Blender Git Commits

Blender Git commits from all branches.

Page: 1899 / 2888

November 14, 2016, 15:32 (GMT)
Merge branch 'multi_previews_id' into temp-sybren-poselib
November 14, 2016, 15:30 (GMT)
Partially revert rBd089bfa9f094.

For now, we do want to keep direct access to raw data as ints/floats.
Time will say whether we want to keep those in final version or not.

Also, minor tweaks to 'set from image' functions.
November 14, 2016, 14:50 (GMT)
Ensure given frame index is valid in drawing icon code.
November 14, 2016, 14:49 (GMT)
Revert "Somewhat gracefully handle frame index out of bound error."

This is not to be handled in BKE, but in calling code, BKE only asserts
values are valid.

This reverts commit e3fd455d3293f7b8bd15491a3e0193002b648abf.
November 14, 2016, 14:49 (GMT)
Merge branch 'master' into multi_previews_id
November 14, 2016, 11:25 (GMT)
Cycles: Store CPU kernel function pointers in objects to clean up the code

With the number of kernel functions that's currently needed, it was just getting to messy.
This commit is based on D1825.
November 14, 2016, 11:25 (GMT)
Cycles: Implement NLM-weight filtering kernel
November 14, 2016, 11:25 (GMT)
Cycles: Split up the estimate_params kernel
November 14, 2016, 10:08 (GMT)
Cycles: Add a NLM filter for testing purposes
November 14, 2016, 10:08 (GMT)
Cycles: Split up the denoising feature row functions
November 14, 2016, 10:07 (GMT)
Cycles: Fix a unused variable warning
November 13, 2016, 20:59 (GMT)
Removed unused util_volume files.
November 13, 2016, 20:42 (GMT)
Merge branch 'cvdb_ray_isect' into openvdb
November 13, 2016, 16:59 (GMT)
Set the primitive type for volume samples to VOLUME.

This is so the primitive_attribute_* functions correctly look for
voxel attributes.
November 13, 2016, 16:40 (GMT)
Cycles: Fix two debugging problems in the Denoiser

The mismatching alloc/dealloc would upset ASan, and the undefined elements would cause FPEs in the NLM code (the result was fine due to masking, but the intermediate values weren't).
November 13, 2016, 16:39 (GMT)
Cycles: Fix CPU Denoising kernels after recent SSE-for-vector-types commit

That change (enabling __KERNEL_SSE__ for AVX(2)) broke the ABI between the kernel and device code when int4 was passed as an argument.
Now, a pointer to the first element is passed instead.
November 13, 2016, 14:14 (GMT)
Fix various cleanup issues and inconsistencies.

All volumes are held in the scene volumes list now, so VolumeManager doesn't need
to free them explicitly.

The ImageManager reference in VoxelAttribute is an artifact from smoke rendering,
this should be phased out eventually and can be assumed to be NULL for Volumes.

Make sure the VolumeManager cleans up the device memory properly, in particular
the ray intersector instances must be freed.
November 13, 2016, 12:40 (GMT)
Proper volume data storage in Cycles when syncing from Blender data.

Previously a viewport render would continuously keep adding Volume instances
to the VolumeManager data, because there was no way to detect existing volume
data. Now the `id_map` model known from meshes and other data types is used for
volume data as well.

Note that the VolumeKey currently simply uses an Object ID pointer, thus
assuming a single Volume per object. In the future volumes could become a
real ID type in Blender, or the specific volume could be narrowed down with
additional info in the key.
November 13, 2016, 10:25 (GMT)
Added an assert to catch non-uniform grids earlier.

OpenVDB ray intersectors only support uniform grids. This should be
checked as early as possible to allow feedback to the Cycles user
(i.e. the Blender app). For now the assert just prevents exceptions
from OpenVDB itself, until a transparent mechanism for this is implemented.
November 13, 2016, 10:24 (GMT)
Removed direct grid lists in VolumeManager, all grids should be in Volumes now.
By: Miika HämäläinenLast update: Nov-07-2014 14:18MiikaHweb | 2003-2021