| Age | Commit message (Collapse) | Author |
|---|
|
|
|
It'll be dependent on time via Time Source -> Rebuild RB World chain.
|
|
There is always an uber eval node on the way. so we can avoid creating
some relations here in order to speed up both construction time and
evaluation.
|
|
|
|
Something funny happened here, there were missing symbols from png
library to math functions.
|
|
This sets forces to zero, when Nabla is zero and a grayscale texture is
used or texture mode is Gradient or Curl.
Nabla equal to zero was causing a zero division, and forces ended up
being set to `nan`.
Reviewed By: mont29
Differential Revision: http://developer.blender.org/D2393
|
|
|
|
|
|
Own error when changing order,
moving experimental features last made some sense,
but causes them to be listed twice.
Reorder and comment to avoid it happening again.
|
|
- Expand overly dense & confusing delta assignments.
- Replace bit shift with multiply.
Also link to 'clipped' version of this function
which may be useful to add later.
|
|
|
|
|
|
These macros conflict and are no longer needed with C99 or C++ anyway.
|
|
The Progress system in Cycles had two limitations so far:
- It just counted tiles, but ignored their size. For example, when rendering a 600x500 image with 512x512 tiles, the right 88x500 tile would count for 50% of the progress, although it only covers 15% of the image.
- Scene update time was incorrectly counted as rendering time - therefore, the remaining time started very long and gradually decreased.
This patch fixes both problems:
First of all, the Progress now has a function to ignore time spans, and that is used to ignore scene update time.
The larger change is the tile size: Instead of counting samples per tile, so that the final value is num_samples*num_tiles, the code now counts every sample for every pixel, so that the final value is num_samples*num_pixels.
Along with that, some unused variables were removed from the Progress and Session classes.
Reviewers: brecht, sergey, #cycles
Subscribers: brecht, candreacchio, sergey
Differential Revision: https://developer.blender.org/D2214
|
|
Quite handy for debugging.
Unfortunately, this doesn't support viewport tweaks yet since those
require GLSL for colorspace conversion. Maybe this will be implemented
as well one day in the future..
|
|
Gives little performance improvement on Linux and gives up to 2%
speedup on koro.blend on Windows.
Inspired by Maxym Dmytrychenko, thanks!
|
|
They are defined for MSVC but seems to be missing in GCC and CLang-3.8.
Maybe some further tweaks to policy when to define those functions is
needed, but should be fine for now.
|
|
Mask primitive adding code was not initializing correctly id_type of
points' parents.
|
|
I can no longer reproduce crash with neither of the files where
the crash was originally visible. This is something where other
changes (light threshold, sampling) had an effect and made code
to work as it is supposed to. Could have been optimizator issue
or something like that.
Let's see if we hit same issue again.
|
|
`IMB_remakemipmap` may 'shrink' the mipmap list without actually freeing
anything, so we need to check all possible levels in `imb_freemipmapImBuf`
to avoid memory leaks, not only those currently used.
|
|
py/oiio/osl/osd/ffmpeg.
|
|
In ccgDM and emDM, looptri array recalculation was being handled
directly by `*DM_getLoopTriArray` (`getLoopTriArray` callback), while
`*DM_recalcLoopTri` (`recalcLoopTri` callback) was doing nothing.
This results in the array not being recalculated when other functions
that depend on the array data called the recalc function.
This moves all the recalculation code to `*DM_recalcLoopTri` and makes
`*DM_getLoopTriArray` call that.
This commit also makes a minor change to the `getNumLoopTri` function,
so that it returns the correct number without having to recalculate the
looptri array.
Reviewed By: mont29
Differential Revision: https://developer.blender.org/D2375
|
|
Quite a few of the operators added for 2.78 were not included in the menus
|
|
the operator is run using wait_for_input=True)
This just brings it in line with the drawing operator
|
|
|
|
|
|
In fact, the issue was caused by light threshold being too high for
certain scenes. Lowered it down to 0.01.
|
|
There were two cases where correlation issues were obvious:
- File from T38710 was giving issues in 2.78a again
- File from T50116 was having totally different shadow between
sample 1 and sample 32.
Use some more simplified version of CMJ hash which seems to give
nice randomized value which solves the correlation.
This commit will break all unit test files, but it's a bug fix
so perhaps OK to commit this.
This also fixes T41143: Sobol gives nonuniform noise
Proper science paper about hash function is coming.
Reviewers: brecht
Reviewed By: brecht
Subscribers: lukasstockner97
Differential Revision: https://developer.blender.org/D2385
|
|
|
|
Most of them are harmless implicit conversions (e.g. Alembic deals with
doubles for storing time information when Blender uses both ints and
floats/doubles) or class/struct mismatch on forward declarations.
|
|
events will not try to reference this deleted class.
|
|
This aims at always ensuring that ID.newid (and relevant LIB_TAG_NEW)
stay in clean (i.e. cleared) state by default.
To achieve this, instead of clearing after all id copy call (would be
horribly noisy, and bad for performances), we try to completely remove
the setting of id->newid by default when copying a new ID.
This implies that areas actually needing that info (mainly, object editing
area (make single user...) and make local area) have to ensure they set
it themselves as needed.
This is far from simple change, many complex code paths to consider, so
will need some serious testing. :/
|
|
Was a ground work for some more improvements here, but got dragged
to some other studio maintenance job here.
The plan would be to enable exposure/gamma control for fallback mode
which will definitely be really handy for development and might be
handy for cases when OCIO config can not be read.
|
|
|
|
Crash is due by mismatching loops and faces counts between the Alembic
data and the Blender derivedmesh which does not appear so
straightforward to fix (the crash happens deep in the derivedmesh code).
So for now, try to detect if the topology has changed and if so, both
only read vertices (vertex colors and UVs won't be read, as tied to face
loops) and add a warning message in the modifier's UI to let the user
know.
|
|
|
|
There are some changes in OIIO includes so now need to do some
things differently.
|
|
The idea is simple: cache PD resolution from cache_point_density() RNA
function because that one is supposed to be called while database is
locked for original synchronization.
Ideally we would also pass array size to the sampling function, but
it turned out to be quite problematic because API only accepts int type
and passing size_t might cause some weird behavior.
|
|
This is a way to avoid possible memory corruption when render threads works
in parallel with UI thread.
Not guarantees complete safe, but makes things easier to check anyway.
|
|
Seems to be a typo in recent commit e1e49fd.
|
|
|
|
for release builds, /opt:noref for debug builds.
|
|
It wasn't all that clear why both pixel-sizes are needed.
|
|
off the executable size
Nobody appears to know why this option was on (and just on for just x64 at that)
|
|
This is useful e.g. for search buttons to quickly clear the filter string. We might want to make this optional for python scripts.
|
|
|
|
causes the factory not to play
Backport of upstream audaspace bugfix (ddd9a4d).
|
|
Currently harmless since the value is always set at its default (32).
Even so, it's confusing since its used with floats after.
|
|
further crashes here
|
|
now true/enabled by default
To bring the API more into line with the UI (and the general expected behaviour of
Blender when it comes to adding stuff), newly created layers and palettes will be
made the active ones by default. It's possible to override this behaviour still
(e.g. in cases where you're auto-generating a large number of them), but otherwise,
this change will help prevent errors like T50123.
|
