This was introduced on 9ad2c0b615.
Although this still doesn't fix the issue, it updates the preview
system to use COLLECTION_DISABLED as intended.
What is missing now is for the flushing to work effectively.
This add the possibility to add screen space raytraced shadows to fix light leaking cause by shadows maps.
Theses inherit of the same artifacts as other screenspace methods.
Two issues here:
- Checking table size to be non-zero is not a proper way to go here. This is
because we first resize the table and then fill it in. So it was possible that
non-initialized table was used.
Trickery with using temporary memory and then doing table.swap() might work,
but we can not guarantee that table size will be set after the data pointer.
- Mutex guard was useless, because every thread was using own mutex. Need to
make mutex guard static so all threads are using same mutex.
Tried 101 but it gives colisions.
I think 257 is enough now that we dont have thousands of uniforms.
This gives some noticeable performance improvement.
Could be refined further.
The issue was caused by light sample being evaluated to nan at some point.
This is root of the cause which is to be fixed, but is very hard to trace down
especially via ssh (the issue only happens on AVX2 release build). Will give it
a closer look when back to my AVX2 machine.
For until then this is a good check to have anyway, it corresponds to what's
happening in regular radiance sum.
This changes quite a few things:
- Drops the allocation of inputs as a chunk.
- Merge the linked list system into the Gwn_ShaderInput.
- Put name buffer into another memory block, easily resizable.
- Use offset instead of char* to direct to input name.
- Add only requested uniforms dynamicaly to the Shader Interface.
This drops some minor optimisation and use a bit more memory for small shaders (which are fixed count).
But this saves a lot of memory when using UBOs because the names and the Gwn_ShaderInput were alloc'ed for every UBO variable.
This also reduce the Shader Interface initial generation.
The lookup time is left unchanged.
Camera clipping was left to default values, which won't work well for
very large (or small) objects. Now recompute valid clipping start/end
based on boundingbox of rendered data, and final location of camera.
This makes brush influence into a tube instead of a sphere.
It can be used along the outline of a mesh to adjust it's silhouette.
Note that all this takes advantage of changes from vertex paint,
from testing this seems useful so exposing from the brush options.
This is an internal structure, and we don't put it to a list for anything else
that hash collision resolution. No need to have dedicated entry here, saves us
from extra allocation and pointer dereference.
This way we reduce number of loops from look-over-all-inputs to
loop-over-collision, which is expected to be much less CPU ticks.
There is still possible optimization: use memory pool of some sort
to manage memory needed for hash entries, but that will only speedup
shader interface construction / deconstruction time.
There are also some trickery happening to speed up process even more
in the case there is no hash collisions detected when constructing
shader interface.
This behavior makes more sense for sculpt, less so for painting.
Restores non PBVH behavior, adding `BKE_pbvh_find_nearest_to_ray` -
similar to ray-cast except it finds the closest point on the surface.
The work size is still very conservative, and this doesn't help for progressive
refine. For that we will need to render multiple tiles at the same time. But this
should already help for denoising renders that require too much memory with big
tiles, and just generally soften the performance dropoff with small tiles.
Differential Revision: https://developer.blender.org/D2856
This was originally done with the first sample in the kernel for better
performance, but it doesn't work anymore with atomics. Any benefit was
very minor anyway, too small to measure it seems.
- Separate the Post Processes settings into sub panel.
- Rename "Viewport Anti-Aliasing" to sampling & super-sampling as it also reduce the noise of other effects.
- Remove Temporal Anti-Aliasing toggle and make it always active unless the number of samples is 1.
Restoring weights is problematic when the stroke overlaps its mirror.
It's better to simply compute the new weight based on the saved data
rather than restoring things, and check that the change is monotonic.
This way is also closer to how things worked before the merge.
This secondary accumulation option accumulated brush falloff.
The same option in image painting accumulates color
as vertex paiht 'Spray' does.
Giving this option different behavior for vertex paint seems strange.
Also this is basically increasing falloff over time.
Remove the new code, expose existing 'Spray' as 'Accumulate'
to match other paint modes.
Notes:
- Changes in paint_vertex.c were simple to merge, mainly related on passing
evaluation context.
- Conflicts in EditDM and drawmesh.c are solved using code from blender2.8
branch. Those areas are deprecated and not to be used in final release.
However, it's possible that some reference code from master is lost, so
keep attention when adding alpha support for vertex painting.
The issue here is that we can not read scale from socket when determining
dependent area of interest. This area will depend on current pixel. Now fall
back to more stupid but reliable thing: if scale size input is connected to some
nodes, we use the whole frame as area of interest.
This makes vertex paint match image painting more closely.
- Add falloff shape option sphere/circle
where sphere uses a 3D radius around the cursor and
circle uses a 2D radius (projected), like previous releases.
- Add normal angle option so you can control the falloff.
- Add Cull option, to paint onto faces pointing away.
Disabling normals, culling and using circle falloff
allows you to paint through the mesh.
When painting with spray disabled - we need to re-apply
on top of the original each time.
Applying the soc-2016-pbvh-painting branch removed this.
While I'd added back a simple previous weight array,
this won't work when multiple groups are painted at once.
This is most pronounced in Auto-Normalize + Multi-Paint. Unlike
vertex paint, the weights being painted on in weight paint mode
don't necessarily correspond to the weight actually stored in
any one vertex group, and may instead be a computed aggregate.
This restores original code behavior lost in rB4f616c93f7cb.
This required some small changes to the data display shaders so that they match the way the object mode renders them.
Strangely enough, I had to remove the normal attribute from the display code because it was being not bound as soon as I created another rendering call in object mode. The problem may be deeper but I did not have time for this so I derive the normal from the sphere pos.
Note that this tool seems like it might need to be rewritten
since results are quite strange.
Projecting on the view vector gives a small improvement though.
paint_vertex.c was getting too big, move all code unrelated to
mode switching and modal painting into their own files.
Also replace vertex-color operators region redraw tag /w notifiers.
GSOC 2017 by Darshan Kadu, see: D2859.
This is a partial merge of some of the features from
the soc-2017-vertex_paint branch.
- Alpha painting & drawing.
- 10 new color blending modes.
- Support for vertex select in vertex paint mode.
This removes a bunch of code that is no longer needed, and running
"make update" will now automatically download the new libraries.
Differential Revision: https://developer.blender.org/D2861
This is done by storing only a subset of PathRadiance, and by storing
direct light immediately in the main PathRadiance. Saves about 10% of
CUDA stack memory, and simplifies subsurface indirect ray code.
This is really convenient for development. Either for profiling the
generated shaders or to check if the generated code is correct.
It writes the shaders to the temporary blender session folder.
2016 GSOC project by @nathanvollmer, see D2150
- Mirrored painting and radial symmetry, like in sculpt mode.
- Volume based splash prevention,
which avoids painting vertices far away from the 3D brush location.
- Normal based splash prevention,
which avoids painting vertices with normals opposite the normal
at the 3D brush location.
- Blur mode now uses a nearest neighbor average.
- Average mode, which averages the color/weight
of the vertices within the brush
- Smudge mode, which pulls the colors/weights
along the direction of the brush
- RGB^2 color blending, which gives a more accurate
blend between two colors
- multithreading support. (PBVH leaves are painted in parallel.)
- Foreground/background color picker in vertex paint
by the transform constraint lines
Ported over e7395c75d5 from the
greasepencil-object branch. I should've fixed this ages ago, but
couldn't figure out why at the time.
This adds TAA to eevee. The only thing important to note is that we need to keep the unjittered depth buffer so that the other engines are composited correctly.
The problem was that orthographic views can have hit position that are negative. Thus we cannot encode the hit in the sign of the Z component.
The workaround is to store the hit position in screenspace. But since we are using floating point render target, we are loosing quite a bit of precision.
TODO: use RGBA16 instead of RGBA16F. But that means encoding the pdf value somehow.
You can change the amount of samples in the user preferences. You do not need to restart blender to see the effect in the new viewport.
This adds another Multisample Framebuffer and textures (so even more memory required).
It works by blitting the default_fb to the multisample_fb each time the renderer need to render one or more "wire" pass.
It it then blit back to the default_fb so that the rest of pipeline is working as expected.
We COULD lower the GPU memory / bandwidth usage to render everything to the same multisample fbo and change the logic depending on if MSAA is enabled or not, but I think it's a bit too much work for now.
One crucial thing here: OpenVDB shoudl be compiled WITHOUT
OPENVDB_ENABLE_3_ABI_COMPATIBLE flag. This is how OpenVDB's Makefile is
configured and it's not really possible to detect this for a compiled library.
If we ever want to support that option, we need to add extra CMake argument and
use old version 3 API everywhere.
Even if pointer assignment may be atomic, it does not prevent reordering
and other nifty compiler tricks, we need a memory barrier to ensure not
only that transferring pointer from wip array to final one is atomic,
but also that all previous writing to memory are “flushed” to
(visible by) all CPUs...
Thanks @sergey for finding the potential (though quite unlikely) issue.
Now uses "Text Selected" theme color of "Menu Back" widget colors. Also
repositioned text slightly to have same margin on top and right (measured
by eye ;) ).
Tested with all bundled themes (contrib and no-contrib) and worked fine.
Considering that different splashes may need different colors for
overlaid text, using theme color may not be the best solution. I would
like to try how this works before adding an ugly way to force a certain
text color though.
Also tried different approaches, but this one I find the least ugly :S
As far as longer term plans go, we wanted to get a redesigned multi-page
splash screen anyway. At this point we can rethink how splash colors work
in general (i.e. auto-contrast, own splash theme colors, etc).
It has been deprecated since at least macOS 10.9 and fully removed in 10.12.
I am unsure if we should remove it only in 2.8. But you cannot build blender with it supported when using a modern xcode version anyway so I would tend towards just removing it also for 2.79 if that ever happens.
Reviewers: mont29, dfelinto, juicyfruit, brecht
Reviewed By: mont29, brecht
Subscribers: Blendify, brecht
Maniphest Tasks: T52807
Differential Revision: https://developer.blender.org/D2333
Adds a FXAA for smoothing out the extracted outlines.
The Post Process Anti Aliasing is only done on the Alpha channel of the outlines.
Because of that we need to add bleed the outline color out of the silouhette so the AA'd alpha can blend the right color and not pick black when the alpha is smoothed out of the silhouette.
Also because of the AA needs to have clear contrast to work with, I decided to ditch the "bluring" or the occluded outlines.
The FXAA adds an overhead of 0.17ms but we gain back 0.22ms * 4 = 0.88ms by removing the blur.
The FXAA Implementation is from Corey Richardson (cmr) (D2717). I had to modify it a bit to only filter the alpha channel.
This introduce some little artifacts on the border of edges because some pixel with very low opacity does not get discarded and then occlude the face rendered behind if it has not been drawn yet.
To fix this. I added an offset in the geometry shader for the edge fixup. This make the artifact only visible on the border of the object if there is a very dense wire region. It's only visible in edge select mode since vertex and face center also hides the artifacts.
We can enable this only if AA is enabled but for now it's always enabled.
Introducing a new header using the Blender socket logo,
commit of the source file will follow soon.
Splash committee: Ton Roosendaal, Dalai Felinto, Pablo Vazquez.
Artwork is a screenshot of 'Wanderer', an Eevee sample file
by Daniel Bystedt, available on blender.org (license: CC-BY-SA)
Textures were bound once. But since it was not unbound it's bind_num would not change and considered still bound next time a shader needed it.
Fix T52866
Fix T52855
Partial revert of 9068c0743e.
This commit tried to do two things:
(1) Fix UBO binding logic [good]
(2) "Improve" texture binding logic [bad]
Don't ever mix different fixes and refactors in the same commit.
Iterate over invisible objects too, so lamps can still lit the scene.
Also, now you can use a collection to set an object to invisible, not
only to visible.
For example:
Scene > Master collection > bedroom > furniture
Scene > View Layer > bedroom (visible)
> furniture (invisible)
The View Layer has two linked collections, bedroom and furniture.
This setup will make the furniture collection invisible.
Note: Unlike what was suggested on D2849, this does not make collection
visibility influence camera visibility. I will keep this as a separate
patch.
Reviewers: sergey
Subscribers: sergey, brecht, fclem
Differential Revision: https://developer.blender.org/D2849
-No more hardcoded python35/36 tokens in the scripts
-disabled python module for boost, was not used
-Updated patches for python to support building with msvc2013
For the first bounce we now give each BSDF or BSSRDF a minimum sample weight,
which helps reduce noise for a typical case where you have a glossy BSDF with
a small weight due to Fresnel, but not necessarily small contribution relative
to a diffuse or transmission BSDF below.
We can probably find a better heuristic that also enables this on further
bounces, for example when looking through a perfect mirror, but I wasn't able
to find a robust one so far.
Similar to what we did for area lights previously, this should help
preserve stratification when using multiple BSDFs in theory. Improvements
are not easily noticeable in practice though, because the number of BSDFs
is usually low. Still nice to eliminate one sampling dimension.
Previously the Sobol pattern suffered from some correlation issues that
made the outline of objects like a smoke domain visible. This helps
simplify the code and also makes some other optimizations possible.
Right now this is exposed in the outliner, though all this
(visible/selectable/enable) should be moved to a new panel soon.
This removes objects from the depsgraph when the collection is disabled.
It allows you to "hide" lamps but still having them lighting the scene.
Same for light probes and other support objects.
Pending tasks:
* Have depsgraph to include invisible objects in the DEG_OBJECTS_ITER, and
then have Eevee and other engines to make a distinction between an
invisible and a visible object.
(for example, we probably want invisible objects to not show in the
viewport, but cast shadows and show up in light probes).
* Change how we evaluate collection settings so that an invisible
collection can force an object to be invisible.
Reviewers: campbellbarton
Subscribers: sergey
Differential Revision: https://developer.blender.org/D2848
Now we replace O(N^2) computational complexity with O(N) extra memory penalty.
Memory is much cheaper than CPU time. Keep in mind, memory penalty is like
4 megabytes per 1M vertices.
Tentative fix, since I cannot reproduce thenissue for some reason here
on linux.
Core of the problem is pretty clear though, thanks to Germano Cavalcante
(@mano-wii): another thread could try to use looptris data after worker
one had allocated it, but before it had actually computed looptris.
So now, we use a temp 'wip' pointer to store looptris being computed
(since this is protected by a mutex, other threads will have to wait on
it, no possibility for them to double-compute the looptris here).
This should probably be backported to 2.79a if done.
The issue was caused by threading conflict around looptris: it was possible
that DM will return non-NULL but non-initialized array of looptris.
Thanks Campbell for second pair of eyes!
Use triple buffer by default now on all platforms, remaing ones where:
* Mesa: seems to have been working well for a long time now, and not using
it gives issues with the latest Mesa 17.2.0.
* Windows software OpenGL: no longer supported since OpenGL 2.1 requirement
was introduced.
* OS X with thousands of colors: this option was removed in OS X 10.6, and
that's our minimum requirement.
Ubo needs to be rebound every times the shader changes.
This simplify the logic a bit.
Also modify texture binding logic to potentially reuse more already bound textures.
When 2x loops have different number of vertices,
the distribution for vertices fan-fill depended on the loop order
and was often lop-sided.
This caused noticeable inconstancies depending on the input
since edge-loops are flipped to match each others winding order.
This fix the crappy binding logic.
Note the current method is doing a lot of useless binding. We should somewhat order the texture so that reused textures are already bound most of the time.
Use more watertight and robust intersection test.
It uses now ray to triangle intersection, but it's all fine because segment was
covering the whole bounding box anyway.
Mainly when object origin is not at the geometry bounding box center.
Seems to be straightforward to fix, hopefully it doesn't break some obscure case
where this was a desired behavior.
The issue here was that removing datablock from main database will poke editors
update, which includes buttons context to free users of texture. Since Cycles
will free datablocks from job thread, it might crash Blender since main thread
might be in the middle of drawing.
Solved by exposing extra arguments to bpy.data.foo.remove() which indicates
whether we want to perform ID user count and interface updates. While scripts
shouldn't be using those normally, this is the only way to allow Cycles to skip
interface update when removing datablock.
Reviewers: mont29
Reviewed By: mont29
Differential Revision: https://developer.blender.org/D2840
The issue was caused by render result identifier only consist of scene name,
which could indeed cause conflicts.
On the one hand, there are quite some areas in Blender where we need identifier
to be unique to properly address things. Usually this is required for sub-data
of IDs, like bones. On another hand, it's not that hard to support this
particular case and avoid possible frustration.
The idea is, we add library name to render identifier for linked scenes. We use
library name and not pointer so we preserve render results through undo stack.
Reviewers: campbellbarton, mont29, brecht
Reviewed By: mont29
Differential Revision: https://developer.blender.org/D2836
Add sanitizer. I wanted to stay away from this because I think we should fix what causes NaNs in the first place. But there can be too much different factor causing NaNs and it can be because of user inputs.
Was caused by numeric overflow when calculating preview dimensions.
Now we try to avoid really insance preview resolutions by fitting
aspect into square.
The branching introduced by the uniform caused problems on mesa + AMD in the resolve stage.
This patch create one shader per sample count without branching.
This improves performance of a single ray per pixel case (3.0ms against 3.6ms in my testing)
The issue was caused by operator redo which frees all object's evaluated data,
including bounding box. This bounding box can not be reconstructed properly
without full curve evaluation (need to at least convert font to nurbs, which is
not cheap already).
There is absolute no reason to have such an indentation level, it only causes
readability and maintainability issues. It is really simple to make code more
"streamlined".
This was cause by a fairly funky unitialize buffer (last frame) that was causing NANs during the SSR resolve stage.
They were then propagated to the whole image during the next swap.
Bypassing the SSR completly if no valid history exists fixes the problem. Also disabling SSR data output in this case so we can have correct reflection in the 1st history buffer.
Rather than treating all ray types equally, we now always render 1 glossy
bounce and unlimited transmission bounces. This makes it possible to get
good looking results with low AO bounces settings, making it useful to
speed up interior renders for example.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D2818
You can now use a transparent shader as a completly transparent bsdf. And use whatever alpha mask in a mix shader between a transparent bsdf and another bsdf.
In fact, any type of baking might have caused holes in mesh.
The issue was caused by zspan_scanconvert() attempting to get order of traversal
'a-priori', which might have failed if check happens at the "tip" of span where
`zspan->span1[sn1] == zspan->span2[sn1]`.
Didn't see anything bad on making it a check when iterating over scanlines and
pick minimal span based on current scanline. It's slower, but unlikely to cause
measurable difference. Quality should stay the same unless i'm missing something.
Reviewers: brecht, dfelinto
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D2837
Previously we used a 1D sequence to select a light, and another 2D sequence
to sample a point on the light. For multiple lights this meant each light
would get a random subset of a 2D stratified sequence, which is not
guaranteed to be stratified anymore.
Now we use only a 2D sequence, split into segments along the X axis, one for
each light. The samples that fall within a segment then each are a stratified
sequence, at least in the limit. So for example for two lights, we split up
the unit square into two segments [0,0.5[ x [0,1[ and [0.5,1[ x [0,1[.
This doesn't make much difference in most scenes, mainly helps if you have a
few large area lights or some types of HDR backgrounds.
This causes render differences in some scenes, for example fishy_cat
and pabellon scenes render brighter in a few spots. This is an old
bug, not due to recent RR changes.
This is in order to use the same texture on multiple sampler.
Also texture counter is reset after each shading group. This mimics the previous behaviour.
Just using same code for distribution for face/volume as the one
changed/used for vertices since some months.
Note that this change is breacking compatibility, in that distribution
of particles over faces/volume may not be exactly the same as
previously.
Previously only the active object was used.
Use coroutines to support baking frames for multiple objects at once,
without having to playback the animation multiple times.
- Replace poisson by concentric samples: Less variance. They are sorted by radius then by angle.
- Separate filtering into 2 blur. First blur is 3x3 box blur. Second is user dependant.
- Group fetches by group of 4.
This brings some data structure changes.
Shared shadow data are stored in ShadowData (in glsl) (aka EEVEE_Shadow in C).
This structure contains the array indices of the first shadow element of this shadow "object".
It also contains how many shadow to evaluate (to be used for Multiple shadow maps).
The filtering is noisy and needs improvement.
- Use only one 2d texture array to store all shadowmaps.
- Allow to change shadow maps resolution.
- Do not output radial distance when rendering shadowmaps. This will allow fast rendering of shadowmaps when we will drop the use of geometry shaders.
- Use indices instead of character args.
- Use numbered macros instead of variadic args.
Parsing using rtags used over 11gb of memory. While this should be
resolved upstream (report as #1053), the extra complexity didn't give
any real advantage.
Disabled forceinline for those architectures, which seems to be compiling
successfully more often.
There might be ~3% slowdown based on quick tests, but better be rendering
something rather than failing to compile kernels again and again.
Those architectures will be doomed for abandon once we'll switch to toolkit 9.
Empty BVH nodes are set to NaN which must be preserved all the way to the
tnear <= tfar test which can then give false for empty nodes. This needs
strict semantices and careful argument ordering for min() and max(), so
the second argument is used if either of the arguments is NaN.
Fixes T52635: crash in BVH traversal with SSE4.1.
Differential Revision: https://developer.blender.org/D2828
There was some invalid state in the screen here, some areas had
sa->full set even though no screen was maximized, which then caused
a restore from the wrong (empty) area, which then led to spacedata
being empty and a crash.
This fix properly clears the sa->full after restore, and also fixes
existing .blend files in such an invalid state.
Only the camera from View3D.localvd is used,
other pointers may be invalid.
Longer term we should probably clear these to ensure no accidents.
For now just follow the rest of Blender's code and don't access.
Baking rigid body cache was broken if some cached frames already
existed.
This is just a band aid for release, the logic need to be looked into
further.
Since the change to prevent shader recompilation at every update, we got
a regression when clearcoat was used.
Basically at the shader build time we would determine if the shader
needed clear coat, and if it didin't, it would build a different GLSL
program.
However if later the user updated the clearcoat value so that it would
then require the full clearcoat shader, the user wouldn't get it until
manually forcing the shader to recompile, or reopening the file.
We now handle the optimization in the GLSL code. That adds a minimum
overhead due to branching. But the overall performance seems unchanged
(tested on linux in AMD and NVidia).
Reviewers: pascal, brecht, fclem
Differential Revision: https://developer.blender.org/D2822
Operators and their properties are two different types
Previously both operators and their properties are added
causing C operators to access the properties, Python the classes.
Favor consistency in this case so only Python classes are added.
A single diagonal axis was used for sorting coordinates,
the algorithm relied on users not having vertices axis aligned.
Use BLI_kdtree to remove doubles instead.
Overall speed varies, it's more predictable than the previous method.
Some typical tests gave speedup of ~1.4x - 1.7x.
For example, if you have two keyframes:
k1 = 1px, k2 = 10px
it was doing:
1px, 9px, 8px, ..., 3px, 2px, 10px
instead of:
1px, 2px, 3px, ..., 8px, 9px, 10px
This reverts commit 134e927965.
Writing into const event is very bad,
but this change broke compositor manipulators.
Will look into better solution eventually.
For some specific pipelines (e.g., holographic rendering) you can easily
need over a million frames (1k * 1k view angles).
It seems a corner case, but there is no real reason not to allow users
doing that.
That said we do loose subframe precision in the highest frame range. Which can
affect motionblur. The current maximum sub-frame precision we have is 16.
While the previous limit of 500k frames has a precision of 32.
Thanks to Campbell Barton for the help here.
To be backported to 2.79
Own error in recent type checks, in many cases the 'idname'
is used for the struct identifier, not the 'identifier'
which is the Python class name in this context.
Mostly internal changes, keeping both manipulators
could have worked but there was no point long term.
There are still some glitches to resolve, will work on those next.
Basically since g_data was malloc'ed (instead of calloc'ed)
g_data->minzbuffer was never initialized.
So when running DRW_framebuffer_init after EEVEE_effects_init, the test
to *g_data->minzbuffer would lead to unpredictable results.
This was caught by valgrind, reported by Sergey Sharybin.
Previous version was trying to do a quick and simple process in the case
we were only considering smooth/flat status of faces.
Thing is, even then, the algorithm was not actually working in all
possible situations, e.g. two smooth faces having a single vertex in
common, but no common edges, would not have split that vertex, leading
to incorrect shading etc.
So now, tweaked slightly our split normals code to be able to generate
lnor spaces even when autosmooth is disabled, and we always go that way
when splitting faces.
Using smooth fans from clnor spaces is not only the only way to get 100%
correct results, it also makes face split code simpler.
One problem is that it was always using __mm_blendv_ps emulation even if the
instruction was supported. The other that the emulation function was wrong.
Thanks a lot to Ray Molenkamp for tracking this one down.
Apparently with Maya in a certain configuration, it's possible to have an
Alembic object without schema in the Alembic file. This is now handled
properly, instead of crashing on a null pointer.
This affects the curve display color setting, but is really intended
for future per-curve options.
The id_data reference in the created rna pointers refers to the object
even if the curve is actually owned by its action, which is somewhat
inconsistent, but the same problem can be found in existing code.
Fixing it requires changes in animdata filter API.
The is following: split copy on write update for node trees, and if we are only
tagging for uniform buffer update we skip whole datablock copy and only invoke
copy default_values form original nodetree to a copied one.
Thing which i'm not sure is: whether we need to use different branches in graph
itself to control such a conditional behavior, or whether we need to store tag
somewhere in the dependency graph. There are obviously cons and pros in both
approaches, and need to think about this. Maybe with more examples it becomes
more obvious which way is better.
This only fixes manual tweaks for now, animation support is coming.
Adding alongside the existing one for now,
but it should eventually replace it.
Uses a matrix instead of (position + scale),
written so rotation can be done more easily.
Currently has a primitive handle for rotation, supports corner scaling.
There were two issues here:
1. material_update did not do anything, because DEG_id_tag_update was storing
update tags in original IDs, which had nothing evaluated. Even more, material
update should have been called with evaluated version of material, Solved
this by copying update tag from original ID to a copied one.
However, perhaps DEG_id_tag_update should tag both original and copied ID,
so updates are never gets lots if some depsgraph is not visible.
2. Tagging material for update should ensure it's copied version of node tree is
up to date, otherwise material will still use old node tree.
This solves missing material updates when changing topology. Tweaking values is
still broken, because of GPUMaterial using pointer to original node's socket
value, which gets broken after copy-on-write of the node tree (pointers of nodes
are changing).
Previously it was returning short, which was really easy to (a) compare against
non-ID type value (b) forget to handle some specific value in switch statement.
Both issues happened in the nearest past, so it's time to tighten some nuts
here.
Most of the change related on silencing strict compiler warning now, but there
is also one tricky aspect: ID_NLA is not in the IDType enum. So there is still
cast to short to handle that switch. If someone has better ideas how to deal
with this please go ahead :)
A single manipulator could only assign a single operator to each part.
Now each part can have it's own.
Also modify 2D selection callback, 2D started at 1, 3D at 0.
Now use -1 for unset value, start both at 0.
- Vertex only meshes never restored their selection history.
- Select history was cleared on the source instead of the target.
Simple Optimizations:
- Avoid O(n^2) linked list looping that checked the entire list before
adding elements (NULL values in the source array to prevent dupes).
- Re-use vert & edge lookup tables instead of allocating new ones.
Issue was nasty hidden one, the dual status (mix of local and linked)
of proxies striking again.
Here, remapping process was considering obdata pointer of proxies as
indirect usage, hence clearing the 'LIB_TAG_EXTERN' of obdata pointer.
That would make savetoblend code not store any 'lib placeholder' for
obdata data-block, which was hence lost on next file read.
Another (probably better) solution here would be to actually consider
obdata of proxies are fully indirect usage, and simply reassign proxies
from their linked object's obdata on file read...
However, that change shall be safer for now, probably good for 2.79 too.
This reverts commit 3888227a7b.
This "Fix" was made while ORCO was broken. Now that orco itself is fixed
this is no longer required, otherwise Tangent node produces different
results in Cycles and Eevee.
Don't use quick sort for small arrays, bubble sort works way faster for small
arrays due to cache coherency. This is what qsort() from libc is doing actually.
We can also experiment unrolling some extra small arrays, for example 3 and 4
element arrays.
This reduces tangent space calculation for dragon from 3.1sec to 2.9sec.
Brings tangent space calculation from 4.6sec to 3.1sec for dragon model in BI.
Cycles is also somewhat faster, but it has other bottlenecks.
Funny thing, using simple `static inline` already gives a lot of speedup here.
That's just answering question whether it's OK to leave decision on what to
inline up to a compiler..
Would be nice to be able to catch this with assert as well, will see what would
be the best way to do this/.\
Need to verify with Mai that this solves crash for her and maybe consider
porting this to 2.79.
Orco should behave the same if it comes from unconnected vec inputs, or
from the Texture Coordinate -> Generated node output.
Fixup for 11e7e0769a.
This is related to T52528. The coordinates are still different between
Eevee and Cycles, but at least it behaves consistent within itself.
In fact the shader should now be correct, but the orco attributes we are
passing the shader seems to be where the problem is. But it's to be
tackled separately.
scripts being "Artwork" which is your sole property and free to license.
I've removed the reference to scripts in this text.
This was from 2002! With our Python scripts becoming part of how Blender runs,
such scripts now are officially required to be compliant with GNU GPL.
For more information; check the FAQ or consult foundation@blender.orghttps://www.blender.org/support/faq/
Some manipulators are used like on-screen buttons,
in this case it doesn't make sense to keep track of their state,
so zero the offset when its unused.
Needed for lamp-target manipulator.
We have a hardcored limit of 1000 images to be baked.
However anything anove 100 would be leading to overflow in the code.
Caught by warning from builder bot (my compiler doesn't even complain
about this, but it should).
Re-use operator return flags for manipulator modal & invoke,
this means manipulators can allow navigation or other events to be
handled as they run - see T52499
Fishy cat benchmark was rendering with wrong shadows. Cause is unclear,
adding printf or rearranging code seems to avoid this issue, possibly a
compiler bug. This reverts the fix and solves the OSL bug elsewhere.
This was needed when we accessed OSL closure memory after shader evaluation,
which could get overwritten by another shader evaluation. But all closures
are immediatley converted to ShaderClosure now, so no longer needed.
While unlikely to have had any serious effects because of limited use, the
previous implementation was not actually atomic due to a data race and
incorrectly coded CAS loop. We also had duplicates of this code in a few
places, it's now been moved to a single location with all other atomic
operations.
We need to make sure we can store all volume closures for all objects in volume
stack. This is a bit tricky to detect what would be the "nestness" level of
volumes so for now use maximum possible stack depth. Might cause some slowdown,
but better to give reliable render output than to fail quickly.
Should be safe for 2.79 after extra eyes.
We were showing "search for unknown menutype WM_MT_button_context" messages in terminal which were not helpful for users, so now they are disabled.
To be backported to 2.79
It is possible to have same image used multiple times at different frames,
which means we can not free it's buffers without any guard. From quick tests
this seems to be doing what it is supposed to.
Need more testing and port this to 2.79.
Although the problem was exposed in 9457715d9a, the problem was in the
original code that was copied over. To have:
```
} else { /* EXPECTED_VALUE */
```
Without an BLI_assert(value == EXPECTED_VALUE); is asking for troubles.
Yet another reason to favour switch statements with:
```
default:
BLI_assert(!"value not implemented or supported");
```
Instead of chained if/else if/else /* expected_value */.
- NOCHECK -> ALL
- ALL -> MAYBE_ALL
Where 'MAYBE_ALL' checks to see if the mesh has changed.
This is clearer that `BKE_MESH_BATCH_DIRTY_ALL` is dirty and
going to be updated without any guess-work.
This was meant to be generic but introduced possible type errors
and unnecessary complication.
Replace with typed PyC_Tuple_PackArray_* functions.
Also add PyC_Tuple_Pack_* macro which replaces some uses of
Py_BuildValue, with the advantage of not having to parse a string.
- WM_manipulatorgrouptype_remove- > free
- WM_manipulator_group -> WM_manipulator_group_type
Naming here is still a bit confusing,
now at least free/remove are differentiated.
This means we have less overall noise for rendered image.
SSR, AO, and Refraction are affected by this change.
SSR still exhibit artifacts because the reconstruction pattern needs to change every frame (TODO).
Regression from rBfed853ea78221, calling this inside thread worker was
not really good idea anyway, and we already have all the code we need in
pre-threading init function, was just disabled for vertex particles
before.
To be backported to 2.79.
This exposes 2 methods for manipulators:
- new_custom_shape
- draw_custom_shape
This can be used for script authors to create and re-use shapes
without dealing with lower level API's.
Python's C-API doesn't provide functions to get
int's at specific integer sizes.
Leaving the caller to check for overflow,
which ended up being ignored in practice.
Add API functions that convert int/uint 8/16/32/64, also bool.
Raising overflow exception for unsupported ranges.
The thing that most often still goes wrong for new users building blender on windows is checking out the libraries, some skip over the wiki, some check out to the wrong folder, in an effort to reduce the time i spend on this, I added detection of svn and misisng libs to make.bat .
When the user has svn installed, and the libdir is missing he'll be asked if he wants to download them
if svn is not installed, or the user chooses 'no' the current error message is shown.
Reviewers: Blendify, sergey, juicyfruit
Reviewed By: sergey
Differential Revision: https://developer.blender.org/D2782
We should only early out with any hit in BVH traversal if the only visibility
bits used are opaque shadow. Not when opaque shadow is one of multiple bits.
Also pass by value and don't write back now that it is just a hash for seeding
and no longer an LCG state. Together this makes CUDA a tiny bit faster in my
tests, but mainly simplifies code.
It's purpose is to limit the amount of light that spread across the screen.
Not entierly sure if it's very usefull, but it sure help to avoid to drown the screen in bloom.
This function was called to recreate the lower mip level of the probe texture. But this is not it's usage and it introduced a stall.
This patch add cubemap mipmap level regeneration in eevee_effects.c
Since we started supporting the (Cycles) Material Output old files
stopped working. There is no reason to keep the original Eevee material
otuput anymore.
It includes doversion for old files.
Move floats around when needed to accomodate vec3 arrays efficiently.
With this we use slightly less memory when possible. Basically vec3s are not
treated as vec4 unless we have no float to use for padding).
Reviewers: fclem, sergey
Differential Revision: https://developer.blender.org/D2800
This includes big improvement:
- The horizon search is decoupled from the BSDF evaluation. This means using multiple BSDF nodes have a much lower impact when enbaling AO.
- The horizon search is optimized by splitting the search into 4 corners searching similar directions to help which GPU cache coherence.
- The AO options are now uniforms and do not trigger shader recompilation (aka. freeze UI).
- Include a quality slider similar to the SSR one.
- Add a switch for disabling bounce light approximation.
- Fix problem with Bent Normals when occlusion get very dark.
- Add a denoise option to that takes the neighbors pixel values via glsl derivatives. This reduces noise but exhibit 2x2 blocky artifacts.
The downside : Separating the horizon search uses more memory (~3MB for each samples on HD viewport). We could lower the bit depth to 4bit per horizon but it produce noticeable banding (might be fixed with some dithering).
Diffuse was not outputing the right normal. (this is not a problem with SSR actually)
Glass did not have proper ssr_id and was receiving environment lighting twice.
Also it did not have proper fresnel on lamps.
This fixes the Principled shader in Eevee, among other nodes.
Basically before we were treating all the vec3 as vec4 as far as memory
goes. We now only do it when required (aka, when the vec3 is not
followed by a float).
We can be even smarter about that and move the floats around to provide
padding for the vec3s. However this is for a separate patch.
That said, there seems to be some strong consensus in corners of the
internet against using vec3 at all [1]. Basically even if we get all the
padding correct, we may still suffer from poor driver implementations in
some consumer graphic cards.
It's not hard to move to vec4, but I think we can avoid doing it as much
as possible. By the time 2.8 is out hopefully most drivers will be
implementing things correctly.
[1] - https://stackoverflow.com/questions/38172696
Deleting the old internal audaspace.
Major changes from there are:
- The whole library was refactored to use C++11.
- Many stability and performance improvements.
- Major Python API refactor:
- Most requested: Play self generated sounds using numpy arrays.
- For games: Sound list, random sounds and dynamic music.
- Writing sounds to files.
- Sequencing API.
- Opening sound devices, eg. Jack.
- Ability to choose different OpenAL devices in the user settings.
This implements Arvo's "Stratified sampling of spherical triangles". Similar to how we sample rectangular area lights, this is sampling triangles over their solid angle. It does significantly improve sampling close to the triangle, but doesn't do much for more distant triangles. So I added a simple heuristic to switch between the two methods. Unfortunately, I expect this to add render time in any case, even when it does not make any difference whatsoever. It'll take some benchmarking with various scenes and hardware to estimate how severe the impact is and if it is worth the change.
Reviewers: #cycles, brecht
Reviewed By: #cycles, brecht
Subscribers: Vega-core, brecht, SteffenD
Tags: #cycles
Differential Revision: https://developer.blender.org/D2730
Use mesh batch cache for mesh selection.
Note that we could create the batches and free immediately
so they don't take up memory.
This resolves a problem where selection was limited
to immediate-mode buffer size.
Since the paste object is pasted in the active collection, and not on
its original one, we need to flush/calculate the new collection base
settings (visibility, selectability, ...).
DEG_id_tag_update() for the scene now. Though it may be better to tag
only the object specific IDs in the future.
Caused by own recent changes in handling of verts/edges/etc. arrays storage
for raycasting (rBe324172d9ca6690e8).
Issue was actually even weirder - there is absolutely no reason at all to
release DM here, those finaldm are stored in Object or EditMesh structs and
handled by general update system, other code shall never try to release them!
Originally we were not respecting the original visibility flags of the
collections. However this is required for Copy-on-write (CoW).
Remember to update the svn lib tests folder. I had to update some of the
json files there.
Also adding a new unittest for this particular issue:
Test render_layer_scene_copy_f
Flag ownership for each index array & vbo's
so we don't have to manually keep track of this and use the right free call.
Instead this can be passed on creation.
See D2676
2.8x branch added bContext arg in many places,
pass eval-context instead since its not simple to reason about what
what nested functions do when they can access and change almost anything.
Also use const to prevent unexpected modifications.
This fixes crash loading files with shadows,
since off-screen buffers use a NULL context for rendering.
I couldn't reproduce either, but calling min() with different argument
data types and indexing vectors with an index not known at compile time
seem likely to cause problems.
Ref T52404, T52404.
We stop using the .zip file and just have all files now in
lib/darwin/python/lib, along with numpy, numpy headers and requests.
This makes it consistent with Linux and simplifies code.
For old libraries the .zip stays, code for that gets removed when we
fully switch to new libraries.
We now have to support more complex copying types, which are controlled
by flags, so all copying logic will need to take those at some point (at
least, all potentially dealing with IDs).
_copy_data() functions shall not do that at all anymore. Kept as option
for now even though that helper is only called from here...
Also moar varnames renaming to standard _src/_dst sufixes.
We do have an history of those pieces of evil in our code, would be nice
to get fully rid of it, but at the very least let's not add more of them
in new code. :)
This can happen with Alembic files exported from Maya. I'm unsure as to the
root cause, but at least this fixes the crash itself.
Thanks to @looch for reporting this with a test file. The test file has to
remain confidential, though, so it's on my workstation only.
FFMPEG & VPX don't handle target with --build parameter, so we need to make sure use of plain configure command
Reviewed by: Brecht Van Lommel
Differential Revision: http://developer.blender.org/D2791
Was complicating general use case, also support for transforming with matrix_space set.
Add matrix_space support for manipulator_window_project_2d too.
This patch adds "Pixel Size" to the performance options, which allows to render
in a smaller resolution, which is especially useful for displays with high DPI.
Reviewers: Severin, dingto, sergey, brecht
Reviewed By: brecht
Subscribers: Severin, venomgfx, eyecandy, brecht
Differential Revision: https://developer.blender.org/D1619
The only reason shutter time was marked as non-animatable is because Blender
Internal render does not support such animation. But this is something what
users are keeping asking for and now Blender Internal is on it's way out.
Enabled animation of this property, but noted in tooltip that Blender Internal
does not support animation of this property.
Bug in new ID copying code, thanks once again to stupid nodetrees, we
ended up wrongly remapping MA node->id pointers to NodeTree when copying
materials using node trees...
Basically, make re-alloc and memcpy from the same lock, otherwise one
thread might be re-allocating thread while another one is trying to
copy data there.
Reported by Mohamed Sakr in IRC, thanks!
Enabled cache for frame accessor and tweaked policy so we guarantee keyframed
images to be always in the cache. The logic might fail in some real corner case
(for example, when doing multiple tracks at once on a system where we can not
fit 2 clip frames in cache) but things are much better now for regular use.
This fix a bug when occluder are on the edge of the screen and occludes more than they should.
Grouped the texture fetches together and clamp the ray at the border of the screen.
Also add a few util functions.
This should fixes the error message that a stall occured because of busy mipmap.
This happened on the minmax buffer generation and introduced a random 0.2ms latency.
I'm not sure of what was happening though.
Creating ngons with multiple axis aligned shapes in the middle of a
single face would fail in some cases.
This exposed multiple problems in BM_face_split_edgenet_connect_islands
- Islands needed to be sorted on Y axis when X was aligned.
- Checking edge intersections needed increased endpoint bias.
- BVH epsilon needed to be increased.
Note: this commit seems to work as expected (also with transform
snapping etc.). However, it is rather unsafe - not enough for 2.79 at
least, unless we get much more testing on it. It also depends on three
previous ones.
Note that using a global lock here is far from ideal, we should rather
have a lock per DM, but that will do for now, whole DM thing is doomed
to oblivion anyway in 2.8.
Also, we may need a `DM_DIRTY_LOOPTRIS` dirty flag at some point. Looks
like we can survive without it for now though... Probably because cached
looptris are never copied accross DM's?
This was... horribly wrong, CDDM will often *not* need to allocate
anything to return arrays of mesh items! Just check whether array
pointer is NULL.
Also, remove `DM_get_looptri_array`, that one is useless currently,
`dm->getLoopTriArray` will always return cached array (computing it if
needed).
The tweakmode flag and the selected-channels flag accidentally
used the same value, due to confusion over where these flags were
supposed to be set. The selected-channels flag has now been moved
to use a different value, so that there shouldn't be any further
conflicts.
To be ported to 2.79.
Old bevel 'Clamp overlap' code was very naive: just limit amount
to half edge length. This uses more accurate (but not perfect)
calculations for the max amount before (many) geometry collisions
happen. This is not a backward compatible change - meshes that
have modifiers with 'Clamp overlap' will likely have larger allowed
bevel widths now. But that can be fixed by turning off clamp overlap
and setting the amount to the desired value.
Own previous fix (rBd5d626df236b) was not valid, curves are actually
supported by SoftBodies. It was rather a mere UI bug, which was not
including Surfaces and Font obect types in those valid for softbody UI.
Thanks to @brecht for the head up!
Also, fix safe for 2.79, btw.
* 255 maximum seems excessive for F-Curve handle vertices; now reduced to 100
* Vertex Size is no longer restricted to the old 10px maximum size limit
(used because Windows limited the maximum vertex size drivers needed to
support)
It's more important that there is some form of feedback that the strips
are muted (i.e. dotted borders) than the fact that those dotted borders
may have slightly rounded corners. So, just use a regular sharp-cornered
rect when the strips need to be muted.
There's no reason to manually iterate over items in a DLRBT_Tree,
as the structure is designed to be able to be safely casted down
to a ListBase and ListBase-like nodes..
Adding structs was checking for duplicates
causing approx 75k string comparisons on startup.
While overall speedup is minimal,
Python access to `bpy.types` will now use a hash lookup
instead of a full linked list search.
See D2774
Changes for 2.8x to use EvaluationContext caused some confusion
- Would use scene layer passed from snap context.
- Would generate duplis from Main eval context.
- Would take context argument and use it to create another eval context.
Adding context args all over and filling in a new eval-context
for every ray-cast test isn't ideal either.
Remove the context argument since the purpose of
SnapObjectContext is to avoid this kind of confusion.
Store the EvaluationContext once and re-use.
This effectivly reduce firefly bleeding all over the place.
We still need the clamp in the resolve pass because the level 0 has not been clamped.
NOTE: I did not clamped each sample individually for performance BUT I did not profile it to know how much it cost.
This was surely cause by float overflow. Limit roughness in this case to limit the brdf intensity.
Also compute VH faster.
Add a sanitizer to the SSR pass for investigating where NANs come from. Play with the roughness until you see where the black pixel is / comes from.
It doesn't seem that useful in practice, was mostly added to match some
other renderers but also seems to be causing user confusing and accidental
long render times. So let's just keep the UI simple and remove this.
Differential Revision: https://developer.blender.org/D2768
We're adding some bias by default, which now I think is the right thing
to do from a usability point of view since you really need to use those
options anyway to get clean renders in a practical time.
Differential Revision: https://developer.blender.org/D2769
Adds thin/default/thick modes to add -1/0/1 to the auto detected line width,
while leaving the overall UI scale unchanged.
Also tweaks the default line width threshold, so thicker lines start from
slightly high UI scales.
Differential Revision: https://developer.blender.org/D2778
Group texture fetches to hide latency. 3.2ms -> 2.2ms (constant time improvement, not depending on scene complexity)
Could optimize further with textureGather (require OpenGL 4.0).
Theses Materials are rendered after the SSR pass.
The only difference with previous method is that they have a depth prepass (less overdraw) and are not sorted.
For the moment the only way to enable this is to:
- enable Screen Space REFLECTIONS.
- enable Screen Space Refraction in the SSR parameters.
- enable Screen Space Refraction in the material tab.
We track the previous ray position offseted by the thickness. If the sampled depth is between this value and the current ray position then we have a hit.
This fixes rays that are almost colinear with the view vector. Thickness is now only important for rays that are comming back to the camera.
As a consequence, this simplify a lot of things.
Also include some refactor.
Since we are working with non power of 2 textures, the mipmap level UV does not line up perfectly.
This resulted in skewed filtering and bad sampling of the min/max depth buffer.
We generate a 3D lut to precompute the btdf intensity.
I decided to use a 64*64*16 (N dot V, ior, roughness) because the btdf varies less with roughness than with IOR.
We also remap the ior to better use the space in the LUT.
*Changed categories of some keywords
*reordered some longer keywords that didn't appear
*Activated another color (reserved builtins) by Leonid
*added some HGPOV and UberPOV missing keywords
Patch by Maurice Raybaud (@mauriceraybaud). Thanks to Leonid for additions, feedback and Linux testing.
Related diffs: D2754 and D2755.
While not a regression, this is new feature and would be nice to have it
backported to final 2.79.
Bug introduced in recent ID copying refactor.
This commit basically sanitizes seq strip copying behavior, by making
destination scene pointer mandatory (and source one a const one).
Nothing then prevents you from using same pointer as source and
destination!
This is a bit confusing, especially when one mixes OpenCL code where ulong equals
to uint64_t with CPU side code where ulong is expected to be something else from
the naming.
This commit makes it so we use explicit name, common on all platforms.
When a mesh changes its number of vertices during the animation,
Blender rebuilds the DerivedMesh, after which the materials weren't
applied any more (causing the default to the first material slot).
Commit b6d7cdd3ce changed how the mesh data
is deformed, which wasn't taken into account yet in this unit test.
Instead of directly reading the mesh vertices (which aren't animated any
more), we convert the modified mesh to a new one, and inspect those
vertices instead.
Problem was that some code checks to see if device_pointer is null or
not and the new allocator wasn't even setting the pointer to anything
as it tracks memory location separately. Setting the pointer to non
null keeps all users of device_pointer happy.
Also internal changes so arrow3d matches grab3d's behavior.
Needed to add WM_MANIPULATOR_DRAW_OFFSET_SCALE flag so
we can optionally apply offset in worldspace or screen scaled values.
This could make output really polluted, where it'll be hard to see actual
issues.
It is still possible to have all backtraces printed using BLENDER_VERBOSE
environment variable.
As part of the fix for T51587, I removed the Depth output for non-Multilayer
images since it seemed weird that PNGs etc. that don't have a Z pass still get
a socket for it.
However, I forgot about non-multilayer EXRs, which are a special case that can
actually have a Z pass.
Therefore, this commit brings back the Depth output for non-multilayer images
just like it was in 2.78.
- Apparently MSVC does not support compound literals
in C++ (at least by the looks of it).
- Not sure how opencl_device_assert was managing to
set protected property of the Device class.
We don't enable global SSE optimizations in regular kernel, and we
keep those disabled on Linux 32bit.
One possible workaround would be to pass arguments by ccl_ref, but
that is quite a few of code which better be done accurately.
Steps to reproduce:
- Create shader Image texture -> Diffuse BSDF -> Output. Do NOT select image yet!
- Start viewport render.
- Select image from the ID browser of Image Texture node.
Thing is: with the memory manager we always need to inform device that memory
was freed.
Common folks, nobody considered master a C++11 only branch. Such decision is to
be done officially and will involve changes in quite a few infrastructure related
areas.
New code was handling correctly ID's internal references to self, but
not references between 'made real' different objects...
Regression, to be backported in 2.79.
It is defined to & for CPU side compilation, and defined to an empty for any GPU
platform. The idea here is to use this macro instead of #ifdef block with bunch
of duplicated lines just to make it so CPU code is efficient.
Eventually we might switch to references on CUDA as well, but that would require
some intensive testing.
The documentation for the bpy.app.handlers.scene_update_{pre,post}
handlers states that they're called "on updating the scenes data".
However, they're called even when the data hasn't changed. Of course
such handlers are useful, but the documentation should reflect the
current behaviour.
Reviewers: mont29, sergey
Subscribers: Blendify
Maniphest Tasks: T46329
Differential Revision: https://developer.blender.org/D1535
Image textures were being packed into a single buffer for OpenCL, which
limited the amount of memory available for images to the size of one
buffer (usually 4gb on AMD hardware). By packing textures into multiple
buffers that limit is removed, while simultaneously reducing the number
of buffers that need to be passed to each kernel.
Benchmarks were within 2%.
Fixes T51554.
Differential Revision: https://developer.blender.org/D2745
Simply disabled python tests, they can't be run anyway (since blender target is
not enabled) and we don't have any player-related tests in that folder.
This will allow much finer controll over how we copy data-blocks, from
full copy in Main database, to "lighter" ones (out of Main, inside an
already allocated datablock, etc.).
This commit also transfers a llot of what was previously handled by
per-ID-type custom code to generic ID handling code in BKE_library.
Hopefully will avoid in future inconsistencies and missing bits we had
all over the codebase in the past.
It also adds missing copying handling for a few types, most notably
Scene (which where using a fully customized handling previously).
Note that the type of allocation used during copying (regular in Main,
allocated but outside of Main, or not allocated by ID handling code at
all) is stored in ID's, which allows to handle them correctly when
freeing. This needs to be taken care of with caution when doing 'weird'
unusual things with ID copying and/or allocation!
As a final note, while rather noisy, this commit will hopefully not
break too much existing branches, old 'API' has been kept for the main
part, as a wrapper around new code. Cleaning it up will happen later.
Design task : T51804
Phab Diff: D2714
Note these are intended for platform maintainers, we do not intend to
support users making their own builds with these. For that precompiled
libraries from lib/ should be used.
Implemented by Martijn Berger, Ray Molenkamp and Brecht Van Lommel.
Differential Revision: https://developer.blender.org/D2753
Since all the shadow catchers are already assumed to be in the footage,
the shadows they cast on each other are already in the footage too. So
don't just let shadow catchers skip self, but all shadow catchers.
Another justification is that it should not matter if the shadow catcher
is modeled as one object or multiple separate objects, the resulting
render should be the same.
Differential Revision: https://developer.blender.org/D2763
Note these are intended for platform maintainers, we do not intend to
support users making their own builds with these. For that precompiled
libraries from lib/ should be used.
Implemented by Martijn Berger, Ray Molenkamp and Brecht Van Lommel.
Differential Revision: https://developer.blender.org/D2753
Since all the shadow catchers are already assumed to be in the footage,
the shadows they cast on each other are already in the footage too. So
don't just let shadow catchers skip self, but all shadow catchers.
Another justification is that it should not matter if the shadow catcher
is modeled as one object or multiple separate objects, the resulting
render should be the same.
Differential Revision: https://developer.blender.org/D2763
This will allow much finer controll over how we copy data-blocks, from
full copy in Main database, to "lighter" ones (out of Main, inside an
already allocated datablock, etc.).
This commit also transfers a llot of what was previously handled by
per-ID-type custom code to generic ID handling code in BKE_library.
Hopefully will avoid in future inconsistencies and missing bits we had
all over the codebase in the past.
It also adds missing copying handling for a few types, most notably
Scene (which where using a fully customized handling previously).
Note that the type of allocation used during copying (regular in Main,
allocated but outside of Main, or not allocated by ID handling code at
all) is stored in ID's, which allows to handle them correctly when
freeing. This needs to be taken care of with caution when doing 'weird'
unusual things with ID copying and/or allocation!
As a final note, while rather noisy, this commit will hopefully not
break too much existing branches, old 'API' has been kept for the main
part, as a wrapper around new code. Cleaning it up will happen later.
Design task : T51804
Phab Diff: D2714
Shows new, reference and diff renders, with mouse hover to flip between
new and ref for easy comparison. This generates a report.html in
build_dir/tests/cycles, stored along with the new and diff images.
Differential Revision: https://developer.blender.org/D2770
* Remove some unnecessary SSE emulation defines.
* Use full precision float division so we can enable it.
* Add sqrt(), sqr(), fabs(), shuffle variations, mask().
* Optimize reduce_add(), select().
Differential Revision: https://developer.blender.org/D2764
I need to use some macros defined in util_simd.h for float3/float4, to emulate
SSE4 instructions on SSE2. But due to issues with order of header includes this
was not possible, this does some refactoring to make it work.
Differential Revision: https://developer.blender.org/D2764
We already detect this automatically based on shading nodes and per shader
settings, and performance of this option is ok now all devices.
Differential Revision: https://developer.blender.org/D2767
- New manipulator tracks lamps to position under cursor.
- Works with multiple lamps, keeping relative offsets.
- Holding Ctrl moves the lamp.
- Access via manipulator or Shift-T.
Code could be improved, but like to get feedback from users.
Render-border & crop-node 2d-cage manipulators where unreasonably
complicated to implement because there was no good way to define
the sub-region the manipulator was transforming in
(render border within the camera's frame for example).
Add matrix-space variable,
remove scale property from cage2d manipulator, use matrix instead.
Was doing this with property get/set but this made view operations
require refreshing manipulator properties.
Simplify by operating on properties in their own space.
Also disable clamping for now since it assumes pixel-space.
Two main things here:
1. Replace all unsafe for #line directive characters into a single loop,
avoiding multiple iterations and multiple temporary strings created.
2. Don't merge token char by char but calculate start and end point and
then copy all substring at once.
This gives about 15% speedup of source processing time. At this point
(with all previous commits from today) we've shrinked down compiled
sources size from 108 MB down to ~5.5 MB and lowered processing time
from 4.5 sec down to 0.047 sec on my laptop running Linux (this was a
constant time which Blender will always spent first time loading kernel,
even if we've got compiled clbin).
Add a safe version of normalize since all uses of normalize
did zero length checks, move this into a function.
Also avoid unnecessary conversion.
Gives minor speedup here (approx 3-5%).
Basically gather lines as-is during traversal, avoiding allocating
memory for all the lines in headers.
Brings additional performance improvement abut 20%.
The idea here is that it is possible to mark certain include statements
as "precompiled" which means all subsequent includes of that file will
be replaced with an empty string.
This is a way to deal with tricky include pattern happening in single
program OpenCL split kernel which was including bunch of headers about
10 times.
This brings preprocessing time from ~1sec to ~0.1sec on my laptop.
The idea is to re-use files which were already processed. Gives about 4x speedup
of processing time (~4.5sec vs 1.0sec) on my laptop for the whole OpenCL kernel.
For users it will mean lower delay before OpenCL rendering might start.
Keyframe handle vertices (the circles on the ends of the handles)
should always be larger than the central vertex. This brings back the
"outer" radius value from the old gluDisk(), and doubles it to get the
necessary diameter, to scale it properly.
TODO's:
- Get rid of all fills inside these circles
- Make the central vertex square-shaped again
* Outlines of keyframes were too thick and ugly
* Size differences between keyframe types was being swallowed
by the pixel-fudge factor, leaving colour as the only distinguishing
factor (bad!)
This changes the Cycles exporting and Cycles/Eevee UI code to support both
output material nodes, giving priority to the renderer native one. Still
missing is Eevee code to prefer the Eevee output node.
This is still far from prefect, but yet much better than what we had so
far (more consistent with inheritent precision available in floats).
Note that this fixes some (currently commented out) units unittests, and
requires adjusting some others, will be done in next commit.
Bug was in RNA nodes code actually, itemf functions shall never, ever
return NULL!
Note that there were other itemf functions there that were potentially
buggy. Also harmonized a bit their code.
* Numbers with units (especially, angles) where not handled correctly
regarding number of significant digits (spotted by @brecht in T52222
comment, thanks).
* Zero value has no valid log, need to take that into account!
It now uses a quality slider instead of stride.
Lower quality takes larger strides between samples and use lower mips when tracing rough rays.
Now raytracing is done entierly in homogeneous coordinate space. This run much faster.
Should be fairly optimized. We are still Bandwidth bound.
Add a line-line intersection refine.
Add a ray jitter between the multiple ray per pixel to fill some undersampling in mirror reflections.
The tracing now stops if it goes behind an object. This needs some work to allow it to continue even if behind objects.
Running undo would notify manipulators to refresh,
but this still allowed for events in the queue to be handled,
where manipulators could be drawn for selection before
their refresh callback runs.
This made Python manipulators raise exceptions
about referencing invalid data (or crash).
Now tag manipulator update on file load (including undo)
and ensure the refresh callback runs
before drawing manipulator selection.
Also split manipulator map refresh flag in two since selection doesn't
perform the same operations as regular drawing.
This will only be noticeable for drawing many instances.
In contrived use-case with many instances, and `USE_PROFILE` disabled
this can close to double playback FPS.
The option to disable this is left in the code in case we want to
debug memory use.
See D2756 for details.
- Each allocation can be a different size
(but should be smaller than the chunk size).
- Result can be looped over in order of allocation.
- Allocations are aligned to pointer size to avoid unaligned reads.
When calling sculpt from Python,
setting 3D 'location' but not 2D 'mouse' stopped working in 2.78.
Now check if the operator is running non-interactively and
skip the mouse-over check.
Regression in D1812: PyDriver variables as Objects
Taking the Python representation is nice in general
but for enums it would convert them into strings,
breaking some existing drivers.
Some users really liked previous behavior,
so making it an option.
Cursor Lock Adjustment can be disabled to give something close to
2.4x behavior of cursor locking.
When lock-adjustment is disabled placing the cursor the view.
This avoids the issue reported in T40353
where the cursor could get *lost*.
Even strands that were excluded by the density texture were being added
to the DM passed to cloth, but these ended up having some invalid data,
because they were not fully constructed.
This simply excludes `UNEXISTED` particles from the DM generation, as
would be expected.
Note that fix is not perfect, systematically make refcounting of all IDs
assigned to node's id pointer, which breaks the 'do not refcount
scene/object/text datablocks' principle...
But besides that principle being far from ideal in general, it becomes
pretty much impossible to apply when using //generic// ID pointer,
unless we add some kind of type data to that pointer somehow.
So for now, better to live with that, than having broken usercount.
This commit is a work forward having less updates during playback, which speeds
things up a lot here. The idea is simple: stop update all copy-on-write
datablocks (which implies full re-evaluation actually) on frame change and
re-use existing evaluated meshes as much as possible.
This brings playback speed to 24 fps on the dino test scene here. Performance
drops down a lot when armature is animated tho, but that's because of need of
tangent space calculation which we can't do much about from just a dependency
graph.
Hopefully this doesn't make copy-on-write too unstable, quick tests here are
surviving fine.
Old performance debug was doing queries for every frame even if not debugging perf.
Also, it did not record when a pass was draw multiple time, leading to incorect measurement.
New module also allows to group the timers to limit infos displayed.
Also fix the background CPU draw timer.
The purpose of the keymap strings is probably for un-embossed menu items
like seen in most pulldowns. I can't see a reason for also adding that
string for regularly drawn buttons within popups, we don't add it
anywhere else in the UI either. So this commit makes sure shortcut
strings are only added to buttons that are drawn like pulldown-menu
items.
The purpose of the keymap strings is probably for un-embossed menu items
like seen in most pulldowns. I can't see a reason for also adding that
string for regularly drawn buttons within popups, we don't add it
anywhere else in the UI either. So this commit makes sure shortcut
strings are only added to buttons that are drawn like pulldown-menu
items.
The Blender text editor's built in python template "Gamelogic" has a reference near the bottom to "objectHitList" as an alleged attribute to the KX_TouchSensor. This name is incorrect, it's correct name is "hitObjectList."
Attempting to access the suggested objectHitList returns error...
```
AttributeError: 'KX_TouchSensor' object has no attribute 'objectHitList'
```
The provided diff corrects this minor error.
Reviewers: kupoman, moguri, campbellbarton, Blendify
Reviewed By: Blendify
Tags: #game_engine, #game_python
Differential Revision: https://developer.blender.org/D2748
`defvert_array_find_weight_safe()` was confusing 'invalid vgroup' and
'valid but totally empty vgroup' cases.
Note that this also affected at least ShrinkWrap and SimpleDeform
modifiers.
This means once an ID is created,
it will keep using the same PyObject instance.
This has some advantages:
- Avoids unnecessary re-creation of instances on UI poll / redraw.
- Accessing free'd ID's gives an exception instead of crashing.
(long standing annoyance!, though this only applies to ID's
and not yet other data that uses the ID's - vertices for eg).
- Allows using instance comparison (a little faster).
Note that the instances won't be kept between undo.
Was doing 2x lookups which is OK for click-select
but this runs on mouse-move and can become slow.
May enable this again if highlighting logic changes.
Also scale hotspot by pixelsize.
- Cleanup array access, move into functions.
- Store allocated size to avoid realloc's on every add/remove.
- Make select editable from Python.
- Rename select callback to select_refresh
(collided with select boolean).
- Call select_refresh when de-selecting as well as selection.
This was cause by some post process not always sampling the highest mipmap.
But if there is no need for mipmapping (i.e. no SSR) these levels will be undefined.
So forcing all Post FX shader to sample level 0 fix this.
This add the possibility to use planar probe informations to create SSR.
This has 2 advantages:
- Tracing is less expensive since the hit is found much quicker.
- We have much less artifact due to missing information.
There is still area for improvement.
BKE_scene_copy explicitly ignores visibility of "source" collections make all
collections visible. This is also tested by regression tests.
While it seems more logical to simply preserve all possible visibility flags
and overrides, don't feel like submitting to a behavior-changes without talking
to author of those guards first.
This commit fixes cycles material preview.
If object is only listed in collection but not added to any of layers we shouldn't create
placeholder for it, because otherwise we'll leave lots of placeholder ID nodes.
Question: can we make this exception to be more reliable?
This commit separate the depth texture into another texture array.
This remove the need to output radial depth into alpha.
Unfortunatly it's difficult to recover position from the non linear depth buffer when applying reflection without adding a bunch of stuff.
This is in preparation of SSR planar reflections.
- Encode normals for other opaque bsdf so they are not rejected by the normal facing test.
- Early out non reflective surfaces.
- Add small offset to raytrace to avoid self intersection.
- Fix fallback probes not appearing.
Output in 2 buffers Normals, Specular Color and roughness.
This way we can raytrace in a defered fashion and blend the exact contribution of the specular lobe on top of the opaque pass.
Ideally need to clean and sane and impossible-to-break way of making sure
evaluation context is fully initialized, but that would need some thoughts
and experimentation.
Otherwise we'll have confused dependency graph builder, which wouldn't be able to
build proper graph.
Didn't find a way to avoid world copy here, we can probably escape with some shallow
copy here, but that will currently complicate code a lot.
Ideas to consider here:
- Use shallow copy of existing world after new ID management API is in place.
Downside would be thread safety, kind of nice to have everything local.
- Switch depsgraph away from ID_TAG and do hash lookup or so.
This will slow down depsgraph builder, but will make code more reliable.
The order of evaluation of function arguments is undefined, and the order
was reversed between these compilers. This was causing regressions tests
to give different results between Linux and macOS.
In the future we should make these two buttons on one line
However because we need `gen_context = 'PAINT_STENCIL'`
this is a little hard and we need to find a proper solution.
One might be using `context_pointer_set`
Patch by @craig_jones with edits by @blendify
Differential Revision: https://developer.blender.org/D2710
In the future we should make these two buttons on one line
However because we need `gen_context = 'PAINT_STENCIL'`
this is a little hard and we need to find a proper solution.
One might be using `context_pointer_set`
Patch by @craig_jones with edits by @blendify
Differential Revision: https://developer.blender.org/D2710
Moving the ray_start_local to the new position does not lose as much precision as moving the ray_org_local to the corresponding position.
The problem of inaccuracy is within the functions: `bvhtree_ray_cast_data_precalc` and` fast_ray_nearest_hit`. And not directly in the values of the rays.
Those functions did not use evaluation context.
Also fixed lots of unused variables warnings caused by commented out code which
needs to be ported away from DerivedMesh and to evaluation context.
Note that some little parts of code have been dissabled because eval_ctx
was not available there. This should be resolved once DerivedMesh is
replaced.
The way we use it, UI_PRECISION_FLOAT_MAX is actually + 1 to get total
number of digits, and float only has 7 meaningful digits, so that define
shall be at 6.
GCC seems to detect uninitialized into function calls now, but then isn't
always smart enough to see that it is actually initialized. Disabling this
warning entirely seems a bit too much, so initialize a bit more now.
The remapping code was creating plkaceholders for objects coming from legacy
bases, but since those objects were never created by dependency graph (since
they are supposed to be ignored) the copy on write relations creation was
confused.
Now we do some special trickery to clear legacy bases on copy on write.
Currently RNA doesn't give us a good way of accessing singleton layers,
for now expose as a layer list (skin & paint-pask do this too).
Noted in T47811 that this should be changed.
This commit unifies the flattened texture slot names for bindless and regular CUDA textures. Texture indices are now identical across all CUDA architectures, where before Fermi used different indices, which lead to problems when rendering on multi-GPU setups mixing Fermi with newer hardware.
Those checks are not always helpful, since id remapping doesn't want to
worry about which components to tag for update. Perhaps in the future we
will introduce special flag which would mean "tag everything possible"/
We get already enough reports of people complaining about crashes on
edit mode unaware that they are in the non-supported Blender Render
engine.
Blender Render is going away, no reason to keep it around. Once we have
a nice fallback on Eevee and fast file loading we can default to Eevee
instead.
This is the fake ID nature of compositor again. Need to discard such
pointers before freeing datablock even for scenes (before it was done
for objects only).
Previously it was possible to run into situation when armature is constructed prior to
objects which are used for it's constraints. This was causing wrong scene evaluation.
Now we create placeholders for objects used by armature in case they don't have ID node
yet, which ensures we have proper mapping from original to copy-on-write ID pointer.
This shows the bug when IK solver doesn't update reliably when targeted an external
object and when that object is handled by build_object() after the armature.
The issue was caused by id_copy_no_main() changing pointers of constraints used
in pose to a newly allocated ID. This is correct, but caused confusion too our
copy on write remapping, because we are mimicing inplace duplication by copying
memory over from a temporarily duplicated ID to a proper placeholder. This was
causing dangling pointers in pose to a temporarily allocated ID.
Now we add special code to remapping callback which replaces temporary ID with
a proper one.
Couple of main things here:
- Properly handle PSYS_UPDATE_* flags from DEG_id_tag_update.
There are still some possible issues here related on the fact
that we don't differentiate different PSYS_UPDATE_* flags here
and handle the mall the same.
Other possibility here is that object level particle settings
evaluation might be forced when particle system evaluation is
tagged for update. Didn't see actual issue here yet, but need
a closer look.
- Don't tag non-object datablocks on visibility changes.
Those don't depend on visibility anyway. This prevents particle
settings IDs from flushing updates to all objects, causing all
cached particles to be lsot.
- Only update translation and geometry components on visibility
changes.
Once again, this prevents particle cache from being invalidated.
We might need to tag material components here still tho.
This commit makes it so that only ID components which correspond to the tag
flag are tagged for update (previously the whole ID would have been updated
in the most of cases).
This allows us to have more granular tag flags and prevent tagging of things
we don't want to be tagged.
Previously tagging particle settings for update will iterate over all objects and
all their particle system to see whether something needs an update or not. Now we
put ParticleSettings as an ID to the dependency graph, so tagging it for update
will nicely flush updates to all dependent particle systems.
Current downside of this is that due to limitation of flush routines it will cause
some extra particle system re-evaluation when it technically not needed, and what's
more annoying currently it will discard point caches more often.
However, this is a good and simple demonstration case to improve tagging/flushing
system to accommodate for such cases (similar issues happens with CoW and shading
components). So let's try to find some generic solution to the problem!
This replaces usage of generic PLACEHOLDEWR with string lookup with more
explicit opcode. This should make it faster to build dependency graph by
avoiding string comparisons when it's not needed.
There should be no user measurable different.
Hopefully fix it actually, at least could not reproduce it anymore with
that changen, but Was already quite hard to trigger before.
We need a memory barrier at this allocation, otherwise it might happen
after preview gets added to done queue, so preview could end up being
freed twice, leading to crash.
Change the implementation so it no longer takes over the mouse cursor motion
from the OS, instead only move it when warping, similar to Windows and X11.
Probably the reason it was not done this way originally is that you then get
a 500ms delay after warping, but we can use a trick to avoid that and get much
smoother mouse motion than before.
While drawing nice 'rounded' values is OK also for 'low precision'
editing like dragging and such, it's quite an issue when you type in a
precise value, validate, edit again the value, and find a rounded
version of it instead of what you typed in!
So now, *only when entering textedit of num buttons*, we always get the highest
reasonable precision for floats (and use exponential notation when
values are too low or too high, to avoid tremendous amounts of zero's).
Tweaked the path radiance summing and alpha to accommodate for possible contribution of
light by transparent surface bounces happening prior to shadow catcher intersection.
This commit will change the way how shadow catcher results looks when was behind semi
transparent object, but the old result seemed to be fully wrong: there were big artifacts
when alpha-overing the result on some actual footage.
Since we added auto DPI on Linux, on some systems the UI draws smaller than before
due to the monitor reporting DPI values like 88. Blender font drawing gives quite
blurry results for such slightly smaller DPI, apparently because the builtin font
isn't really designed for such small font sizes. As a workaround this clamps the
auto DPI to minimum 96, since the main case we are interested in supporting is
high DPI displays anyway.
Differential Revision: https://developer.blender.org/D2740
This is a different fix for the issue from D2088, preserving backwards compatibility
for IK stretching. The main problem with this patch is that this new behavior has
been there for a year, so it may break rigs created since then which rely on the new
IK stretch behavior.
Test file for various cases:
https://developer.blender.org/diffusion/BL/browse/trunk/lib/tests/animation/IK.blend
Reviewers: campbellbarton
Subscribers: maverick, pkrime
Differential Revision: https://developer.blender.org/D2743
Revise the logic here to be more robust when keyframes with
similar-but-different frame numbers (e.g. 70.000000 vs 70.000008)
would cause the search to go into an infinite loop, as the same
keyframe was repeatedly found (and skipped).
Tweak the bias from the previous fix a bit to be more backwards compatible in
some scene. In the end which way we round is quite arbitrary, but keeping the
case where the texture coordinate is exactly zero the same seems better.
For users that means you can tweak shaders in the nodetree and things
are way faster. This is a huge improvement, particularly in
systems that have no shader cache.
From the code perspective it means we are no longer re-compiling the
shader every time a value is tweaked in the UI. We are using uniforms
for those values.
It would be slow to add that many uniforms for all the shaders. So
instead we are using UBO (Uniform Buffer Objects).
This fixes the main issue of T51467. However GWN_shaderinterface_create() still
needs to be improvedi. When opening a .blend all shaders are compiled once, so
optimizing it will bring a measurable impact.
========================================================================
NOTE: This breaks update of Cycles material upon nodetree nodes
tweaking. It will be fixed separately by depsgraph, once tackling T51925
(Animated Eevee values slowdown).
The idea is to make Depsgraph update more granular. The XXX TODO in
rna_nodetree.c will be tackled at that time as well.
========================================================================
Reviewers: sergey, brecht, fclem
Differential Revision: https://developer.blender.org/D2739
This commit makes simple cases to work, for example:
- IK solver to an external object
- Object with Armature modifier, "parented" to the deforming armature
(via animation).
More complicated setups (like agent rig) are crashing still.
If you merge from master to blender2.8 after a commit, remember to test in 2.8, otherwise what is the point?
Specially if it's a non-essential commit such as c9817c67fc.
Bug introduced on: f4155d3778 (the merge
that included the above commit).
This commit effectively makes workspace switching useless as far as the
active scene layer goes.
The functions from the scene layer API to get the correct scene layer
from "context" were a placeholder to be addressed by the workspace
commit.
When workspace was merged, however G.main was used as a replacement to pass the
correct argument for the functions. As it turned out (surprise!) this
leads to crash on render preview.
We need to get rid of:
* BKE_scene_layer_context_active_ex_PLACEHOLDER
* BKE_scene_layer_context_active_PLACEHOLDER
And either use SceneLayer explicitly or replace it by:
* BKE_scene_layer_from_workspace_get
Goal is to make them more modular, to allow more variants (variable
single-color, thickness, ...) to be added without having to
copy-and-change-one-line of whole chain of shaders.
The option has always (un)set the "Overwrite" flag on all strips. Calling
it "Override" seems misleading, since even when unchecking it, it overrides
whatever was set on the selected strips. It really just (un)sets the
"Overwrite" flag, and now it is also labeled as such.
This is a first step towards proper depsgraph "ownership", where
we would allow scene to be in multiple states dependent on active
workspace or scene layer.
This commit introduces a basic API to get proper dependency graph
for a given scene layer. It also renames scene->depsgraph to
depsgraph_legacy, so it's easier to search0-n-replace in the future.
In the reported example it seemed reasonable to apply this change.
But it causes a much more common case (selecting projections)
to be split into 2x islands.
Resolves T50970
This makes sense when we want to avoid float precision error
for near co-linear edges. OTOH, this is an arbitrary decision,
so keep functions separate.
That problem occurs because of the imprecision of `short int` (16 bits).
The 3d coordinates are converted to 2d, and when they are off the screen, their values can exceed 32767! (max short int value)
One quick solution is to use float instead of short
The snap code is actually a little tricky. I want to make some arithmetic simplifications in it
The only similarity between these functions is that both serve to snap.
However their codes are totally different from one another.
So by separating these functions, it:
- removes the need to put several conditions;
- simplifies and
- optimizes the code
* Display a warning above the pose list if the pose library is in an invalid
state (i.e. when it has keyframes but no pose-markers associated with those
keyframes). This warning prompts users to run the "Sanitize Pose Library Action"
operator, which should fix up such issues.
* "Sanitize" operator now creates unique names for each newly create pose
marker it generates, including the frame on which it found the pose
The problem here was that the "frame_start" and "frame_end" RNA properties of
the Stepped FModifier were shadowing/overriding "frame_start" and "frame_end"
properties of the base FModifier. As a result, when the range() callback
for the In/Out parameters (defined as part of the base FModifier) checked
it's start/end properties, they were always still zero, meaning that the
acceptable range for the In/Out parameters was 0 -> 0 = 0.
Note:
If you've got old files with this problem, you'll need to manually click on
the frame_start/end properties to flush out the old values. It's probably
not worth the effort of applying a version patch for this (given that this
modifier is not one of the most often used ones AFAIK).
As with Strip Time, the updates here would get triggered before the
autokeying had a chance to record the unkeyed values, making it impossible
to autokey.
Hashed Alpha transparency offers a noisy output but has the benefit of being correctly ordered. Noise can be attenuated with Multisampling / AntiAliasing.
This is something which was reported to work fine by Mai, Benjamin and
confirmed by myself. Disabling this workaround gains us some speedup:
Before Now
bmw27 04:28.42 04:07.79
classroom 09:26.48 08:54.53
fishy_cat 08:44.01 08:18.70
koro 09:17.98 08:57.18
pavillon_barcelone 12:26.64 11:52.81
Test environment is:
- Ubuntu 16.04, with all updates installed
- AMD RX 480 GPU
- amdgpu pro driver version 17.10-450821
The issue was caused by combination of following factors:
- Clipboard cleanup function will pass node tree as NULL to node free
function.
This is fine on it's own, we don't have tree in clipboard.
- Node free function will call node storage cleanup only when there is
a non-NULL node tree.
This is somewhat weird, because storage cleanup does not take node
tree as argument.
So the solution here: move node storage cleanup outside of check that
node tree is not NULL.
imapaint's clone and canvas are refcounting Image usages.
And particle's editsettings' object and scene seem to be pure runtime
data (they are reset to NULL in readcode), so resetting them to NULL
here as well.
The "fix" happened due to a mistake in copy-on-write commit, that mistake solved
preview render but broke something else.
This reverts commit 45720922f7.
Really, really, really need to get rid of this usercount handling
everywhere, hopefully incomming ID copying rewrite will help sanitize
that mess. But fix was needed for 2.79 release!
UVs need specific data in the VBO, which is not computed unless the
shaders assigned to the mesh actually use UVs. When adding UVs to the
shader, the VBOs were not being recomputed to include the required data.
This adds a DEG relation between the shader and the mesh, and recomputes
the required data if the shader changed.
Thanks Sergey, for all the DEG stuff...
The code was somewhat weird: it was first copying border/crop settings from
the "source" scene, then was checking border settings of the current scene
and only then was copying border from "source" scene.
Now we first copy border/crop flags, then copy border from source and then
check whether border is a full-frame.
A pointer to the uniform data for the empty drawing was being freed
before the actual draw call, which invalidates the uniform.
This makes the data only be freed after drawing.
Auto & aligned handles wouldn't restore to their correct locations.
Note that a more direct fix for the bug is possible
(storing the handle locations to restore on cancel).
But that still gives some odd behavior, see code-comments for details.
The default anisotropic tangent computation could fail in some cases,
leading to NaNs and artifacts. Use a simpler formulation that doesn't
suffer from this.
Unfortunately this means disabling the code that ensures the title
bar is properly scaled with DPI, however better to have that as a
cosmetic issue than Blender being unusable with a lot of Intel GPUs.
The issue was caused by combination of following factors:
- Blender Internal viewport render can not distinguish between which parts of
main database changed, so it does full database re-sync when anything is
tagged for an update.
This way, if any NodeTree (including compositor) is changed, Blender Internal
viewport is tagged for full render database update.
- With old dependency graph, scene-level drivers are evaluated on every
iteration of scene_update_tagged, even if nothing is tagged for an update.
This causes compositor drivers be evaluated quite often.
- Driver evaluation checks whether value was changed, and if so it tags
corresponding ID type as updated (this is what was telling viewport to do
render database update).
This check was quite stupid: current property value was checked against the
one coming from driver expression. This means, if driver value is outside
of the hard limit range of the property, the property will always be
considered updated.
The fix is to compare current property value against clamped value from the
driver.
Up until now users had to switch to Blender Internal to set the dimension of their openGl renderings. This simple change adds the panel at the top of the scene render tab.
It's cause by degenerate triangle having normals set to (0,0,0) which may not be compressed like it should.
This fix the problem in the final indirect lighting evaluation which might be costly.
The issue was caused by updates being flushed for all scenes, while actual update
was only called for an active one.
Not sure why do we need to flush updates for all scenes, so now we only flush
scenes which are updated.
New dependency graph is tacking root bone into account when building the graph.
This is required in order to get proper dependencies between bones. so we can
reliably use bones as targets from the same rig (and even indirect relations
via external objects). This forces us to tag relations for update when we change
root IK chain bone.
Since relations rebuild is not fully trivial operation, we only do it for
the new dependency graph. In the future it'll be nice to avoid whole graph
rebuild for such cases, but that's mentioned as a TODO.
This situation happens when a file with a text effect sequencer strip is
loaded in Blender < 2.76 and saved. This destroys the effect data, causing
a crash in Blender ≥ 2.76.
d2f748a222 prevented the crash when opening such a file, but accessing
the strip still caused a crash. This commit fixes that by actually
initialising the invalid strip. Of course this still causes data loss, but
that already happened by opening & overwriting the file in Blender < 2.76.
This makes no sense to add theses. In cycles, each volume node is computed separatly. In eevee only the combined parameters are evaluated and phase should be averaged in this case.
Using an arbitrary face as the source of the UV data is mostly fine, as
vertices on seams will generally map to different parts of the texture
that have the same color.
This is regarding fed853ea78
Some of the functions might have been inlined, but others i don't see
how that was possible (don't think virtual functions can be inlined here).
In any case, better be explicitly optimal in the code.
Clearing of custom bones outline's line thickness was not done at proper
point, wireframe drawing never changes line thickness, only solid draw
with outline does...
Last fix only accounted for direct changes to the RB settings, but
failed for, say, object transformations. This fix accounts for any
change that might invalidate the RB cache.
Fix 9cd6b03187 introduced a bug that
prevented simulation after a cache invalidation (for instance when
changing a setting after simulating). This fixes that.
This allow specialized shaders to redefine the closure interface to fit their needs.
For instance, Volumetric closure needs to pass more than one vec4 (absorption vec3, scattering vec3, anisotropy float).
The problem here was that when a "invalid" path is generated by the panoramic camera, it was tagged
as RAY_TO_REGENERATE with the intention of generating a new path in kernel_buffer_update.
However, since that state was not handled in kernel_queue_enqueue, kernel_buffer_update did not
process the path which resulted in an infinite loop.
Things like missing directories are now properly checked for, rather than
crashing Blender.
This also adds support for relative paths when opening an ABC file.
At the moment libblock_remap_data_preprocess is using
FOREACH_SCENE_OBJECT to iterate over all the objects of the scene and
unlink them.
However we were storing a reference to the Base of the removed object.
Anyways, the loop is now sanitized so that this crash no longer happens.
Also now we have an unittest for this.
This makes the last time (`ltime`) stored in the rigid body world (`rbw`)
only be updated once a simulation step actually occurs, this prevents
another simulation step from being solved unless the current time is
exactly one frame after the last cached frame. Thus this prevents the
formation of gaps in the cache, such as seen in T50230.
Reviewers: mont29, sergey, angavrilov
Tags: #physics
Maniphest Tasks: T50230
Differential Revision: https://developer.blender.org/D2458
Now computers that support OpenGl3.3 (but not 4.5) can run Blender 2.8.
For any given HDC, you may only call SetPixelFormat *ONCE* any future
calls for the same HDC will fail. And computers that would support only
OpenGL 3.3 wouldn't have a change to get a valid OpenGL context because
the pixelformat was already set while trying to probe the supported
contexts.
We fix this by splitting the final context creation from the query of
supported OpenGL versions.
Patch by Ray Molenkamp (bzzt_ploink/LazyDodo) with code style fixes and
comments by me.
D2729 by @IgorNull
Currently, trackball rotation sequentially applies rotation across x axis and y axis,
which produces a strange/unusable result on diagonal pointer motion.
This change fixes the problem by using a single axis which is orthogonal
and proportional to mouse delta - matching view-port trackball.
To recompile all the shaders is expensive. And something to be avoided at all costs.
It was needed before because for every new lamp in the file we needed to
recompile the shaders. Now this is no longer required since we are using
UBOs for the sahders.
This avoids using GWN_vertbuf_attr_set which needs to calculate the
offset and perform a memcpy every call.
Exposing the data directly allows us to avoid a memcpy in some cases
and means we can write to the vertex buffer's memory directly.
Was doing O(n^2) list lookups with blender-render drawing & transform.
Also missing NULL checks would crash.
Use Object.base_flag (already used by new draw manager in places)
to avoid list lookup.
Note, transform still performs inefficient lookups,
but only for selected parents (like 2.7x), not all parents.
As the title says, the normal wasn't set for the Hair BSDF because it wasn't
needed before. However, the denoiser uses it to store the feature passes, so
it needs to be set now.
Now that some node types may have custom context, we need to handle that
in the (convoluted :| ) UI code of nodes as well.
Reported in T43295 by Gabriel Gazzán (@gab3d), thanks.
`BMO_iter_as_array()` may fill less items than requested in given array,
so we have to update number of items to work on from its returned value,
otherwise code might try to use uninitialized memory.
Read from the GPUMaterial to find custom-data layers used for drawing.
This resolves problem where having UV's would always calculate tangents
causing noticeable slow down compared to 2.7x.
The original code was doing a sanity check to see if existing index was
out of range. However the comparison was wrong.
So if the previous ct->user (active index of texture node) was larger
than then number of available texture nodes + 1 in the other material,
we would never re-set the index to 0.
Bug introduced on c31f74de6b.
There was an early attempt of fixing this (2b2ac5d3cc) but it was just working
by pure, luck. And failing in cases like the one from this bug report.
Add a safe version of normalize since all uses of normalize
did zero length checks, move this into a function.
Also avoid unnecessary conversion.
Gives minor speedup here (approx 3-5%).
Convenience makefile now uses CMAKE_BUILD_TYPE_INIT,
this means you can change the build type of an existing build
and it won't be overwritten when running `make`.
Useful if you want to add debug info to a release build for profiling.
This is a very important, potentially deadly side-effect of this
operator. If something goes wrong, it can save a broken .blend file.
Ideally we could get rid of that operation anyway, once ID management if
fully renewed, but for now would rather keep it around.
Related to T51902.
Fix is a bit ugly, but cannot think of another solution for now, at
least this **should** not break anything else.
And now I go find myself a very remote, high and lonely mountain, climb
to its top, roar "I hate proxies!" a few times, and relax hearing the echos...
This commit makes the fullsample option for viewport renderings always
on: Render > OpenGL Render Options > Full Sample.
(The UI still allows users to set this, so we will need to revisit this
before 2.8 releases).
Even in computers that can handle MSAA we had issues.
The way Blender gpu_* implementation is handling anti-aliasing is buggy.
For example, in Blender 2.7x if you have depth of field in a viewport
with multi-sampling, the DoF gives us jagged edges.
Since Eevee uses framebuffers for a lot of things, this issue was
leading to very visible buggy render in some computers, and more subtle
inconsistent buggy renders in others (easy to test with the depth of
field in Eevee).
Some code for freeing shaders on scene updates (because of previous
dependency shaders had on lamps) had been disabled, as it is no longer
required. This removes that code altogether.
Object shaders no longer depend on lamp data at compile time, thus they
don't need to be invalidated when lamps change. Disabling shader
recompilation allows fast viewport updates when changing lamp settings.
Note that even though shaders for lamps are currently not being used,
`lamp_changed` is still freeing the lamp shaders, as at some point we
might want to use shaders for lamps...
Based on D2578, now you can install JACK audio server and use it in
Blender build without having to specify the `--with-all` option (that
one still enables also JACK of course).
Reviewers: mont29
Maniphest Tasks: T51033
Differential Revision: https://developer.blender.org/D2578
This also renames some flags/variables to be more generic for updating
purposes. The call used here was previously only used for updating
paint data, but as it was reused here, flags and variables were renamed
to accomodate more clearly to the new usages.
Shadow catching was disabled for hair in Eevee, because of an issue in
the shadow maps. The issue has since been resolved, so this re-enables
shadow catching for hair.
Not a single node based shader was working since a recent merge from
master.
The merge brought changes from the principle bsdf shader where
unsupported gl_ProjectionMatrix was still being used.
Initial support for Python/Manipulator integration
from 'custom-manipulators' branch.
Supports:
- Registering custom manipulators & manipulator-groups.
- Modifying RNA properties, custom values via get/set callbacks,
or invoking an operator.
- Drawing shape presets for Python defined manipulators (arrow, circle, face-maps)
Limitations:
- Only float properties supported.
- Drawing only supported via shape presets.
(we'll likely want a way to define custom geometry or draw directly).
- When to refresh, recalculate manipulators will likely need
integration with notifier system.
Development will be continued in the 2.8 branch
Note that this drops all changes from master,
useful commits will need to be cherry-picked.
Did this since most commits were relating to UV/textures
which is complicated by texface being removed in 2.8
@campbell Barton: Why is this declaration needed at all in stubs.c?
Further up the file collada.h is imported and that already decalres
the function and results in a duplicate declaration.
avoids wrong texture data when multiple objects are exported. Note: This
commit might possiblyt not work fully. The full feature is added with the
next commit)
Technically this enables the use of MinmaxZ pyramid inside the probe captures.
Also Disable AO for cubemap probes because it shows big discontinuities at cubeface limits.
The problem was that the depth prepass was using the clip plane but not the shading pass.
During the clipping stage, the triangle is converted to a quad clipped to the given clip plane.
But this introduce subtle changes in the depth when this new geometry is rasterized. Since the shading pass was using an EQUAL depth test, the depth values from the shading pass were not always equal to the depth prepass.
Enabling clipping in the shading vertex shader has a too small impact to require a dedicated shader.
This fixes an issue introduced by the new output node system, where "No
output node" was displayed in the material panel even when an output
node was present.
Naming was confusing, while technically correct -
the result is no scaling (manipulator ignores zoom-level).
Also remove 3D from name since this can be supported for 2D views too.
Although we are calling all of them light probes, there are a lot of
differences between them. This commit does the following:
* Prevent user from changing the probe type once added
* Unify "sphere" and "cube" probes into reflection cubemap
(as before you can switch between them from the probe UI)
To be done
==========
* Don't show add probe menus unless we are on Eevee
* Light probes panels should not be visible in Clay. Light probe objects
should not be visible in Clay viewport (nor on Cycles).
Notes
=====
* We need icons for the different light probes, and for lightprobes as a
whole (we are using RADIO for now).
Changes from custom-manipulator branch.
- use property type definitions.
- add property free callback.
- move properties into the wmManipulator struct (over alloc).
- use array length from property types instead of arg passing.
Reorder keyingsets registration order, since it also afects order of I
menu options, better show most used ones first, pure alphabetical order
is not great here... Will likely break some muscle memory though. :|
Based on D2720 by Carlo Andreacchio (@candreacchio), thanks.
Although the original report was about the docs, the real issue was in
the API.
My original commit started from a copy-paste from the Switch
Node. However I don't use custom1 for thew Switch View node.
The docs is slightly incomplete since it would be nice to mention the
views here. Or maybe even expose them via Python. But honestly they are
generated depending on the scene multi-view settings.
Implement GTAO (Ground Truth Ambient Occlusion) which is a special case of Horizon Based Ambient Occlusion that is more physically accurate.
Also add a bent normal option to sample indirect irradiance (diffuse lighting) with the least occluded direction.
- Unify variations between default shaders and material shader.
- Only create default shader passes if needed.
Downside is that we have a big array of passes and shading grp in the vedata ... And it will double in size each time a new variation flag is added.
This commit introduce the computation of a depth pyramid containing min and max depth values of the original depth buffer.
This is useful for Clustered Light Culling but also for raytracing on the depth buffer (SSR).
It's also usefull to have to fetch higher mips in order to improve texture cache usage.
As of now, 1st mip (highest res) is half the resolution of the depth buffer, but everything is already done to be able to make a fullres copy of the depth buffer in the 1st mip instead of downsampling.
Also, the texture used is RG_32F which is a too much but enough to cover the 24bits of the depth buffer. Reducing the texture size would make things quite faster.
This special case function enables rendering to a miplevel while using the miplevels above as texture input.
This is needed for some algorithm (i.e. creating a min-max depth pyramid texture).
If there was any specularity in the Principled BSDF, it would get a sampling
weight of one regardless of its actual impact.
This commit makes Cycles estimate the contribution of the component and adjust
the weighting accordingly, which greatly improves the noise characteristics of
the Principled BSDF in many cases.
Note that this commit might slightly change the brightness of areas when using
MultiGGX and high roughnesses, but the new brightness is more accurate and
closer to the result of Branched Path Tracing. See T51836 for details.
Differential Revision: https://developer.blender.org/D2677
The PDF of the MultiGGX sampling is approximated by the singlescattering GGX
term as well as a scaled diffuse term that makes up for the energy in the
multiscattering component that's missed by GGX.
However, there were two problems with the glossy terms: The diffuse term missed
a normalization factor, and the singlescattering term was not properly scaled
down based on the albedo estimate.
The glass term was completely wrong and has been rewritten. It uses the fresnel
factor to weight reflection vs. refraction and uses the glossy MultiGGX model
for reflection.
For refraction, the correct singlescattering term is now used, and a new
albedo approximation is used that was derived by evaluating GGX albedo for
roughnesses from 0 to 1 and IORs from 1 to 3 and fitting numerical
approximations to it. The resulting model has a mean relative error of 9e-5,
but could probably be simplified without losing noticable accuracy in the
final render.
The improved PDFs help with glossy highlights (due to better light sampling vs.
closure sampling MIS) and fix the situation described in T51836 where mixing
MultiGGX with other closures (as it happens in e.g. the Principled
BSDF) causes incorrect darkening.
This is annoying especially for exporters who do use mesh name, since it
broke any relation with actual Mesh naming in original Blend file.
Unfortunately, we cannot avoid the extra .xxx digits. ;)
This makes manipulator access closer to operators,
and allows Python access.
This adds RNA for manipulators, but not Python registration yet.
- Split draw style into 2x settings:
`draw_style` (enum) & `draw_options` (enum-flag)
- Rename wmManipulator.properties -> properties_edit,
Use wmManipulator.properties for ID-properties.
Note that this area of the API will need further work since
manipulators now have 2 kinds of properties & API's to access them.
Temporary screens caused a crash here,
if we can't assume an area will never have its screen changed
better not store a back-pointer.
This makes `rna_ManipulatorProperties_find_operator` search even more
exhaustive. Not nice but hard to avoid :S
Quicker to find the ID-subtype then use direct pointer comparisons.
Also modify `ID_code_to_RNA_type` so missing ID-codes raise a warning
(handy when adding new ID types).
We should be able to differentiate between OpenGL render (viewport
render), offline render (F12), and view render (viewport draw).
This allows for us preventing offline render to skip mode drawings, grid, ...
Even OpenGL render can benefit from this forcing a higher quality
anti-alias and sampling than the viewport drawing.
I'm not sure if it's clever to keep the memset(0x00) outside the render
loop function as it is in this patch. An alternative is to just pass the
render "type" as a flag to the render function, and set DST.options
inside it. (I may change this tomorrow, I will wait to hear from
Campbell on that).
This makes the node menus aware of the Cycles/Eevee distinction, and
only show the relevant nodes for the current engine. Names have also
been changed to accomodate for the new output node system.
This makes Eevee consistent with Cycles, by having a single output node,
and multiple shader nodes that connect to it.
Note that node systems for Eevee saved before this will be missing the
output node, and thus will show an invalid material. This is easily
resolved by connecting the shader output to a new output node.
It tried to assert that
addons/io_blend_utils/blender_bam-unpacked.whl/__init__.py was loaded when
the io_blend_utils module was imported. However, this happens only on
demand, and not directly when importing the add-on.
This makes it possible to select objects in the viewport.
Selection in edit mode is still tricky, mainly because currently such update
would require tagging batch as updated, which is not possible with using just
and original object because it will never have batch.
Possible solution here would be to introduce some "batch" or "render" component
to depsgraph ID node for objects, so we can tag batch for update via depsgraph
tagging API.
< Dependency graph Copy-on-Write >
--------------------------------
\ ^__^
\ (oo)\_______
(__)\ )\/\
||----w |
|| ||
This is an initial commit of Copy-on-write support added to dependency graph.
Main priority for now: get playback (Alt-A) and all operators (selection,
transform etc) to work with the new concept of clear separation between
evaluated data coming from dependency graph and original data coming from
.blend file (and stored in bmain).
= How does this work? =
The idea is to support Copy-on-Write on the ID level. This means, we duplicate
the whole ID before we cann it's evaluaiton function. This is currently done
in the following way:
- At the depsgraph construction time we create "shallow" copy of the ID
datablock, just so we know it's pointer in memory and can use for function
bindings.
- At the evaluaiton time, the copy of ID get's "expanded" (needs a better
name internally, so it does not conflict with expanding datablocks during
library linking), which means the content of the datablock is being
copied over and all IDs are getting remapped to the copied ones.
Currently we do the whole copy, in the future we will support some tricks
here to prevent duplicating geometry arrays (verts, edges, loops, faces
and polys) when we don't need that.
- Evaluation functions are operating on copied datablocks and never touching
original datablock.
- There are some cases when we need to know non-ID pointers for function
bindings. This mainly applies to scene collections and armatures. The
idea of dealing with this is to "expand" copy-on-write datablock at
the dependency graph build time. This might introduce some slowdown to the
dependency graph construction time, but allows us to have minimal changes
in the code and avoid any hash look-up from evaluation function (one of
the ideas to avoid using pointers as function bindings is to pass name
of layer or a bone to the evaluation function and look up actual data based
on that name).
Currently there is a special function in depsgraph which does such a
synchronization, in the future we might want to make it more generic.
At some point we need to synchronize copy-on-write version of datablock with
the original version. This happens, i.e., when we change active object or
change selection. We don't want any actual evaluation of update flush happening
for such thins, so now we have a special update tag:
DEG_id_tag_update((id, DEG_TAG_COPY_ON_WRITE)
- For the render engines we now have special call for the dependency graph to
give evaluated datablock for the given original one. This isn't fully ideal
but allows to have Cycles viewport render.
This is definitely a subject for further investigation / improvement.
This call will tag copy-on-write component tagged for update without causing
updates to be flushed to any other objects, causing chain reaction of updates.
This tag is handy when selection in the scene changes.
This basically summarizes ideas underneath this commit. The code should be
reasonably documented.
Here is a demo of dependency graph with all copy-on-write stuff in it:
https://developer.blender.org/F635468
= What to expect to (not) work? =
- Only meshes are properly-ish aware of copy-on-write currently, Non-mesh
geometry will probably crash or will not work at all.
- Armatures will need similar depsgraph built-time expansion of the copied
datablock.
- There are some extra tags / relations added, to keep things demo-able but
which are slowing things down for evaluation.
- Edit mode works for until click selection is used (due to the selection
code using EditDerivedMesh created ad-hoc).
- Lots of tools will lack tagging synchronization of copied datablock for
sync with original ID.
= How to move forward? =
There is some tedious work related on going over all the tools, checking
whether they need to work with original or final evaluated object and make
the required changes.
Additionally, there need synchronization tag done in fair amount of tools
and operators as well. For example, currently it's not possible to change
render engine without re-opening the file or forcing dependency graph for
re-build via python console.
There is also now some thoughts required about copying evaluated properties
between objects or from collection to a new object. Perhaps easiest way
would be to move base flag flush to Object ID node and tag new objects for
update instead of doing manual copy.
here is some WIP patch which moves such evaluaiton / flush:
https://developer.blender.org/F635479
Lots of TODOs in the code, with possible optimization.
= How to test? =
This is a feature under heavy development, so obviously it is disabled by
default. The only reason it goes to 2.8 branch is to avoid possible merge
hell.
In order to enable this feature use WITH_DEPSGRAPH_COPY_ON_WRITE CMake
configuration option.
*Sigh* One more example of why we should keep ID management handling in
as few places as possible! It's impossible to keep more than a few
places in sync regarding which ID pointer is refcounted etc.
Maybe not very useful but it's here for feature parity with the other probes.
The fragment shader clipping is here because I'm lazy and don't want to creating a proper quad Batch.
Children where always getting at least one segment of fixed length...
Now fully hidden ones (zero length) get no segment at all.
Note that even very short ones keep getting one 'unit' length segment - would
rather avoid changing that at this point, given how complex children
particles 'length' can get with all kind of modifiers... Think we can
live with that for now anyway.
This avoids having to use manipulator-type specific functions
to set the orientation.
And will make it simpler to access transformation from Python.
Currently the matrix is still used as an offset in places.
Also per-type orientation values still need to be removed.
The crash did not happen yet because we always had proper vmemh defined in
the parent scope.
Patch by Ivan Ivanov (aka obiwanus), thanks!
Differential Revision: https://developer.blender.org/D2715
- Option to start helper angle lines based on a vector
instead of the initial mouse coords (useful for bisect & spin).
- Option to show 2x helper lines
useful when dial is used to rotate an axis value.
Calculation was done in screen-space giving inaccuracy,
making the angle incorrect for tool code.
Cast mouse coords onto the dial plane to calculate the angle instead.
Instead use generic 'WM_manipulator_new', adding a new 'setup'
callback (like wmManipulatorGroup.setup) used to initialize type vars.
This moves conventions closer to wmOperator and simplifies exposing to
Python.
This is not enough to mutex-guard modification code of integer values,
since this operation is NOT atomic. This is not even safe for a single
byte data types.
For now guarded the getter functions, similar to other functions in
this module.
Ideally we want to switch modification to an atomic operations, so we
wouldn't need any locks in the getters.
This is needed so manipulators can tag themselves for removal
without causing problems from freeing data within a callback.
Also use properties within the dial manipulator and fix an error where
removing a wmManipulatorGroupType didn't remove its keymap.
Since we're unlikely to use this, it's tedious to keep so many
differences from 'blender2.8' branch.
Currently the only difference between this
branch and 2.8 is RNA/Python work-in-progress integration.
This is needed so manipulators can tag themselves for removal
without causing problems from freeing data within a callback.
Also use properties within the dial manipulator and fix an error where
removing a wmManipulatorGroupType didn't remove its keymap.
'Convert To...' Object operation has very weird effect of actually
working at obdata level, not object level, which means *all* objects
(even unselected/hidden/in other scenes/...) using same obdata will be
converted to new selected type.
IMHO this is very bad behavior, but... not a bug really, so do not
change this for now.
But at least, do not do that when working on some linked data, else it
leaves Blend file in invalid (incoherent) state until next reload.
So workaround for now is to enforce the 'Keep Original' option when some
linked object/obdata is affected by the operation.
Also fixed somewhat broken usercount handling in Curve->Mesh part.
That one was probably not an actual issue, except maybe in some corner
cases (like deleting a linked scene also used by some other linked scene).
Again, better not try to do smart & complex freeing logic outside of
BKE_library area, let's keep spaghetti nitghmare in a single place!
While this is work-in-progress from custom-manipulators branch
its stable so adding into 2.8 so we don't get too much out of sync.
- ManipulatorGroupType's are moved out of the manipulator-map and are now
global (like operators, panels etc) and added into spaces as needed.
Without this all operators that might ever use a manipulator in the 3D
view would be polling the viewport.
- Add optional get/set callbacks for non-RNA properties
Needed so re-usable manipulators can control values that
don't correspond to a single properly or need conversion.
- Fix divide by zero bug in arrow manipulator (when moving zero pixels).
Previous code (same as what `BKE_libblock_free_us()` is doing when
usercount reach 0) was probably OK in that specific case, but still not
good idea, and potentially risky.
Even though in that specific it was probably safe-ish, there is no
guarantee at this point Brush we want to remove are not used somewhere,
better take the slightly slower, much safer `BKE_libblock_delete()` path here.
Eeeeeek!^2 Calling unconditionnaly ID freeing `BKE_libblock_free()` on a
datablock (ob->data, i.e. Curve) that may be used elsewhere...
Veryveryvery bad!
Until 3d-viewport manipulators change functionality we need to it
working how it already is.
So add modal operator which is called from draw when there are no
properties applied to the manipulator.
This will be needed to have a general rotation manipulator.
*tryed "#" as preprocessor used in POV-Ray for language keywords best behaviour was to have it as a punctuation symbol
*moved "finish" to its proper category
*changed order of some POV-Ray ini files keywords to have them work better
*added a few keywords from latest pov version
*Fixed C-style closing of multiline comments (*/)
Reviewers: campbellbarton, mont29
Reviewed By: campbellbarton, mont29
Subscribers: mont29
Differential Revision: https://developer.blender.org/D2707
Early implementation. Slow and still has quality
3 ways of storing irradiance:
- Spherical Harmonics: Have problem with directionnal lighting.
- HL2 diffuse cube: Very low resolution but smooth transitions.
- Diffuse cube: High storage requirement.
Also include some name change.
Noisy change, but safe, and better do it sooner than later if we are to
rework copying code. Also, previous commit shows this *is* useful to
catch some mistakes.
- Add a different kind of properties that use function callbacks
instead of RNA.
Needed for situations when there isn't 1:1 correspondence
between the manipulator's position and the internal value.
- Move manipulator properties into their own file.
This is similar to some other datablocks. Mainly applies to collections,
so freeing scene does not involve changing any non-directly owned data.
There are two main usecases foreseen for the future:
- Less CPU ticks on bmain free, where everything is freed anyway and
there is no need to preserve id counters.
- Easier freeing of temporary data, including data which is used by
depsgraph's copy-on-write mechanism.
Neither of those are currently implemented, but will be shortly.
New implementation of hair for Eevee.
Note: A hard coded "transmission" property is being used. This should
eventually be exposed to the UI, possibly in the form of SSS
properties.
This stops the memory breakage caused by the material preview jobs,
to enable the new hair drawing implementation.
(This should be reverted once T51796 is actually fixed!)
These changes are intended for operators to register their own
widget-types temporarily, so _every_ operator that uses manipulators
doesn't need to keep them continuously polling the view to check if the
operator is running.
- Register `wmManipulatorGroupType` globally (like all other RNA types).
- Add `wmManipulatorGroupTypeRef` for type-maps to reference groups.
- Remove `wmManipulatorMapType.idname` (spaceid & regionid are enough info).
- Add PERSISTENT flag for `wmManipulatorGroupType`
intended for use with operators so loading a new file for eg doesn't
keep the manipulator type around.
Avoids using copy-constructor invoked every time we pass function
to the builder functions.
Should lower number of CPU ticks spent during DEG construction.
Was rather weird and only used for time source. It is simpler to make depsgraph
to keep track of time source directly.
No need to introduce extra entitites without actual need.
There is no reason to be special for workspace and go against other design
decision in Blender. If something like this is going to become a common
practice in Blender it should be well thought and well tested, including
tests of all supported compilers and configurations.
This feature was relying on type re-definition, which is not only confusing
but also available in C11 only.
Was some mismatch in address space. Seems to be caused by recent additions.
Additionally, moved decoupled ray marching functions under ifdef, so they
don't try to use malloc() functions.
Thanks Mai for testing the patch!
I considered just copying the evaluated data from the LayerCollection.
However we need to run the evaluation so Depsgraph can handle overrides.
I will double-check with Sergey Sharybin.
We need to traverse the same exact objects for both nodes and relationship
builder. We were using FOREACH_SCENE_OBJECT for relationships, which
would lead to plenty of warnings in multiple situations.
In the future we will need to change this to build the depsgraph
relations and nodes to one single render_layer.
Fix T51780: If an object is in two collections and I do a full copy of
scene things go bad
Now, when there is no usable neighboring pixel for denoising, the noisy value
is preserved instead of producing a NaN.
Also, negative results are clamped to zero.
Note that there are just workarounds that don't fix the underlying problems,
but these issues are very rare and I'm not sure if it's even possible to fix
the underlying problems without introducing a significant slowdown or quality
decrease in other situations.
Because of that and since 2.79 is happening very soon, I just went for these
workarounds for now.
Replaces the placeholder 'emtpy' icons of "Force Field" and "Group
Instance" entries in object-add menu with proper new ones.
Icons by @zlsa, thanks a lot!
Maniphest task T51291.
Technically not passing all buffers used by a kernel is undefined
behavior. We haven't had any issues with this so far on AMD or
Nvidia, but it's known to be a problem with Intel and we received
a report from AMD that this is a problem on newer hardware, so we
need to make this change at some point.
Unfortunately there a cost to being correct, about 5% for the
benchmark scenes. For low sample counts it's even worse, I've
seen up to 50% slowdown. For the latter case I think adjusting
tile updating logic can help, but not sure what that would look
like yet (it would be just a few lines change however).
Unlike regular path tracing, branched path tracing is usually used with lower
sample counts, at least for primary rays. This means that are less samples for
the GPU to work on in parallel and rendering is slower. As there is less work
overall there is also more inactive threads during rendering with BPT. This
patch makes use of those inactive rays to render branched samples in parallel
with other samples.
Each thread that is preparing for a branched sample will attempt to find an
inactive thread and if one is found the state for the sample is copied to that
thread. Potentially, if there are enough inactive threads, 100s of branched
samples could be generated from the same originating thread and ran in
parallel giving large speed ups.
Gives 70% faster render for pavillion midday scene. 20-60% faster on BMW
with car paint replaced with SSS/volumes.
Due to various driver issues with AMD GCN 1 cards we can no longer support
these GPUs. This patch makes them unavailable to select for Cycles rendering.
GCN cards 2 and higher are still supported. Please use the most recent
drivers available to ensure proper functionality.
See here for a list to check which GPUs are supported:
https://en.wikipedia.org/wiki/List_of_AMD_graphics_processing_units
Sync with custom-manipulators branch
- Use identifiers for properties.
- Property array index access.
- Remove operator from manipulators
(wasn't used and will likely add in a different way).
One 1024px² octahedral probe (according to the GPU mem stats):
- RGBA16F = 17Mb
- RGB_11_11_10 = 4Mb
For the time being I prefer to maximize the number of probe possible in the scene for users to test.
This is obviously a temporary solution and the final choice of cubemap format should be exposed to the user.
After using it for like 30 sec, the min max bound box is absolutely not practical. Reverting into using a unit cube with object transform.
This also simplify the code.
In the future of center probes will be implemented using another object matrix (via an object pointer).
Probes were completely out of depsgraph, so tagging them could not work at all.
For now using some placeholder operations just to ensure order of updates.
We still do it a few times, but that helps already. Related to T51718.
Note that it also reinforces the idea that any geometry datablock will
have a generated copy-on-write Mesh provided by Depsgraph.
Reported by Andy Goralczyk (@eyecandy) over IRC, thanks!
Simply nuke all that poor broken custom one-by-one handling in
object_relations.c code, and use highly complex but powerful and
well-tested BKE_library_make_local() in all cases of MakeLocal!
ID management, especially related to linking, is very hairy matters,
better to have as few as possible core functions managing all the dirty
details. ;)
Regular camera lens have a ratio of 1.0. The anamorphic bokeh can be
used freely as an artistic decision, but it's nice to clamp it to 2.0,
to match real cameras.
That end up giving a reference for artistis playing with the
parameters unaware of the more realistic limits.
Edge collapse was using bounding box center as the point to collapse to.
When collapsing multiple adjacent edges together, this caused
inconsistencies in placement of the collapsed point, depending on the
orientation of the edges in relation to the space axis.
This makes edge collapse use the mean point instead.
The previous outlier heuristic only checked whether the pixel is more than
twice as bright compared to the 75% quantile of the 5x5 neighborhood.
While this detected fireflies robustly, it also incorrectly marked a lot of
legitimate small highlights as outliers and filtered them away.
This commit adds an additional condition for marking a pixel as a firefly:
In addition to being above the reference brightness, the lower end of the
3-sigma confidence interval has to be below it.
Since the lower end approximates how low the true value of the pixel might be,
this test separates pixels that are supposed to be very bright from pixels that
are very bright due to random fireflies.
Also, since there is now a reliable outlier filter as a preprocessing step,
the additional confidence interval test in the reconstruction kernel is no
longer needed.
This was introduced in 23c93873f4.
This could be moved to deg_flush_base_flags_and_settings but since we
only need this for duplis I think it's fine to be handled separately.
Use indirect access to it via object.
It was already flushing from base to object, now we can avoid such flushing.
Still weird to have selection color filled in by dependency graph, but now
there is no synchronization going on at least.
As with operators, the window-manager has the API for defining,
the editor can implement and register its own manipulators.
This exposes wmManipulator, keeping it opaque isn't
practical if editors and Python are to implement their own.
UNIFORM_NONE should never match a valid uniform (builtin or custom).
The logic for UNIFORM_CUSTOM was just wrong, since it returned the first custom uniform. This function should only accept builtin (non-custom) uniforms.
Quick hash rejection instead of string comparison. Uniform lookups already work this way. I don't expect a major overall speedup since attributes are looked up less frequently than uniforms.
Having the type in mixed in with instance
made it hard to expose types to RNA/Python.
This disables face-map select,
we could enable it but it looks like face maps will be made to work
differently.
The problem was that the strokes in the copy-paste buffer could be keeping
dangling pointers to colors that were already freed. Therefore, this commit
makes it so that when copying the strokes, we now make copies of the colors
and put them in a hashtable beside the stroke buffer. This is convenient,
as it saves us having to look up what colours need to be copied over each
time when pasting.
The problem was that newly pasted strokes were still using colours from
the original datablock. As a result, you'd either get an immediate crash,
or if you managed to save the file before it crashed, each stroke would get
reloaded with a dummy colour.
This commit fixes makes it possible to copy/paste strokes between datablocks
again. However, there are still problems when trying to paste across file
boundaries (i.e. copy strokes in one file, paste in another), which the next
commit will address.
faces.
This was requested by script writers. Especially needed if beveling
wire edges with vertex_only.
Should be backward compatible as just adds two new keys to returned
dict in python ('edges' and 'verts').
Was internally a no-op operation, which only caused extra work
to be done during depsgrpah traversal and evaluation, without
making any measurable improvement.
The issue was caused by usage of address of dupli-object (which will vary
from iteration process to iteration process) as something denoting whether
we've got the data synchronized to Cycles or not.
For now solved by using address of original object (the one DupliObject
points to) as a pointer for the map.
Need to do more thoughts about this.
This is supposed to help catch bugs if referrencing stack data out of
the draw loop context.
No change is suppose to happen for users (specially because the changes
here happens mostly on debug).
It includes a change in the logic for render loop, to make sure DST is
not accessed before we enter it - contribution by Campbell Barton.
This commit extends the work from Dalai made around scene iterators to
support iterating into objects from dupli-lists.
Changes can be summarized as:
- Depsgraph iterator will hold pointer to an object which created current
duplilist. It is available via `dupli_parent` field of the iterator.
It is only set when duplilist is not NULL and guaranteed to be NULL
for all other cases.
- Introduced new depsgraph.duplis collection which gives a more extended
information about depsgraph iterator. It is basically a collection on top
of DEGObjectsIteratorData.
It is used to provide access to such data as persistent ID, generated space
and so on.
Things which still needs to be done/finished/clarified:
- Need to introduce some sort of `is_instance` boolean property which will
indicate Python and C++ RNA that we are inside of dupli-list.
- Introduce a way to skip dupli-list for particular objects.
So, for example, if we are culling object due to distance we can skip all
objects it was duplicating.
- Introduce a way to skip particular duplicators.
So we can skip iterating into particle system.
- Introduce some cleaner API for C side of operators to access all data such as
persistent ID and friends.
This way we wouldn't need de-reference iterator and could keep access to such
data really abstract. Who knows how we'll be storing internal state of the
operator in the future.
While there is still stuff to do, current state works and moves us in the proper
direction.
We only show object center if object is selected, active or if viewport
has the "All Object Origins" options.
The viewport display options can migrate to renderlayer options.
However, we can mimic 2.7x as a compromise while the final design is
finalized.
This way we are not afraid of recursion being too deep.
That could have happened when having two collections which
are sharing same list of 1000s of objects.
- Move callbacks into type struct.
- Rename render_3d_intersection -> draw_select.
- Add header for function signatures (needed for types and api headers).
- Add WM_manipulatormaptype_find
- move callbacks into type struct
- rename render_3d_intersection -> draw_select
- add header for function signatures (needed for types and api headers).
Replace completly random noise by Blue noisen, giving a better aspect.
Also randomize the distance in the sample direction to cover the whole distance even with 1 sample. Using another blue noise for this.
Replace spiral samples (that had tendency to align if the number of samples was near the chosen constant) with Hammersley samples that have good coverage even for low number of samples.
Use a UBO instead of Texture (a bit less latency) making things a tiny bit faster.
Move the noise data to the SceneLayerData, because each render layer can have a different sample count.
Separate material handling inside another file.
Make use of enums to identify shader variations.
Group all 64*64 LUTs into one array texture.
Only update world probe if world changes.
Cycles is using rna_depsgraph, not rna_scene. Duplis are still not working there
but now at least it shows it was showing before the commit.
To show duplis in Cycles do:
- data->flag = DEG_OBJECT_ITER_FLAG_SET;
+ data->flag = DEG_OBJECT_ITER_FLAG_ALL;
Now dupli groups, objects, particles, ... are all working.
This introduces a flag for the iterator to determine whether we go over
Set and dupli objects or not.
Important to remember to keep the iteration of DEG_ as readonly.
Cycles is not working well for dupli groups, and it's memleaking
for dupli particles. So for now we iterate over main objects and set
only, not dupli.
To change that go in rna_scene.c and:
-DEG_OBJECT_ITER(graph, ob, DEG_OBJECT_ITER_FLAG_SET)
+DEG_OBJECT_ITER(graph, ob, DEG_OBJECT_ITER_FLAG_ALL)
Review and suggestions by Sergey Sharybin
This is a minor update add more information on how Blender handles modal
operators. The existing docs provide a good overview, but might not be
as helpful to those unfamiliar with modal programming. This patch also
corrects a few small grammar issues.
This commit moves the list of transform orientations from scenes to workspaces.
Main reasons for this are:
* Transform orientations are UI data and should not be stored in the scene.
* Introducion of workspaces caused some (expected) glitches with transform orientations. Mainly when removing one.
* Improves code.
More technically speaking, this commit does:
* Move list of custom transform orientations from Scene to WorkSpace struct.
* Store active transform orientation index separate from View3D.twmode (twmode can only be set to preprocessor defined values now).
* Display custom transform orientation name in header when transforming in it (used to show "global" which isn't really correct).
This commit does the main integration of workspaces, which is a design we agreed on during the 2.8 UI workshop (see https://wiki.blender.org/index.php/Dev:2.8/UI/Workshop_Writeup)
Workspaces should generally be stable, I'm not aware of any remaining bugs (or I've forgotten them :) ). If you find any, let me know!
(Exception: mode switching button might get out of sync with actual mode in some cases, would consider that a limitation/ToDo. Needs to be resolved at some point.)
== Main Changes/Features
* Introduces the new Workspaces as data-blocks.
* Allow storing a number of custom workspaces as part of the user configuration. Needs further work to allow adding and deleting individual workspaces.
* Bundle a default workspace configuration with Blender (current screen-layouts converted to workspaces).
* Pressing button to add a workspace spawns a menu to select between "Duplicate Current" and the workspaces from the user configuration. If no workspaces are stored in the user configuration, the default workspaces are listed instead.
* Store screen-layouts (`bScreen`) per workspace.
* Store an active screen-layout per workspace. Changing the workspace will enable this layout.
* Store active mode in workspace. Changing the workspace will also enter the mode of the new workspace. (Note that we still store the active mode in the object, moving this completely to workspaces is a separate project.)
* Store an active render layer per workspace.
* Moved mode switch from 3D View header to Info Editor header.
* Store active scene in window (not directly workspace related, but overlaps quite a bit).
* Removed 'Use Global Scene' User Preference option.
* Compatibility with old files - a new workspace is created for every screen-layout of old files. Old Blender versions should be able to read files saved with workspace support as well.
* Default .blend only contains one workspace ("General").
* Support appending workspaces.
Opening files without UI and commandline rendering should work fine.
Note that the UI is temporary! We plan to introduce a new global topbar
that contains the workspace options and tabs for switching workspaces.
== Technical Notes
* Workspaces are data-blocks.
* Adding and removing `bScreen`s should be done through `ED_workspace_layout` API now.
* A workspace can be active in multiple windows at the same time.
* The mode menu (which is now in the Info Editor header) doesn't display "Grease Pencil Edit" mode anymore since its availability depends on the active editor. Will be fixed by making Grease Pencil an own object type (as planned).
* The button to change the active workspace object mode may get out of sync with the mode of the active object. Will either be resolved by moving mode out of object data, or we'll disable workspace modes again (there's a `#define USE_WORKSPACE_MODE` for that).
* Screen-layouts (`bScreen`) are IDs and thus stored in a main list-base. Had to add a wrapper `WorkSpaceLayout` so we can store them in a list-base within workspaces, too. On the long run we could completely replace `bScreen` by workspace structs.
* `WorkSpace` types use some special compiler trickery to allow marking structs and struct members as private. BKE_workspace API should be used for accessing those.
* Added scene operators `SCENE_OT_`. Was previously done through screen operators.
== BPY API Changes
* Removed `Screen.scene`, added `Window.scene`
* Removed `UserPreferencesView.use_global_scene`
* Added `Context.workspace`, `Window.workspace` and `BlendData.workspaces`
* Added `bpy.types.WorkSpace` containing `screens`, `object_mode` and `render_layer`
* Added Screen.layout_name for the layout name that'll be displayed in the UI (may differ from internal name)
== What's left?
* There are a few open design questions (T50521). We should find the needed answers and implement them.
* Allow adding and removing individual workspaces from workspace configuration (needs UI design).
* Get the override system ready and support overrides per workspace.
* Support custom UI setups as part of workspaces (hidden panels, hidden buttons, customizable toolbars, etc).
* Allow enabling add-ons per workspace.
* Support custom workspace keymaps.
* Remove special exception for workspaces in linking code (so they're always appended, never linked). Depends on a few things, so best to solve later.
* Get the topbar done.
* Workspaces need a proper icon, current one is just a placeholder :)
Reviewed By: campbellbarton, mont29
Tags: #user_interface, #bf_blender_2.8
Maniphest Tasks: T50521
Differential Revision: https://developer.blender.org/D2451
It was never actually used apart from being stored at a construciton time.
This caused some redundancy and ncertanty about which relation type to use
during construciton (often existing types were not close enough to particular
use case).
Before this change Gawain was doing list lookup twice,
doing string comparison of every and each input which
is not efficient and not friendly for CPUs with small
cache size.
Now we store hash of input name together with actual
name and compare hashes first. Additionally, we do
everything in a single pass which is much better from
cache coherency point of view.
This brings Eevee cache population time from 80ms to
60ms on my desktop and from 800ms to 400ms for Clement
when navigating in a file from T50027.
Reviewers: merwin, dfelinto
Subscribers: fclem
Differential Revision: https://developer.blender.org/D2697
Made this resilient to unknown types, for now. Supporting specific INT
sockets (through implicit conversion to GPU_FLOAT ones) is considered nice TODO.
Was just keeping the default '1' user from `BKE_libblock_alloc()`,
instead of using correct way to handle extra virtual user needed when we
want to keep unused datablocks around...
Do not call invoke ops from outliner's operations menus. Invoke op would
search again for item under mouse coordinates... when it is invoked!
Means often entry menu you would have clicked would not be over target
item, leading to either nothing or operation being applied to wrong item.
Note: about groups, there is another minor annoyance leading to some
assert - groups have an annoying virtual fake user which breaks
usercount, will see whether this is easily fixable. :|
The idea is to accumulate all new tasks in a thread local queue
first without doing any thread synchronization (aka, locks and
conditional variables) and move those tasks to a scheduler queue
once they are all ready. This way we avoid per-task-pool lock
and only have one lock per bunch of tasks.
This is particularly handy when scheduling new dependency graph
node children. Brings FPS of cached simulation from the linked
below file from ~30 to ~50.
See documentation for BLI_task_pool_delayed_push_{begin, end}
and for TaskThreadLocalStorage::do_delayed_push.
Fixes T50027: Rigidbody playback and simulation performance regression with new depsgraph
Thanks Bastien for the review!
If users wanted to bake only a few of the mesh materials, they would
still need to create dummy textures for the other parts.
This commit report (as RPT_INFO) the materials with no texture, but move
on to bake the others materials.
This was causing proxies updates on every frame, even if they
do not really change. Additionally, it was causing second round
of armature update when used from inside dupligroup (viewport
ensures all objects from dupligroup are up to date before draw).
It's now less confusing (for example, using nr_of_samples directly,
instead of using 1 / 1 / nr_of_samples). Might also have fixed a bug.
Also added unittests.
The scale matrix must have its homogeneous 'w' (at mat[3][3]) set to the
scale in order to also scale the translations along with it. However, this
also scales the transform matrix's 'w' component, which is not supposed
to happen.
The old default values (start/end frame = 1) could have been an actually
desired setting (for example when exporting a non-animated model). To
make this worse, this was only interpreted as "start/end of the scene" by
the export operator when running interactively, but not when run from
Python.
By choosing INT_MIN as default it's highly unlikely that the interval
[start, end) was intended as actual export range.
This way we always have predictable behavior, especially from the
performance point of view. Additionally, if some bottleneck is found
in stack implementation it'll be easier for us to address.
Add face maps, needed for face-map widgets,
only data structure, widgets will be separate commit.
This comes from 'custom-manipulator' branch with only minor changes.
Yep, that got reported... Was slightly more involved than UI message
fixing though: RNA string length getter shall return exact lentgh of
string (same as strlen), not size of allocated buffer to contain it!
Otherwise, NULL final char leaks in and...
Poly index was used for a short time when switching to BMesh.
This is practically the same kind of data,
any inconsistency between object face-maps will need to be supported
anyway - since object & mesh might span blend-files.
Manually setting up links per bone to an object & facemap is tedious.
It's also going to cause quiet a bit of book-keeping internally to
ensure its always valid (linking, adding removing objects proxies).
Instead match the face-maps names to bones (as with vertex groups).
The DNA members have been moved into runtime cache.
This also adds an option for bones to use facemaps,
to avoid excessive searching on heavy armatures that might use very few
(if any) face-maps.
Denoising was setting session parameters for every frame, which was detected as
a change and therefore caused a resync.
Since the parameter modification change is only needed for viewport rendering
(which doesn't support denoising anyways) and resyncing after a frame change
(which isn't affected by denoising settings), an easy fix is to just ignore
the denoising parameters like it's currently done with the samples.
This enables us to use 2D texture arrays for multiple probes.
There is a little artifact with very high roughness caused elongated pixel due to the projection (along every 90° meridian).
Follow up to 9f044cb422
These comments described the difference between Microsoft & MinGW's struct definition. Now that we dropped MinGW we don't need to go into these details.
DM evaluation code was simply never clearing the `deformedOnly` flag
when evaluating a generative modifier...
Quite astonishing this never got catched before, a lot of particle code
relies on valid value of this flag!!!
We now floor the corner position and use this position as origin.
This gives us perfect derivatives in all cases even if very far from the origin.
Unfortunately this won't fix the low precision of coordinates used for computing the actual grid size, resulting in thick, non-smoothed lines.
Also did a bit of refactor how the axes are drawn.
Do not use the inverse perspective matrix inside the shader to recover world positions.
That leads to severe float imprecision leading to nasty artifacts.
Instead we compute the world view vector for each pixels and do a ray to plane intersection.
We are still getting low precision derivatives when going far away from the origin, and thus artifacts.
This commit also fixes the non-appearing negative Z axis in 3D view.
The fact that we can end with uninstantiated objects is not expected
currently, but would rather not start chasing all corner cases that may
lead to that situation.
User shall be able to delete uninstantiated objects from Outliner, though!
Properly remap nodes' pointers to copied IDs in copied ntrees.
Note that this only affects root trees, node groups are not concerned
here, since they are assumed to be reusable chunks and hence *not*
duplicated.
The operator is indeed not adding frames but inserting them at the current frame (shifting all subsequent ones). Changed the operator name and description.
Approved by Antonio.
This operator relies on a rather specific context setup, so it shall not
be exposed to user in 'operator search' menu etc.
Based on D2528 by Vuk Gardašević (lijenstina).
The Issue
=======
For a long time now MinGW has been unsupported and unmaintained and at this point,
it looks like something that we should just leave behind and move on.
Why Remove
==========
One of the big motivations for MinGW back in the day is that it was free compared to MSVC which was licensed based.
However, now that this is no longer true we have basically stopped updating the need CMake files.
Along with the CMake files, there are several patches to the extern libs needed to make this work. For example, see:
https://developer.blender.org/diffusion/B/browse/master/extern/carve/patches/mingw_w64.patch
If we wanted to keep MinGW then we would need to make more custom patches to the external libs and
this is not something our platform maintainers are willing to do.
For example, here is the patches needed to build python: https://github.com/Alexpux/MINGW-packages/tree/master/mingw-w64-python3
Fixes T51301
Differential Revision: https://developer.blender.org/D2648
We had handling of fully duplicated polygons already, but... absolutely
nothing to sanitize partially merged polygons! This were giving us
totally invalid geometry, with duplicated vertices in single poly,
invalid edges, etc.
Now we do check for invalid loops inside polys, and generate new edges
as needed to get only valid polys.
For some reason this was a nightmare to get running fully OK, playing
with old and new indices is really, really mind breaking.
Note for users
==============
The active_layer option used for the filebrowser operators is now called
active_collection.
If there is no collections in the scenelayer or if this option is not selected
we automatically create a new collection for the new objects.
This is the same behaviour of trying to add a new object when there is
no collection.
Note for developers
===================
For those cases I moved the object user count handling from readfile to
the scene collection system. It's working fine for those, but we still
need to re-visit this for Add objects, and Duplicate - In those cases
the usercount is 2 when it should be 1.
Reviewers: mont29, sergey
Differential Revision: https://developer.blender.org/D2686
This feature got lost with new auto-track API,
Added it back by extending frame accessor class. This isn't really
a frame thing, but we don't have other type of accessor here.
Surely, we can use old-style API here and pass mask via region
tracker options for this particular case, but then it becomes much
less obvious how real auto-tracker will access this mask with old
style API.
So seems we do need an accessor for such data, just matter of
finding better place than frame accessor.
It's a bit ugly but I couldn't find a better way to keep fast installs and
correct handling of switching between master and blender2.8 with different
lib directories.
- Derived-mesh drawing.
- All non UV members of TexFace structs.
MTexPoly is now redundant but keeping with a dummy member,
will check on complete removal later.
TexFace complicates the now more popular shading pipeline by having
per-face images, see: T51382 for details.
To keep the ability to select a per-material edit-image
(used with UV-mapping workflow), the material now stores an image
which will be set when changing images in edit-mode.
This is used as a bake-target when not using Cycles too.
This makes it possible to have an animated / procedurally generated mesh
that starts empty and obtains data in later frames.
Fixes the export of an empty mesh with an Ocean Modifier, as described in
issue T51351.
This allows you to put any kind of animation data on the mesh, and its
shape will be exported on each timekey. Note that this timekey is unrelated
to the animation data (so we don't export on each keyframe, for example).
A practical example is the addition of an animated custom property to
trigger the export of animated mesh data. The mesh data can then be created
from any source, like Python scripts.
Not only is this useful in itself, it also provides a workaround for one
of the two issues described in T51351.
This is written in a custom metadata key, so it isn't shown by utilities
like abcecho or abcls. However, it's still something that's useful to
have available.
Storing edit-mesh data as indices for passing
only to get back edit-mesh data.
This also complicated checks for real edges in tessellation tris.
Simpler to pass data directly.
Houdini writes vertex data in a different format than Blender does; Houdini
uses "face-varying scope", which means that the vertex colours are indexed
by an ever-increasing number over all vertices of all faces instead of the
vertex index.
I've also merged the read_custom_data_mcols() and read_mcols() functions,
because the latter was only called from the former, and the changes in this
commit would add yet more function parameters to pass.
A big chunk of code was copied between the if and else bodies. By using
a boolean to store whether the c3f_ptr or c4f_ptr should be used, the
in-loop condition is kept as simple as possible.
Note: the angle in bug isn't really reflex - using the vertex normal
for this test isn't always right, but usually is. At any rate,
shouldn't try to put vertex on edge between if a reflex angle.
This was two-fold.
1) The export used viewport settings to obtain the particle cache, rather
than render settings.
2) The child hair writer tried to obtain UV-coordinates from the parent
chair, without checking whether those were available in the first place.
This function is not performance critical, but I prefer the branch-free code and no hack needed to appease gcc.
Follow-up to recent 23035cf46f and f637145450.
Since we already have a rather advanced PovRay exporter, makes sense to
also nicely display generated 'code'.
Patch by Maurice Raybaud (@mauriceraybaud), thanks!
Cleanup (mostly styling) by @mont29.
Brightness/contrast node was changing color but did not modify alpha
or ensured colors are premultiplied on the output. This was giving
artifacts later on unless alpha was manually converted.
Compositor is supposed to work in premultiplied alpha (except of
some really corner cases) so it makes sense to ensure premultiplied
alpha after brightness/contrast node.
This is now done as an option enabled by default, so we:
(a) Keep compatibility with old files.
(b) Have correct behavior for newly created files.
Later on we can get rid of this option.
We were looping over all vgroups in destination mesh and making string
comparison, for every vgroup of every vertex of merged mesh! Crazy!
Now we simply create a temp mapping of vgroup indices, seriously
simplifies things (and gives significant speedup when merging huge meshes
with lots of vgroups, here with quick stupid test went from 120ms in
vgroup merging to less than 5ms, 25 times quicker!).
Root of the issue here was that two stupid modifiers could create named
vgroup CD layers (vgroup editing ones... shame on me :") ).
Fix that, and added some versionning code to also fix 'corrupted' blend
files created by those so far.
Seems re-loading module invalidates memory pointers by the looks of it,
which gives an error on the next kernel call.
Not sure how to move memory pointer from one CUDA module to another one,
so for now simply disabling kernel re-load for CUDA devices. Not ideal,
but better than failing render.
Feature-selective option for CUDA is not an official feature anyway.
`screen_findedge()` is not expected to return NULL in that case, but
checking against that does not hurt (we do it in all its other call
cases anyway), better than crashing.
For some reason GCC-6 successfully compiles test program with
-Wno-implicit-fallthrough passed via command line. It just
silently ignores the unknown arguments which are starting with
-Wno-.
The issue is, if some other waning happens in the code, then
GCC will complain about unknown -Wno- argument which is not
supported by current GCC version.
This makes some misleading warning prints about unknown
command line argument when any other warning happens in code
from extern/.
Goal is to make most of the API independent of OpenGL, Vulkan, any other backend.
Able to remove default case from ElementList_size because IndexType only covers index types. Not that and *everything else* like GLenum.
Volume shaders without anything connected to the surface output are treated
as if they had a transparent BSDF as the surface shader in Cycles, so the
denoiser should skip feature pass writing for them just as it does with an
actual transparent BSDF.
If the central pixel is an outlier, the denoiser is supposed to predict its
value from the surrounding pixels. However, in some cases the confidence
interval test would reject every single surrounding pixel, which leaves the
model fitting with no data to work with.
We render linear distance to the light in a R32 texture and store it into an octahedron projection inside a 2D texture array.
This render the sampling function much more simpler and without edge artifacts.
- Some arguments were inapproriatry tagged as unused
using (void)foo semantic.
Only use such semantic in tricky casses, when something
needs to be ignored in release builds or something is
dependent on tricky ifndef policy.
For rest of the cases just use void foo(int /bar*/)
semantic, which ensures variable is not used. Solves
confusion and code running out of sync with later
development.
- Used proper unused semantic to some arguments.
- Added braces to make code easier to follow, tricky
indentation with ifdef, uh.
For example, when using a radius of 1, only 9 pixels (due to weighting maybe
even less) will be used, but the transform code may still decide to use a
5-dimensional (or even higher) fit.
This causes severe overfitting and therefore weird pixel values.
To avoid this, this commit limits the amount of dimensions to a third of the
pixel number. For a radius of 3 or more, this doesn't change anything, but
for 1 and 2 it can prevent fireflies and/or negative values being produced.
Once again, numerical instabilities causing the Cholesky decomposition to fail.
However, further increasing the diagonal correction just because of a few
pixels in very specific scenes and settings seems unjustified.
Therefore, this commit simply falls back to the basic NLM-filtered pixel
if the more advanced model fails.
Shader stages need to agree about interpolation qualifiers. Apparently implicit smooth (the default) and explicit smooth are considered different by some GLSL compilers. Found by @letterrip on Linux + Intel.
Follow-up to 941e739d70
prefer vector math over scalar
prefer * over /
shorten vec3(x, x, x) to vec3(x)
use clamp, max, etc. instead of custom logic
declare loop vars as part of for loop
spacing
vertexColor output was not being written --> shader failed to link --> assert hit while setting that shader's uniforms.
Vertex attribs are smooth by default, so I shortened the declaration.
@fclem or @dfelinto: is color = 0 ok here?
This still has a couple of issues:
* Instancing is not working when multiple particle systems use the same
primitive. Only the last particle system to be drawn with a particular
primitive shows up.
* Because of colors being passed as uniforms with static variables, the
color of the collection of the last object to be evauluated is used for
all particles being displayed.
Also, note that while this is being drawn in the clay engine, this might
be moved to the object mode later intead.
Part of T51378
This was introduced by a removed line in
rB0eb32ab22809d9c0c41bfff5082f58b1a7aa9965
If we want to change the value to a differen default, it should be done
in a separated commit.
We only keep this as a way to get GPU_stubs to run, in case we want to do a
throughout cleanup in the codebase and want code using legacy calls to
fail to build.
There is no more point of keep those around. ES20 may need special case
when/if we dabble with it again. Meanwhile no point on polluting the
code with this.
(ghost still has reference for the PROFILE, but that's reasonable)
I wouldn't mind switching fully to Google style, but i am against of
mixing two different styles in same project. So just stick to brace
at the new line after function definition.
There were following issues with ccl_restrict_ptr:
- We already had ccl_restrict for all platforms.
- It was secretly adding `const` qualifier to the declaration,
which is quite weird since non-const pointer can also be
declared as restricted.
- We never in Blender are using foo_ptr or FooPtr type definitions,
so not sure why we should introduce such a thing here.
- It is absolutely wrong from semantic point of view to put pointer
into the restrict macro -- const is a part of type, not part of
hint for compiler that some pointer is never aliased.
Denoise commit introduced kernel_write_result() which saves light passes, so
no need to call both kernel_write_result() and kernel_write_light_passes() from
the split kernel.
Weirdly enough. kernel_write_result() does not take care about debug passes.
The issue is coming from some weird semi-finished canvas feature, which
was remapping coordinate without applying any differential on the sampling
ellipse (in fact, there is no ellipse, sampling think is always a single
pixel).
The whole thing is just weak in the compositor, for now just bring behavior
back to how it was prior to optimization (multithreading) commit.
The problem was that Cycles implicitly uses a transparent surface shader when only
volume nodes are used, but since the black emission shader gets optimized away,
it was no longer detected and therefore no transparent surface was used.
Therefore, the shader now stores whether volume nodes were connected before
optimizing.
In the move to OpenGL 3.3 core profile, we drop support for compatibility profile and older versions.
OpenSubdiv was the only user; I'll update OSD next.
Now we always use GLSL 3.3, AKA #version 330. Most of the extensions we used are built into OpenGL 3.3 so we don't need them anymore.
Cleaned up comments related to GLSL version.
Part of T49012
These are always supported now
- instancing as of GL 3.1
- geometry shaders as of GL 3.2
The change to rna_scene.c could use some cleanup, since we don't really need a runtime query function.
Extremely bright pixels in the rendered image cause the denoising algorithm
to produce extremely noticable artifacts. Therefore, a heuristic is needed
to exclude these pixels from the filtering process.
The new approach calculates the 75% percentile of the 5x5 neighborhood of
each pixel and flags the pixel if it is more than twice as bright.
During the reconstruction process, flagged pixels are skipped. Therefore,
they don't cause any problems for neighboring pixels, and the outlier pixels
themselves are replaced by a prediction of their actual value based on their
feature pass values and the neighboring pixels.
Therefore, the denoiser now also works as a smarter despeckling filter that
uses a more accurate prediction of the pixel instead of a simple average.
This can be used even if denoising isn't wanted by setting the denoising
radius to 1.
The implementation originally handled four different cases:
Regular glossy, glass, metallic fresnel glossy and diffuse.
However, only the first two are actually used currently. Therefore, this commit
removes the other two, which allows to simplify the code.
Additionally, due to the Principled BSDF, the function arguments are now
identical for glossy and glass, which allows to get rid of some ugly #ifdefs.
The "Modern Viewport" was an option in the Viewport panel that would use
the new fancy wire frame code and some depth debugging functionality.
This was introduced as a quick and dirty way to get the new drawing
system showing in the viewport.
Overtime we built a complete Draw Manager system, leaving this original
approach deprecated which adds clutter to Blender code since it would be
removed sooner or later.
Note: The new fancy wireframes and related shaders (white out other
objects) are still in Blender code, though you can't use them anymore.
If we are to have any of those drawing options they should be integrated
in the new draw manager, instead of integrated in the old drawing
pipeline as it was originally coded.
This commit fixes two issues:
- UV/Image editor uvs menu did not match the 3D View's which was changed in rB2b240b043078
- Circle select tool was missing in particle edit mode
Reviewers: Severin
Differential Revision: https://developer.blender.org/D2329
It works with Cycles and Blender Internal for now.
The Blender Internal support will disappear, but we should be able to
integrate it with the Draw manager and use it with Clay, Eevee, ...
Mainly adding 'wire' suffix to wire/distance drawing func and shader.
Also, match wire vertex shader behavior with solid one regarding
head/tail only drawing (i.e. alwas expect head bone mat, never tail one,
and assume that if a radius is negative, then we only draw on the other
end of the bone).
Envelope bones are now pretty much identical to old drawing code.
Note that currently new DwM drawing code does not seem to care about
wire/solid drawing modes at all, guess this is still TODO... For now we
hence just get both wire and solid for envelope bones, this can be
refined later.
Use smarter check of where the file is coming from instead of
attempting to replace same source twice with different settings.
Brings down processing time from 3.6sec to 1.8sec.
This is not complete, it does not implement 3D solid drawing of
envelope bones. 2D wire is hence always drawn for now.
Some notes:
I did not try to implement the 'capsule' approach suggested by @fclem, because:
1. I spent enough time on this already, and finally got something working.
2. I managed to get rid of geometry shader completely.
3. Current approach allows us to use same shader for
distance outline and envelope wire.
It's working fine, except for one glitch - superpositions of envelope
outlines do not work as expected, not sure what's wrong here, tried to
disable zbuff, enable GL_BLEND, no luck so far...
I think we need our own 'background' drawpass to get them working (also
to avoid them drawing over the wire lines).
Was using intermediate functions to store each data type.
This is still done in some places (for more involved access).
For basic data: coords, normals, colors..etc this is unnecessarily indirect.
-Use 11_11_10 buffers for hdr content.
-Eevee compositing share 1 buffer if bloom and DOF are both activated.
-Fix slowdown when resizing EEVEE viewport.
-Removed DRW_BUF_*** enums causing confusion.
We should avoid casting from an int to a flag.
In this case it is probably fine, but if the OB_MODE_SCULPT
was too high it can overflow the bool.
(also adding explicit parenthesis to a flag check)
It is not a good idea to:
1. Duplicate metadata to self
2. Ignore the fact that something might have had metadata already.
Also moved metadata copy to a preparation function, so it is
never lost.
The mesh interpolation code had an edge case where one of two
adjacent edges to a vertex has 0 length. This caused an assert
failure indexing the vertex mesh for splash Blenderman.blend.
Although this is working fine, there are two changes expected in the new
future once depsgraph copy on write is implemented:
1) To call ED_info_stats_clear a callback from depsgraph, instead of the
notifier system. (that would also allow us to clear only one
SceneLayer).
2) To store/get stats from the evaluated SceneLayer, as well as iterate
over the evaluated objects as well.
Found when was looking into T49864. The issue is caused here
by render_copy_renderdata() doing a copy of views with
BLI_duplicatelist() so we can not just zero the pointers out.
Similar thing is happening for layers as well.
Fix T50882: VSE: Blend Modes on Scenes do not layer properly
Fix T51002: Scene strip with Alpha over not working as expected
The byte-to-float conversion was being skipped if the color spaces of the sequence and the scene
are the same, which is the default, resulting in any non-float strips becoming invisible.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D2635
Normally, segments up to 50 can be quite enough for most cases.
However, when dealing with things like braids,
the current limit can sometimes be quite a pain.
Revert 7a18ee62eb and 1ff97bbfff after discussing with @fclem.
VertexBuffer_size should always report the same buffer size, but without asking/calling OpenGL.
Deactivated by default.
All shading attribs can be packed to take less VRAM at the cost of precision (not noticable in this case).
UVs can be packed into I16 but that limits their positions into the [-1, +1] range.
This could be a setting option in the future.
This commit fixes crash, but user feedback can be improved here to
inform artist that one can't use Render Result as a texture since that
will cause feedback loop.
Basically upon invoking cycles baking we could canell it which would
leave G.is_break hanging as true. Since we were not setting is_break to
false before exec baking, it would misbehave.
Follow-up to 0722e41f9e which removed all uses of this struct. Gawain VertexFormat auto aligns attributes as described in @nicholasbishop's performance note.
Only mask are handled by sculpt mode engine and are multiplied on top of the render.
There is room for improvement:
- Shaded meshes don't have correct tangents or uvs.
- Masks are in range 0.8 - 0.2 thus always darkening at least 20% the render.
- It only uses the first material slot of the mesh.
Physics panels are not all back, and the material related panels (e.g.,
hair render panel) should go be changed since there should be no
material for those.
The issue was caused by stupid workaorund for libav. Now things works for
FFmpeg. There might need some tweaks needed for Libav, but that one is
not really priority for support.
The original idea was to use Clay as the default. But since clay may not
even be in the final release, we can't afford to have a startup.blend
that doesn't deliver a reliable result regardless of build options.
Previous method was based on face-area, giving un-even results
based on topology and gave issues with zero area faces.
This method gives matching results for concave ngons and the same geometry triangulated.
Adds a version of template_ID that can be used for non-ID properties.
The property to search for and the collection to search in has to be
passed to it.
Like template_ID it also takes arguments to define a 'new' and 'unlink'
operator. They will be displayed as icon-only buttons then.
Also added a version that can display preview thumbnails.
Had to do some additional changes to make text-buttons support
displaying/modifying empty RNA properties.
This will be needed for workspaces, see D2451.
Reviewed By: campbellbarton
Differential Revision: https://developer.blender.org/D2666
This removes a few options from CMake:
* WITH_LEGACY_OPENGL
* WITH_GL_PROFILE_COMPAT
* WITH_GL_PROFILE_CORE
We still have WITH_GL_PROFILE_ES20. So you can still alternate between
ES20 and CORE profile (when es20 is disabled).
If you want to explicitly see the stubs errors just define
WITH_LEGACY_OPENGL.
Nothing terribly exciting, just a hackish implementation of solid mode CDDM
drawing for cases when we are building with core profile.
The goal is to be able to see SOMETHING with new OpenGL for thew guys who
tests alembic and friends in blender2.8 branch. This is a temporary solution,
the whole drawing will be reworked. Limited to hardcoded material and lighting.
Selection outline color is also hardcoded because of lack of glColor() in the
core profile.
After this commit there is no big stoppers from stop worrying about legacy
OpenGL, so we can go core profile as the only way for Blender.
While some areas will become broken after that, we'd better just fix them
after that and for now just focus on more important design work.
The code only updated nodes in the nodetree of the scene to which the render layer belongs. Therefore, when using scene B in the compositor setup of scene A, A's node wouldn't be updated.
With this fix, the update function loops over all scenes and checks them for relevant nodes.
Was preventing update in 3DView etc. when changing something in the
World's NodeTree, especially annoying in blender2.8 branch (since legacy
depsgraph has been removed there), but also affecting master.
We still need to update the RNA interface to access those. But since
there is no RNA_def_property_float_array_funcs I'm not sure how many
changes this will require.
Old code was working quite unreliable in combination with fast math
flag, especially when compiling with Clang. It seems we were hitting
result of the following bug submitted to Clang [1].
Basically, it was happening so that (int)sqrtf(64) was 7 when Cycles
is built with Clang but was correct 8 when built with GCC.
This commit works this around. Annoying, but don't see other way to
keep sampling pattern the same for Clang and GCC.
[1] https://bugs.llvm.org//show_bug.cgi?id=24063
A bit too early to disable workaround for particles.
Guess it was commented out after particles removal but was not brought back
after particles restore commit.
The idea here is to keep things in a logical order to match the order of ones worflow.
This concept can be seen in Graph > Dope Sheet > NLA. This issue is mainly affecting the manual.
Fixes T50709
Differential Revision: https://developer.blender.org/D2630
The issue was happening when display device is set to None, which makes it so
all the color transformation is a no-op which does not really require any LUT.
This is something we can not know from Blender side easily, because LUT sampling
and related logic is fully done in OCIO library itself. The following happens:
- OCIO sees that no LUT is needed and uses simple pass-through logic in the
color conversion function.
- GLSL compiles sees that uniform used for LUT is unused in the GLSL code and
strips it out.
We can not know this from Blender side because technically any conversion to
the same space might be a no-op and that we wouldn't know without some tricky
parse of the OCIO configuration.
So for now we simply avoid crash but are disabling checks for existence of the
uniform.
Ideally would be nice to have some GLSL-code parses which gets the uniforms
from the code itself, so we can distinguish between typo in the uniform name
and uniform being optimized out.
The goal is to reduce wasted space and improve clarity in the 'N' panel of the VSE through layout changes.
The changes are intentional conservative to avoid making people re-learn anything.
Author: @mpan3
Differential Revision: https://developer.blender.org/D2439
* "Filmic" and "False Color" view transforms added (sRGB display device only).
* "Very Low/Low/Base/High/Very High Contrast" looks added.
* Added filtering so that Filmic only shows look names prefixed with "Filmic - ".
Filmic Dynamic Range LUT configuration created by Troy James Sobotka with
special thanks and feedback from Guillermo, Claudio Rocha, Bassam Kurdali,
Eugenio Pignataro, Henri Hebeisen, Jason Clarke, Haarm-Peter Duiker, Thomas
Mansencal, and Timothy Lottes.
Differential Revision: https://developer.blender.org/D2659
This commit contains the first part of the new Cycles denoising option,
which filters the resulting image using information gathered during rendering
to get rid of noise while preserving visual features as well as possible.
To use the option, enable it in the render layer options. The default settings
fit a wide range of scenes, but the user can tweak individual settings to
control the tradeoff between a noise-free image, image details, and calculation
time.
Note that the denoiser may still change in the future and that some features
are not implemented yet. The most important missing feature is animation
denoising, which uses information from multiple frames at once to produce a
flicker-free and smoother result. These features will be added in the future.
Finally, thanks to all the people who supported this project:
- Google (through the GSoC) and Theory Studios for sponsoring the development
- The authors of the papers I used for implementing the denoiser (more details
on them will be included in the technical docs)
- The other Cycles devs for feedback on the code, especially Sergey for
mentoring the GSoC project and Brecht for the code review!
- And of course the users who helped with testing, reported bugs and things
that could and/or should work better!
Would cause pointer to allocated memory to be overridden.
Steps to recreate were:
* Start Blender
* Save
* Open saved .blend
* Close Blender -> should report unfreed memory
This is routinely mis-used to continuously run scripts,
causing performance problems.
This should be replaced with more specific handlers, or possibly timers.
See: T47811
Caused by rBe87ddda149b. Selecting objects shouldn't be done in such a
low level function anyway (IMHO).
Made sure duplicating objects still works correctly.
This add a new set of (possible) render settings that can be defined at
the scene level and overridable at the scene layer level.
Once we get workspaces we can either add workspace inbetween scene and
scene layer evaluation. Or to replace layer settings, to avoid extra
confusion to users.
An example of this setting is "samples", as implemented now for the clay
engine.
Those shall not be considered while checking whether a to-be-made-local
ID will end up fully local, or still be partially used by linked data...
Even less since we already do have special handling of proxies later.
Fixes main remaining issue found with 04_01_H.lighting.blend Agent327
file, and allows us to switch back to optimized post-processing in
make_local code.
That one tags those ugly little 'from' ID pointers (shape keys and
proxies), which point back from used to user ID, and require a lot of
special care in data-block management...
Again, Agent327's 04_01_H.lighting.blend shows some problem here, it
triggers several times the 'not used at all' assert in step 5 of secure
code, and with optimized version we lose the connection between
rigs and the main characters!
Will keep investigating on this, but for now let's try to give something
working to the studio.
This should not be needed imho, we already set POSE_RECALC flag
correctly there, but it still is missing actual update of poses in some
(complex and convoluted) cases. So at least for now, let's go with this
hack, it's not really harming anyone anyway.
Fixes crash in Agent327's 04_01_H.lighting.blend when making all local.
Not sure how this happens, but in some cases we can evaluate
deformations of an armature which pose is not valid, at least put a
warning here to help identifying the issue quickly.
Adds 2 static unsigned int to Gawain and GPUTexture to roughly keep track of the memory stats on the GPU.
Stats can be view at the bottom of the GPU stats with debug value > 20.
This join them with \n separators.(right now they clash on top of each
other).
Note: We still need to change "ED_region_info_draw" to accept multi-line strings.
Disabled by default. Set ENABLE_EFFECT_MOTION_BLUR to 1 to enable.
No fancy motion blur. Use depth and camera matrix to get the motion vectors. Then blur in this direction.
Only available in camera view.
Only Camera animation is supported, does not take into account the parents motion
This can be used by any engine actually. For example to give feedback
regarding probe caching in Eevee, ....
Unline master, we try to conciliate both the original viewport info
(FPS, view name) and the render info.
Note: I had to split the function in two because the camera alpha
passepartout is handled inside the view3d_draw_region_info function.
Review: Campbell Barton and Clement Foucault.
Thanks to Sergey Sharybin and Pablo Vazquez for some of the design
discussions.
The issue was caused by unlimited textures commit, root of the issue is that
displacement code updates some of the image slots directly, so it needs to
ensure device vectors are all proper size.
This fix doesn't really fix the situaion now because at the moment we
only flush the evaluated flags at DEG_OBJECT_ITER time, which is silly.
But it's all we can do until we have CoW.
So the parent visibility flag is not flushed. And even if we flushed its
visibility flag, we would need to rely on having the parent flag flushed
before we get to the object in question.
Anyways, take this commit as a future fix for when depsgraph comes to
rescue all our souls.
This was set to maxcpu which in an 8 core box would be 8, each project would then spawn
8 instances of cl.exe, making a possible of 64 simultaneously running compiler instances
slowing the compile down instead of speeding it up.
This implements weight rendering with the draw manager, with all drawing
options (Shading, wire, face masking, vertex masking).
This is part of T51208
Reviewers: campbellbarton
Subscribers: dfelinto
Differential Revision: https://developer.blender.org/D2654
Note: We should test if the armature is the "armature modifier" parent of the
object.
If any good samaritan feels like tackling this, the function to change
is: DRW_pose_mode_armature
We unfortunately cannot fix this for previous versions of Blender, but
at least the issue (Blender crashing on unknown IDProp types) should now
be addressed for future.
Simply reset unknown IDProp types to integer one, and reset its value to zero.
Previously, every RenderPass would have a bitfield that specified its type. That limits the number of passes to 32, which was reached a while ago.
However, most of the code already supported arbitrary RenderPasses since they were also used to store Multilayer EXR images.
Therefore, this commit completely removes the passflag from RenderPass and changes all code to use the unique pass name for identification.
Since Blender Internal relies on hardcoded passes and to preserve compatibility, 32 pass names are reserved for the old hardcoded passes.
To support these arbitrary passes, the Render Result compositor node now adds dynamic sockets. For compatibility, the old hardcoded sockets are always stored and just hidden when the corresponding pass isn't available.
To use these changes, the Render Engine API now includes a function that allows render engines to add arbitrary passes to the render result. To be able to add options for these passes, addons can now add their own properties to SceneRenderLayers.
To keep the compositor input node updated, render engine plugins have to implement a callback that registers all the passes that will be generated.
From a user perspective, nothing should change with this commit.
Differential Revision: https://developer.blender.org/D2443
Differential Revision: https://developer.blender.org/D2444
Reduce thread divergence in kernel_shader_eval.
Rays are sorted in blocks of 2048 according to shader->id.
On R9 290 Classroom is ~30% faster, and Pabellon Barcelone is ~8% faster.
No sorting for CUDA split kernel.
Reviewers: sergey, maiself
Reviewed By: maiself
Differential Revision: https://developer.blender.org/D2598
Since offscreen drawing now uses draw engine, this doesn't need to be
considered legacy.
Note that there are some calls into view3d_draw_legacy.c from view3d_draw.c
this is generally not accepted, so its only dont where there are checks
for new/old engine.
Functions exposed to do this use a VP_deprecated prefix.
We still need to bring pose bone drawing for the pose engine.
Note: In the future, once we get face-widget selection, we will need to
make sure the pose engien only draws then if "weight paint mode" is not
enabled too.
In the future this will allow us to implement things like face manipulators.
For now however, this is getting on the way of using pose mode
(mesh objects are drawn green).
NOTE: pose engine should also be the one taking care of pose bones
(and for that matter, X-Ray armature option should/could be a property
of the Pose engine even).
Previously the logic was different for duplis and regular objects: regular objects
were using render visibility when Render Layer option is enabled which duplis were
always using viewport visibility when rendering from the viewport.
This was quite confusing because caused different results in viewport and render
when artists were expecting them to match 1:1.
8b2640f08 introduced the error,
caused by external-engine needing depth mask to be disabled,
which happened to be the case before changing state changing logic.
With this we also do not need IDP_MergeGroupValues anymore.
If this causes problems in the future we can always make recursion an
option (like overwrite is).
Most of this was copying/adapting code from octahedral existing draw
code, tough part was to get BBone matrices to behave as expected, they
are using a fairly specific setup...
Addresses T51365 and T51362.
Changing states didn't properly reset between shading groups
causing the GL state to be wrong based on draw order.
States are now only set when changed.
This implements branched path tracing for the split kernel.
General approach is to store the ray state at a branch point, trace the
branched ray as normal, then restore the state as necessary before iterating
to the next part of the path. A state machine is used to advance the indirect
loop state, which avoids the need to add any new kernels. Each iteration the
state machine recreates as much state as possible from the stored ray to keep
overall storage down.
Its kind of hard to keep all the different integration loops in sync, so this
needs lots of testing to make sure everything is working correctly. We should
probably start trying to deduplicate the integration loops more now.
Nonbranched BMW is ~2% slower, while classroom is ~2% faster, other scenes
could use more testing still.
Reviewers: sergey, nirved
Reviewed By: nirved
Subscribers: Blendify, bliblubli
Differential Revision: https://developer.blender.org/D2611
Negative scale on camera is a nice trick to invert render image on one
axis at no extra CPU cost. It was implemented in the Decklink branch but
I introduced a typo when porting it to master. It is now fixed.
The change was initially needed for Blender 2.8 branch but the actual
function was reverted in there. So no reason to keep dead unused
placeholder in the dependency graph.
This reverts commit fd69ba2255.
We can now use object and other modes on top of Cycles.
Since we are now always on "render_to_view" (old Rendered mode), the
pause button is always visible.
Note: In one case (rna_Scene_editmesh_select_mode_set) I had to resort
to the temporary solution of BKE_scene_layer_context_active.
This is fine for now, but it will be a problem once workspaces are
introduced. In this case the fix is to use a temporary value in the set
routine, and actually only set the active object in the update function
(which can take bContext).
- code_generate_fragment : Making sure that shaders uses the new uniforms.
- code_generate_vertex_new : create require attribute based on their names and not their id (see draw_cache_impl_mesh.c).
- add support for ramp textures.
This is a bad implementation waiting for some improvement that are :
- Gawain support for multiple names. -> will remove duplication of attribute alloc
- Glsl safe name string escaping. -> will remove any risk of hash colision.
Old glsl drawing was relying on attribute being choosen by DerivedMesh at drawing time.
For this reason, we declare all possible attrib "name" for each data layer inside the batches.
In the glsl code, we declare required data by type and name.
Then Gawain only bind vbos if names correspond.
This is way cleaner as we do not need to access the CD itself when drawing.
One other problem is that the hash maybe 11 caracters long and rapidly overload gawain's attrib name buffer.
- Unify GPUMaterial creation (world/mesh).
- Support for multiple shader variations (not used for now).
- Convert GPUInputs to DRWUniforms to be used with the draw manager.
- Nodetree Update is not supported. The only way to refresh the shaders is to change render engine.
- Cleanup in GPUPass.
- Add new temporary Node Compatibility type. Compatibility types should be removed in the future.
Note that smooth-shaded outlines of muted NLA strips should also be
dashed, not sure how to do that currently (in a simple way), so kept
them solid for now.
Using geometry shader allows us to get rid of the 'line origin' extra
vertex attribute, which means dashed shader no longer requires fiddling
with those vertex attributes definition, and, most importantly, does not
require anymore special drawing code!
As you can see, this makes code much simpler, and much less verbose,
especially in complex cases.
In addition, changed how dashes are handled, to have two 'modes', a
simple one with single color (using default "color" uniform name), and a
more advanced one allowing more complex and multi-color patterns.
Note that since GLSL 1.2 does not support geometry shaders, a hack was
added for now (which gives solid lines, but at least does not make
Blender crash).
There was a bug in older GLEW version which kept glGenVertexArrays as NULL
when 3.2 core context is requested [1].
Didn't find a way to check GLEW version from CMake, it seems the version
is not really exposed to the header.
[1] https://sourceforge.net/p/glew/bugs/124/
Needed a bit of twisting in generic private func behind the imm_draw_circle
helpers, but think it's fine.
Note that this demonstrate yet another downside of new dashed shader
compared to olde line style: not only does it needs more complex setup,
and can only work with PRIM_LINES type of primitives, but it also
behaves totally wrong with chained short segments!
We really need to find a better way to do this effect at some point. :(
Avoid calculating a new split-index when re-fitting.
While checking if a knot can be removed, the index with the highest error
can be used as a candidate to replace the knot
(in the case it can't be removed).
Also restore drawing of local action markers in NLA strips...
Note: there is another usage of setlinestyle in this file, not sure how
to hanlde it for now, so will leave it for later...
Not sure if this is a proper fix, but was getting frequent crashes, so
committing this real quick just to make master sable again. Can be
reverted later if there's a better fix. The changes to images really
need a closer look...
Hi Guys,
as one of my clients needs the possibility to have custom menu entries in the general right click menu (all over Blender: in the node editor, properties, toolbars,..) I talked with Campbell about expanding our hard coded menu a bit. This is the outcome. As I only need those two, I support currently a button_prop and a button_pointer.
{F540397}
I tested the changes with a custom script where I added a custom entry and executed an operator on click - it seems to work exactly how it's intended to. The script: {F540435}
As I'm not too experienced in rna stuff I would really appreciate any review.
Thanks very much Campbell for his open ears & help on this issue!
Reviewers: campbellbarton, mont29
Reviewed By: campbellbarton, mont29
Subscribers: sybren, mont29
Tags: #addons
Differential Revision: https://developer.blender.org/D2612
This upgrade the drawing code to use latest opengl calls.
Also, it adds a fallback shader for opencolorio.
Reviewers: sergey, brecht
Subscribers: merwin, fclem
Differential Revision: https://developer.blender.org/D2652
This one has been painful, it's doing lots of nifty conditional drawing
of all kind... Also, very nice illustration of how cumbersome it is to
use dashed shader currently, we really need to find a simpler way at
some point imho. But this is not critical issue either, for now we can
live with some uglyness like that. :/
Previous fix did not work for mixed textures. This one will over-allocate
information array, but it's better than not being able to render at all.
Some more cleanup and improvement is coming.
- enabled lights
- alpha test
- texture environment
- point sprites (always enabled in modern GL)
Moved is_clip_plane for better struct packing, no functional change there.
Part of T51164
These parts will not be part of final viewport, but are called indirectly during the transition. To avoid runtime errors on core profile, exit early -- functions effectively do nothing.
I put the early exits inside the functions to avoid cluttering the code that calls these. But (long term) the calling functions need to change.
Basic shader's detect_options function was unused and full of old, so I deleted it.
Part of T51164
GL_LOGIC_OP is a synonym for GL_INDEX_LOGIC_OP, from OpenGL 1.0's indexed color drawing. It's not part of modern GL and causes an error in core profile.
GL_COLOR_LOGIC_OP is still supported in the latest GL. It's the last stage before writing to the framebuffer.
Part of T51164
There are multiple ways to orphan a buffer resource, and this code picks the oldest/safest way no matter which OpenGL version we're running.
TODO: use other (more recent) methods after thorough testing
Follow up to 8e0c57a812
Texturing is always enabled in GLSL. Simply use a sampler in the shader.
Replaced gpu_generate_mipmap with glGenerateMipmap since the former just Enabled/Disabled the texture target and called the latter.
Part of T51164
Note that I also made 'dash anchor point' consistent (the static one,
not the mouse one), in previous code somtimes dashed were anchored to
the static center point, in others, to the moving mouse position, the
later was rather disturbing imho...
This patch allows for an unlimited number of textures in Cycles where the hardware allows. It replaces a number static arrays with dynamic arrays and changes the way the flat_slot indices are calculated. Eventually, I'd like to get to a point where there are only flat slots left and textures off all kinds are stored in a single array.
Note that the arrays in DeviceScene are changed from containing device_vector<T> objects to device_vector<T>* pointers. Ideally, I'd like to store objects, but dynamic resizing of a std:vector in pre-C++11 calls the copy constructor, which for a good reason is not implemented for device_vector. Once we require C++11 for Cycles builds, we can implement a move constructor for device_vector and store objects again.
The limits for CUDA Fermi hardware still apply.
Reviewers: tod_baudais, InsigMathK, dingto, #cycles
Reviewed By: dingto, #cycles
Subscribers: dingto, smellslikedonkey
Differential Revision: https://developer.blender.org/D2650
Previously canceling a render done by the split kernel could cause artifacts
such as very bright or dark tiles. This was caused by unfinished samples
being included in the output buffer. To avoid this we now wait till all the
currently rendering samples have finished, up to a limit of twice the
expected time for them to finish (currently this is no more than 20 seconds,
but usually its much less). If samples still haven't finished by then we
stop anyways in case there's an endless loop occurring.
The scale version was working(ish), but it was not really extendable to
a 3D line version of the shader.
Also note that sequencer view still keeps its 'UI scale' adaptation
(dashes grow together with UI scale setting). Would be nice to do that
everywhere ultimately imho, but nothing urgent here.
Previously canceling a render done by the split kernel could cause artifacts
such as very bright or dark tiles. This was caused by unfinished samples
being included in the output buffer. To avoid this we now wait till all the
currently rendering samples have finished, up to a limit of twice the
expected time for them to finish (currently this is no more than 20 seconds,
but usually its much less). If samples still haven't finished by then we
stop anyways in case there's an endless loop occurring.
It was totally unclear whether the device is enabled or disabled.
Lots of people got fully lost in the current interface.
While the solution is not fully ideal, it is at least solves
ambiguity in the interface.
Simple child hairs don't have a face index number assigned, so the
call to dm->getTessFaceData(dm, num, CD_MFACE) would cause a crash. To
work around this, UV and normal vectors are copied from the parent
hair.
I've also removed an unnecessary call to dm->getTessFaceArray(dm);
Reviewers: kevindietrich
Differential Revision: https://developer.blender.org/D2638
Not much to add here, except that it needs the scale of its MVP matrix
to be taken into account here...
Reviewers: merwin, dfelinto
Differential Revision: https://developer.blender.org/D2647
The function doesn't return whether the object is a shape at all, since
it also returns true for camera objects (and soon also for empties). It
returns true when objects of this type can be exported to Alembic at all.
This is now reflected in the name.
Use new 2D dashed line shader in 3DView camera view.
Note that this also involved converting UI_draw_safe_areas() to this
dashed shader, which means it cannot be used anymore with other shaders.
Part of D2647.
Even though in some cases this does not seems useful, in others (like
zommed 2D views) we have to correct MVP matrix scaling to get fixed
dashes size.
Note that we could do that differently (commented about it in shader),
would also have been cleaner to extract that MVP scale from within the
shader, but there does not seem to be a way to initialize uniform values
from within a shader, and would rather avoid recomputing the scale for
every run (especially since 3DViewport does not need it e.g.).
Part of D2647.
That still not ideal, but so be it. IS_VIEWPORT_LEGACY should be
removed anyways, this is code from pre-draw manager and the
"modern viewport" as well as legacy will go away entirely once the draw
manager is feature complete.
Selection code needs to iterate over DEG_OBJECT_ITER otherwise we won't
get modifiers, dupli objects, ...
Also make selection respect selectability flag.
Review by: Sergey Sharybin
- use DRW_engine prefix for engine types.
- use engine suffix for files that define a draw engine.
- remove engines from include path (they're only referenced once)
- Use uniform color world for the world probe.
- Refactored the Fresnel expression to be better with Area Lights.
- Squared the roughness for default materials.
This works around a long outstanding issue T50176 with cycles on msvc2015/x86 . root cause is still unknown though,feels like a game of whack'a'mole
Reviewers: sergey, dingto
Subscribers: Blendify
Tags: #cycles
Differential Revision: https://developer.blender.org/D2573
Needed to remove old draw code entirely.
Object mode selection support, pose and armature still need to be added.
Enabled when 'use_modern_viewport' is set.
Using -cl-fast-relaxed-math assumes no NaN/Inf values in any expression.
This causes problems on overflow, division by zero, square root of negative number.
Comparisons with NaN or infinite value are affected as well.
This patch causes <2% slowdown on benchmark scenes.
Fix T50985: Rendering volume scatter with GPU OpenCL comes to an halt after a few seconds
- Added New Batch cache routine to get the split mesh for each material. (optimization needed)
- Did some shader modifications to create default_frag and support a somwhat correct fresnel for lamps (to finish)
Otherwise if we create a new collection, and drag a collection into it we either
get a crash, or at the very least we dont get the visibility flags correct.
This is part of the new draw manager design. Any engine (even clay, eevee, ...) should be able to draw to the viewport, as well as render to an image directly.
Changing the API names to conform to that.
Now we just iterate over all objects. It may be a better idea to use a
smarter approach for performance. However, since this function is not
called regularly (only when changing Alembic file paths), it's not a very
big deal.
Was performing context access per object for scene, region etc.
While not very slow they don't change during drawing so cache in a struct.
This also makes it possible to draw in views besides the current context.
HDF5 Alembic files are not officially supported by Blender. With this
commit, the HDF5 format is detected even when Blender is compiled without
HDF5 support, and the user is given an explanatory error message (rather
than the generic "Could not open Alembic archive for reading".
The final goal to reach is to make vectorized types much easier to maintain
and the previous design had following issues:
- Having all types and methods implementation made the source file rather
bloated and unfun to navigate in.
- It was not possible to quickly glance available API for the type you are
interested in.
- Adding more vectorization types will bloat the file even more, making
things even more tricky to follow.
Compatibility profile was working fine, this is mostly to get the highest GL core profile version available.
Our minimum requirement is 3.3 core profile. When we request a specific GL version:
- AMD and Intel give us exactly this version
- NVIDIA gives at least this version <-- desired behavior
so we ask for 4.5, 4.4 ... 3.3 in descending order to get the best version on the user's system.
Accept OpenGL 3.0 on Mesa instead of 3.3+ compatibility profile. (requested by @LazyDodo) This will be removed after we finish moving to core profile.
Part of T49012 and T51164
Blender subsystems that care about OpenGL use GL_DEFINITIONS, which now includes the newest (temporary) WITH_LEGACY_OPENGL.
Also updated Gawain's CMake to use this instead of its own logic.
Fixes performance issues of C++ one with Windows MSVC debug builds...
Merely a translation from msgfmt.cc code by @sergey, using BLI libs intead of C++'s stdlib.
Reviewers: sergey, campbellbarton, LazyDodo
Subscribers: sergey
Differential Revision: https://developer.blender.org/D2605
QtKit was removed in macOS Sierra, this patch disables WITH_CODEC_QUICKTIME
in Sierra and greater versions of macOS.
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D2645
MX (Multiple conteXt) support was dropped from the GLEW 2.0 library to make core profile support cleaner.
Our WITH_GLEW_MX build option was OFF by default already; this commit removes the inactive code paths.
I'm working on a plan for multiple GPUs, contexts, resource sharing, etc. This commit gives us a cleaner starting point for that upcoming work.
Tested on Mac, will test on Linux & Windows immediately after pushing.
This was blocking core context setup on Mac, since accumulation buffers are obsolete. With this (and appropriate CMake options) I now get
renderer: 'Intel HD Graphics 4000 OpenGL Engine'
vendor: 'Intel Inc.'
version: '4.1 INTEL-10.24.45'
in system-info.txt intead of
version: '2.1 INTEL-10.24.45'
By mistake, the code relied on ALEMBIC_ROOT_DIR being defined by the user
running the tests. Now CMake macros are used to correctly find the Alembic
root directory.
Matrix.decompose() should either return "location, orientation, size" or
"translation, rotation, scale". Since there are constructors for the former,
I've replaced "location" in the documentation with "translation".
The code is still the same, I just changed the documentation.
The original code was failing because the base to object flushing was
only happening as part of the depsgraph. However we can use the
evaluated values to set the initial values of the base.
In this particular case, we couldn't set the new object visible because
its selectability flag was not set yet.
We were not updating depsgraph when collection just "became" selectable,
but although this doesn't change object selection, it change its
selectability.
The idea is to have osme geenric BSDF node which is subclassed by
"regular" BSDF nodes and uber shaders.
This way we can access special type and closure type for making
decisions somewhere else.
No longer passing time as float and constructing ISampleSelectors all
over the place. Instead, just construct an ISampleSelector once and
pass it along.
It is disabled by default, so should not affect existing configurations.
Main benefits of this goes as:
- Linux distros can use that to avoid libraries duplication and link
blender package against gflags package from the system.
- It it easier to test whether Blender works with updated version of
Gflags prior to re-bundling the library.
Pending: Include the set in the rna depsgraph.objects
In fact it would be nice to unify them both. However this will likely
change once Depsgraph incorporate this iterator, so I'm not sure we
should bother with that.
Related to T51203
This switches the internal color representation of the eye dropper from display space to linear. Any time a linear color is requested and the color is picked from a linear object, the result is now precise to the bit as the color gets patched through directly. Color space conversion now only happens when a color is picked from non-linear display space objects or when the color is requested to be returned in non-linear space.
In addition, this patch changes the DifferenceMatte node to interpret a tolerance of 0.0 to accept colors that are identical bit by bit, as apposed to simply refusing all colors.
Replaced some STREQ(snode->tree_idname, ...) calls with ED_node_is_*() calls for improved readability, fixed one case where the STREQ was used the wrong way
Iterator to return the objects evaluated from depsgraph. The
iterator itself will change once CoW is implemented.
This reproduces the behaviour of DEG_OBJECT_ITER but for Python.
Even though this will have to change once we get workspaces, we will
still have a depsgraph for the Scene.
This is required for the upcoming depsgraph.objects RNA iterator.
Example, imagine an object Cube in collections 1 and 2 where both
collections are nested to A. Now we set a "color" property as follow:
```
Scene -> GREEN
--
A -> RED
↳ 1 -> BLUE
↳ 2 -> -
```
In this case the object will be RED, because of A↳ 2.
Now if we have:
```
Scene -> GREEN
--
A -> RED
↳ 1 -> -
↳ 2 -> PINK
1 -> -
--
The object will be PINK because of A↳ 2.
Note that the (top level) collection 1 doesn't influence the object color
because there are no overrides on it. The scene render settings (GREEN
in this case) are only used as fallback if an override is not set at
all.
That's a quick hack to address that specific case, new pointer IDProp
actually enlights a generic problem - datablocks using themselves - which
is not really handled by current code, would consider this not-so-urgent
TODO though.
Immediate mode built-in shader is used for this drawing already.
So there is no reason to try building basic shader which is not
supported in the core profile.
This replaces all direct usage of:
- FIRSTBASE
- LASTBASE
- BASACT
- OBACT
Some usages still remain in legacy utility functions which are called
all over the place.
With the explicit calls we don't need to worry about current state
outside of the GPU module now. In fact. we don't need to worry about
current matrix mode in core profile at all.
Legacy OpenGL now has some code which ensures current matrix mode
when using explicit calls to push/pop matrix.
There is no need to break assumptions of what's being modified
by this call and what's restored. The changes in this function
simply spread crappyness outside of the UI block.
There are two major things in this commit.
First one is to have proper stack for projection matrices. This is
something what OpenGL specification grants to have at least 2 elements
for and what is required to have for proper editor drawing without
refactoring the way how we restore projection matrix.
Supporting this stack have following advantages:
- Our GPU stack is closer to OpenGL specs, making it easier to follow
by other developers who are always familiar with OpenGL.
- Makes it easier to port all editors to a new API.
- Should help us getting rid of extra matrix push/pop added in
various commits to 2.8 branch.
The new API follows the following convention:
- gpuPushMatrix/gpuPopMatrix ALWAYS deals with model view matrix
and nothing more.
While this name does not fully indicate that it's only model view
matrix operator, it matches behavior of other matrix operations
such as transform which also doesn't indicate what matrix type
they are operating on.
- Projection matrix has dedicated calls for push/pop which are
gpuPushProjectionMatrix/gpuPopProjectionMatrix.
There is only two places which are using texture matrix mode:
- Tiled tface support.
- Texture shading mode for texture mapping.
Both cases are subject for reconsideration: it is likely that we'll be getting
rid of tface, which means game properties like tiles needs to be revisited
anyway. As for texture shading it is using basic shader which is also not
supported by core profile anyway.
The ABC_export and ABC_import functions both take a as_background_job
parameter, and return a boolean.
When as_background_job=true, returns false immediately after scheduling
a background job. This was the old behaviour of this function, which makes
it very hard for scripts to do something with the data after the import
or export completes.
When as_background_job=false, performs the export synchronously, and
returns true when the export was ok, and false if there were any errors.
This allows further processing.
The Scene.alembic_export() function is deprecated, and will be removed from
Blender 2.8 in favour of calling the bpy.ops.wm.alembic_export() operator.
As such, it has been hard-coded to the old background job behaviour.
It is required for our new matrix stack to only have projection and model view
matrix modes. This commit ports paint cursor overlay to model view matrix.
The export is still slower than needed, as the particle systems themselves
aren't disabled during the export. It's only the writing to the Alembic
file that's skipped.
We could not edit them, but still could delete them, which makes no
sense, API-defined properties are similar to class members, removing
them from single instances is pure garbage. And it was broken anyway.
Found by @a.romanov while checking on T51198, thanks.
This is an OpenGL 3.3 feature, currently being used for surface normals. It's 1/3 the size of floating point triples and perfect for normals or colors.
See T49012 for why Windows can use this feature regardless of core vs compatibility.
This combines the Mesa fix (16e929e6ff by @brita_) with the Mac fix (89e23c743e by @merwin).
And uses the same fix for another call to glFramebufferTexture introduced in f1fb605ec9.
The issue was going to the fact that GL_ALPHA was deprecated in core profile
and common solution online is to use GL_RED instead. That is what is done in
this commit.
Curve resolution isn't natively supported by Alembic, hence it is stored
in a user property "blender:resolution". I've looked at a Maya curves
example file, but that also didn't contain any information about curve
resolution.
Differential Revision: https://developer.blender.org/D2634
Reviewers: kevindietrich
Those tokens are still valid. Although we should not being using them,
they should not prevent us from building with core profile.
In the future we can have a new flag WITH_GL_CORE_STRICT for those.
But only after we build with core profile by default
Object Info node can be useful to give some variation to a single material assigned to multiple instances. This patch adds support for Viewport and BI.
{F499530}
Example: {F499528}
Reviewers: merwin, brecht, dfelinto
Reviewed By: brecht
Subscribers: duarteframos, fclem, homyachetser, Evgeny_Rodygin, AlexKowel, yurikovelenov
Differential Revision: https://developer.blender.org/D2425
This replaces access to the first scene layer (which was wrong) with a linear
lookup of scene layer by it's active index.
This seems to be a better temporary solution to have things working for until
we've got proper workspace, depsgraph and it's per-layer storage in place.
The order number written to Alembic is the same as we use in memory, so
the +1 wasn't needed, at least according to the reference Maya exporter
maya/AbcExport/MayaNurbsCurveWriter.cpp, function
MayaNurbsCurveWriter::write(), in the Alembic source code.
Furthermore, when writing an array of nurb orders, the curve type should
be set to kVariableOrder, otherwise the importer will ignore it.
For now only compute GGX convolution. The GGX LUT used for the split sum approximation (UE4) is merged with the LTX mag LUT that uses the same parameters (theta and roughness)
First: we settle on core profile 3.3, not 4.1. Second, GL_CURRENT_PROGRAM still exists in 4.1 and 4.2. But it is recommented to use GL_ACTIVE_PROGRAM instead.
Following the logic of gl-deprecated.h we should define it to USE_GL_ACTIVE_PROGRAM instead. That said, removing it for now.
commit 90778901c9
Merge: 76eebd93bf0026
Author: Schoen <schoepas@deher1m1598.emea.adsint.biz>
Date: Mon Apr 3 07:52:05 2017 +0200
Merge branch 'master' into cycles_disney_brdf
commit 76eebd9379
Author: Schoen <schoepas@deher1m1598.emea.adsint.biz>
Date: Thu Mar 30 15:34:20 2017 +0200
Updated copyright for the new files.
commit 013f4a152a
Author: Schoen <schoepas@deher1m1598.emea.adsint.biz>
Date: Thu Mar 30 15:32:55 2017 +0200
Switched from multiplication of base and subsurface color to blending
between them using the subsurface parameter.
commit 482ec5d1f2
Author: Schoen <schoepas@deher1m1598.emea.adsint.biz>
Date: Mon Mar 13 15:47:12 2017 +0100
Fixed a bug that caused an additional white diffuse closure call when using
path tracing.
commit 26e906d162
Merge: 0593b8c223aff9
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Feb 6 11:32:31 2017 +0100
Merge branch 'master' into cycles_disney_brdf
commit 0593b8c51b
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Feb 6 11:30:36 2017 +0100
Fixed the broken GLSL shader and implemented the Disney BRDF in the
real-time view port.
commit 8c7e11423b
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Fri Feb 3 14:24:05 2017 +0100
Fix to comply strict compiler flags and some code cleanup
commit 17724e9d2d
Merge: 379ba34520afa2
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Jan 24 09:59:58 2017 +0100
Merge branch 'master' into cycles_disney_brdf
commit 379ba346b0
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Jan 24 09:28:56 2017 +0100
Renamed the Disney BSDF to Principled BSDF.
commit f80dcb4f34
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Fri Dec 2 13:55:12 2016 +0100
Removed reflection call when roughness is low because of artifacts.
commit 732db8a57f
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed Nov 16 09:22:25 2016 +0100
Indication if to use fresnel is now handled via the type of the BSDF.
commit 0103659f5e
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Fri Nov 11 13:04:11 2016 +0100
Fixed an error in the clearcoat where it appeared too bright for default
light sources (like directional lights)
commit 0aa68f5335
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Nov 7 12:04:38 2016 +0100
Resolved inconsistencies in using tabs and spaces
commit f5897a9494
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Nov 7 08:13:41 2016 +0100
Improved the clearcoat part by using GTR1 instead of GTR2
commit 3dfc240e61
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Oct 31 11:31:36 2016 +0100
Use reflection BSDF for glossy reflections when roughness is 0.0 to
reduce computational expense and some code cleanup
Code cleanup includes:
- Code style cleanup and removed unused code
- Consolidated code in the bsdf_microfacet_multi_impl.h to reduce
some computational expense
commit a2dd0c5faf
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed Oct 26 08:51:10 2016 +0200
Fixed glossy reflections and refractions for low roughness values and
cleaned up the code.
For low roughness values, the reflections had some strange behavior.
commit 9817375912
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Oct 25 12:37:40 2016 +0200
Removed default values in setup functions and added extra functions for
GGX with fresnel.
commit bbc5d9d452
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Oct 25 11:09:36 2016 +0200
Switched from uniform to cosine hemisphere sampling for the diffuse and
the sheen part.
commit d52d8f2813
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Oct 24 16:17:13 2016 +0200
Removed the color parameters from the diffuse and sheen shader and use
them as closure weights instead.
commit 8f3d927385
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Oct 24 09:57:06 2016 +0200
Fixed the issue with artifacts when using anisotropy without linking the
tangent input to a tangent node.
commit d93f680db9
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Oct 24 09:14:51 2016 +0200
Added subsurface radius parameter to control the per color channel
effection radius of the subsurface scattering.
commit c708c3e53b
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Oct 24 08:14:10 2016 +0200
Rearranged the inputs of the shader.
commit dfbfff9c38
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Fri Oct 21 09:27:05 2016 +0200
Put spaces in the parameter names of the shader node
commit e5a748ced1
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Fri Oct 21 08:51:20 2016 +0200
Removed code that isn't in use anymore
commit 75992bebc1
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Fri Oct 21 08:50:07 2016 +0200
Code style cleanup
commit 4dfcf455f7
Merge: 243a0e32cd6a89
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Thu Oct 20 10:41:50 2016 +0200
Merge branch 'master' into cycles_disney_brdf
commit 243a0e3eb8
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Thu Oct 20 10:01:45 2016 +0200
Switching between OSL and SVM is more consistant now when using Disney
BSDF.
There were some minor differences in the OSL implementation, e.g. the
refraction roughness was missing.
commit 2a5ac50922
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Sep 27 09:17:57 2016 +0200
Fixed a bug that caused transparency to be always white when using OSL and
selecting GGX as distribution of the Disney BSDF
commit e1fa862391
Merge: d0530a87f76f6f
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Sep 27 08:59:32 2016 +0200
Merge branch 'master' into cycles_disney_brdf
commit d0530a8af0
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Sep 27 08:53:18 2016 +0200
Cleanup the Disney BSDF implementation and removing unneeded files.
commit 3f4fc826bd
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Sep 27 08:36:07 2016 +0200
Unified the OSL implementation of the Disney clearcoat as a simple
microfacet shader like it was previously done in SVM
commit 4d3a0032ec
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Sep 26 12:35:36 2016 +0200
Enhanced performance for Disney materials without subsurface scattering
commit 3cd5eb56cf
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Fri Sep 16 08:47:56 2016 +0200
Fixed a bug in the Disney BSDF that caused specular reflections to be too
bright and diffuse is now reacting to the roughness again
- A normalization for the fresnel was missing which caused the specular
reflections to become too bright for the single-scatter GGX
- The roughness value for the diffuse BSSRDF part has always been
overwritten and thus always 0
- Also the performance for refractive materials with roughness=0.0 has
been improved
commit 7cb37d7119
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Thu Sep 8 12:24:43 2016 +0200
Added selection field to the Disney BSDF node for switching between
"Multiscatter GGX" and "GGX"
In the "GGX" mode there is an additional parameter for changing the
refraction roughness for materials with smooth surfaces and rough interns
(e.g. honey). With the "Multiscatter GGX" this effect can't be produced at
the moment and so here will be no separation of the two roughness values.
commit cdd29d06bb
Merge: 02c315ab40d1c1
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Sep 6 15:59:05 2016 +0200
Merge branch 'master' into cycles_disney_brdf
commit 02c315aeb0
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Sep 6 15:16:09 2016 +0200
Implemented the OSL part of the Disney shader
commit 5f880293ae
Merge: 630b80eb399a6d
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Fri Sep 2 10:53:36 2016 +0200
Merge branch 'master' into cycles_disney_brdf
commit 630b80e08b
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Fri Sep 2 10:52:13 2016 +0200
Fresnel in the microfacet multiscatter implementation improved
commit 0d9f4d7acb
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Fri Aug 26 11:11:05 2016 +0200
Fixed refraction roughness problem (refractions were always 100% rough)
and set IOR of clearcoat to 1.5
commit 9eed34c7d9
Merge: ef29aaeae475e3
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Aug 16 15:22:32 2016 +0200
Merge branch 'master' into cycles_disney_brdf
commit ef29aaee1a
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Aug 16 15:17:12 2016 +0200
Implemented the fresnel in the multi-scatter GGX for the Disney BSDF
- The specular/metallic part uses the multi-scatter GGX
- The fresnel of the metallic part is controlled by the specular value
- The color of the reflection part when using transparency can be
controlled by the specularTint value
commit 88567af085
Merge: cc267e5285e082
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed Aug 3 15:05:09 2016 +0200
Merge branch 'master' into cycles_disney_brdf
commit cc267e52f2
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed Aug 3 15:00:25 2016 +0200
Implemented the Disney clearcoat as a variation of the microfacet bsdf,
removed the transparency roughness again and added an input for
anisotropic rotations
commit 81f6c06b1f
Merge: ece5a087065022
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed Aug 3 11:42:02 2016 +0200
Merge branch 'master' into cycles_disney_brdf
commit ece5a08e0d
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Jul 26 16:29:21 2016 +0200
Base color now applied again to the refraction of transparent Disney
materials
commit e3aff6849e
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Jul 26 16:05:19 2016 +0200
Added subsurface color parameter to the Disney shader
commit b3ca6d8a2f
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Jul 26 12:30:25 2016 +0200
Improvement of the SSS in the Disney shader
* Now the bump normal is correctly used for the SSS.
* SSS in Disney uses the Disney diffuse shader
commit d68729300e
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Jul 26 12:23:13 2016 +0200
Better calculation of the Disney diffuse part
Now the values for NdotL und NdotV are clamped to 0.0f for a better look
when using normal maps
commit cb6e500b12
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Jul 25 16:26:42 2016 +0200
Now one can disable specular reflactions again by setting specular and
metallic to 0 (cracked this in the previous commit)
commit bfb9cb11b5
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Jul 25 16:11:07 2016 +0200
fixed the Disney SSS and cleaned the initialization of the Disney shaders
commit 642c0fdad1
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Jul 25 16:09:55 2016 +0200
fixed an error that was caused by the missing LABEL_REFLECT in the Disney
diffuse shader
commit c10b484dca
Author: Jens Verwiebe <info@jensverwiebe.de>
Date: Fri Jul 22 01:15:21 2016 +0200
Rollback attempt to fix sss crashing, it prevented crash by disabling sss completely, thus useless
commit 462bba3f97
Author: Jens Verwiebe <info@jensverwiebe.de>
Date: Thu Jul 21 23:11:59 2016 +0200
Add an undef for sc_next for safety
commit 32d348577d
Author: Jens Verwiebe <info@jensverwiebe.de>
Date: Thu Jul 21 00:15:48 2016 +0200
Attempt to fix Disney SSS
commit dbad91ca6d
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed Jul 20 11:13:00 2016 +0200
Added a roughness parameter for refractions (for scattering of the rays
within an object)
With this, one can create a translucent material with a smooth surface and
with a milky look.
The final refraction roughness has to be calculated using the surface
roughness and the refraction roughness because those two are correlated
for refractions. If a ray hits a rough surface of a translucent material,
it is scattered while entering the surface. Then it is scattered further
within the object. The calculation I'm using is the following:
RefrRoughnessFinal = 1.0 - (1.0 - Roughness) * (1.0 - RefrRoughness)
commit 50ea5e3e34
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue Jun 7 10:24:50 2016 +0200
Disney BSDF is now supporting CUDA
commit 10974cc826
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue May 31 11:18:07 2016 +0200
Added parameters IOR and Transparency for refractions
With this, the Disney BRDF/BSSRDF is extended by the BTDF part.
commit 218202c090
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon May 30 15:08:18 2016 +0200
Added an additional normal for the clearcoat
With this normal one can simulate a thin layer of clearcoat by applying a
smoother normal map than the original to this input
commit dd139ead7e
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon May 30 12:40:56 2016 +0200
Switched to the improved subsurface scattering from Christensen and
Burley
commit 11160fa4e1
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon May 30 10:16:30 2016 +0200
Added Disney Sheen shader as a preparation to get to a BSSRDF
commit cee4fe0cc9
Merge: 4f955d06b5bab6
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon May 30 09:08:09 2016 +0200
Merge branch 'cycles_disney_brdf' of git.blender.org:blender into cycles_disney_brdf
Conflicts:
intern/cycles/kernel/closure/bsdf_disney_clearcoat.h
intern/cycles/kernel/closure/bsdf_disney_diffuse.h
intern/cycles/kernel/closure/bsdf_disney_specular.h
intern/cycles/kernel/closure/bsdf_util.h
intern/cycles/kernel/osl/CMakeLists.txt
intern/cycles/kernel/osl/bsdf_disney_clearcoat.cpp
intern/cycles/kernel/osl/bsdf_disney_diffuse.cpp
intern/cycles/kernel/osl/bsdf_disney_specular.cpp
intern/cycles/kernel/osl/osl_closures.h
intern/cycles/kernel/shaders/node_disney_bsdf.osl
intern/cycles/render/nodes.cpp
intern/cycles/render/nodes.h
commit 4f955d0523
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue May 24 16:38:23 2016 +0200
SVM and OSL are both working for the simple version of the Disney BRDF
commit 1f5c41874b
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue May 24 09:58:50 2016 +0200
Disney node can be used without SVM and started to cleanup the OSL implementation
There is still some wrong behavior for SVM for the Schlick Fresnel part at the
specular and clearcoat
commit d4b814e930
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed May 18 10:22:29 2016 +0200
Switched from a parameter struct for Disney parameters to ShaderClosure params
commit b86a1f5ba5
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed May 18 10:19:57 2016 +0200
Added additional variables for storing parameters in the ShaderClosure struct
commit 585b886236
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue May 17 12:03:17 2016 +0200
added output parameter to the DisneyBsdfNode
That has been forgotten after removing the inheritance of BsdfNode
commit f91a286398
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue May 17 10:40:48 2016 +0200
removed BsdfNode class inheritance for DisneyBsdfNode
That's due to a naming difference. The Disney BSDF uses the name 'Base Color'
while the BsdfNode had a 'Color' input. That caused a text message to be
printed while rendering.
commit 30da91c9c5
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed May 4 16:08:10 2016 +0200
disney implementation cleaned
commit 30d41da0f0
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed May 4 13:23:07 2016 +0200
added the disney brdf as a shader node
commit 1f099fce24
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue May 3 16:54:49 2016 +0200
added clearcoat implementation
commit 00a1378b98
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Fri Apr 29 22:56:49 2016 +0200
disney diffuse und specular implemented
commit 6baa7a7eb7
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Apr 18 15:21:32 2016 +0200
disney diffuse is working correctly
commit d8fa169bf3
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Apr 18 08:41:53 2016 +0200
added vessel for disney diffuse shader
commit 6b5bab6cec
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed May 18 10:22:29 2016 +0200
Switched from a parameter struct for Disney parameters to ShaderClosure params
commit f6499c2676
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed May 18 10:19:57 2016 +0200
Added additional variables for storing parameters in the ShaderClosure struct
commit 7100640b65
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue May 17 12:03:17 2016 +0200
added output parameter to the DisneyBsdfNode
That has been forgotten after removing the inheritance of BsdfNode
commit 419ee54411
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue May 17 10:40:48 2016 +0200
removed BsdfNode class inheritance for DisneyBsdfNode
That's due to a naming difference. The Disney BSDF uses the name 'Base Color'
while the BsdfNode had a 'Color' input. That caused a text message to be
printed while rendering.
commit 6006f91e87
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed May 4 16:08:10 2016 +0200
disney implementation cleaned
commit 0ed0895914
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Wed May 4 13:23:07 2016 +0200
added the disney brdf as a shader node
commit 0630b742d7
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Tue May 3 16:54:49 2016 +0200
added clearcoat implementation
commit 9f3d39744b
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Fri Apr 29 22:56:49 2016 +0200
disney diffuse und specular implemented
commit 9b26206376
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Apr 18 15:21:32 2016 +0200
disney diffuse is working correctly
commit 4711a3927d
Author: Pascal Schoen <pascal_schoen@gmx.net>
Date: Mon Apr 18 08:41:53 2016 +0200
added vessel for disney diffuse shader
Differential Revision: https://developer.blender.org/D2313
This is possible to use surface-only nodes and connect them to volume output.
If there was something connected to surface output those extra connections
will not change anything visually but will force volume features to be included
into feature-adaptive kernels.
In fact, this exact reason seems to be causing slowdown of Barcelone file
comparing AMD OpenCL to NVidia CUDA.
Currently only supported by the final F12 renders because of the current design
of what gets optimized out when and how feature-adaptive kernel accesses
list of required features.
Reviewers: dingto, nirved, maiself, lukasstockner97, brecht
Reviewed By: brecht
Subscribers: bliblubli
Differential Revision: https://developer.blender.org/D2569
Remapping to itself is nonsense here (was triggering an assert in
BKE_library code actually), just make it a bail out early in RNA
callback in that case.
glFramebufferTexture() is only available from OGL 3.2, it is also not part of the ARB_framebuffer_object extension.
I don't know if there is a better way to check for mesa drivers or if using the 2D version will have issues with f2f16a256, but so far I had no problems
Sanitize a bit how cache path is handled by fluidsim (there is much more
to be done here though :( ), and forbid empty path (we reset to default
path relative to current .blend file in case it's empty).
If people really, really want to use current OS-wise directory, they can at
least use '.' as path. ;)
Follow-up to 3bc6b831d8
Initial code was getting the number of ShaderInterfaces that could fit between two pointers (which makes no sense). I really meant number of bytes.
Surprised it worked on Mac, glad I tested on Windows before bed :D
ShaderInterface stores names of custom uniforms and all attributes in an internal buffer, but it doesn't know how large to make this buffer until all inputs are scanned.
This commit shrinks the buffer to the exact size needed.
ShaderInterface_uniform searches custom uniforms first, then builtin uniforms if needed.
This reduces the amount of string matching since you're almost certainly looking for one of those custom uniforms. Otherwise you would've called ShaderInterface_**builtin**_uniform.
- test for 2D textures first since it's the most common case
- declare variables close to where they're used
- fix compiler warning for proxy (uninitialized use)
- safe return if n != 1, 2, 3 (should never happen)
- white space
The fragment shader expects a normal, but the vertex shader was not providing one.
Fix: added a new vertex shader that has normals + smooth color interpolation.
I also split gpu_shader_3D_vert in two:
- one with just position
- one with position + normal
For each of the builtin shaders, we should be able to look at the GLSL and tell exactly what it's doing. Using #defines and #ifdefs for rendering options makes the shaders hard to read and easy to break.
The specialized color functions are better in every way:
- faster lookup (don't have to match "color" string)
- flexible inputs (RGB with separate alpha)
- automatic alpha = 1.0 if not specified
Sort of related to T49043
When moved the options to toolsetting, this part was missing. The problem was not the pointer as suggested in D2629.
Thanks Arvīds Kokins for his help fixing this bug
attrib_binding.c uses NULL, so include stddef.
Also swapped stdlib for stddef in vertex_format.c, since it only needs NULL.
Not sure why MSVC 2017 and clang/Mac work without this...
Thanks to @youle for reporting!
This eliminates tons of glGetAttribLocation calls from the drawing loop. Vast majority of code can keep making the same function calls. They're just faster now!
- remove 2D-specific variants of BuiltinUniform enum
- rename remaining builtins to exclude "_3D" since they can be used by 2D or 3D shaders
Follow up to D2626
See GPU_matrix.h & gpu_matrix.c for the important changes. Other files are mostly just updated to use the latest API.
- remove unused functions, defines, enums, comments
- remove "3D" from function names
- init to Identity transform (otherwise empty stack)
- gpuMatrixReset lets outside code return to initial state
Part of T49450
Follow up to D2626 and 49fc9cff3b
This avoids the unnecessary creation of bvhtree, which can be highly inefficient in some cases
(for example: in the `operator_modal_view3d_raycast.py` template)
Unfortunately this does break compatibility in that the viewport will look a
bit different depending on the settings, but the old behavior was simply not
usable for higher distances.
* Brings us closer to core profile, all matrices are working, and apart
from a problem with text drawing, Blender is working fine.
* Reduce the coding overhead of having to setup/teardown when
alternating between 2D and 3D drawing sessions.
* Gives us fewer modes and states we need to keep track of.
Unfortunatelly this also "rejects a fundamental change" the original
design was trying to make - that 2D is different from 3D and
deserves its own best implementation.
That said, it is still aligned with the function API design as
originally implemented (i.e., it still uses gpuTranslate2D, ...).
Finally, if you build with core profile and this patch you get:
https://developer.blender.org/F545352
[The text glitch is an unrelated issue].
Reviewers: merwin, sergey, brecht
Differential Revision: https://developer.blender.org/D2626
Object Info node can be useful to give some variation to a single material assigned to multiple instances. This patch adds support for Viewport and BI.
{F499530}
Example: {F499528}
Reviewers: merwin, brecht, dfelinto
Reviewed By: brecht
Subscribers: duarteframos, fclem, homyachetser, Evgeny_Rodygin, AlexKowel, yurikovelenov
Differential Revision: https://developer.blender.org/D2425
On Apple we use OpenGL 2.1 + an ARB extension for framebuffers.
The glFramebufferTexture function is part of OpenGL 3.0 but not part of the ARB extension. This commit fills that gap.
All other platforms are using GL 3.0+ already so it's not an issue there. All platforms (Mac too) will use GL 3.3+ soon so this workaround will be removed.
The U-resolution of the imported curves was kept at the default value
of 12, which is way too high for imported hair. We export hair at a
fairly high resolution already, so it's not needed to subdivide even
further when importing.
Of course this may have an impact on other curves that do require this
U-resolution to be higher. In that case the resolution can be
increased after importing.
I removed the default nu->orderu = num_verts, as that allowed every
point to influence the entire spline, which was more expensive for the
CPU, and unlikely to be needed. The orderu computations had off-by-one
errors in the curve importer, which are now also fixed. The correct
values are:
- Linear: orderu = 2
- Quadratic: orderu = 3
- Cubic: orderu = 4
These values are also what is stored in the Alembic file for curves of
type kVariableOrder, according to the reference Maya exporter
maya/AbcExport/MayaNurbsCurveWriter.cpp, function
MayaNurbsCurveWriter::write(), in the Alembic source code.
The result is a frame rate increase of roughly 100x (tested with one
100-hair test on one machine, so take with grain of salt).
This test checks that a set of cubes are exported with the correct
transform, both with flatten=True and flatten=False.
This commit also adds an easy to use superclass for upcoming Alembic
unit tests.
These were the last few glGetUniformLocation calls in source/blender.
The new system gets uniform information once when a shader is created, then uses this cached info every time after that.
This removes glGetBooleanv queries for GL_LIGHTING. This has been #ifdef'd
out with legacy opengl disabled. Thus a false positive still shows up in
the gl queries. Also, note that this removes support for wireframes in
opensubdiv, when desabling legacy opengl, which should be fixed later.
Part of T49043
Some elbeem debug functions had old gl drawing calls. I removed the
functions instead of converting, because Blender doesn't even build
with them enabled anymore, because some elbeem debug libs got removed
at some point.
Part of T49043
This supports our common character animation workflow, where a character,
its rig, and the custom bone shapes are all part of a group. This group
is then linked into the scene, the rig is proxified and animated. Such
a group can now be exported. Use "Renderable objects only" to prevent
writing the custom bone shapes to the Alembic file.
The absence of datablock properties "will certainly be resolved soon as the need for them is becoming obvious" said the [[http://wiki.blender.org/index.php/Dev:Ref/Release_Notes/2.67/Python_Nodes|Python Nodes release notes]]. So this patch allows Python scripts to create ID Properties which reference datablocks.
This functionality is implemented for `PointerProperty` and now such properties can be created with Python.
In addition to the standard update callback, `PointerProperty` can have a `poll` callback (standard RNA) which is useful for search menus. For details see the test included in this patch.
Original author: @artfunkel
Alexander (Blend4Web Team)
Reviewers: brecht, artfunkel, mont29, campbellbarton
Reviewed By: mont29, campbellbarton
Subscribers: jta, sergey, campbellbarton, wisaac, poseidon4o, mont29, homyachetser, Evgeny_Rodygin, AlexKowel, yurikovelenov, fjuhec, sharlybg, cardboard, duarteframos, blueprintrandom, a.romanov, BYOB, disnel, aditiapratama, bliblubli, dfelinto, lukastoenne
Maniphest Tasks: T37754
Differential Revision: https://developer.blender.org/D113
... even though 3x3 feels better.
This is a compromise to get core profile up & running sooner. Eventually I'd like to finish the original 3x3 plans, but this commit will let us get on with other tasks.
External API stays (almost) the same. Our GLSL shaders can use this without any changes.
Part of T49450 and T51164
These functions mark the beginning and end of 3D drawing, causing the matrix API to use our own implementation instead of legacy OpenGL.
Part of T49450
Look up "name[0]" when asked for "name", since that marks the beginning of the array.
We're comparing to the name stored in ShaderInterface which comes from glGetActiveUniform.
This eliminates tons of glGetUniformLocation calls from the drawing loop. Vast majority of code can keep making the same function calls. They're just faster now!
- Batch_Uniform*
- immUniform*
- gpuBindMatrices
- and others
- builtin uniforms match what Blender needs
- set input counts in struct (stupid mistake)
- look up uniforms by name
- look up builtin uniforms by enum
- check attrib/uniform locations for error
Same as MEM_SAFE_FREE macro,
checks for NULL, runs free then sets NULL.
Blocks of code that do this many times are noisy and likely
errors here wouldn't be noticed immediately.
Also NULL's static vars which were being left set.
Problem was an edge case where vertex_ct logic and draw logic disagreed on how many dividers to draw.
Fix: copy draw logic to earlier vertex_ct
I also skip any drawing or setup if vertex_ct = 0, and set color attribute only for each line's provoking vertex. Small optimizations but these things add up.
We can not re-use anything for such pools, because we will know nothing about whether
the main thread is sleeping or not. So we identify such threads as 0, but we don't
use main thread's TLS.
This fixes dead-locks and crashes reported by Luca when doing playblasts.
This removes MAX_STORAGE, MAX_BUFFERS, MAX_TEXTURES, MAX_PASSES limits.
Actual memory saving isn't so important, it just means we don't need to
manually bump these based on changes to engines.
The issue was caused here by usage of deprecated GL_CURRENT_PROGRAM
which was returning rubbish value.
Now we use imm API and create vertex format prior to immBindProgram.
This made us required to have some sort of state passed from setup
function to actual drawing.
Note: I’d assume gawain equivalent to glDrawArrays would be batches? But
for two lines drawn twice this looks totally overkill anyway, so
switched back to basic immediate-mode-like API.
A bit frustrating to work on this code, since afaict you cannot check
the results in Blender, being mostly unused currently...
Global size depends on memory usage which might change during rendering.
Havent seen it happen but seems possible that this could cause the global
size to be different than what was used for allocating buffers.
The issue here is that the preferences are still used because both can be accessed from the 3D View, view menu. In the future, it is likely that the old mode will be removed (maybe 2.8?) but for now we want to keep both operational.
Differential revision: https://developer.blender.org/D2320
This was slightly less obvious, especially particle billboard drawing.
That one now requires 1.5 times more space in VBOs and ParticleDrawData
buffers, since we have to draw two triangles per particle, instead of a
single quad... And diagonal of each quad is now drawn in wire mode, not
sure how much we consider this an issue (as in, will this particle draw
code change a lot in future?).
From quick check on the web seems there is no other way to do anyway. :/
Do not recompute both points's 2D coordinates for each segments, we can
copy over from previous one... Does not gives any measurable speedup off
hands, though.
Couple of things here:
- Boost is not necesserily compiled into your /opt/lib and system-wide
version might have been used. The recent change in Alembic did not
take this into account.
- Alembic needs some extra component of Boost.
This part might be missing now for other distros than DEB.
The issue was caused by recent change in inline policy.
There is some sort of memory corruption happening here, ASAN suggests
it's stack overflow issue. Not quite sure why it is happening tho and
was not able to solve anything here yet in the past hours.
Committing fix which works with a big TODO note.
The issue is visible on AVX2 machine when rendering cycles_reports_test.
Using Texture Arrays to store shadow maps so less texture slots are used when shading. This means a large amount of shadows can be supported.
Support Projection Shadow Map for sun like in old BI/BGE.
Support Cube Shadow Map for Point/Spot/Area lights. the benefit of using it for spot light is that the spot angle does not change shadow resolution (at the cost of more memory used). The implementation of the cubemap sampling is targeted for 3.3 core. We rely on 2D texture arrays to store cubemaps faces and sample the right one manualy. Significant performance improvement can be done using Cubemap Arrays on supported hardware.
Shadows are only hardware filtered. Prefiltered shadows and settings comming next.
This is an auto-generated list, crossing gl-deprecated.h, glew.h and the
Blender code. It allows Blender to build with core profile.
WITH_OPENGL_LEGACY=ON: nothing changes
WITH_OPENGL_LEGACY=OFF and WITH_GL_PROFILE_CORE=OFF:
It stubs deprecated legacy calls.
WITH_OPENGL_LEGACY=OFF and WITH_GL_PROFILE_CORE=ON:
It stubs deprecated legacy calls thus allowing Blender to build with
core profile only.
Technically you only want to use WITH_OPENGL_LEGACY=OFF when
WITH_GL_PROFILE_CORE=ON. But it doesn't hurt to have it working for both
scenarios.
Reviewed by: merwin
Differential Revision: https://developer.blender.org/D2610
Doing this in a fully 'clean' way is far from obvious, especially
unregister, you often end up leaving nasty 'orphanned' keymap items
referring to unregistered operators...
See gpu_shader.c for the main changes.
EXT_geometry_shader4 brought this feature to GL versions < 3.2, but now it's just cluttering up our code.
Soon all platforms will be on version 3.3 so we won't even have to check support at runtime!
Callers now have to use Gawain's PRIM enum to specify geometric primitives.
This makes the API more bullet-proof (at least less vulnerable) since GLenum covers waaay more than GL_POINTS, GL_LINES, etc.
Also prepares us for Vulkan.
Callers now have to use Gawain's COMP enum to specify vertex attributes.
This makes the API more bullet-proof (at least less vulnerable) since GLenum covers waaay more than component types.
Also prepares us for Vulkan.
Single quads are drawn as a TRIANGLE_FAN, with 4 verts in the same order.
Multiple quads now use PRIM_QUADS_XXX and will need further work. Only 8 places still use this.
Part of T49043
Quads are not part of modern GL or Vulkan, so we should avoid them. XXX makes coders think "hmm how could I draw this without using quads?"
Quads will be removed during the transition to core profile.
Part of T49043
We concatenate #defines and #extensions into these, and can count the max string lengths needed. 256 is enough to hold today's strings; we can adjust later if needed.
When WITH_LEGACY_OPENGL = OFF.
This is our final target for Blender 2.8, all previous versions will be dropped in the future. GLSL 3.3 is richer so we don't require as many extensions.
This seems to happen on Windows only, happened to Thomas and Nathan already.
Similar patch Thomas was showing, but i do not see it committted. So comitting
now in order to get more developers and users happy.
Ever since we merged the extra texture types (half etc) and spit kernel the compile time for cycles_kernel has been going out of control.
It's currently sitting at a cool 1295.762 seconds with our standard compiler (2013/x64/release)
I'm not entirely sure why msvc gets upset with it, but the inlining of matrix near the bottom of the tri-cubic 3d interpolator is the source of the issue, this patch excludes it from being inlined.
This patch bring it back down to a manageable 186 seconds. (7x faster!!)
with the attached bzzt.blend that @sergey kindly provided i got the following results with builds with identical hashes
58:51.73 buildbot
58:04.23 Patched
it's really close, the slight speedup could be explained by the switch instead of having multiple if's (switches do generate more optimal code than a chain of if/else/if/else statements) but in all honesty it might just have been pure luck (dev box,very polluted, bad for benchmarks) regardless, this patch doesn't seem to slow down anything with my limited testing.
{F532336}
{F532337}
Reviewers: brecht, lukasstockner97, juicyfruit, dingto, sergey
Reviewed By: brecht, dingto, sergey
Subscribers: InsigMathK, sergey
Tags: #cycles
Differential Revision: https://developer.blender.org/D2595
It's possible that cancellation occured between the creation of the reader
and the creation of the Blender object, in which case reader->object()
returns a NULL pointer.
BKE_libblock_free_us() was called on the object data, which decrements
its user count, after which the same function was called on the object,
which decrements the user count of the object data again. This double
decrement was too much.
Original code from @Severin with changes from @dfelinto & @hypersomniac.
This doesn't cause many functional changes
besides using new transform manipulators.
Submitted as D2604
The state mask wasnt applied before comparison giving false results. It
shouldnt really happen that a ray state contains any flags that need to
be masked away, but if it does happen its better to not get stuck.
This fixes T51143.
gl_ClipDistance is part of GLSL version 1.3 but Mac is stuck on 1.2 for now.
This workaround uses GPU_SHADER_3D_UNIFORM_COLOR for the entire rotation widget, ignoring any clipping plane. The CLIPPING shader only works on GLSL 1.3+ so I removed its 1.2 cruft.
A legacy implementation using gl_ClipVertex might be possible, but is not worth the effort. This problem (and workaround) goes away when all platforms move to 3.3 core profile.
changes:
- exact vertex count
- take bool (filled vs outline) instead of GLenum
This function has some flexibility that is not currently used. I left that in.
Each function takes a bool (filled vs outline) and a color. We already had multiple ways of passing color in; these are still here. Special variant for anti-aliasing.
- took GLenum out of interface
- removed UI_RB_ALPHA flag (only one place really used it)
- use exact vertex count
- removed redundant state changes (BLEND, LINE_SMOOTH)
This introduces a new CMake option - WITH_LEGACY_OPENGL. Without this option
things may not draw perfectly, however, we should soon be able to build with
OpenGL core profile.
The matrix-related api calls are (still) not handled here (glTranslate, ...).
There seems to be no consensus on whether to make this build option the
default. We can talk about this later. For now two things are the
priority:
(1) To get rid of deprecated calls when WITH_LEGACY_OPENGL is ON
(2) To make core profile work for Mesa/Mac when WITH_LEGACY_OPENGL is OFF
Reviewers: merwin, sergey, campbellbarton
Differential Revision: https://developer.blender.org/D2603
Many thanks for Sergey Sharybin for the help.
Previously, a GHash was used to store a flattened mapping of parent
information based on the Alembic hierarchy, and then that hash was used to
set parent pointers on Blender objects. This resulted in errors and
some duplicate objects. The new approach stores parent pointers while
traversing the Alembic hierarchy, which means that there is much more
information about the actual context of the Alembic object itself,
producing a more stable import.
Also replaced the bool param "to_yup" with "AbcAxisSwapMode mode", so that
it's more explicit that axes are swapped.
Also added unittests for create_swapped_rotation_matrix.
There was a problem with parent-child relations not getting set up
correctly when an Alembic object was both the transform for a mesh object
and the parent of other mesh objects.
convert_matrix() now only converts from Imath::M44d to float[4][4] (taking
different camera orientations into account). Import-time scaling is now
performed by the caller.
Also renamed AbcObjectReader::readObjectMatrix to
setupObjectTransform, as it does more than just reading the object
matrix; it also sets up an object constraint if the Alembic Xform is
animated.
Alembic is an interchange and caching format, that can contain custom
object schemas. Blender shouldn't crash (because of failing asserts) just
because it doesn't know such an object type.
It's a mapping from full path of an Alembic object to an AbcObjectReader*.
The fact that at some point it is used to construct parent-child relations
doesn't matter.
The importer was guessing whether an Alembic IXform object was part of a
child object, or should be represented as an Empty in Blender. By reversing
the order in which objects are visited, the children can now claim their
parent as part of the same object (so IPolyMesh claims its parent IXform
as part of the same Blender object). This results in much less guesswork.
I've also removed similar guesswork from the code that sets parent pointers,
by simply searching for the parent in a hierarchical way, instead of trying
to predict (again) which IXforms were turned into empties.
Also, visit_object() now actually visits the object -- previously it only
visited its children, and assumed the object it was called on was already
handled by a previous call.
create_transform_matrix(float[4][4]) did mostly the same as
create_transform_matrix(Object *, float[4][4]), but more elegant.
However, the former has some inconsistencies with the latter (which
are now merged and made explicit, turned out one was for z-up→y-up
while the other was for y-up→z-up), and was renamed to
copy_m44_axis_swap(...) to convey its purpose more clearly.
Furthermore, "loc" has been renamed to "trans", as matrices don't
store locations but translations; and more variables now have a src_
or dst_ prefix to denote whether they contain a matrix/vector in the
source or destination axis orientation.
AbcExporter::createTransformWriter() tries to predict the parent Xform
name, but if it cannot be found has multiple ways of creating it, possibly
under a different name than originally searched for.
See intern/gawain for the API change. Other files are updated to use the new name. Also updated every call site to the recommended style:
unsigned int foo = VertexFormat_add_attrib(format, "foo", COMP_ ... )
This decoupled setting draw flags from calculating the manipulator center.
I can't see any advantage to doing this with the current code
It just increases the chance the two get out of sync.
Reference document: http://docs.gl/gl3/glPushAttrib
This patch only tackles the bits that are set by Blender with the
following exceptions:
1) Deprecated states (e.g., GL_STIPPLE) are not saved/restored
2) The exception being GL_ALPHA_TEST, which will be removed, but it may
affect drawing too much now. To be removed once we no longer set GL_ALPHA_TEST
elsewhere.
3) paint_cursor will be tackled separated, since it was abusing
glPush/PopAttrib in the first place.
4) Despite what the glPushAttrib page above may suggest, GL_DEPTH_WRITEMASK needs glGet, not glIsEnabled
5) BGE is still a problem since it relies on GL_ALL_ATTRIB_BITS which
would lead to a way more complete/lenghty solution. Since the BGE has
other (OpenGL deprecated) problems anyways, it can be handled on its own
time.
Finally, the original design for 2.8 was to implement a proper stack
system. However we need to move to core profile sooner than later. So
this is a pragmatic temporary (that may be permanent) solution.
Reviewers: merwin, campbellbarton
Differential Revision: https://developer.blender.org/D2600
These bits became obsolete with the new layer system, so we can
simplify some code around them or avoid existing workarounds which
were trying to keep things working for them.
There are still work needed to be done for on_visible_change to
avoid unnecessary updates, but that can also happen later.
Core of the issue was that some of our Theme colors are RGB-only, but
were loaded as RGBA.
Note that tracking all possible cases is pretty impossible, so we'll
have to tackle those as they get reported am afraid.
- use 'imm_draw_' prefix for functions that draw.
- use '_3d' suffix for 3d functions, no suffix for 2d functions.
- use terms fill/wire (shorter than filled / lined).
Also add `imm_draw_circle_fill_3d` (only had wire version)
The idea is to have a system where we properly log error messages and
let the users know that errors occured redirecting them to the console
for explanations. This is only implemented for the exporter since the
importer already has similar functionalities; however they shall
ultimately be unified in some way.
Reviewers: sybren, dfelinto
Differential Revision: https://developer.blender.org/D2541
Added new shader for clipping, also cleaned up rotation manipulator
drawing code a bit.
This code should be replaced by new transform manipulators soon, just
keeping this working in the meanwhile.
I started cleaning up UI_view2d_scale_get where the y scale was unused, then got carried away...
- for loop scope
- declare variables closer to where they are used
- move early exits closer to function start
- unsigned --> unsigned int
This avoids write access happening in non-atomic manner in
Shader::tag_update which modifies the global managers. Even
for 1 byte data types it's quite dangerous.
The issue is coming from the fact that float3 is actually 16 bytes aligned
data type and the "padding" was not initialized. This caused memcmp() to
access non-initialized memory.
This provides us with a clearer API (so I don't have to use const_cast<>
in upcoming code). It also allows layering of different Alembic files,
so you can have a base file and load a separate file containing overrides.
Verbally approved by Dr. Sergey.
Alembic requires one of ALEMBIC_LIB_USES_BOOST, ALEMBIC_LIB_USES_TR1, or
C++11, and silently defaults to the latter if the former two are OFF.
Before this change, Alembic was only built without C++11 of OpenEXR
was built at the same time. This dependency was both unnecessary and
undocumented.
Fixed a few issues
- wrong transform matrices
- assert on node with no sockets
- color of collapsed nodes, some title areas
Also includes minor cleanup.
Objects that were using the same lamp data were having the same display matrices.
This is fixed by allowing engine to store a memory block inside the object itself.
Using Linear Transform Cosines to compute area lighting. This is far more accurate than other techniques but also slower.
We use rotating quad to mimic sphere area light. For a better approximation, we use a rotating octogon.
This moves selectability/visibility flag flush from some hardcoded
places in the code to depsgraph. This way it is possible to simply
tag depsgraph to update those flags and rest it'll do on its own.
Using depsgraph for such flush is an overkill: those flags are fully
static and can not be animated, so it doesn't really make sense to
hook only those to depsgraph.
However, in the future we will have overrides on collections, which
ideally would need to be animatable and drivable and easiest way
to support this is to do this on depsgraph level, so it ensures
proper order of evaluation for animation and drivers. And it seems
logical to do both overrides and flags flush from depsgraph from
this point of view.
This commit also includes the evaluation of IDProperty for collections,
which basically are just another form of override. So once we implement
the other kind of overrides the flushing and collection evaluation won't
change.
Patch by Sergey Sharybin and Dalai Felinto
This should have been done earlier. But now that we may see more activity in master due to bcon3 merges, it is even more important to stress out the separation between master and 2.8.
Also I needed the version bump for idproperties doversion (they needed
to happen in _after_linking, and we don't have access to
DNA_struct_elem_find there).
Last but not least, every time I posted a video or image from 2.80, I
got someone confused about the version showing in the infobar.
This function was modifying texture datablock, which makes the call
unsafe for call from multiple threads. Now we pass the argument that
we don't need nodes to the underlying functions.
There will be still race condition in noise texture, but that should
at least be free from crashes. Doesn't mean we shouldn't fix it tho.
This removes graph 3d render feature, also use same variables as master
for SpaceNode.
The feature (could be added back if needed),
but didn't get such a positive functionality review, see D1781,
although this is still an open topic.
In this case the Pyobject gets lost from pybm, and bm.free() does not invalidate the PyElem.
This will cause the destructor of python to read invalid memory and crash.
The solution is to make a copy of the pyobjects pointers before overwriting.
First this replace a custom data struct with IDProperty, and use
IDProperty group merge and copying functions. Which means that a collection
property setting is only created if necessary.
This implements the "Layer Collection settings" override system, as
suggested in the "Override Manifesto" document.
The core is working, with Scene, LayerCollection and Object using a
single IDProperty to store all the render settings data. Next step is to
migrate this to depsgraph.
Note: Clay engine "ssao_samples" was hardcoded to 32 for now. It will come
back as part of "Workspace Settings" later.
Many thanks for Bastien Montagne for the help with the UI template
nightmare ;)
Differential Revision: https://developer.blender.org/D2563
This area is a subject of reconsideration, so for now used simplest
way possible -- ensure depsgraph's nodes have proper layer flags
when going in and out of local mode.
The issue was apparently caused by -fno-finite-math-only added to kernel.cpp
CFLAGS. For now just removed this flag from the kernel (we don't really want
it there at this point, and we don't have it for SSE/AVX optimized kernels).
But surely more investigation is needed here.
GPU_framebuffer no longer handles transform matrices, which this code was relying on. Made screen_preview_draw responsible for its own ModelView matrix.
This restores the feature for legacy viewport only. Modern viewport, Clay, Eevee, etc. will need further work.
Eventually we should rename this something other than "OpenGL".
I was hoping this would fix the issue of the object not moving after you copy it (right now you need to manually grab the object afterwards). But unfortunatelly it does not
The idea is to make include statements more explicit and obvious where the
file is coming from, additionally reducing chance of wrong header being
picked up.
For example, it was not obvious whether bvh.h was refferring to builder
or traversal, whenter node.h is a generic graph node or a shader node
and cases like that.
Surely this might look obvious for the active developers, but after some
time of not touching the code it becomes less obvious where file is coming
from.
This was briefly mentioned in T50824 and seems @brecht is fine with such
explicitness, but need to agree with all active developers before committing
this.
Please note that this patch is lacking changes related on GPU/OpenCL
support. This will be solved if/when we all agree this is a good idea to move
forward.
Reviewers: brecht, lukasstockner97, maiself, nirved, dingto, juicyfruit, swerner
Reviewed By: lukasstockner97, maiself, nirved, dingto
Subscribers: brecht
Differential Revision: https://developer.blender.org/D2586
rB870440dee910c9 just did NULL-check for Main pointer, actual issue is
that bContext pointer was NULL. This can be fixed by ensuring
PROP_ENUM_NO_CONTEXT flag is not set by calling
WM_operator_properties_sanitize when creating RNA buttons. Now, layout
previews are visible in keymap editor too.
The intention of this commit it to address issues mentioned in the
reports T43865,T50164 and T50452.
The code is based on Embree code with some extra vectorization
to speed up single ray to single triangle intersection.
Unfortunately, such a fix is not coming for free. There is some
slowdown for AVX2 processors, mainly due to different vectorization
code, which caused different number of instructions to be executed
and different instructions-per-cycle counters. But on another hand
this commit makes pre-AVX2 platforms such as AVX and SSE4.1 a bit
faster. The prerformance goes as following:
2.78c AVX2 2.78c AVX Patch AVX2 Patch AVX
BMW 05:21.09 06:05.34 05:32.97 (+3.5%) 05:34.97 (-8.5%)
Classroom 16:55.36 18:24.51 17:10.41 (+1.4%) 17:15.87 (-6.3%)
Fishy Cat 08:08.49 08:36.26 08:09.19 (+0.2%) 08:12.25 (-4.7%
Koro 11:22.54 11:45.24 11:13.25 (-1.5%) 11:43.81 (-0.3%)
Barcelone 14:18.32 16:09.46 14:15.20 (-0.4%) 14:25.15 (-10.8%)
On GPU the performance is about 1.5-2% slower in my tests on GTX1080
but afraid we can't do much as a part of this chaneg here and
consider it a price to pay for more proper intersection check.
Made in collaboration with Maxym Dmytrychenko, big thanks to him!
Reviewers: brecht, juicyfruit, lukasstockner97, dingto
Differential Revision: https://developer.blender.org/D1574
That one was:
* Resetting non-ID pointers (lib_link_xxx funcs should only affect ID
pointers, everything else shall be done in direct_link_xxx func).
* Even worse, always calling lib_link_animdata, even when
LIB_TAG_NEED_LINK tag was unset...
We do not need any special handling anymore for usercount of images used
by faces/polygons (tpage stuff), since we have the 'real_user' handling,
which will gracefully cope with all possible situations.
So better not keep that ugly confusing useless special case.
Mainly:
* Add missing `IDP_LibLinkProperty()` calls for many ID types
(harmless currently, but better be consistent here!).
* Bring lib_link_xxx functions more in line with each other.
* Replace some long if/else by switch.
Simplifies code quite a bit, making it shorter and easier to extend.
Currently no functional changes for users, but is required for the
upcoming work of shadow catcher support with OpenCL.
It uses an idea of accumulating all possible light reachable across the
light path (without taking shadow blocked into account) and accumulating
total shaded light across the path. Dividing second figure by first one
seems to be giving good estimate of the shadow.
In fact, to my knowledge, it's something really similar to what is
happening in the denoising branch, so we are aligned here which is good.
The workflow is following:
- Create an object which matches real-life object on which shadow is
to be catched.
- Create approximate similar material on that object.
This is needed to make indirect light properly affecting CG objects
in the scene.
- Mark object as Shadow Catcher in the Object properties.
Ideally, after doing that it will be possible to render the image and
simply alpha-over it on top of real footage.
Client vertex array state is deprecated, and these are the default values anyway.
No need to bind any basic shader. Let drawing code decide which shader it wants to use.
Part of T49165 (general OpenGL upgrade)
- use in/out instead of attribute/varying
- use named output instead of gl_FragColor
- use texture() instead of the multitude of older texture sampling functions
The #if __VERSION__ == 120 paths (needed on Mac) will be removed after we switch to 3.3 core profile.
Part of T49165 (general OpenGL upgrade)
Before, Cycles would first sync the shader exactly as shown in the UI, then determine and sync the used attributes and later optimize the shader.
Therefore, even completely unconnected nodes would cause unneccessary attributes to be synced.
The reason for this is to avoid frequent resyncs when editing shaders interactively, but it can still be avoided for noninteractive renders - which is what this commit does.
Reviewed by: sergey
Differential Revision: https://developer.blender.org/D2285
This change looks small, but it switches the entire 3D viewport from legacy OpenGL functions to our own code.
Kept non-modern viewport on legacy path so we can compare easily (via the Modern Viewport checkbox).
Part of T49450
Builtin names staring with gl_ will not be available in core profile. Same with the ftransform function. New matrix API provides the same names minus the gl_ prefix.
Part of T49450
- init projection matrices with identity
- fix copy/paste mistake in GetProjectionMatrix3D
- add extra matrices needed by material GLSL
Working toward T49450
New matrix API does not support texture matrices. Not sure what the final code will look like, but this at least avoids interference with new ModelView matrix.
Marked each line with TEXTURE so they can be disregarded during searches.
Related to T49450
Moving to manual class registration means its easier to accidentally
miss registering classes.
Now detect missing class registration
and warn when running with `--debug-python`
Removed infinite details and went for decreasing details with the distance instead (like the axis aligned ortho view auto sized grid).
Separate drawing of the Z axis into 2 pass (using shading group) to render ordered transparency with the main grid pass.
For Windows 8.1 and X11 (Linux, BSD) now use the DPI specified by the operating
system, which previously only worked on macOS. For Windows this is handled per
monitor, for X11 this is based on Xft.dpi or xrandr --dpi. This should result
in appropriate font and button sizes by default in most cases.
The UI has been simplified to a single UI Scale factor relative to the automatic
DPI, instead of two DPI and Virtual Pixel Size settings. There is forward and
backwards compatibility for existing user preferences.
Reviewed By: brecht, LazyDodo
Differential Revision: https://developer.blender.org/D2539
This adds the ability to switch between different application-configurations
without interfering with Blender's normal operation.
This commit doesn't include any templates,
so its mostly to allow collaboration for the Blender 101 project
and other custom configurations.
Application templates can be installed & selected from the file menu.
Other details:
- The `bl_app_template_utils` module handles template activation
(similar to `addon_utils`).
- The `bl_app_override` module is a general module
to assist scripts overriding parts of Blender in reversible way.
See docs:
https://docs.blender.org/manual/en/dev/advanced/app_templates.html
See patch: D2565
Use fast-math friendly version of this function.
We should probably avoid unsafe fast math, but this is to be done with
real care with all the benchmarks properly done.
For now comitting much safer fix.
Ideally we need to find a way to remove such a static limit here, but it's not so
trivial to implement for texture nodes. Requires some bigger system redesign there.
Just raising limit for now, which is fine for modern systems.
It is unused now and if we want similar function we should use
Pluecker intersection which is same performance with SSE optimization
but which is more watertight.
The title says it all actually. Gives up to 10% speedup on test scenes here
on i7-6800K.
Render times on GPU are unreliable here, but there might be some slowdown
caused by watertight nature of intersections.
Avoid construction of temporary array and make utility function force-inlined.
Additionally avoid calling float4_to_float3 twice.
This brings render times to the same values as before current patch series.
This is a preparation work for the followup commit which wil l move
remaining parts of Woop intersection logic to an utility file.
Doing it as a separate commit to keep changes more atomic and easier
to bisect when/if needed.
There are following benefits:
- Modifying intersection algorithm will not cause so much re-compilation.
- It works around header dependency hell and allows us to use vectorization
types much easier in there.
This removes the goal springs, in favor of simply calculating the goal forces on the vertices directly. The vertices already store all the necessary data for the goal forces, thus the springs were redundant, and just defined both ends as being the same vertex.
The main advantage of removing the goal springs, is an increase in flexibility, allowing us to much more nicely do some neat dynamic stuff with the goals/pins, such as animated vertex weights. But this also has the advantage of simpler code, and a slightly reduced memory footprint.
This also removes the `f`, `dfdx` and `dfdv` fields from the `ClothSpring` struct, as that data is only used by the solver, and is re-computed on each step, and thus does not need to be stored throughout the simulation.
Reviewers: sergey
Reviewed By: sergey
Tags: #physics
Differential Revision: https://developer.blender.org/D2514
Take advantage of 2D functions, rotation about the X Y or Z axis, uniform scale factors.
We no longer need to call gpuMatrixBegin_legacy() before using the new API locally in functions.
related to T49450
This is already called by wm_init_userdef, in old code
different initialization methods were used but now it's not needed.
Confusing since prefs are loaded in this function that don't initialize temp.
Mass replacing the common code of all tests
```
echo "Fixing $1"
ed "$1" <<'EOF'
1,/Testing/d
i
from render_layer_common import *
import unittest
import os
import sys
sys.path.append(os.path.dirname(__file__))
.
w
q
EOF
```
Using line-width of 120
There seems to be a compiler bug of MSVC2013. The issue does not happen on Linux and
does not happen on Windows when building with MSVC2015.
Since it's reallly a pain to debug release builds with MSVC2013 the AVX2 optimization
is disabled for curve sergemnts for this compiler.
Utility to get a file/dir in the path by index,
supporting negative indices to start from the end of the path.
Without this it wasn't straightforward to get
the a files parent directory name from a filepath.
When the new window didn't end up using the size stored in the preferences
the splash would not be centered (even outside the screen in some cases).
Now centered popups listen for window resizing.
Pretty sure source/blender is now finished, with all legacy matrix calls confined to gpu_matrix.c.
This was the easy part, but doing it first makes the next part much easier. TODO and XXX notes describe what is left.
glMatrixMode is still in place, since the new API does not share this concept of modes. Similar for glOrtho and glFrustum which I'll tackle very soon.
Part of T49450
For functions that expect a 4x4 matrix, you can pass in that, or array[16], or float*, or... Casting at each call site can get annoying, and obscures the logic.
The C11 section still needs work, but the non-C11 macros help on the system I tested on (Mac/clang).
Part of T49450
For example, for RX480 you'll no longer see "Ellesmere" but will see
"AMD Radeon RX 480 Graphics" which makes more sense and allows to easily
distinguish which exact card it is when having multiple different cards
of Ellesmere codenames (i.e. RX480 and WX7100) in the same machine.
This is used to send latest matrix values to shader when drawing.
Previously handled by calling OpenGL matrix functions, followed by gpuMatrixUpdate_legacy. With this change that function is no longer needed.
Part of T49450
For the sake of forward progress on T49450
We can now replace legacy gl* matrix function calls with their gpu equivalents. "Inactive" in this code means we're using the legacy matrix stacks, not our own. Setting up the proper gpuMatrixBegin2D/3D/End calls can be done afterward.
Most or all of this will be removed after the transition to core profile.
Previously, the code would only update the status string if the main status changed.
However, the main status did not include the remaining time, and therefore it wasn't updated until the amount of rendered tiles (which is part of the main status) changed.
This commit therefore makes the BlenderSession remember the time of the last status update and forces a status update if the last one was more than a second ago.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D2465
We render selected meshes into another buffer and use a screen space shader to expand the color out of the mesh silouhette.
Pros: only one additionnal render pass is needed (like old outline code), and we have occluded informations.
Cons: memory usage is a problem. This method needs 2 color buffer to ping pong when expanding the outline and 1 depth buffer to test occluded fragments. This gives a 88 bits/pix memory footprint.
Idea: Since we don't need all color range but only some uniform colors (theme colors) we could manipulate only the color ID instead of the whole color this could cut the color buffer size and lower the memory footprint to 58 bits/pix.
Although this wasn't so obvious since it
only showed up for factory settings and in the preferences window.
Panel display order depends on registration order,
Sorry for the noise. On the bright side we no longer need to move
classes around to re-arrange panels.
It also adds nice occluded silouhette information for selected objects that are behind visible objects.
This methods is really heavy because it needs to render the wires twices.
This attempt at TLS was leftover from an earlier prototype. It never worked with Blender's build system, and was defined to just exist, not to actually do anything.
NOTE: This is really a backend-only implementation, nothing is changed in the UI
Adds a tab button-type and the basic drawing and handling code for it.
More work needs to be done on it, but idea is to get in ready for usage in the
topbar.
Differential Revision: https://developer.blender.org/D1371
Instead of calling a function looping over whole list of a given ID
type, make whole loop over Main in parent function, and call functions
writing a single datablock at a time.
This design is more in line with all other places in Blender where we
handle whole content of Main (including readfile.c), and much more easy
to extend and add e.g. some generic processing of IDs before/after
writing, etc.
From user point, there should be no change at all, only difference is
that data-block types won't be saved in same order as before (.blend
file specs enforces no order here, so this is not an issue, but it could
bug some third party users using other, simplified .blend file reader maybe).
Reviewers: sergey, campbellbarton
Differential Revision: https://developer.blender.org/D2510
Before now it lived in source/blender/gpu for convenience. Only a few files in the gpu module use Gawain directly.
Tested on Mac, time to push and test on Windows.
Todo: some CMake magic to make it easy to
#include "gawain/some_header.h"
from any C or H file. Main problem here is the many editors that include GPU_immediate.h which includes Gawain's immediate.h -- is there a way to avoid changing every editor's CMakeLists?
Meshes w/o modifiers wouldn't have their derived mesh applied.
Check was to avoid crash but its in fact meaningless,
since the modifier might be disabled, or there may be virtual modifiers.
Not sure why I did not put those from start... Actually *not* having an
undo point here can be problematic, since undoing some previous action
was trying to restore from bad pointer (I think) in UI, generating
asserts.
Note however that it's not a 'pure' undo, in that you may not find your
linked data in exact same state as before deleting it, after an undo,
since it actually implies *reloading* the deleted libraries (and not
restoring from a previously stored memory dump).
Reported by @sergey, thanks.
There is no way currently to prevent the option from showing in menu, so
instead report a warning to user (and curse again current nightmarish
system of operation in outliner...).
Reported by @sergey, thanks.
Pass globals as a bare pointer, same as it sued to be prior to split kernel rework.
AMD CPU platform and Intel OpenCL were complaining about this.
Perhaps we shouldn't pass globals as pointer at all, this isn't something what is
really portable and can cause issues on 32 bit perhaps.
Now that we're almost done with T49043, let's run immediate mode at full speed. Debug builds will still do strict checks.
Developers should still test their changes before committing! Recommended:
$ make debug (or make lite debug)
$ blender --debug-gpu
We concluded this is going to be the display mode users will need to work
with the most, so makes sense to make it the default one.
Also, if the opened file only has one collection in the active render
layer, we expand it (almost empty list would be misleading).
What I had to do to make the expanding work is a bit ugly, but didn't
find a better way. During do_version we don't have access to the
TreeElement instances, and including ED_outliner.h to share code here
should be avoided too.
Internal change needed for template support.
Loading the user preferences first so it's possible
for preferences to control startup behavior.
In general it's useful to load preferences before data-files,
so we know security settings for eg.
The range is controlled using the following command line arguments:
--cycles-resumable-start-chunk
--cycles-resumable-end-chunk
Those are 1-based index of range for rendering.
GL_SELECT is really slow in this branch and will be removed.
For now we simply change AUTO behavior to avoid possible conflicts with merges
and upcoming color-id-based selection.
Talked with Pablo Vazquez (venomgfx) and Julian Eisel (Severin), and we came up with this solution instead.
Basically, if the file has only one layer, it is converted to a collection named "Default Collection". Otherwise we name the collections: "Collection 1 [converted from 2.75]"
Just do early output and don't bother with any GLSL program bind when
there is not enough points of lasso to draw.
This could have happened at the very beginning of the stroke.
The thing i'm really starting to hate is the requirement to specify both
operation code and node type. Seems to be duplicated enums without real
need for that.
This addresses an issue raised by D2453 -
that there was no way to check if operators are run
multiple times in a row.
Actions are still ignored that don't cause an UNDO event.
Was using some threaded queue on top of task pool, tssk...
Now using properly task pool directly to crunch chunks of smooth fans.
No noticable changes in speed.
Tried to completely get rid of the 'no threading with few loops' code,
but even just creating/freeing the task pool, without actually pushing
any task, is enough to make code 50% slower in worst case scenario (i.e.
few thousands of simple cube objects).
The root of the issue was in custom normal code, so far it assumed that
we could only have one cyclic smooth fan around each vertex, which is...
blatantly wrong (again, the two cones sharing same vertex tip e.g.).
This required a rather deep change in how smooth fans/clnor spaces are processed,
took me some time to find a 'good' solution.
Note that new code is slightly slower than previous one (maybe about 5%),
not much to be done here, am afraid.
Tested against all older report files I could find, seems OK.
- Add optional 'display_name' callback
so callers can construct own names.
- Add optional 'prop_filepath' argument
(for operators that don't use "filepath").
- Add doc-string.
- Use keyword only arguments.
So we can store (for example) vertex positions in one buffer and normals + colors in another buffer. Not super exciting right now, but very useful once we start changing some attribute values.
Supports future work by Clément et al. Only tested with one VBO per Batch since that's all our current code uses.
glMaterial
glColorMaterial
glPixelTransfer
and glEnable/Disable for
GL_FOG, GL_LIGHTING, GL_COLOR_MATERIAL
All of these were just setting default values, so I don't expect any visible change.
Part of T49165 (general OpenGL upgrade)
GPU_viewport_bind & unbind should be called as pairs. A recent commit -- 3b91989a09 -- called unbind only for some code paths, overflowing OpenGL's matrix & attrib stacks.
These are from the ARB_vertex_type_2_10_10_10_rev extension that became part of OpenGL 3.3.
So they are new, but only exist for compatibility with immediate mode, which is old.
Related to T49165 (general OpenGL upgrade)
A user doesn't want to necessarily create a new Screen only because she
wants a new window.
This patch allows the user to pick the screen to use for the new Window.
If the screen picked is the active one, it duplicates it (as the old
behaviour in Blender).
Patch with contributions and fixes by Julian Eisel (Severin)
Subscribers: venomgfx
Differential Revision: https://developer.blender.org/D2555
While this compiler is not officially supported yet, getting it to work is
a nice thing because more and more AMD cards will fall under MESA driver.
It's also nice to use explicit comparison with NULL, which makes it more
clear whether variable is a boolean or pointer. Even Rust enforces this!
Patch by Ian Bruce with own modifications.
The mesh convert operator can 'freeze' a mesh
(WYSIWYG, modifiers, shape keys etc).
However its not very obvious that the way to perform this
operation is to convert a mesh to a mesh.
Expose this as 'Visual Geometry to Mesh' in the 'Apply' menu,
since this is where users might expect to see it.
Use expanded names for bmesh primitive operations
(urmv jvke semv jfke).
Use 'bmesh_kernel_' prefix,
these functions aren't intended for wide use so favor readability.
Remove BM_face_vert_separate,
it wasn't used and only skipped step of finding correct loop of face.
It might be useful to keep the search string stored in some cases, but
in most it's not useful but confusing. Especially if the string is taken
from a menu showing a different enum.
When the loop region passed in had no loops to edge-split from,
it was assumed nothing needed to be done.
This ignored the case where loops share a vertex
without any shared edges.
Now BM_face_loop_separate_multi behaves like BM_face_loop_separate.
Fixed error where faces remained connected by verts in BM_mesh_separate_faces.
This commit adds new features to the breakdowner, giving animators more
control over what gets interpolated by the breakdowner. Specifically:
"Just as G R S let you move rotate scale, and then X Y Z let you do that
in one desired axis, when using the Breakdower it would be great to be
able to add GRS and XYZ to constrain what transform / axis is being
breakdowned."
As requested here:
https://rightclickselect.com/p/animation/csbbbc/breakdowner-constrain-transform-and-axis
Notes:
* In addition to G/R/S, there's also B (Bendy Bone settings and C (custom properties)
* Pressing G/R/S/B/C or X/Y/Z again will turn these constraints off again
Turned out to be a quite easy fix. I thought the issue was that we
couldn't identify the TreeStoreElem when (re)creating its TreeElement item
correctly, because for non-ID elements that would be index dependent (=
bad for drag & drop). Turns out that we're actually allowed to store
custom data within the TreeStoreElem, the thing is just that it gets
stored as ID pointer (highly ugly and highly misleading).
Anyway, seems to work now so I won't complain too much :)
This way we can ensure the overlay to indicate where the item would be
placed if it was dropped now is always at the correct place and doesn't
mislead the user.
Two issues are remaining, they'll be fixed separately:
* Graphical feedback when dragging within the master collection is wrong
* There's some bug where collections swap places instead, Dalai will investigate
- Connectivity length was overwritten by distance to closest selected.
- Vertices used the 'island' center of the closest vertex,
even if it wasn't connected.
Now optionally keep track of the original index of used as the closest
connected distance.
To support this needed to add optional support for islands of 1 vertex.
The changes introduced in rB3e628eefa9f55fac7b0faaec4fd4392c2de6b20e
made the non-subframe frame change behaviour less intuitive, by always
truncating downwards, instead of rounding to the nearest frame instead.
This made the UI a lot less forgiving of pointing precision errors
(for example, as a result of hand shake, or using a tablet on a highres scren)
This commit restores the old behaviour in this case only (subframe inspection
isn't affected by these changes)
Selection loop would draw the selection ignoring xray.
Now draw in a separate pass after clearing the depth buffer,
as with regular drawing.
Also disable depth sorting,
caller can sort the hit-list by depth if needed.
Note 1: renamed draw_cursor to draw_text_decoration, since it was drawing
cursor, margin, selection and line highlight
Note 2: commented out code update coming next
Part of T49043
Single program generally compiles kernels faster (2-3 times), loads faster,
takes less drive space (2-3 times), and reduces the number of cached kernels.
Reduces memory allocation for split kernel.
This allows for faster rendering due to bigger global size,
specially when GPU memory is limited.
Perfromance results:
R9 290 total render time
Before After Change
BMW 4:37 4:34 -1.1 %
Classroom 14:43 14:30 -1.5 %
Fishy Cat 11:20 11:04 -2.4 %
Koro 12:11 12:04 -1.0 %
Pabellon Barcelona 22:01 20:44 -5.8 %
Pabellon Barcelona(*) 15:32 15:09 -2.5 %
(*) without glossy connected to volume
Decoupled ray marching is not supported yet.
Transparent shadows are always enabled for volume rendering.
Changes in kernel/bvh and kernel/geom are from Sergey.
This simiplifies code significantly, and prepares it for
record-all transparent shadow function in split kernel.
This also fixes a little bug, which caused `draw_fcurve_samples` to
never be called, and thus sampled curve range boundaries were not drawn.
Part of T49043
All engines are now called by the draw manager. Engines are separate entities that cannot interfer with each others.
Also separated draw_mode_pass.c into the mode engines.
This is to be used from the Outliner, when dragging and dropping
collections from the Active Render Layer
It also includes a cleanup on the outliner so it calls the new
functions. Note: the outliner still needs fix to allow all the
functionality here exposed.
But this will be tackled by Julian Eisel later.
Intended to replace legacy GL_SELECT, without the limitations of
sample queries which can't access depth information.
This commit adds VIEW3D_SELECT_PICK_NEAREST and VIEW3D_SELECT_PICK_ALL
which access the depth buffers to detect whats under the pointer,
so initial selection is always the closest item.
The performance of this method depends a lot on the OpenGL
implementations glReadPixels.
Since reading depth can be slow, buffers are cached for object picking
so selecting re-uses depth data, performing 1 draw instead of 3
(for 24, 18, 10 px regions, picking with many items under the pointer).
Occlusion queries draw twice when picking nearest,
so worst case 6x draw calls per selection.
Even with these improvements occlusion queries is faster on AMD hardware.
Depth selection is disabled by default, toggle option under select method.
May enable by default if this works well on different hardware.
Reviewed as D2543
The issue was caused by sometimes negative color returned by the filter node.
Seems to be caused by precision issues. Don't see any reason why we would want
negative colors in output. Those only causing issues later on.
By calculating the size of the state buffer in the kernel rather than the host
less code is needed and the size actually reflects the requested features.
Will also be a little faster in some cases because of larger global work size.
Because the split kernel can render multiple samples in parallel it is
necessary to have everything initialized before rendering of any samples
begins. The code that normally handles initialization of
`rng_state` (`kernel_path_trace_setup()`) only does so for the first sample,
which was causing artifacts in the split kernel due to uninitialized
`rng_state` for some samples.
Note that because the split kernel can render samples in parallel this
means that the split kernel is incompatible with the LCG.
This was only needed for the previous implementation of parallel samples. As
we don't have that any more it can be removed.
Real reason for removal tho is this: `per_sample_output_buffers` was being
calculated too small and artifacts resulted. The tile buffer is already
the correct size and calculating the size for `per_sample_output_buffers`
is a bit difficult with the current layout of the code. As
`per_sample_output_buffers` was only needed for `sum_all_radiance`,
removing that kernel and writing output to the tile buffer directly
fixes the artifacts.
This is to help debug and track memory usage for generic buffers. We
have similar for textures already since those require a name, but for
buffers the name is only for debugging proposes.
Simple workaround for some issues we've been having with AMD drivers hanging
and rendering systems unresponsive. Unfortunately this makes things a bit
slower, but its better than having to do hard reboots. Will be removed when
drivers have been fixed.
Define CYCLES_DISABLE_DRIVER_WORKAROUNDS to disable for testing purposes.
This does a few things at once:
- Refactors host side split kernel logic into a new device
agnostic class `DeviceSplitKernel`.
- Removes tile splitting, a new work pool implementation takes its place and
allows as many threads as will fit in memory regardless of tile size, which
can give performance gains.
- Refactors split state buffers into one buffer, as well as reduces the
number of arguments passed to kernels. Means there's less code to deal
with overall.
- Moves kernel logic out of OpenCL kernel files so they can later be used by
other device types.
- Replaced OpenCL specific APIs with new generic versions
- Tiles can now be seen updating during rendering
Suspended pools allows to push huge amount of initial tasks
without any threading synchronization and hence overhead.
This gives ~50% speedup of cached rigid body with file from
T50027 and seems to have no negative affect in other scenes
here.
The idea is to allow some amount of tasks to be pushed from working
thread to it's local queue, so we can acquire some work without doing
whole mutex lock.
This should allow us to remove some hacks from depsgraph which was
added there to keep threads alive.
This allows us to avoid TLS stored in pool which gives us advantage of
using pre-allocated tasks pool for the pools created from non-main thread.
Even on systems with slow pthread TLS it should not be a problem because
we access it once at a pool construction time. If we want to use this more
often (for example, to get rid of push_from_thread) we'll have to do much
more accurate benchmark.
Basically move all thread-specific data (currently it's only task
memory pool) from a dedicated array of taskScheduler to TaskThread.
This way we can add more thread-specific data in the future with
less of a hassle.
This feature was adding extra complexity to task scheduling
which required yet extra variables to be worried about to be
modified in atomic manner, which resulted in following issues:
- More complex code to maintain, which increases risks of
something going wrong when we modify the code.
- Extra barriers and/or locks during task scheduling, which
causes extra threading overhead.
- Unable to use some other implementation (such as TBB) even for
the comparison tests.
Notes about other changes.
There are two places where we really had to use that limit.
One of them is the single threaded dependency graph. This will
now construct a single-threaded scheduler at evaluation time.
This shouldn't be a problem because it only happens when using
debugging command line arguments and the code simply don't
run in regular Blender operation.
The code seems a bit duplicated here across old and new
depsgraph, but think it's OK since the old depsgraph is already
gone in 2.8 branch and i don't see where else we might want
to use such a single-threaded scheduler.
When/if we'll want to do so, we can move it to a centralized
single-threaded scheduler in threads.c.
OpenGL render was a bit more tricky to port, but basically we
are using conditional variables to wait background thread to
do all the job.
This slightly changes SDef behavior, by now respecting object transforms
at bind time, thus not requiring the objects to be aligned in their
respective local spaces, but instead using world space.
When rendering multi-view in side-by-side or top-bottom mode, we squash
the UI to half of its size and draw it twice on screen. That means the
cursor coordinates used for UI interaction don't match what's visible on
screen.
This commit is a little event system hack (tm) to fix this. It has some
small glitches with cursor grabbing, but nothing to bad.
We'll also use it for viewport HMD support.
D1350, thanks for the feedback @dfelinto!
It was only possible to separate all geometry from an intersection or none.
Made this into an enum with a 3rd option to 'Cut', (now default)
which keeps each side of the intersection separate
without splitting faces in half.
There was a bug in the intended code behaviour to always seek with a
pitch of 1.0 regardless of pitch/pitch animation/doppler effects.
Check the bug report for a more detailed explanation of problems
concerning pitch and seeking.
This helper function was marked DEPRECATED since it uses old OpenGL calls.
Switched last 2 uses to imm_draw_checker_box, which does the same thing, only awesome.
Part of T49043
Comments said that function was supposed to 'stop worker threads', but
it absolutely did not do anything like that, was merely wiping out TODO
queue of tasks from given pool (kind of subset of what
`BLI_task_pool_cancel()` does).
Misleading, and currently useless, we can always add it back if we need
it some day, but for now we try to simplify that area.
Freeing pool was calling `BLI_task_pool_stop()`, which only clears
pool's tasks that are in TODO queue, whithout ensuring no more tasks
from that pool are being processed in worker threads.
This could lead to use-after-free random (and seldom) crashes.
Now use instead `BLI_task_pool_cancel()`, which does waits for all tasks
being processed to finish, before returning.
The paint slot name was not the same as what is displayed on the texture properties panel.
Instead, the slot type (e.g. "Diffuse Color") was used as the name.
Patch by Suchaaver (@minifigmaster125) with minor changes from @mont29.
Reviewers: mont29, sergey
Maniphest Tasks: T50704
Differential Revision: https://developer.blender.org/D2523
Can't say enough how much I hate those proxies... their duality (sharing
some aspects of both direct *and* indirect users) is a nightmare to handle. :(
Renamed existing getter/setter that only FX shaders use. We could convert FX code to use the richer new interface or leave it as is.
Removed unused GPUShader fields. ShaderInterface tracks the same information.
It wasn't using old immediate mode, but was using
- client vertex arrays (obsolete)
- quads (obsolete)
- state attrib stack (obsolete)
- polygon mode (still allowed, but gross)
After a GLSL program is linked we can get all its inputs & never have to ask it again.
Several uniforms are considered "built-in". Nothing special about these to OpenGL itself, they just follow conventions of our built-in shaders.
This will help the matrix API, immediate & batch drawing APIs, and allow extra error/compatibility checking.
PrimitiveClass will help match shaders to draw calls & detect usage errors. For example a point sprite shader only makes sense with PRIM_POINTS.
Updated other files to include new primitive.h
Nothing happen yet when it's supposed to insert the collection into
another one, that part will be handled by @dfelinto.
See gif for demo of how it works UI wise: {F500337}
Also fixed off-by-one error in utility function.
There are now only referenced in:
* drawobject.c
* particle_edit.c
* space_image.c (a single case to be handled on workspace branch)
* rigidbody_constraint.c (to be handled in the following commit)
Issue was that the VIEW_OT_manipulator operator calls the transform
operators and passes them it's own operator properties. That means the
transform operator got properties passed that it doesn't have.
The custom poll function for surfacedeform_bind seems to have caused
issues when calling it from Python. Fixed by using the generic modifier
poll function, and setting the button to be active or not in the
Python UI code instead. (there might be a better way, but for now this
works fine)
The issue was introduced by 4df75e5 and seems we just need to explicitly
add new keymap item now.
There is still some difference from old behavior, which is planar transform
is using precision movement since e138cde and here i don't see nice solution
currently: the change was requested here in the studio and it's just a
conflict in picking shift key for something which is not supposed to be
accurate.
At least now it's possible to invoke planar constraint and simply unhold
shift.
Not really happy of per-pool threads limit, need to find better
approach to that. But at least it's possible to get rid of half
of the nastyness here by removing getter which was only used in
an assert statement.
That piece of code was already well-tested and this code becomes
obsolete in the new depsgraph and does no longer exists in blender
2.8 branch.
This was only used for progress report, and it's wrong because:
- Pool might in theory be re-used by different tasks
- We should not make any decision based on scheduling stats
Proper way is to take care of progress by the task itself.
There were some issues with how we store outliner tree elements:
Apparently the only removable elements have been data-blocks so far.
When recreating the TreeElements, their TreeStoreElem instances were
mainly identified by their ID pointer. However non-data-blocks mostly
depend on an index. For collections, such an index isn't a reliable
measure though if we want to allow removing items. Depending on it for
identifying the TreeStoreElem instance would cause some quite noticeable
glitches (wrong highlights, two elements sharing highlight, etc).
For now I've solved that by actually removing the TreeStoreElem that
represents the removed element. A little limitation of this is that
after undoing the removal, some information might get lost, like
flags to store selection, or opened/closed state.
A better solution that would also fix this issue would be having a real
unique identifier for each non-data-block element, like an idname or even
its data-pointer. Not sure if we can get those to work reliable with
file read/write though, would have to investigate...
Also added a general Outliner tree traversal utility.
Artifacts still remained when 2 vertices were linedup in orthographic mode or with very acute vertex angles.
This commit fix that by adding even more geometry.
ifdef out most debugging code since Apple does not implement any debug extensions. We can revisit this if they ever do (don't expect that).
Changed output message so Mac users don't think --debug-gpu is broken.
I was removing deprecated/obsolete state from this function. About halfway through I started questioning the need for the whole thing.
GPU_state_print is not called anywhere, but is (was) available as a development aid.
External GL debugging tools are really good these days! We should use those to examine state & not roll our own.
Updated shader names and code that uses them.
All of these shaders produce round points that are anti-aliased and blended against the background.
These were initially named SMOOTH because they replace glEnable(GL_POINT_SMOOTH). But SMOOTH in shader-land refers to vertex attribute interpolation (like glShadeModel(GL_SMOOTH)).
Using SMOOTH to mean two things is confusing, so we now use AA to mean "the point is anti-aliased".
- Size parameter is total size of the shape, not its radius (half size). Updated hard-coded sizes to match this.
- Shader expands size to include outline.
- Fixed fringe between outline color and transparent background.
We do a strip of triangles around the triangle of interest and modify the variying vars to "trick" the fragment shader into rendering what we need.
This keeps the compatibility for both shaders.
Vertex still get half displayed when angle at the vertex is very acute.
This patch is only for the simple, no vert clipped case.
This is 2.5x slower than without the trick (on my hardware with 1million tris).
Do a "full" update on leaving sculpt mode, so we are sure scene will be brought
to a consistent state.
Ideally we'll only do that when there are objects which depends on geometry
without re-calculating self geometry, but that's a bit tricky currently.
The idea is to make it simpler to remove noise from scenes when some prop uses
Sharp glossy closure and causes noise in certain cases. Previously Sharp Glossy
was not affected by Filter Glossy at all, which was quite confusing.
Here is a file which demonstrates the issue: {F417797}
After applying the patch all the noise from the scene is gone.
This change also solves fireflies reported in T50700.
Reviewers: brecht, lukasstockner97
Differential Revision: https://developer.blender.org/D2416
Based on the previous overlay shader from merwin.
This shader takes care of clipped vertex cases and do all edit mode face info in one pass (except face centers).
As the shading is done one the triangle itself the visual can't go beyond the surface of the mesh. That leads to half displayed edges on the outline of the mesh.
This problem can be fixed by a second pass.
This is work in progress.
Can't see any reason to call AUD exit early in WM_exit, that's a
low-level module that has no dependency on anything else in Blender, but
is dependency of some other parts of Blender, so it should rather be
exited late in the process!
This was causing blender to segfault.
We now add create a new collection and link to the layer before adding
the new object
(also included unittests, and requires updated lib/tests)
This adds an option to force fields of type "Force", which enables the
simulation of gravitational behavior (dist^-2 falloff).
Patch by @AndreasE
Reviewers: #physics, LucaRood, mont29
Reviewed By: #physics, LucaRood, mont29
Tags: #physics
Differential Revision: https://developer.blender.org/D2389
- for rc/release: /api/2.79c/, zip file named blender_python_reference_2.79c_release.zip
- for dev: /api/master/, zip file named blender_python_reference_2_79_4.zip
Note that since there is no (efficient) ways to get arrays of
MVert/MEdge/etc. out of a BMesh, I refactored quite heavily internals of
BKE_mesh_render.
Now, when you do need acess to mesh data to generate cached batches, you
create an abstract struct from mesh (either Mesh or BMesh if available),
and then use advanced helpers to extract needed data, on a per-item
basis (no more handling of arrays of verts/edges/... in batches code).
This allows to:
* Avoid having to create arrays of BMesh elements.
* Take advantage of existing advanced BMesh topology and connectivity data.
Reviewers: dfelinto, fclem, merwin
Differential Revision: https://developer.blender.org/D2521
Please never, ever use same DNA var for two different things. Even worse
if they do not have same type and ranges!
This is only ensuring issues (as described in report, but also if
animating both RNA props using same DNA var... yuck).
And we were not even saving any byte in DNA, could reuse some padding
there to store the two new needed vars (yes, two, since we cannot re-use
existing one if we want to keep backward *and* forward compatibility).
Replaced all calls to `imm_draw_line` by plain `immVertex2f` calls, and
removed `imm_draw_line`, as that function was not supposed to exist in
the first place, and causes unnecessary calls to `immBegin`/`immEnd`.
Part of T49043
* Fix several wrong coordinates, causing things to be drawn in the wrong places.
* Remove unexpected return.
* Slight peformance improvement, by reducing number of shader binds.
* Minor code style stuff.
Part of T49043
Was actually harmeless and not crashing, but I'd say more or less only
by luck: the NULL-check for region data would only evaluate to true for
the correct 3D View region. However, if we were to add region data to a
different region type in future, this would lead to undefined behavior
if executed in the wrong region.
So... Curve+shapekey was even more broken than it looked, this report was
actually a nice crasher (immediate crash in an ASAN build when trying to
edit a curve shapekey with some viewport rendering enabled).
There were actually two different issues here.
I) The less critical: rB6f1493f68fe was not fully fixing issues from
T50614. More specifically, if you updated obdata from editnurb
*without* freeing editnurb afterwards, you had a 'restored' (to
original curve) editnurb, without the edited shapekey modifications
anymore. This was fixed by tweaking again `calc_shapeKeys()` behavior in
`ED_curve_editnurb_load()`.
II) The crasher: in `ED_curve_editnurb_make()`, the call to
`init_editNurb_keyIndex()` was directly storing pointers of obdata
nurbs. Since those get freed every time `ED_curve_editnurb_load()` is
executed, it easily ended up being pointers to freed memory. This was
fixed by copying those data, which implied more complex handling code
for editnurbs->keyindex, and some reshuffling of a few functions to
avoid duplicating things between editor's editcurve.c and BKE's curve.c
Note that the separation of functions between editors and BKE area for
curve could use a serious update, it's currently messy to say the least.
Then again, that area is due to rework since a long time now... :/
Finally, aligned 'for_render' curve evaluation to mesh one - now
editing a shapekey will show in rendered viewports, if it does have some
weight (exactly as with shapekeys of meshes).
This adds initial support for reordering collections from the Outliner
using drag & drop.
Although drag & drop support is limited to collections for now, this
lays most foundations for general drag & drop reordering support in the
Outliner. There are some design questions to be answered though:
* Would reordering of other data types (like objects) be a purely visual change or would it affect the order in which they are stored? (Would that make a difference for the user?)
* Should/can we allow mixing of different data types? (e.g. mixing render layers with objects)
* How could we realize this technically?
Notes:
* "Sort Alphabetically" has to be disabled to use this ("View" menu).
* Reordering only works with collections on the same hierarchy level.
* Added some visual feedback that should work quite well, it's by far not a final design though: {F493806}
* Modified collection orders are stored in .blends.
* Reordering can be undone.
* Did minor cleanups here and there.
This was fixed ages ago for the interface case but not for the
command line. The thing here is that currently external engines
are relying on reports system to indicate that error happened
so suppressing reports storage in the background mode prevented
render pipeline from detecting errors happened.
This is all weak and i don't like it, but this is better than
delivering black frames from the farm.
This should give the overall direction to whom wants to finish it.
- Renamed EDIT mode engine to EDIT_MESH mode engine
- Introduce EDIT_ARMATURE mode engine
- Started to port legacy drawarmature.c to draw_armature.c
- Added runtime display matrices to EditBone and bPoseChannel
- Added Object space instance vertex shader and modified the simple lighting shader accordingly
New logic of split_faces was leaving mesh in a proper state
from Blender's point of view, but Cycles wanted loop normals
to be "flushed" to vertex normals.
Now we do such a flush from Cycles side again, so we don't
leave bad meshes behind.
Thanks Bastien for assistance here!
Finding which loop should share its vertex with which others is not easy
with regular Mesh data (mostly due to lack of advanced topology info, as
opposed with BMesh case).
Custom loop normals computing already does that - and can return 'loop
normal spaces', which among other things contain definitions of 'smooth
fans' of loops around vertices.
Using those makes it easy to find vertices (and then edges) that needs
splitting.
This commit also adds support of non-autosmooth meshes, where we want to
split out flat faces from smooth ones.
The issue seems to be caused by vertex normal being re-calculated
to something else than loop normal, which also caused wrong loop
normals after re-calculation.
For now issue is solved by preserving CD_NORMAL for loops after
split_faces() is finished, so render engine can access original
proper value.
This way we can point submodules to different branches.
There are two side-effects to this:
- Git 1.8.2 becomes the minimal required version now
to support this feature.
- Not sure how doing local changes in submodules followed
by `make update` will behave. We don't use explicit rebase
now.
Perhaps this is not so bad, since it was already quite
dangerous thing to do.
Basically DEG_OBJECT_ITER (or rather,
BKE_scene_layer_engine_settings_update) wasn't creating
Object->collection_settings data for invisible objects.
Now I'm removing those objects from the loop entirely. If we are to
bring them back we need to either create CollectionEngineSettings dat
from them or to skip them in DRW_mode_cache_populate.
Logic of handling shapekeys when entering and leaving edit mode for
curves was... utterly broken.
Was leaving actual curve data with edited shapekey applied to it.
test_evaluation_visibility_a failed before, but it is now fixed
test_evaluation_visibility_b passed before and was used as control to make sure it was not broken
Based on D2371 from @ryry. Mostly T49043, a little T49042.
Deleted some unused drawing code instead of updating it.
I have a few more things in mind for this file... tomorrow!
The release of these arrays should be the programmer's discretion since these arrays can continue to be used.
Only the expanded functions `bvhtree_from_mesh_edges_ex` and `bvhtree_from_mesh_looptri_ex` are currently being used in blender (in mesh_remap.c), and from what I could to analyze, these changes can prevent a crash.
[Note: this patch functionality was implemented in parallel, independently at bf83f097ad
That said, the original patch was also removing an unnecessary include,
so here it is]
Replaced the one call to `glutil_draw_lined_arc`.
This seems to be the only draw call in this file.
Reviewers: merwin
Tags: #bf_blender_2.8
Maniphest Tasks: T49043
Differential Revision: https://developer.blender.org/D2497
This is a more complex approach, which makes me really want to use
IDProperty instead (assuming we handle their merging nicely).
In fact I would expect this to happen in readfile.c, not during
doversion, but I can revert this later.
For now this allow for demo files saved with 2.8 to keep working even
after we add/remove engine settings properties.
The values are merged. There is no purge though, so
old CollectionEngineSettings and CollectionEngineProperty will live
forever (for the time being).
This prevents crashes when a file was saved with 2.8, but a new
engine settings property was created.
In those cases any previous collection settings are wiped out. We can do
an elegant merge soon.
A group of object groups can be formed by means of the dupli_group option in
the Object properties window. The present revision extends the Selection by
Group option in the Freestyle Line Set so as to support not only flat object
groups but also nested groups.
We need to first split all vertices before we can reliably
check whether edge can be reused or not.
There is still known issue happening with a edge-fan mesh
with some faces being on the same plane.
Other than the general conversion:
* Made some slight aesthetic improvements.
** Removed gradients.
** Replaced stipples with transparency for hidden strips.
** Made strip borders less harsh.
** Removed stripes from offsets and made them brighter.
* Made only the visible parts of waveforms be drawn.
* Fixed a few drawing bugs.
** Background was not being drawn when buffer is NULL, and no
grease pencil is being drawn.
** Offset drawing ignored strip visibility.
Also, note that diagonal stripes for locked and error strips, are still
being drawn with the old api, as they await a new shader in order to
be converted.
Part of 49043
`SEQUENCER_OT_slip` was calling `draw_sequence_extensions` to redraw the
extensions during modal operation, but that is redundant, as it is
already called by the regular draw loop. Because it was called on top of
the draw loop, it was actually obscuring other parts of the strip that
would normally be drawn on top of it.
Somewhat part of 49043
This commit adds a way to debug Cycles motion blur issues which
are usually happening due to something crazy happening in between
of frames. Biggest trouble was that artists had no clue about
what's happening in subframes before they render. This is at
least inefficient workflow when dealing with motion blur shots
with complex animation.
Now there is an option in Time Line Editor which could be found
in View -> Show Subframe. This option will expose current frame
with it's subframe to the time line editor header and it'll allow
scrubbing with a subframe precision in time line editor.
Please note that none of the tools in Blender are aware of
subframe, so they'll likely be using current integer frame still.
This is something we don't consider a bug for now, the whole
purpose for now is to give a tool for investigation. Eventually
we'll likely tweak all tools to be aware of subframe.
Hopefully now we can finish the movie here in the studio..
Now new edges will be properly created between original and
new split vertices.
Now topology is correct, but shading is still not quite in
some special cases.
The idea is to iterator over the active layer of the current scene and
then over the active layer of the set scenes.
In the future, once we get workspace we will get the initial renderlayer
from context, while the background sets will still use their active
renderlayer.
Doesn't currently change anything, but would need for some future
work here.
It uses existing padding in kernel BVH structure, so there is
nothing changed memory-wise.
The issue here was mainly coming from minimal pixel width feature
which is quite commonly enabled in production shots.
This feature will use some probabilistic heuristic in the curve
intersection function to check whether we need to return intersection
or not. This probability is calculated for every intersection check.
Now, when we use multiple BVH nodes for curve primitives we increase
probability of that primitive to be considered a good intersection
for us. This is similar to increasing minimal width of curve.
What is worst here is that change in the intersection probability
fully depends on exact layout of BVH, meaning probability might
change differently depending on a view angle, the way how builder
binned the primitives and such. This makes it impossible to do
simple check like dividing probability by number of BVH steps.
Other solution might have been to split BVH into fully independent
trees, but that will increase memory usage of all the static
objects in the scenes, which is also not something desirable.
For now used most simple but robust approach: store BVH primitives
time and test it in curve intersection functions. This solves the
regression, but has two downsides:
- Uses more memory.
which isn't surprising, and ANY solution to this problem will
use more memory.
What we still have to do is to avoid this memory increase for
cases when we don't use BVH motion steps.
- Reduces number of maximum available textures on pre-kepler cards.
There is not much we can do here, hardware gets old but we need
to move forward on more modern hardware..
The change was delivering broken topology for certain cases.
The assumption that new edge only connects new vertices was
wrong.
Reverting to a commit which was giving correct render results
but was using more memory.
This reverts commit af1e48e8ab.
Some tests may break Blender, which makes the entire unittest routine to fail.
They are now I isolate the tests into individual files
Kudos to Sybren Stüvel and Sergey Sharybin for the advice.
Note: at the moment test_link.py is failing (since a41bbfb7)
Instead of bothering with matrix transformations, I am simply adding the jitter to the vertices.
Related to: rB31a21135cf72c8623be7f5aee2bfdac983ceae2e
The bug T46099 no longer applies since the addition of `dist_squared_to_projected_aabb_simple`
Has also been added comments that relates to an occlusion bug with the ruler. I'll investigate this.
Note: This makes the jittering to not work :/
@merwin, would you know how to use gpuMatrixBegin2D for this case? I
think it must be the reason behind the lack of jittering. But I couldn't
get it to work (the 2D shader is asking for a 3D Matrix).
Part of T49043
When importing an Alembic file with grouped transforms, it would badly name the transforms, taking the name of the parent instead of its own.
Patch by @maxime.robinot
Differential Revision: https://developer.blender.org/D2507
All (except for stick and wire) are now rendered using batch or immediate API.
Using simple front facing lighting for now, waiting for gpu_basic_shader.c to be recoded.
This involved some refactoring. The original code was relying on a cache of pre-allocated arrays which in turn were still re-populated every redraw loop.
We now ditch those arrays, and make the draw "on the fly".
Part of T49043
@merwin can you just check if you foresee any performance impact with this approach?
Quite simple fix for now which only deals with this case. Maybe we want to do
some "clipping" on image load time so regular textures wouldn't give NaN as
well.
This allows us to use faster math and still have reliable
isnan/isfinite tests.
Only do it for host side, kernels stays unchanged.
Thanks Lukas Stockner for the tip!
How to test this drawing: create and edit a curve and press shift + drag your mouse (or tablet). The Curve needs a Bevel Depth > 0.0.
Note: The ideal solution would be to use a different shader, that takes
no lighting. However according to Clément Foucault there is an assert preventing the same batch to
me used with different attributes (or something like that). Il wait
until the end of such resolution before revisiting this. That said, it
is working fine.
Part of T49043
Differential Revision: https://developer.blender.org/D2501
Note: we are now merging all the collection engines (mode and render), which eventually may get slow. But as stated before, this is to expose the functionality, while waiting for proper depsgraph integrated solution.
I didn't manage to get the proper object context in the collection
properties editor. That said I got it working for now in a temporary way
since this will change once we get workspaces anyways
(see changes in buttons_context.c and
rna_scene.c::rna_LayerCollection_mode_settings_get)
I still need to handle the merging of the settings. I will find a
provisory solution while we wait for depsgraph.
(also layer_collection_create_mode_settings_object and layer_collection_create_mode_settings_edit could probably be elsewhere - under draw/engines likely)
The issue was caused by usage of non-initialized image user, which
could have different settings, causing some random image being loaded
or not loaded at all.
This caused non-deterministic behavior of Cycles image loading because
it was querying image information from several places.
This fixes crash reported in T50616, but it's not a complete fix
because preview rendering in material is wrong (same wrong as in
2.78a release).
Also some conversion to new imm mode (T49043).
Multiple editors affected.
We could push this even further & draw all keyframes in an editor with a single draw call.
Something is strange with keyframe markers in blender2.8 -- they're not showing up before or after this commit. They do appear in master. This commit probably needs some follow-up work after keyframes are showing again. Better to share this code now instead of sitting on it.
Now we can draw keyframe markers as point sprites, with fewer draw calls and state changes.
Based on the builtin shader for round points with anti-aliased outline. This one is more pointy.
Changed the color function to output a float[4] to use with shaders.
Temporary leaving the UI_ThemeColor/glColor function.
Porting draw_bone_octahedral (wire) to Batch API.
ui_draw_but_TRACKPREVIEW
Changed stipple shader usage to a bunch of GL_LINES with flat color
OpenGL immediate mode: interface_draw.c (end)
Eradicate leftover legacy functions.
Fix a crash with histogram resolution.
coding style, float literals
Converted some (expr / literal) to (expr * literal) where it does not impact readability.
Simplified a few function calls.
ui_draw_but_COLORBAND
Introduced a new checker shader to be used mostly on transparent areas.
OpenGL immediate mode: interface_draw.c (cont)
ui_draw_but_UNITVEC
Introduced a new shader to be used for simple lighting.
Some `UI_ThemeColor` calls were left in converted files, in some cases
because they were just overlooked, and in the case of text drawing,
because the new BLF color functions were not yet implemented at the
time of conversion.
Also converted one `drawcircball` call that was left in
transform_constraints.c
Part of T49043
Still has one `UI_ThemeColor` call, because drawing is happening in a DM
drawing callback which hasn't been converted yet.
Also has some old gl calls that are #ifdef'ed out.
This also changes active face drawing in UVs from stippled to a solid
color, which makes active faces much more visible, and also looks nicer.
The same should probably be done for active face drawing in the 3d view.
(has been discussed with merwin on IRC)
Part of T49043
We now assert that we now file version of libraries (needed for
do_version after linking step), so for missing libraries, set dummy
numbers (using version of main .blend file actually).
Works similar to regular Cycles tests, just does OpenGL render to
get output image.
Seems to work fine with the only funny effect: Blender window will
pop up for each of the tests. This is current limitation of our
OpenGL context. Might be changed in the future.
The parent_selected() function mixed semantics of "needs to be exported"
and "is selected", which is confusing. Now just selected objects are
exported to Alembic; any parent transforms that are required were already
taken care of by other code.
Things like selection outlines didn't work at all.
Caused by rBc973e8d2da5cf3f.
When splitting up bitflags, the equivalent to `foo->flag & (bar1 + bar2)` is
`(foo->flag1 & bar1) || (foo->flag2 & bar2)`, *not*
`(foo->flag1 & bar1) && (foo->flag2 & bar2)`.
Also, let's please avoid using '+' operator for bitwise operations, a
binary addition is a binary OR *with* cary, which can cause quite some damage.
Fix outliner related crashes. Basically in some functions bContext was
not passed around, so CTX_data_scene_layer(C) was crashing.
Right now we still rely on ob->flag SELECT in some places. In order to use the base flag we will need to bring back the Bases to the outliner.
Adds a new outliner display mode "Collections" which draws the active
collection. We might want to rename it to "Active Collection" if we don't
plan to support showing other collections there.
Also added the buttons for restricting visibility and selectability.
@dfelinto, code in restrictbutton_collection_hide_cb and
restrictbutton_collection_hide_select_cb is duplicated from
rna_LayerCollection_hide_update and
rna_LayerCollection_hide_select_update, maybe utility functions would be
handy for this?
(and replace more instances of BaseLegacy/scene->base with Base/sl->object_bases)
Still need mouse selection, box selection, and menu selection
Also, there is still a problem with BA_WAS_SEL, at the moment only the
objects centers are highlighted.
Seems CUDA failed to de-duplicate the array across multiple inlined
versions of the shadow_blocked(). Helped it a bit with that now.
Gives about 100MB memory improvement on a scenes after previous
commit and brings up memory "regression" to only 100MB comparing to
the master branch now.
This commit enables record-all behavior of transparent shadows
rays.
Render times difference goes as following:
GTX 1080 render time
BMW -0.5%
Fishy Cat -0.0%
Pabellon Barcelona -11.6%
Classroom +1.2%
Koro -58.6%
Kernel will now use some extra VRAM memory to store the intersection
array (200MB on my configuration). This we can optimize out with some
further commits.
The idea is to record all possible transparent intersections when
shooting transparent ray on GPU (similar to what we were doing on
CPU already).
This avoids need of doing whole ray-to-scene intersections queries
for each intersection and speeds up a lot cases like transparent
hair in the cost of extra memory.
This commit is a base ground for now and this feature is kept
disabled for until some further tweaks.
Now we break the traversal cycle and then perform volume attenuation
and check with zero throughput. Not sure it makes any measurable sense
at this moment, but in the future it might help de-duplicating some
extra logic here.
Removed unnecessary call to DM_update_tessface_data(). This call is
already performed by DM_ensure_tessface(dm). The call being performed
twice caused a failing BLI_assert().
Reviewed by: Kévin Dietrich
With the new names the arguments (yup, zup) are in the same order as
they appear in the function name. The old names used copy_src_dst(dst,
src), which I found very confusing. Furthermore, now it is clear from
where to where the copy is made.
This makes the function names a little bit longer, though. If that is
a real issue, we can just name them zup_from_yup(zup, yup).
Reviewed by: Kévin Dietrich
To bring this fix a step further we need to address all the BA_WAS_SEL instances, and make sure they follow the new design.
This commit allow you to see the object selected (its center anyways) when you do select all.
Note: in the clay engine selection (a) was already working fine.
Speedup is mainly gained by multi-threading. Gives about 3x
fps gain on an edit shot file.
There is still some room for improvements, will happen in one
of the upcoming commits.
Every time:
* A collection settings is set
* A collection visibility changes
* An object is added/removed/ ...
We need to recalculate the "accumulated" CollectionEngineSettings that
the render engine should use for an object.
This is to be handled by the depsgraph. Meanwhile this code should allow
us to start using those settings in the render engines.
Note: We are storing this in the objects, which means we can only have
one active calculated option every time.
This is intended to get the conversation with the Depsgraph department
going.
Derived mesh for particles did not include tessellated faces when it
was expected to. Now added explicit function to copy CDDM with tess
faces without need to re-tessellate the result.
Initial work by Clément Foucault with contributions from Dalai Felinto
(mainly per-collection engine settings logic, and depsgraph iterator placeholder).
This makes Blender require OpenGL 3.3. Which means Intel graphic card
and OSX will break. Disable CLAY_ENGINE in CMake in those cases.
This is a prototype render engine intended to help the design of real
render engines. This is mainly an engine with enphasis in matcap and
ambient occlusion.
Implemented Features
--------------------
* Clay Render Engine, following the new API, to be used as reference for
future engines
* A more complete Matcap customization with more options
* Per-Collection render engine settings
* New Ground Truth AO - not enabled
Missing Features
----------------
* Finish object edit mode
- Fix shaders to use new matrix
- Fix artifacts when edge does off screen
- Fix depth issue
- Selection sillhouette
- Mesh wires
- Use mesh normals (for higher quality matcap)
- Non-Mesh objects drawing
- Widget drawing
- Performance issues
* Finish mesh edit mode
- Derived-Mesh-less edit mode API (mesh_rende.c)
* General edit mode
- Per-collection edit mode settings
* General engines
- Per-collection engine settings
(they are their, but they still need to be flushed by depsgraph, and
used by the drawing code)
Design Documents
----------------
* https://wiki.blender.org/index.php/Dev:2.8/Source/Layers
* https://wiki.blender.org/index.php/Dev:2.8/Source/DataDesignRevised
User Commit Log
---------------
* New Layer and Collection system to replace render layers and viewport layers.
* A layer is a set of collections of objects (and their drawing options) required for specific tasks.
* A collection is a set of objects, equivalent of the old layers in Blender. A collection can be shared across multiple layers.
* All Scenes have a master collection that all other collections are children of.
* New collection "context" tab (in Properties Editor)
* New temporary viewport "collections" panel to control per-collection
visibility
Missing User Features
---------------------
* Collection "Filter"
Option to add objects based on their names
* Collection Manager operators
The existing buttons are placeholders
* Collection Manager drawing
The editor main region is empty
* Collection Override
* Per-Collection engine settings
This will come as a separate commit, as part of the clay-engine branch
Dev Commit Log
--------------
* New DNA file (DNA_layer_types.h) with the new structs
We are replacing Base by a new extended Base while keeping it backward
compatible with some legacy settings (i.e., lay, flag_legacy).
Renamed all Base to BaseLegacy to make it clear the areas of code that
still need to be converted
Note: manual changes were required on - deg_builder_nodes.h, rna_object.c, KX_Light.cpp
* Unittesting for main syncronization requirements
- read, write, add/copy/remove objects, copy scene, collection
link/unlinking, context)
* New Editor: Collection Manager
Based on patch by Julian Eisel
This is extracted from the layer-manager branch. With the following changes:
- Renamed references of layer manager to collections manager
- I doesn't include the editors/space_collections/ draw and util files
- The drawing code itself will be implemented separately by Julian
* Base / Object:
A little note about them. Original Blender code would try to keep them
in sync through the code, juggling flags back and forth. This will now
be handled by Depsgraph, keeping Object and Bases more separated
throughout the non-rendering code.
Scene.base is being cleared in doversion, and the old viewport drawing
code was poorly converted to use the new bases while the new viewport
code doesn't get merged and replace the old one.
Python API Changes
------------------
```
- scene.layers
+ # no longer exists
- scene.objects
+ scene.scene_layers.active.objects
- scene.objects.active
+ scene.render_layers.active.objects.active
- bpy.context.scene.objects.link()
+ bpy.context.scene_collection.objects.link()
- bpy_extras.object_utils.object_data_add(context, obdata, operator=None, use_active_layer=True, name=None)
+ bpy_extras.object_utils.object_data_add(context, obdata, operator=None, name=None)
- bpy.context.object.select
+ bpy.context.object.select = True
+ bpy.context.object.select = False
+ bpy.context.object.select_get()
+ bpy.context.object.select_set(action='SELECT')
+ bpy.context.object.select_set(action='DESELECT')
-AddObjectHelper.layers
+ # no longer exists
```
Marks matrix state as dirty so shader will use the latest values from glScale, glTranslate, etc.
We'll remove this after transitioning 100% to the new matrix API, which handles this sort of thing automatically.
Part of T49450
It is not necessary to add MOTO library dependency when we use
WITH_IK_SOLVER (now it uses Eigen) or we use WITH_MOD_BOOLEAN (it was
used by bsp intern library some time ago but it is not present in the
code anymore).
Reviewers: mont29, sergey
Subscribers: mont29, sergey
Differential Revision: https://developer.blender.org/D2477
No reason to not make this private to this file, and it gave conflict
when using bpy as module and loading it in a GLib application (which
also has a g_atexit var).
The title says it all actually. Use BLI task to loop over vertices
and distort their locations. Gives 2x FPS increase in a file with
just time-dependent displace modifier on my desktop.
This version will give less spin locks and now well-tested by render engines.
This should reduce amount of threading overhead when having multiple objects
with displace modifier enabled.
In the future this will also help us threading the modifier.
There are more modifiers which could benefit from this, but let's first
investigate the new behavior with one of them.
We (the Microsoft C++ team) use the Blender project as part of our "Real world code" tests.
I noticed a place in WIN32 specific code (dvpapi.cpp:85) where a string literal is losing
its const-ness when being passed to BLI_dynlib_open(). This is not permitted when using the
/permissive- conformance compiler switch (see our blog
https://blogs.msdn.microsoft.com/vcblog/2016/11/16/permissive-switch/)
My suggested fix is to add const and propagate it where needed. Another possible fix would be
to explicitly cast away the const.
Reviewers: mont29, sergey, LazyDodo
Subscribers: Blendify, sergey, mont29, LazyDodo
Tags: #platform:_windows
Differential Revision: https://developer.blender.org/D2495
We had two versions of several BLF functions -- one for a specific font ID & another for the default font.
New BLF_default function lets us simplify this API & delete the redundant code.
There are still many places to fix. I'll miss the bright yellow!
This commit also uses the new BLF_default function where possible.
Part of T49043 since we call glColor less often.
Minimum target is still 3.3
On AMD pro driver, asking for a 3.3 context gives us *exactly* 3.3, not 3.3+ as desired.
Have not tested proprietary NV or Intel drivers, but this fix should work on all vendors.
Instead of reference the vertex first and test the bitmap afterwards. Test the bitmap first and reference the vertex after.
In a mesh with 31146 vertices and the entire bitmap disabled, the loop time is 243% faster
With all bitmap enabled, the time becomes 463473% faster!!!
One possible reason for this huge difference in peformance is that maybe the compiler is not putting the function "BM_vert_at_index" inline (I dont know if buildbot do this, but it's good to investigate).
Looks like `object_map` and `mem_arena` may be NULL sometimes...
Also, cleaned up function pointers declaration of Nearest2dUserData,
those were warning out in gcc. Please, *always* use typdef defined
prototypes for function pointers, it is sooooo much cleaner and clearer
that way. And easy to convert from compatible functions too.
BKE_lamp_free was somehow missing the refactor of datablocks handling
(which, among other things, completely separated ID refcounting and
linking management from ID freeing itself).
Either forgot during development, or lost during merge...
The code looks for the closest element between its centers. In the case of islands, the center of each vertex is the center of the island.
The solution here is to skip the search for islands when the operation is translation
-Remove NPOT check as it should be supported by default with OGL 3.3
-All custom texture creation follow the same path now
-Now explicit texture format is required when creating a custom texture (Non RGBA8)
-Support for arrays of textures
Reviewers: dfelinto, merwin
Differential Revision: https://developer.blender.org/D2452
Pretty straight forward actually, just do not bother about obdata part
of vgroups in that case, only copy object part of it.
And let's curse once again those stuff spread accross several types of
data-blocks...
The order was wrong from the semantic point of view, caused
by some legacy workarounds in Libmv. Didn't realize it's was
not how things were expected to be used.
Issue was indeed in join operation, mesh in which we join all others
could be re-added to final data after others, leading to undesired
re-ordering of CD layers, and existing vertices etc. being shifted away
from their original indices, etc.
All kind of more or less bad and undesired changes, fixed by always
re-inserting destination mesh first.
Also cleaned up a bit that code, it was doing some rather
non-recommanded things (like allocating zero-sized mem, doing own
coocking to remove a data-block from main, etc.).
Tricky issue caused by CDDM_copy() coying MFACE array but not MTFACE which
confused logic later on.
Now we don't copy ANY tessellation unless it is requested to.
Thanks Bastien for help and review!
blenderplayer uses BLF but not Editor UI, so we got a link error for the missing UI_GetThemeColor function.
Moved the new function from BLF to UI.
@Blendify reported problem in IRC
For anything fancier than regular Theme colors (shading, alpha, etc.) do this:
unsigned char color[4]
UI_GetThemeColor[Fancy]4ubv(... color)
BLF_color4ubv(fontid, color)
That way the BLF color API stays simple.
'page' prop of scroll up/down operators would get stuck once set once by
pageup/down keys... Now only take this prop into account if explicitely
set, not when its value is inherited from previous run.
Strangely this change does not affect the performance very much.
Suzanne subdividide 6x (ortho view):
Before:0.00013983
After :0.00013920
But it makes it easier to read the code
When the function that tests snap on multiple elements starts from the face and ends at the vertex, the transition between elements becomes much smoother.
Better to have clear way to tell whether flag is parameter for
BKE_library_foreach_ID_link(), parameter for its callback function, or
return value from this callback function.
This code was already disabled.
We might be able to simplify GPU_check_scaled_image even further. Maybe even delete it? Just removing the obvious stuff now.
Keeping is_power_of_2_resolution function since it is still used in DXT logic.
Taking advantage of the area, the depth is decreased 0.01 BU to each loop to give priority to elements in order: Vertice > Edge > Face. This increases the threshold and improves the snap to multiple elements
The previous solution took arbitrary values to determine if the mouse was near or not to the Bound Box (it simply scaled the Bound Box).
Now the same function that detected the distance from the BVHTree nodes to the mouse is used in the Bound Box
This revision extends the functionality of the "Fill Range by Selection" button in
the "Distance from Camera/Object" modifiers so that only selected mesh vertices
in the edit mode are taken into account (instead of considering all vertices when
in the object mode) to compute the min & max distances from the reference.
This will give users much finer control on the range values.
Use new Main->relations ID usages mapping in BKE_library_make_local().
This allows a noticeable simplification in code, and can be up to twice
quicker as previous code (Make Local: All from 2 to 1 minute e.g. in a
huge production file with thousands of linked data-blocks).
Note that new code has been successfuly tested with several complex cases
(production files from Agent327), as well as some testcases from recent
bug reports related to that function. But as always, nothing beats real
usage by real users, so please check this before we release 2.79. ;)
Main areas that would be affected: Make Local operations (L shortcut in
3DView), and append from libraries.
Use Main->relations in BKE_library_foreach_ID_link(), when possible
(i.e. IDWALK_READONLY is set), and if the data is available of course.
This is quite minor optimization, no sensible improvements are expected,
but does not hurt either to avoid potentially tens of looping over e.g.
objects constraints and modifiers, or heap of drivers...
The new MainIDRelations stores two mappings, one from ID users to ID
used, the other vice-versa.
That data is assumed to be short-living runtime, code creating it is
responsible to clear it asap. It will be much useful in places where we
handle relations between IDs for a lot of them at once.
Note: This commit is not fully functional, that is, the infamous, ugly,
PoS non-ID nodetrees will not be handled correctly when building relations.
Fix needed here is a bit noisy, so will be done in next own commit.
This provides a slight improvement in performance in specific cases, such as when the observer is inside a high poly object and executes snap to edge or vertex
Distance calculation performed by the "Fill Range by Selection" button of the
"Distance from Camera" color, alpha and thickness modifiers was incorrect,
limiting the usefulness of the functionality.
The problem was that the distance between the camera and individual vertex
locations was calculated in the world space, which was inconsistent with the
distance calculation done by the modifiers in the camera space.
The new `isect_ray_aabb_v3_simple` function replaces the `BKE_boundbox_ray_hit_check` and can be used in BVHTree Root (first AABB). So it is much more efficient.
In order to simplify the reading of these functions, the parameters: `snap_to`, `mval`, `ray_start`, `ray_dir`, `view_proj` and `depth_range` are now stored in the struct `SnapData`
Checking only whether mverts is same as base mesh one is not enough in
all cases, some modifiers (deform ones) can only generate new mvert
data, while keeping others from original mesh.
Now checking both mvert or medge, hopefully this will be enough to catch
all problematic cases this time.
Thanks @gaia for finding that problem. :)
Although the "BLI_bvhtree_find_nearest_to_ray" function is more practical than the generic "BLI_bvhtree_walk_dfs", it does not work to snap in perspective view. This makes it necessary to add "ifs" and functions that make the code difficult to understand
patch: D2474
This is a speed up option which is mainly useful for viewport. Gives nice speedup in
the barbershop scene of 2x when replacing GI with AO after 2nd bounce without loosing
too much details.
Reviewers: brecht
Subscribers: eyecandy, venomgfx
Differential Revision: https://developer.blender.org/D2383
We are not bumping file version, but we cannot have the doversion code running twice.
In this particular case it was crashing files, since we were setting node->storage to NULL, and later on accessing it.
At least for initial merge we don't need anything in DNA (and I'm not sure how DNA structs are handled which are not explicitely saved to disk... does it still store some info about it?).
Had to do some changes to allow this, basically the wmManipulator.mgroup pointer to get the manipulator-group from the manipulator is back which is a bit ugly. Think that's fine though, only a minor annoyance.
Updating and drawing works closer together now, makes it easier to define what to draw (3D vs 2D vs depth-culled manipulators) and to avoid unnecessary updates. Totally untested :)
Instead of having multiple manipulator-maps for 2D and 3D manipulators, manipulator-group-types should be tagged as being either 2D or 3D. When it comes to drawing these can be drawn separately.
Note that more work needs to be done here (added some todo marks), but there are probably going to be quite a few issues when merging into custom_manipulators, so will handle these first.
Will continue work on them in 2.8 based custom_manipulators branch, there I can also work on making manipulator drawing based on OpenGL 3.2 core profile.
This commit adds a new 'custom-manipulators' branch in which changes of the wiggly-widgets branch are applied onto the blender2.8 branch. I've done it so I can start porting manipulator drawing code to use the new abstractions for OpenGL 3.2 core profile.
At some point I had to do it anyway - better earlier than later to reduce loss of git history.
From now on this is the main branch for the custom manipulators project, I'll delete the wiggly-widgets branch in a bit (but keep it available at https://github.com/julianeisel/blender/tree/wiggly-widgets). It's still possible to merge some manipulators for pre-2.8 by rewriting the drawing of those to use OpenGL <= 2.1 (or copy & pasting from earlier state).
COMMAND${CMAKE_COMMAND}-Ecopy"${PYTHON_OUTPUTDIR}/python${PYTHON_SHORT_VERSION_NO_DOTS}${PYTHON_POSTFIX}.lib"${BUILD_DIR}/python/src/external_python/run/libs/python${PYTHON_SHORT_VERSION_NO_DOTS}.lib#missing postfix on purpose, distutils is not expecting it
COMMAND${CMAKE_COMMAND}-Ecopy"${PYTHON_OUTPUTDIR}/python${PYTHON_SHORT_VERSION_NO_DOTS}${PYTHON_POSTFIX}.lib"${BUILD_DIR}/python/src/external_python/run/libs/python${PYTHON_SHORT_VERSION_NO_DOTS}${PYTHON_POSTFIX}.lib#other things like numpy still want it though.
Some files were not shown because too many files have changed in this diff
Show More
Reference in New Issue
Block a user
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
