Blender Git Loki

Support for image sampling in the LLVM backend.

This reintroduces the "globals" concept for mapping ID datablocks with
a consistent permanent identifier key.

Merge branch 'master' into soc-2016-multiview

Fix invalid function signature when passing derivatives for types that don't support them.

WIP: Select shortest path - Computing the shortest path using regular Dijkstra algorithm

partial solution for FM autoexecute, auto disables when sim runs and is enabled again on startframe

note: cache will always be overwritten in autoexec sim mode, because keeping it may lead to incorrect sim results
furthermore a couple of changes for external mode (attempt to be more BCB compatible)

Improve (fix a few bugs in) the transform hierarchy creation.

Now an empty is created for every transform node except for the cases
when the transform node only has a single child that is not a transform
itself.

There are still a few issues in complex hierarchies where multiple
objects have the same name (which is legal in Alembic since it is the
whole path to the object that counts, not just the name).

Half float: Half EXR now uses a half4 slot on CPU.

Code in util_half.h inspired by work from Lukas Stockner.

Still a lot missing (GPU, Data Textures...).

Half Textures: Fix some todos from last commit.

Implemented the hilbert spiral

Decklink: fix output on some DeckLink cards.

Opening a DeckLink card for output with the DeckLink object
sometimes failed because some types of card report no support for BGRA pixel
format, yet it works. The test is maintained but a warning is printed
instead of throwing an exception.

Optional cache size suffix added in Decklink output format string: may be
needed for some card that needs more than the default 8 frames.

Decklink: Fix bug: Decklink.right attribute was not unitialized.

tile manager, start with hilbert spiral

better liquid inflow function

make an independent multiview reconstruction pipeline for multiview

Merge branch 'master' into alembic_basic_io

Minor UI fix.

Add a points/particles reader.

Similarly to the hair being written as curves, this does not create a
particle system, but rather reads the points as Mesh which only has
vertices.

Unify code generation through the llvm API.

Also standard functions for copying and zeroing values are now
implemented per type rather than relying on a simple load/store.
This will become more important with larger aggregate types such as matrix44.

Cleanup: de-duplicate modifier addition.

Add a flag to indicate that a cache is made of sequence of file or not.