Blender Git Commits

Blender Git commits from all branches.

Page: 1970 / 2888

July 11, 2016, 20:41 (GMT)
Cycles/Render API: Implement interactive updating while denoising

The tile highlighting is still a bit random and the progress bar isn't showing either,
but the basic live update works.
To avoid lots of duplicated code, editors/render and editors/space_image now share two functions.
July 11, 2016, 19:08 (GMT)
Alternative implementation for strand fiber interpolation without a geometry shader.

Geometry shaders easily become performance bottlenecks, for in-depth explanation see e.g.

http://gamedev.stackexchange.com/a/48434
http://rastergrid.com/blog/2010/09/history-of-hardware-tessellation/

The geometry shader was used so far for generating the fiber geometry (line strips)
from just the root vertices. Without a geometry shader every vertex has to be put into
a buffer and uploaded to the GPU (we could use a tesselation shader to cut down on size,
but that remains to be decided later if OpenGL 4.x is available).

To limit the size of this vertex buffer, the vertices contain only minimal necessary
information: an index for the fiber curve, and an interpolation curve parammeter in [0.0, 1.0].
The locations of vertices are interpolated in the vertex shader using texture lookups.
A range of textures encode all the necessary per-fiber attributes, in particular the
interpolation indices and weights wrt. control curves. These control curves are stored in
yet another texture (as is the case with the geometry shader too).

Initial performance seems to improve drastically without the geometry shader.
July 11, 2016, 17:37 (GMT)
Cycles: Fix Memory leak when denoising renderred images
July 11, 2016, 17:36 (GMT)
Render API: Implement the denoising process as a job to keep the UI responsive
July 11, 2016, 17:02 (GMT)
Cleanup layer writing code
July 11, 2016, 16:52 (GMT)
Get file read/write and undo to work - finally
July 11, 2016, 16:25 (GMT)
Cycles: Support denoising after rendering on CUDA
July 11, 2016, 15:22 (GMT)
bugfix for NFP computation

Also introduces some shiny new debug prints.
July 11, 2016, 14:42 (GMT)
attempt to make fast bisect usable with mousebased fracture... now the user has a bit control how the halving is performed via the helpers
July 11, 2016, 13:53 (GMT)
Merge branch 'master' into soc-2016-layer_manager

And make manual adjustments of merged parts to fit changes in branch.
Conflicts:
source/blender/blenkernel/intern/library_remap.c
source/blender/blenkernel/intern/object.c
source/blender/editors/animation/anim_filter.c
July 11, 2016, 10:06 (GMT)
Prevent infinite loops and make "scale around origin" it's own function
July 11, 2016, 08:53 (GMT)
Depth limited binary search to determine the scale factor of items which don't fit
July 11, 2016, 07:22 (GMT)
Added "effects" to the hair modifier for testing deformation of the hair in the shader.

These features will ultimately be implemented through nodes, but they are quite useful
for testing the shading itself and the performance.
July 11, 2016, 03:50 (GMT)
Change some tooltips.
July 11, 2016, 03:18 (GMT)
Merge branch 'master' into alembic_basic_io
July 11, 2016, 02:47 (GMT)
Merge branch 'master' into soc-2016-multiview
July 11, 2016, 02:46 (GMT)
Add keyframe selection in multi-view solver mode
July 11, 2016, 01:12 (GMT)
Merge branch 'master' into temp-cycles-microdisplacement
July 11, 2016, 00:04 (GMT)
Prepare for merge with master

Sergey's bvh work made use of and removed the same unused bytes that
were being used for storing patch info, so this info now has to be
moved elsewhere. Unfortunately we use up two more textures for this.
July 10, 2016, 23:12 (GMT)
Fix performance regression during mesh sync

The switch from std::vector to ccl::array in Mesh caused quite a big slowdown
during mesh sync. Using a quick fix for now. Proper thing to do would be to
avoid most memory allocation in subd code.
By: Miika HämäläinenLast update: Nov-07-2014 14:18MiikaHweb | 2003-2021