* Added CUDA Binaries to Windows scons config files, disabled by default.
Also the path to nvcc is commented out, best to set this in the user config, as it varies on each system probably.
- editmesh smooth & subdivide were using old mirror axis flag still.
- removed colbits from outliner and object code.
- commented some other parts of the code which access deprecated members and aren't called anywhere.
this means use of deprecated struct members gives a warning.
- makesdna.c preprocessor skips this.
- DNA_DEPRECATED_ALLOW is used so readfile.c can do versioning without warnings.
- this exposes some use of deprecated struct members, will deal with this after.
as with the HSV node the OSL code is relying on the (yet to be implemented) autorename.
Also the svm code could use mix (svm_lerp) instead:
32 . float3 color_inv = make_float3(1.0f, 1.0f, 1.0f) - color;
35 . . stack_store_float3(stack, out_color, svm_lerp(color_inv, color, factor));
I have a feeling that each node 'program' should have the least program as possible. I'll see with Brecht later.
But overall I don't know if that's any fast. And apart from that I think we will need this kind of function to move to a library if multiple functions linked in are not a problem.
add it as an option (under "Hybrid") in the tracking settings. The
region tracker is a combination of brute force tracking for coarse
alignment, then refinement with the ESM/KLT algorithm already in
libmv that gives excellent subpixel precision (typically 1/50'th
of a pixel)
This also adds a new "brute force" region tracker which does a
brute force search through every pixel position in the destination
for the pattern in the first frame. It leverages SSE if available,
similar to the SAD tracker, to do this quickly. Currently it does
some unnecessary conversions to/from floating point that will get
fixed later.
The hybrid tracker glues the two trackers (brute & ESM) together
to get an overall better tracker. The algorithm is simple:
1. Track from frame 1 to frame 2 with the brute force tracker.
This tries every possible pixel position for the pattern from
frame 1 in frame 2. The position with the smallest
sum-of-absolute-differences is chosen. By definition, this
position is only accurate up to 1 pixel or so.
2. Using the result from 1, initialize a track with ESM. This does
a least-squares fit with subpixel precision.
3. If the ESM shift was more than 2 pixels, report failure.
4. If the ESM track shifted less than 2 pixels, then the track is
good and we're done. The rationale here is that if the
refinement stage shifts more than 1 pixel, then the brute force
result likely found some random position that's not a good fit.
----------------------------
reviewed and approved by Brecht
Important note:
the camera Z is reverted compared to Blender render.
Now it goes from zero (camera) to positive (in front of the camera)
.........................
note, the OSL code has a problem.
In the original node the input and output nodes have the same name (Color).
So this will be fixed here once Brecht come up with a nice autorenaming (or we do a doversion patch) for that.
Array indexing doesn't work there.
I'm yet to setup my CUDA computer, in the meantime this proved to work (tested by Daniel Salazar).
If I found other ways of doing it I get back to that.
Talked with Brecht and Campbell and they both agreed that bpy.types should match bpy.props
In the ideal world we would rename bpy.props to BooleanProperty. This would break scripts though. So we go for a compromise and at least have some consistency.
MSVC is not C99-compliant, but with the macro as it was, the "args" part was consuming and dropping one of the params, so the format string wanted N extra args, and there were only N-1... boom, crash! Without "args" this works correctly (checked on MSVC10 release build)
Crash was caused by non-initialized original index passing to compareDrawOptions
callback. Due to in some cases it's enough to know indices of faces in final DM
(like for comparing if the same texture used for mesh drawing) assume this
callback receives index in final DM and if it's needed it will make conversion
to original index itself.
This should help reaching extra speed if textured object is affected by
bevel modifiers, i.e.
* Compile all of cycles with -ffast-math again
* Add scons compilation of cuda binaries, tested on mac/linux.
* Add UI option for supported/experimental features, to make it
more clear what is supported, opencl/subdivision is experimental.
* Remove cycles xml exporter, was just for testing.
