Blender Git Commits

Include all module functions in a single header for convenience.

Simple coordinate input node for texture nodes.

Handle all node type kinds appropriately.

The LLVM backend also supports PASS nodes, even though these are usually
removed from the node graph in a preprocessing step for clarity.

ARG nodes have their output values mapped in advance, so no extra code
needs to be generated for them.

Fix incorrect argument order for vector/color constructor functions.

Replaced mix of bool type flags for node types with a single enum for special kinds of nodes.

Bindless Textures: Some cleanup.

1) Rename extended_images to bindless_textures

2) Remove kepler textures slots, they will no longer be needed.

Nothing to see here yet. Just some preparations for further work.

Implemented math functions for the LLVM backend.

Fix for missing notifier application on texture node trees.

The pynodes tree does not have the static NTREE_TEXTURE type set.

Fix node socket update callbacks exiting early because of missing output connection.

Outputs are currently identified in the nodes via the DO_OUTPUT flag. This is not
nicely accessible for pynodes and needs to be reimplemented in a proper way.
Until then this optimization for node recalculation should just be ignored for texnodes,
so nodes now have a callback for checking usage, and will always be considered used by default.

Merge branch 'master' into cycles_hair_bvh

Remove redundant dependency declaration from pynodes, which creates a circular dependency.

Previously the "parameters" depsnode was the POST operation for node trees, but has now
been supplanted by the cache invalidation node. Making an explicit dependency on the component
thus creates a cycle.

Switch texture nodes to a pynodes implementation.

This also removes the static bNodeTree type definition C, because
BVM requires some functions that are best defined in python (and it
would have been switched eventually anyway).

Removed texture nodes.

The node tree type is left alone for now, only the node types are removed.

Change default geometry cache size to 1 GB

256 MB was a little small, and a larger cache will preform better. This will
probably change at some point in the future but for now its better to have a
larger default cache size.

fixed broken CMake for mantaflow directory

Lower default max_T from 128 to 12

This improves cache performance by reducing the maximum size of subpatches,
making tessellation both less likely to occur and faster. Difference in
performance is theoretically 100 times faster during thrashing as maximum
subpatch surface area is 100 time smaller, tho actual performance will vary.

Define the base opcode functions for LLVM and use util_ files for types.

Split opcode macro into thematic sections to make it easier to handle them in parts.

Merge remote-tracking branch 'origin/master' into temp-cycles-microdisplacement

Conflicts:
intern/cycles/kernel/kernel_globals.h

More meaningful dependencies between node tree parameters and cache-invalidation operations, and textures that use them.