Blender Git Loki

cache GPUx batches for performance++

WIP â?? developed on Mac, want to test on Linux & Windows so pushing
unfinished.

Everything needed for drawing is collected into a GPUxBatch. These are
generated per-object the first time drawn, then reused for subsequent
draws. Regenerated when DerivedMesh or draw mode changes.

TODO:
- actively reclaim memory (like gpu_buffers.câ??s pool)
- store a list of batches per DM, for complex drawing
- find cause of low perf on high poly OB_SOLID rendering

WIP work to pass asset data to operator.

Asset engine ID, asset/variant/revision uuids, root path...
First goal is link code!

BGE Decklink: upgrade to Desktop Video 10.4

Merge branch 'cycles_memory_experiments' into gooseberry

Merge branch 'alembic' into gooseberry

Revert recent commits, no time to improve things now.

Merge branch 'asset-experiments' into asset-engine

Merge branch 'master' into asset-experiments

Merge branch 'master' into cycles_memory_experiments

Merge branch 'master' into gooseberry

Conflicts:
source/blender/editors/space_view3d/drawobject.c

Don't write dupli instances to caches in the Render pass.

Duplilists are always generated for viewport/realtime settings, render
settings only affect the mesh geometry but not the duplis.

Cleanup: Removed unused writer/reader classes for Alembic.

These classes were used previously to write parts of the Blender DNA
directly to Alembic. Their code is shared with other classes that are
still in use, so none of the really substantial parts have been deleted.

Cleanup: Removed unused parts of the C API for caching.

This only removes trivial C function wrappers for parts of the C++ code
that are not used directly any more from outside the bl_pointcache
library. Some internal parts may be removed later, this cleanup is
mostly to reduce code noise.

Renamed passes for cache writing to reflect their origin with mesh
modifier evaluation.

Smoke: Fix compilation error with smoke disabled

Merge branch 'asset-experiments' into asset-engine

Merge branch 'master' into asset-experiments

Remove the dual "root" object at the top of Alembic archives.

Using two full scene nodes for render/realtime data creates a lot of
duplication in both storage and processing time. Instead storing
specific data types like meshes in different versions keeps duplication
to the necessary minimum and allows easier fallback where available.

Fix OpenCL on Apple again, double spaces in compile arguments.

Initial support for icon/preview/thumbnail of files, including py API.

All this is rather rough at the edges still, but seems to be working from quick tests.
Basically, it:
* Rework a bi BKE_icon, adding a special kind of icons, which type is 0, and which
own a PreviewImage in their obj field.
* Add helper to generate such icon from a file path.
* Add RNA API (to bpy.data currently).

So now, one can request a 'Preview' for a given path, that preview has an icon_id
that can be passed as 'icon_value' parameters of UILayout functions.

Main issue currently is, if you request several time same path, you'll get different
icons/previews each time, maybe we want to store paths in another ghash instead?