On 12/19/2011 10:10 AM, Dave Airlie wrote:
Hi guys,
So we have some piglit tests
generated_tests/spec/glsl-1.10/execution/interpolation/interpolation-none-gl_FrontColor-smooth-none.shader_test
that piglit fails
This seems to be due to its linear interpolation not producing the
correct answers compared to the hardware, on the rendered triangle,
there is a obvious circle of darkness that isn't there on hw rendered.
Looking at tri_linear_coeff I notice a large comment explaining that
this could be done better, llvmpipe also uses the same code, is this
what is causing the fail?
Quite possibly. It shouldn't be hard to hack softpipe to use
perspective-corrected color interpolation to see if that helps.
-Brian
_______________________________________________
mesa-dev mailing list
mesa-dev@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/mesa-dev
