Blender Git Loki

Cycles:
* GPU kernel can now be compiled without __NON_PROGRESSIVE__ again, was broken after my last commit. Also add a check for have_error(), in case the GPU kernel comes without Non-Progressive, to avoid a crash.

* Don't compile progressive kernel twice on CPU, if __NON_PROGRESSIVE__ would be disabled there.

Fix cycles not rendering with 1x1 resolution anymore, not so common for users
but nice for debugging.

Fix cycles passes UI panel to line up better.

OpenGL VBO's: free VBO pool before redraw, otherwise this just holds onto memory
after objects are deleted until another big object is added. There's no good reason
to do this, or to think that our pool is somehow much faster than using the OpenGL
API to allocate and free buffers.

Cycles / Non-Progressive integrator:
* Non-Progressive integrator is now available on the GPU (CUDA, sm_20 and above).

Implementation details:
* kernel_path_trace() has been split up into two functions:
kernel_path_trace_non_progressive() and kernel_path_trace_progressive().

* We compile two CUDA kernel entry functions (in kernel.cu) for the two integrators, they are still inside one .cubin file but due to the kernel separation there should be no performance problem. I tested with the BMW file on my Geforce 540M and the render times were the same for 100 samples (1.57 min in my case).

This is part of my GSoC project, SVN merge of r59032 + manual merge of UI changes for this from my branch.

Merged revision(s) 58994-59032 from trunk/blender into soc-2013-dingto.

Cycles / Non-Progressive integrator:
* Code refactor to split the GPU kernel into two, one for each integrator.
This way we can enable Non-Progressive integrator on GPU in trunk without a performance drop.

Thanks to Brecht for some help and review!

Fix #36146, Cycles SSS Shader Node - Scale Value Input Box - to sensitive. Changed the ui limit from 1,000,000 to 1,000. Not sure if such huge values are needed in some cases, but they can always be
entered explicitly, bypassing the ui limit.

NOTE: this only applies to newly added nodes. For legacy reasons the limit values are stored directly in each socket, so changing them for existing nodes would require a do_versions, seems unnecessary
for such a minor change.

Fix #36417: vertex weight mix modifier crash with texture mask and subsurf modifier
preceding it, was writing past end of array.

Fix #36298 and #36359: timeline no longer did redraw during animation render.

This got broken in 58217, but before that it only worked by accident because
of a missing break statement in a switch.

Fix #36274: empty particle texture slots should not have any influence, same
as for materials, world, lamps, etc.

Fix incorrect display of constraint names on retina / high DPI.

Fix #36317: keyframing metaball elements did not work yet, now you can keyframe
properties like stiffness.

Fix #36243: one screen should never be used by multiple windows, but this could
happen when a screen was maximized to a single editor, now that case is disabled
too.

Fix #36394: rendering a point density texture with a hair particle system did
not work correct, the positions would change over time and child particles did
not render.

Linking Brushes did not link the mask texture properly

pass compiler defines such as __FLT_MIN__ to smatch and sparse (they fail without them).

disable cycles when building without python, also use const for function args in more places.

Fix for a formatting error in the docstring of SVertex.viewvertex().

Fix for docstrings of CurvePoint.first_svertex() and CurvePoint.second_svertex().