Blender Git Loki

Fix crash in own recent rB776a8548f03a049.

New ID may be null, have to check this too!

Reported by @panzergame over IRC, thanks.

To be ported to 2.78 as well.

Cycles: Use correct light sampling PDF for MIS calculation with Branched Path Tracing

The light sampling functions calculate light sampling PDF for the case that the light has been randomly selected out of all lights.
However, since BPT handles lamps and meshlights separately, this isn't the case. So, to avoid a wrong result, the code just included the 0.5 factor in the throughput.

In theory, however, the correction should be made to the sampling probability, which needs to be doubled. Now, for the regular calculation, that's no real difference since the throughput is divided by the pdf.
However, it does matter for the MIS calculation - it's unbiased both ways, but including the factor in the PDF instead of the throughput should give slightly better results.

Reviewers: sergey, brecht, dingto, juicyfruit

Differential Revision: https://developer.blender.org/D2258

Cleanup: remove unused function.

Alembic: only interpolate vertices when their number don't differ
between samples.

This could cause some crashes.

Cycles: Fix T49411: Multiscatter GGX with zero roughness when Filter Glossy is enabled

Volume slice settings: tweak/fix UI enabling conditions.

It was still possible to have some properties enabled when it does not
make sense to have them so.

Smoke debug draw: take adaptive domain position into account.

Eventually the various functions that deals with adaptive domain
bounding box shall de-duplicated.

some improvement for dynamic fracture, it now takes location and size of impact object into account with "Limit Impact"

overhaul gpuMatrix API

Complete (for our needs) 2D & 3D transformation API. Should be easy to port legacy OpenGL matrix stack-based code to this. Still needs testing.

Ported ortho, frustum, lookAt functions from Viewport FX (rB194998766c65). Kept plenty of Viewport FX code from previous commit.

Stack API and 2D routines ported from Gawain. This version uses BLI_math library so everything is licensed under GPL instead of the usual MPL.

Part of T49450

port math function from Viewport FX

mul_v4_m4v3(r, M, v) means r = M * vec4(v, 1.0)
Based on rB194998766c65

specialize mat3 multiply routines

Same as 5c6baf1ee9754e39392926cd077bb2097baf39b9 but for 3x3 matrices.

Viewport smoke: fix a couple of issues in the new display settings.

- WITH_SMOKE macro was not defined so some code was not compiled, though
it was still accessible from the UI
- some UI elements were disappearing due to bad indentation, also rework
the UI code to not hide but rather disable/grey out button in the UI
- Display thickness was not used due to bad manual merge of the code
from the patch.

specialize mat4 multiply routines

mul_m4_m4m4(R, A, B) gives us R = AB in general. Existing code assumed the worst, that A and B both alias the output R. For safety it makes internal copies of A and B before calculating & writing R.

This is the least common case. Usually all 3 matrices differ. Often we see M = AM or M = MB, but never M = MM.

With this revision mul_m4_m4m4 is called in exactly the same way but copies inputs only when needed. If you know the inputs are independent of the output use the "uniq" variant to skip the saftety checks.

Fix T49310: incorrect Cycles standalone normals with negative scale.

Curve Fitting: re-fitting heap pre-allocated too much

Curve Fitting: 'offset' method used sign when it shouldn't

Curve Fitting: inline dot-product (avoid temp vector)

Curve Fitting: de-duplicate cubic evaluation

Quiet shadow warning

No need to declare new iterator for second loop.

Modiffied vertex paint's done step to be more efficient. WPaint could use a fast update mode like vpaint currently has...