Blender Git Commit Log

fix T50004: Removed check for empty mesh and adjusted the vertex import function to accept meshes without vertices as well

Fix Make Vertex Parent operator missing from vertex/curve/lattice menus.

Reviewed By: brecht

Differential Revision: https://developer.blender.org/D2346

Fix T49904: Cycles standalone missing default generated texture coordinates.

Fix Cycles OSL compilation based on modified time not working.

Fix Cycles standalone not finding CPU device after recent changes.

Fix T49985: cycles standalone XML missing distant lights.

Fix T49985: cycles standalone using wrong socket names for XML reading.

Sculpt UI: Make DynTopo constant detail a resolution value

This should make it easier to sculpt in high resolutions, downside is that the new way to calculate maximum edge length is a bit less intuitive. Maximum edge length used to be calculated as blender_unit * percentage_value, now it's blender_unit / value.

Reused old DNA struct member, but had to bump subversion to ensure correct compatibility conversion. Also changed default value slightly (would have had to set to 3.333... otherwise).

Was Requested by @monio (see https://rightclickselect.com/p/sculpting/zpbbbc/dyntopo-better-scale-input-in-constant-detail-mode) and I think it's worth testing.

BMesh: update comments, ifdef'd code

Improved thread data handling for the OpenVDB Cycles node.

OpenVDB grid accessors, samplers, and ray intersectors must be stored per thread.
Previously this data was stored in a per-thread map in each grid. This causes problems
because the pthread_id keys can become invalid, and it also creates a lot of unused
accessors.

The patch switches thread data storage around, so that thread-local data is now passed
along KernelGlobals for every work task as needed (tasks only run in one thread).
This system is much more reliable and similar to other cases of thread-local data,
such as OSL shaders.

The patch also removes the virtual base class for "volumes", which might give a minor
performance improvement. If a generic volume type for Cycles is needed it can be
abstracted from scratch.

Removed the unused thread_id list in Cycles CPU tasks, which were only needed
for the now deprecated OpenVDB grid accessor lookup.

Gawain: strict error checking in debug builds

Not release builds. Fix 66d7f01eccdd9f82a4929f2873639e4012362090 to match what I meant to do.

Gawain: new enum type for vertex attrib components

Motivations:
1) GLenum is too broad; tightly-defined enum just for this is safer.
2) enable a Vulkan future

New code should use these instead of GL_FLOAT etc. When all existing code has been updated to use new enum, we can drop compatibility with GLenum values.

Early work towards 10_10_10 format, more to come soon.

removed obsolete high res grid access function

added numObstacle grid in manta script and access functions

Fix renaming error in last commit

BMesh: Minor improvement to face-join

Pass in loops instead of edge & faces.
Nearly all callers have the loop-pairs to pass in.

added missing grids to liquid/smoke grid import/export and also cleaned up scripts a bit

BKE_library_make_local(): some minor improvements.

Do not set 'real user' to groups every time we run the first clearing loop.

And do fully clear properly LIB_TAG_DOIT (this is not yet enforced in
existing code, but would love to get to that stage in future, so let's
do it at least with new code!).

Minor naming cleanup.

Enhance BKE_library_make_local() to make it much quicker in complex cases.

Basic idea is to split first loop in two, and run checks before making
anything actually local, to detect data-blocks that we can directly make
local (because we are sure they are only used by already/future local
datablocks).

This allows to avoid a lot of overhead in later 'cleanup' steps of this
function, here with barbershop shot it's four times quicker (from 190s to 48s).

We are still far from the instantaneous results of MakeLocal in 2.77,
but in that version main characters lose their connection to their
armature and remain static after makelocal, so guess new code is still
better. ;)

There are probably more optimizations possible here, but would rather
polish this area of code once we get rid of proxies, those really
make it a nightmare to work on.