Blender Git Loki

Protect against OSL alloc's null pointer dereference

(I got this one in MSVC)

Ref T54796

Fix typo with Chiang's roughness modifier

Ty nirved on IRC for noticing

Ref T54796

Add primary refraction roughness modifier

(Does it need a better parameter name?)

Ref T54796

Remove extra comment in Spanish

Fix Direct coloring calculation

/facepalm

Ref T54796

OSL: initialize geometry as part of the setup

This needs a custom OSL closure. Now it renders, but the colors don't make sense

Ref T54796

Fix more of nirved's OpenCL errors

Thank you nirved for finding and explaining them!

Ref T54796

Add nirved's OpenCL compatibility patch

Ref T54796

A first attempt towards a OSL shader

Obviously, it doesn't work, but at least it renders something non-black...

Ref T54796

Enable Filter Glossy for the hair shader

Ref T54796

Merge branch 'master' into soc-2018-hair-shader

First conflict: IES vs our shader

Fix h value not being computed correctly.

The new code only works for thick curves, not ribbons, but should hopefully
work correctly for that case. If we already know the normal we don't need to
look at the curve geometry, which simplifies things.

Fix sd->dPdu normalization, renamed dPdCD to Y, increase assert epsilon.

Cleanup: use kernel_assert() instead of assert() in kernel.

Cleanup: keep BSDF local frame computation out of curve code.

Cleanup: add some comments, guard against NaNs in hair_ap.

Use two variables to store azimuthal and longitudinal roughness

It makes debugging those pesky NaNs far easier

Ref T54796

Fix improper optimization in the trimmed logistic function

This was the source of the noise that Ronan mentioned.

Ref T54796

Ringfence possible NaNs with degenerate hair geometry

There are two ways: radiance (F) and intersect position (sd->P)

F comes from invalid core distance, sd->P from invalid intersect position
(which is usually caused from a NaN in ray->t, outside our shader)

I knew I had a leftover printf somewhere...

Ref T54796