Blender Git Loki

Merge branch 'usd-importer-T81257-merge' of https://gitlab-master.nvidia.com/omniverse/blender into omniverse

USD Import format fixes.

Add a little todo list

USD Import: not importing subdiv by default.

Changed the Import Subdive Scheme option default to false,
because this option can make large scenes considerably heavier
and because it needs to be further debugged.

Move component implementations to separate files

Merging latest USD Import changes.

USD Import: made USDPrimReader destructor virtual.

USD Import: USD Preview conversion.

Added experimental Import USD Preview option to convert
UsdPreviewSurface shaders to Principled BSD shader networks,
for an approximate preview. Added new USDMaterialReader
class, which implements this feature. Also added a
Set Material Blend option to automatically set the material
blend method based on the UsdPreviewSurface shader's opacity
and opacityThreshold inputs. Finally, updated the logic so that
no materials will be imported if the Materials option is
not checked.

Merge branch 'master' into temp-gpencil-bezier-v2

Merge branch 'master' into temp-lineart-contained

Merge branch 'master' into temp-gpencil-io

Merge branch 'master' into greasepencil-object

Fix various CUDA issues fater recent change

- Strict compiler warnings in the CUDADeviceQueue
- Missing install file for the megakernel

fix missing updates when only the transformations are animated

USD Import: material blend bug.

Fixed bug setting the material blend mode due
to an uninitialized variable.

Fix wrong samples stop criteria in Cycles path trace

With multiple samples scheduled at once the stop criteria needs to be
adjusted.

On a first glance seemed like things will be much faster, but no: the
way how scheduling worked was just requesting 0 samples for many loop
iterations. Still stupid but does not cause that measurable of speedup.

Cycles: return number of active paths from queue

Cycles: track number of queued kernels with atomics on GPU

Also change logic to assume integrator state is zero initialized, so that
the init_from_camera will be able to check if a path is terminated.

Cycles: use array of queued/terminated paths to schedule less work for GPU

Passing along a path index now that may be different than the work index.

Cycles: add megakernel for new integrator, to finish last path faster

Also always used for CPU since it's a bit faster.