tessellated and recorded
> >     as triangles (the order of tessellation within a primitive is
> >     undefined). Individual lines or triangles of a strip or fan
> >     primitive will be extracted and recorded separately."
> >
> > This test verifies the correctness of the tessellation and extraction
> > from strips and fans.  It does so by feeding the output of transform
> > feedback back into the GL pipeline and verifying that the rendered
> > image is the same.
> >
> > Verified using the nVidia proprietary driver for Linux.  The nVidia
> > driver passes all tests except "tessellation polygon flat_last",
> > "tessellation quad_strip flat_last", and "tessellation quads
> > flat_last".  These tests fail because the order in which the driver
> > tessellates polygons and quads fails to preserve the correct provoking
> > vertex, leading to different results from flatshading.
>
> Ensuring that provoking vertex is respected seems like a reasonable
> requirement, and at least our hardware is built to support it it looks
> like.
>
> I like the test idea, and have just a few little style comments.  I find
> the vertex transformations a bit unintuitive compared to most of our
> tests, but I'm not really worried.
>
> > +#include "piglit-util.h"
> > +
> > +#ifndef M_PI
> > +#define M_PI 3.141592653589793
> > +#endif
>
> M_PI?
>
>
I'm afraid I don't understand the question.  Can you elaborate?


> > +#define BUFFER_SIZE 20
> > +
> > +#define Elements(array) (sizeof(array)/sizeof(array[0]))
>
> There's ARRAY_SIZE available so you don't have to redefine this.
>

Ah, ok.  Thanks.


>
> > +/**
> > + * Check that two strips of the window match.  Strips are numbered
> > + * from the top from 0 to 3.
> > + */
> > +static GLboolean
> > +match_strips(int reference, int compare)
> > +{
> > +     int reference_offset = (3 - reference) * (piglit_height / 4);
> > +     int compare_offset = (3 - compare) * (piglit_height / 4);
> > +     int x, y;
> > +     for (x = 0; x < piglit_width; ++x) {
> > +             for (y = 0; y < piglit_height/4; ++y) {
> > +                     float probed[4];
> > +                     glReadPixels(x, y + reference_offset, 1, 1,
> > +                                  GL_RGBA, GL_FLOAT, &probed);
> > +                     if (!piglit_probe_pixel_rgba(x, y + compare_offset,
> > +                                                  probed))
> > +                             return GL_FALSE;
> > +             }
> > +     }
> > +     return GL_TRUE;
> > +}
>
> It would be nice if you could use piglit_probe_rect_halves_equal_rgba()
> or a derivative here.  glReadPixels()ing each pixel one by one sucks for
> performance.
>
>
Yeah, that makes sense.  I'll rework the test to use a single
glReadPixels() call to read the reference image, and then
piglit_probe_image_rgba() to check the other two images.


> > +                     printf("Test requires GL_EXT_provoking_vertex "
> > +                            "or GL_ARB_provoking_vertex\n");
> > +                     piglit_report_result(PIGLIT_SKIP);
> > +                     exit(1);
>
> piglit_report_result() implies exit()
>

Ok, thanks.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.freedesktop.org/archives/piglit/attachments/20111109/8f482df6/attachment.html>