Safe for 2.73...
This revert rB9b0ab890676790bb1e8e77797629b889ea66f69e - needed to set that threshold to a small
negative value to remove the last artefacts reported in T39735, but now I could not reproduce
any with the previous 0.0f value, so restoring it for the time being.
If this 'shadowed neighbor face' case re-appears, we can always choose a value in-between, like -1e-18f...
Murmur2a is a very fast hashing function generation int32 hashes.
It also features a very good distribution of generated hashes.
However, it is not endianness-agnostic, meaning it will usually generate
different hashes for a same key on big- and little-endian architectures.
Consequently, **it shall not be used to generate persistent hashes**
(never store them in .blend file e.g.).
This implementation supports incremental hashing, and is a direct
adaptation of reference implementation (in c++):
https://smhasher.googlecode.com/svn-history/r130/trunk/MurmurHash2.cpp
That cpp code was also used to generate reference values in gtests file.
Reviewers: sergey, campbellbarton
Reviewed By: campbellbarton
Projects: #bf_blender
Differential Revision: https://developer.blender.org/D892
The root of the issue comes to the way how we sample the gaussian filter in
RE_filter_value(). We need to scale x to -3*sigma..3*sigma segment in order
to get the whole bell.
The old code tried to do it, but failed dramatically, plus it used some weird
gaussian sampling formula. Replaced it with much more clear one, which gives
proper blur now.
There's no visible different in AA sampling in BI render tho.
Other filters like Mitchell still tends to give wrong square shaped blurs,
but they're much more difficult to resolve because they're just wrong in
the code -- for some reason smaller kernel size means more blur. Let's solve
this later.
Noise function's significant bits are up to 31st bit. This should now
give the same visual result as before, minus the stripes.
Issue pointed out by Anthony Edlin, thanks!
Two fixes here (only the second one is strictly needed to fix the issue,
but both make the system better).
First is introduction of a random generator array for use with threaded
systems where each thread needs to access its own number generator.
The random texture now uses this so it should not be influenced by other
random generator reseedings of the main random generator like it did
before.
Second, I reshuffled the texture code to resample the upper bits of the
random number first. According to Numerical Recipes, this is where the
most variance can be found, so by sampling those we avoid correlation
issues. Also, multiplying here is not ideal because if a pair of bits
are zero, then the whole result will also be zero.
Overall this is much more random (tm) than before, however result will
also be brighter, since we now have less black spots. Tweaking the
brightness/contrast should somewhat fix that, generally having the same
result as before is not possible anyway if we are to really fix this.
Also, seems like exposing procedural depth might be nice here since it
influences the precision of the texture lookup.
Currently only summed number of traversal steps and intersections used by the
camera ray intersection pass is implemented, but in the future we will support
more debug passes which would help checking what things makes the scene slow.
Example of such extra passes could be number of bounces, time spent on the
shader tree evaluation and so.
Implementation from the Cycles side is pretty much straightforward, could only
mention here that it's a build-time option disabled by default.
From the blender side it's implemented as a PASS_DEBUG with several subtypes
possible. This way we don't need to create an extra DNA pass type for each of
the debug passes, saving us a bits.
Reviewers: campbellbarton
Reviewed By: campbellbarton
Differential Revision: https://developer.blender.org/D813
Node was simply ignored by occ shading (noted as TODO), though it's a mere matter
of a very few lines of code, nowadays... Just copied from similar task in bake code.
The title says it all, move the EWA filter to BLI (currently it's
math_interp.c) and use the function from both BI renderer and the
compositor.
This makes more central place of the algorithm, allowing to have
fixes and optimizaitons synchronized across the two usages.
This also fixes T41440: Displacement in compositing creates holes
Reviewers: campbellbarton, lukastoenne
Reviewed By: lukastoenne
Maniphest Tasks: T41440
Differential Revision: https://developer.blender.org/D748
