Blender Git Loki

Cycles: Workaround for performance loss with the CUDA 9.0 SDK.

CUDA 9.0.176 apparently caused some slow down on high-end Pascal cards that can be mitigated by increasing the number of registers. See https://developer.blender.org/F1142667 for a detailed comparison.

Fix T53692: OpenCL multi GPU rendering not using all GPUs.

Ensure each OpenCL device has a unique ID even if the hardware ID is not
unique for some reason.

Fix T53755: Cycles OpenCL lamp shaders have incorrect normal.

Cleanup: SGI format, avoid overflow

Harmless but causes warnings

Cleanup: SGI format, use uint/short/char types

Fix SGI foramt reader CVE-2017-2901

Integer Overflow Code Execution Vulnerability.

Reader no longer crashes on corrupt images (from own fuzz testing).

Cleanup: SGI format, remove unused struct members

Fix background_job template

Update attribute and use empty file operator

Fix T53311: transform edge/normal orientation

When the edge is aligned with it's own normals,
transform orientation wasn't aligned with the edge.

Fix bmesh.utils.face_join arg parsing

Fix T53752: Draw curve fails w/ stylus

Additional fix after cherry-picking rB870840e8b7dd (in rB72151f3e36b).

Add an 'atomic cas' wrapper for pointers.

Avoids having to repeat obfuscating castings everywhere...

Fix scalability issue in threaded code of Mesh normals computation.

We tried to do as much as possible in a single threaded callback, which
lead to using some nasty tricks like fake atomic-based spinlocks to
perform some operations (like float addition, which has no atomic
intrinsics).

While OK with 'standard' low number of working threads (8-16), because
collision were rather rare and implied memory barrier not *that* much
overhead, this performed poorly with more powerful systems reaching the
100 of threads and beyond (like workstations or render farm hardware).

There, both memory barrier overhead and more frequent collisions would
have significant impact on performances.

This was addressed by splitting further the process, we now have three
loops, one over polys, loops and vertices, and we added an intermediate
storage for weighted loop normals. This allows to avoid completely any
atomic operation in body of threaded loops, which should fix scalability
issues. This costs us slightly higher temp memory usage (something like
50Mb per million of polygons on average), but looks like acceptable
tradeoff.

Further more, tests showed that we could gain an additional ~7% of speed
in computing normals of heavy meshes, by also parallelizing the last two
loops (might be 1 or 2% on overall mesh update at best...).

Note that further tweaking in this code should be possible once Sergey
adds the 'minimum batch size' option to threaded foreach API, since very
light loops like the one on loops (mere v3 addition) require much bigger
batches than heavier code (like the one on polys) to keep optimal
performances.

Subsurf: Avoid possible use of partially initialized edge hash

Subsurf: Avoid global lock for loops and orig index layers

This is a bit annoying to have per-DM locking, but it's way better (as in, up to
4 times better) for playback speed when having lots of subsurf objects,

install_deps: Fix OCIO always re-cloning its whole git repository.

No groping in Blender's Tracker!

This a small cleanup of something which I think is just a typo anyway.

With all the recent talks of harrassment and groping, I think we better avoid
that within our source code! :)

Reviewers: sergey

Reviewed By: sergey

Tags: #motion_tracking

Differential Revision: https://developer.blender.org/D2979

Fix T53612: Blender crashes on CleanTracks with 'DELETE_SEGMENTS' and a disabled marker

Simple fix, which is totally safe for 2.79a!

Fix T53509: Datablock ID Properties attached to bpy.types.Material are not loaded.

Stupid mistake in material reading code, thanks to Simon Wendsche (@BYOB) for the investigation and fix!

To be backported to 2.79a.