Tests random vectors, and makes sure that encoding/decoding
round trip only introduces precision errors below a certain
threshold.
Pretty confident that the implementation is correct now.
This way the executable code can be worked with directly, instead
of via the python file that generates the rust code.
Also introduced some small optimizations.
Rust 1.27 stablized a variety of cpu intrinsics, including SIMD
on x86/64 platforms. This commit moves to using those intrinsics
for the optimized Float4 implementation. This means Psychopath
now compiles on stable Rust with all optimizations. Yay!
The sampling method used before is numerically unstable for very
small lights. That sampling method is still used for large/close
lights, since it works very well for that. But for small/distant
lights a simpler and numerically stable method is used.
It *seemed* to fix the problem I was running into, but it actually
made the SphereLight ray intersection code incorrect, and wa just
avoiding intersections that should have happened.
I should test better before committing. :-)
Thanks to a discovery by Petra Gospodnetic during her GSOC
project, I was able to substantially improve light tree sampling
for lambert surfaces. As part of this, the part of the surface
closure API relevant to light tree sampling has been adjusted to
be more flexible.
These improvements do not yet affect GTR surface light tree
sampling.
More specifically: prior to this, SurfaceLights returned the
shadow ray direction vector to use. That was fine, but it
kept the responsibility of generating proper offsets (to account
for floating point error) inside the lights.
Now the SurfaceLights return the world-space point on the light
to sample, along with its surface normal and error magnitude.
This allows the robust shadow ray generation code to be in one
place inside the renderer code.
