Fix broken behavior in case of texcoords mapping (we do need texcoords
node in all cases, then, even for UV coords...).
Use nodes by default when generating new write-allowed wrapper around a
material.
Do not try to find nodes in existing tree all the time, do it only once,
even for lazy-initialized nodes (through accessors).
Fix ugly spacing in property accessors (since it looks like some people
do not like a single 'block' with both getters, setters and prop
definition, at least use one sep line everywhere (and two sep lines to
separate properties)...
New iterator CTX_DATA_BEGIN_FOR_ID, to restrict the loop to the specified object only.
This is not super efficient, but it should be fine for now.
I will talk to other developers. A more elegant solution would be to
have something like "active_object_selected_pose_bones" in the context.
This reverts commit 178530cc50e2fb4bd3faf4fff22392a58688bed1.
Those operators are called from the UI, which only takes into
consideration the active object for the parameters and whether to call
reset, update or calculate.
We can re-revert if needs be.
* POSE_OT_paths_clear
* POSE_OT_paths_calculate
* POSE_OT_paths_update
Despite my personal opinion on the matter, those operators were listed
as to be converted (see T54650).
They are called from the UI, where you only see the parameters for the
active object/armature.
I will commit and revert soon after, so we can quickly bring it back
once we re-visit this design.
The automatic highlighting of constant curve areas was checking that
the bezier handles are horizontal even if a non-bezier interpolation
mode was active. Conversely, it was highlighting based on just handles
with Elastic interpolation, which always generates movement.
This reverts commits:
* 2a2858b30d
* 7cf8eed5e0
Similar to ARMATURE_OT_layers_show we should keep this operator
single-object oriented. This one is a bit more tricky since we
may want to quickly see all the layers of all the armatures.
I will check with animators what is the best way to proceed here.
But overall I think it makes sense to have this only for the active
object.
This reverts commits:
* 29a281f9ef
* 3a8b56ce24
This operator shouldn't behave multi-objects. This would only work
if all the selected objects had the same logic for their layers,
which may be likely for characters, but it won't be for mixing props and
characters.
Using CTX_DATA_BEGIN_WITH_ID here.
Unlike the edit mode counter-part this operator is well served with the
macro above. ARMATURE_OT_autoside_names needs to treat mirrored bones
separately, while for pose we don't handle those cases.
Be ware that using this macro makes the code smaller, however it also
tags every single (selected) pose to update, regardless of whether we
changed the names of the bones.
In this case it is fine since we most likely renamed all the bones.
But In other cases I'm still a bit wary of using CTX_DATA_BEGIN_WITH_ID
instead of BKE_view_layer_array_from_objects_in_mode_unique_data.
To be seen in upcoming commits. Stay tuned.
This operator doesn't need any changes. Following the guideline of
multi-objects behaving alike joined armatures, there is no need
to change any armature that is not the active one.
That said I will revert the behaviour of ARMATURE_OT_SELECT_hierarchy to
follow the same rule (i.e., revert dcc623e7e7).
This decouple the vertex display from the face+edges.
This is to reduce the number of triangles required to fix the edges
artifacts (aliasing) and increase viewport reactivity when not actively
navigating (ie. mouse scroll).
Also it makes all vertices visible (not cut-off) even when navigating.
However it makes the navigation drawing a bit slower because it has to
render twice.
Also add a depth bias to the wires to avoid depth fighting when previewing
final mesh (modifiers applied).
This makes is easy to create nested drawcalls that will inherit all the
parents properties. This is usefull if only a few uniforms changes for that
drawcall.
Best to give its actual name so it's clear that the Principled BSDF does
not have a diffuse color exactly, and does not have an alpha component.
Also image textures use UVs by default, so avoid creating a texture
coordinate node for that.
This is to solve an issue where a blend file could be compressed
unbeknownst to the artist. This happened in the following situtation:
- Artist edits an uncompressed blend file.
- Some script saves a compressed blendfile to a separate location.
- When the artist saves the file (s)he is editing (File>Save, or Ctrl+S),
it was silently compressed.
Before, the list kept the reverse order enabled in the filter. Now the filter is reset when the option is disabled and don't need any user change or restart.
Now, it was possible to invert the order of the UIlist using the filter, but it was impossible to know if the list was inverted or not.
The problem with this is that any other element depending of this value could not be adjusted.
See https://devtalk.blender.org/t/how-to-access-uilist-properties/2268
The new parameter allows to set the reverse order by default. When the list is set as reverse, it cannot be inverted again, so the invert button is removed of the filter.
This change is needed to fix a requested feature for Grease Pencil (T56985) and because a lot of 2D softwares use the drawing layers in the inverse order used in Blender.
This option defines whether or not menu items drawing code must
separate drawstring in two, keeping right part at all cost.
This is used to show shortcuts of operators' enum entries usually.
Previous way to decide that was based on button having an RNA prop
pointer, assuming those without it were not 'data buttons' and hence
needed the shortcut special handling thingy.
That's wrong is many, many cases (especially since search templates
often generate more than one buttons, not all linked to actual RNA
data). So instead now checking whether a button has an optype set or
not, hopefully this will be much more accurate...
Other solution if thsi also fails, is to add new flag to buttons, and
explicitely set it when needed, instead of trying to guesstimate...
Encountered on Nvidia + Linux, it seems that doing everything all at once
can make the driver give up the whole command list and return nothing as
the output of the render.
Loop triangles are tessellated triangles create from polygons, for renderers
or exporters that need to match Blender's polygon tesselation exactly. These
are a read-only runtime cache.
Tessfaces are a legacy data structure from before Blender supported n-gons,
and were already mostly removed from the C code.
Details on porting code to loop triangles is in the release notes.
Differential Revision: https://developer.blender.org/D3539
- mesh_calc_modifiers & editbmesh_calc_modifiers
now follow similar naming.
- me and mesh were too easily confused in mesh_calc_modifiers
(remove 'me', access ob->data).
There is no active bone data in the cowed edit armature.
This was introduced on rBe12df10120a0, back when we hoped the depsgraph
would fullfill all our hopes and dreams.
Before the single-context-editing depsgraph reality came crashing on us.
This only worked on new objects, which is no longer needed.
It didn't account for:
- Quaternion or axis-angle rotation.
- Parenting.
- Constraints.
If we support object rotation alignment, it might be best to
make it a more general set of alignment operators
(align transform, orientation - not limiting to view).
This property (even in 2.7) was not working. In fact it behaves as extended
when extended was FALSE.
Besides all that, the operator is not affected my multi-objects, so it
is good to go.
Before when press . (view_select) the object was centered at the dummy, but now it's centered with the strokes bounding box size.
Also fixed some problems in edit mode when the object origin was not in view origin.
- new "Align to View" option when loading a new image
- automatically align to view when dropping an image into a viewport
- larger default size for image empties
- fix image empty gizmo in orthographic view
- new "Align Objects to View" operator
Reviewer: brecht
Differential: https://developer.blender.org/D3778
Since my temporary buffer commit (about a month ago), the OpenCL device was zeroing the wrong buffer, leading to
completely wrong filtered feature passes and therefore significantly lower-quality results than CPU and CUDA.
* Move all copy & paste operators into the menus. There was no real reason
why these particular operators should be in the header and not in the menus,
like all other operators
* Move ‘Update Automatically’ toggle from UV/Image Editor header into menu.
* Move the pin toggle next to the ID blocks, because it is related.
* Move OpenGL render from sequence header into View menu.
* Sequence editor display mode and channels are now not expanded.
For now we have categories collection, object, object data, modifiers &
constraints, and shading. The icons can be categorized by adding e.g.
DEF_ICON_OBJECT() in UI_icons.h.
Light themes will need to be updated to use darker colors to keep icons
visible in the outliner.
Now by default the UVs are calculated with a fixed size and this makes easier to add patterns for drawings like Manga. Before, the texture changed depending of the stroke size.
The appearance is a bit different than 2.79 where the flame was just added
on top of the smoke without correct blending.
Now it's much more realistic and using volumetric integration. You can see
the smoke actually masking the flame.
The other difference is that the flame color was not using proper color
managed blending. Now with the use of filmic it shows bright yellow.
This could be adjusted and displayed as a user parameter in the future.
Includes the following fixes
- Fix smoke texture creation: data was interpreted as Byte instead of Floats.
- Fix Velocity texture not being free after draw: also was causing crashes.
- Fix display_thickness not being copied during COW.
- Fix Blending and general volume rendering algorithm.
- Add Volume Shadowing support.
Add an option to display contour lines tracing through points with the
same interpolated weight value in weight paint mode. This can be useful
for working on gentle gradients over a relatively high resolution mesh,
where the difference in color between adjacent vertices is very small.
The contour grid has 3 levels of detail going down to step 0.001,
which automatically fade in or out based on the weight gradient.
Fade out works by capping both screen space and weight space line
width, and reducing alpha when the screen space width becomes too
small for moire and noise-less rendering.
Reviewers: fclem
Differential Revision: https://developer.blender.org/D3749
Using 'format' prefix made this read as if t was for string formatting.
Use 'PyC_StructFmt' prefix instead since these values are compatible
with formatting from Python's 'struct' module.
while it is possible to permanently flip the colors using the 'X' button, this makes it consistent with Texture Paint.
fixes T56994
Reviewers: campbellbarton, brecht
Reviewed By: brecht
Subscribers: JulienKaspar
Tags: #sculpting_and_painting
Maniphest Tasks: T56994
Differential Revision: https://developer.blender.org/D3753
It now works like 2.7x. That said this operator is kind of incompatible with
x-mirror (the way it is implemented anyways). But if people were happy with
this in 2.7x they will be multi-object happier.
That said, do turn off x-mirror before using this or select all bones, or just
stop using this operator altogether ;)
Jokes aside, the operator can also be fixed. It shouldn't be hard.
Also removed the insane insanity check. This would fail in the
poll function. No point in that level of paranoia.
And return OPERATOR_FINISHED regardless. Since this has a parameter
(though not used) you may get want to tweak REDO (in the future).
There were two issues here:
* cutsom_shape_transform was not properly tagged as not-owned pointer.
* custom_shape_transform and bbone_start/_end could be set to a pchan from
another data-block (not from UI, but RNA access code itself allowed it).
Those two issues were specificaly breaking complex rigs in static
override case.
EDIT: since yesterday, bbone_start/_end are mere accessors to edit_bone
data in RNA, so most of fix related to those was ditched (they still needed
to be 'deactivated' from RNA diffing with PROPOVERRIDE_NO_COMPARISON).
Both original handle types are based on location, and Absolute uses it
in a weird way: the Start handle uses the head, while End uses the tail.
This makes controlling the shape of the B-Bone via control bone rotation
really non-intuitive, especially if trying to add a single control for
the tangent in the middle of a B-Bone chain.
To remedy this, add a new custom handle type that uses the orientation
of the control bone, while completely ignoring location. It is even
possible to control both ends of one B-Bone with the same handle bone,
resulting in an S shape.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D3769
* Return OPERATOR_FINISHED only if at least one object succeeded.
* Have a per object fork value, otherwise an object may influence the next.
(which is why we should use `const foo = ...; foo_iter = foo;`)
The patch itself was fine (save for sending notifiers for objects instead of ob).
But I couldn't apply it, so I re-did from scratch.
Based on D3394 by @codemanx
Specifically, it should always use the position of the custom handle
bone head, even when affecting the handle at the tail of the main bone,
and shouldn't apply the special handling for joining two B-Bones.
This handle type was unusably broken before a bug fix included in
recent changes, so it should be safe to break backward compatibility.
The grid now can be configured by object because this helps to identify objects and allows to define diferent grid parameters for each objects.
Also added a color option.
The rest shape of B-Bones is actually affected by custom handles or
the default connected parent/child mechanism. Ignoring these effects
thus leads to the edit mode shape being different from the actual
rest pose.
This splits the b_bone_spline_setup function that is used to compute
the correct rest and pose shape from pose channels into two parts,
and applies the data structure independent half to edit mode.
In order to access the custom handle settings in Edit Mode, they are
moved to Bone and EditBone, while the bPoseChannel fields are downgraded
in status to a cache for performance. Also, instead of flags, introduce
an enum to specify the handle operation modes, so that new ones could
be added later.
Reviewers: aligorith, brecht
Differential Revision: https://developer.blender.org/D3588
Custom handle settings actually affect the B-Bone rest shape,
so they should be changed in Edit mode rather than Pose mode.
This is necessary to be able to display the correct rest shape
of the bone in Edit Mode.
Also, instead of flags, introduce an enum to specify the handle
operation modes, so that new ones could be added later.
Differential Revision: https://developer.blender.org/D3588
The changes are:
- The shader now is passed as a parameter of the batch `draw` method (batch.draw(shader)). Since the batch always has to set a shader before drawing;
- The batch methods to specify a value to a uniform have been removed. Uniforms are parameters of the program (here called shader). If you change a uniform, it changes in all batchs that use the same program;
- New methods were added to set uniforms by the shader;
- The `batch.program_set_builtin` was removed. It is a duplicate of `program_set` but without a shader object. We need the shader object to configure the uniform;
Differential Revision: https://developer.blender.org/D3752
Avoid accessing object mode via the view layer active object
when drawing since it's done everywhere we store the active object mode
in the draw-context.
The problem is that string width computing is performed in integers
(pixels), which can generate a rather annoying error (a few pixels)...
Simply work around that for now, by trimming an extra middle char when
needed.
`to_strlen` just benefits from using pre-computed kerning table for
ascii chars (gives about 30% speed improvements).
`to_rstrlen` was re-written and heavily simplified, basically using same
logic as `to_strlen`, and `BLI_str_find_prev_char_utf8()` to loop
backward in the string, instead of looping forwards the whole string,
storing each gliph's width in temp array, and looping backward on that
temp array to find final string matching expected width. Gives about 70%
speed improvements!
And both functions can now share their core logic.
With new jemalloc versions memory allocated by threads that then become
inactive is not longer automatically freed. Instead we have to enable a
background thread to do it.
Some testing is needed to find out of this is sufficient, because the
background thread only runs periodically.
This parameter allows to define the percentage of offset of a stroke when uses surface mode.
Before, this was a fixed value, but for some artists' purposes, it's good to have the option to change it.
Users can select the main unit they want to use now.
Previously the displayed unit always depended on the magnitude of the value.
The old behavior can be restored by switching to the "Adaptive" mode for length, mass and time units.
Meters, kilograms and seconds are the default units for new and old scenes.
The selected unit is also the default unit for user input.
E.g. if cm is selected, whenever the user inputs a unitless number into a field of type length, it will be interpreted as cm.
Reviewer: brecht
Differential: https://developer.blender.org/D3740
Currently some modes share tool keymaps, we might want to disable
this since it's confusing editing one thing in multiple places.
However this should be resolved in the tool definitions.
This includes all the modes that don't need multi-object support:
* SIMEDBONE_CHILDREN
* SIMEDBONE_CHILDREN_IMMEDIATE
* SIMEDBONE_SIBLINGS
* SIMEDBONE_GROUP
* SIMEDBONE_SHAPE
Unlike the mesh select similar operator, in this case we take
only the active bone in consideration, not all the selected ones.
I decided not to change that, and keep this 2.79 behaviour.
We could though if we want to.
It is crucial that scripts do not write, and even more importantly, do
not store any references to those. Otherwise, that’s a rather straight
path to crash.
This is quite an obscure option only useful in rare cases,
when enabled by accident it's confusing since single objects
can't be transformed.
Also, w/ 2.8x you don't see the option change in the header.
This allows gizmo groups to store properties in the tool.
This makes sense for gizmo options which only control gizmo display and
don't control operator execution.
Unlike similar kinds of properties,
this isn't accessible via the gizmo-group-type instance.
For now the it's only stored in the workspace tool as can be done for
operator properties, so each instance doesn't have different settings
which would be confusing from a user perspective and complicate access
from the top-bar.
Later we could add gizmo-group properties if needed.
Currently it's not showing the subdivided mesh (if there is a subdiv mod)
and there is some sync issue if there is multiple uv image space opened.
But thoses will be tackled later on. The purpose of this commit is to fix
the overflow issue of IMM and speed issue.
NOTE: This commit only concern edit UVs and not the "shadow" mesh displayed
when texture painting. This will be address in a future commit.
We now cache the uv mesh in the mesh batch cache and only reupload data on
changes.
Update could be more granular (and a bit faster) but it's not our main
concern ATM.
This should fix problem caused by the IMM api used to draw large meshes.
This makes performance skyrocket compared to previous implementation.
There is still a big CPU bottleneck when not in sync selection mode but it
is not related to the drawing function directly.
Allow raycasting in two directions and culling front or back faces.
Also implement a new Invert Cull option in both constraint and
modifier that can be used to aim for faces aligned with the project
axis direction when raycasting both ways.
Reviewers: mont29
Differential Revision: https://developer.blender.org/D3737
Was using 'selected_editable_bases', which used to save a lookup.
This is no longer the case and complicates access from Python
which cant yet easily access Bases.
Now it's possible to disable the lasso effect of the fill color while drawing with a fill color.
This sometimes is required to see the lines and it's the first step to implement a new lasso brush.
Now it's possible to define what material is used in the generated strokes and if the strokes are put in front of the original (default) or keep the original in front.
Before, the generated strokes have been always on top of the original because they were drawn later.
Following the vertice delete menu logic here, where delete needs no verb
and the special case (dissolve here) gets the full verb + noun)
Note: This operator wasn't even listed in the T54646 task. I suspect other
operators may share the same fate.
I'm storing the name of the vertex groups in a gset. This way
we can select vertex groups with the same name across different
objects.
Also this is the last select similar mode that needs porting o/
The crash was caused by BVH traversal stack being overflowed.
That overflow was caused by lots of false-positive intersections
for rays originating on a non-finite location.
Not sure why those rays will be existing in the first place,
this is to be investigated separately.
This commit moves pre-SSE4.1 check to a higher level function
and enables it for all miroarchitectures.
It's too easy to miss at the bottom of the screen, and if the status bar is
collapsed critical messages may not be seen at all. There might be a better
looking solution, but for now this solves the main usability problem.
Seems there is something wrong with ternary operator
and type deduction.
Copying matrix to a variable with cleaner name seems
to be better solution than to force cast to an array.
The option WITH_C11 is gone, and C++ is defaulting to C++11 now,
so guess it's fine to assume we need C11 now.
This is technically what we use anyway, with all the re-definitions
of structs (like when we typedef anonymous struct in a header file
first, and them define it to a proper structure in implementation
file).
That new bpy_extras' node_shader_utils module is remotely based on
existing addons' cycles_shader_compat module. It has some key
differences though:
- It is based on Principled shader, instead of the noodle combination
of half a dozen simpler shaders.
- It does not do any value conversion (like e.g. clamping, which was
only suited for FBX). Any conversion/adaptation is to be done
by each IO add-on.
- It extensively uses accessors to offer:
- Easy read/write of values, hiding all the hairy noodly nodes
(principled shader helps a lot here).
- lazy creation of most nodes on accessing (when generating a
material), which avoids getting unused nodes in final shader.
- We only use Principled BSDF so far though, which means some features
supported by previous system are not yet implemented in new one.
Note that, even though we support 'exporting' side of things, this will
only work for a very specific (and basic) kind of shader. This will have
to be clearly explained in end-user documentation.
Also, that code has had some basic testing, but most certainly needs a
lot more refinement.
As proof-of-concept, OBJ IO script will be updated to use that new
system after that commit.
I had to make the viewport navigation icons a bit smaller in the SVG since
the edges were being clipped off, we only support 16x16 icons currently.
They are a bit blurry because of this.
Move the weight paint drawing to the fragment shader. The shader
uses a texture that uses the U.coba_weight custom color band, or
an internal color band.
In addition to actual weights, the shader has to display two
alert colors: missing vertex group, and zero weight. The zero
weight alert has to be blended with regular weight colors,
so that a single alert vertex surrounded by weighted ones is
still visible.
Reviewers: campbellbarton, fclem
Differential Revision: https://developer.blender.org/D3675
New entry in the Add Object menu.
Opens a file selector and creates a new empty object from the selected image.
Previously more steps were needed to archieve the same.
Differential: https://developer.blender.org/D3708
Reviewer: brecht
This was caused by the object center being selected by the wireframe pass.
The wireframe pass is now only drawn if we search for the nearest surface.
This makes all wire objects (like non-meshes objects) higher priority than
mesh surfaces.
Previously copy and paste was handled within the same function using a mode switch, which is quite ugly in most cases.
This patch separates the function into two and splits up those in even smaller functions to increase maintainability.
Finally one new feature was added: You can copy from disabled fields now. (requested by hjalti)
Reviewers: brecht
Differential Revision: https://developer.blender.org/D3738
* Panels now use single column layout.
* Footage Info was moved into Footage Settings.
* Display settings are now in a popover in the header.
* Graph view shows tracking controls in the header center.
Differential Revision: https://developer.blender.org/D3643
- Use ctrl for de-select since we may want to use shift for extend.
- Use cursor tool by default,
it's agreed lasso select isn't accessible from a shortcut by default.
This commit make the Xray option for the wireframe different from the other
shading mode. This makes it possible to rapidly switch between wireframe +
Xray and Solid mode without Xray.
Xray alpha is also decoupled.
Both variables are duplicated and exposed separately through RNA.
* Proportional pie menu at shift+O.
* Snap pie menu at shift+S.
* Pivot pie menu at comma. Previous comma, ctrl+comma, period
and ctrl+period shortcuts for specific pivot types were removed.
Ref T56881.
This commit includes several performance, stability, and reliability
improvements to cloth collisions.
Most notably:
* The implementation of a new self-collisions system.
* Multithreading of collision detection.
* Implementation of single sided collisions and normal overrides.
* Replacement of the `plNearestPoints` function from Bullet with a
dedicated solution.
Further, this also includes several bug fixes, and algorithmic
improvements.
Reviewed By: brecht
Differential Revision: http://developer.blender.org/D3712
In addition to the original map to surface and Keep Above Surface,
add modes that only affect vertices that are inside or outside
the object. This is inspired by the Limit Distance constraint,
and can be useful for crude collision detection in rigs.
The inside/outside test works based on face normals and may not be
completely reliable near 90 degree or sharper angles in the target.
Reviewers: campbellbarton, mont29
Differential Revision: https://developer.blender.org/D3717
Select minimum of render subdivision levels and 3 for
the initial value of quality. This way we don't force
too much quality for meshes which were not supposed to
be too much quality :)
Need to resize `have_functions` arrays when adding more functions.
Not sure there is more reliable way which will prevent accidents
like this.
Was caused by c47be43674.
Unlike 2.79 I am not using the comparison type to get coplanar planes that are above/below the plane.
That seems arbritary at best, and a bit of a challenge to get it right.
Also, I'm doing some shenanigans to use a single tree to store the face direction and the distance.
Here in my tests it is working fine though.
Now handles are drawn using index buffer instead of duplicating memory
requirement.
Also make use of shader tricks to draw handles antialiased, and respond to
UI scalling.
Make vertex point match edit mesh vertex size.
Note: Not doing worldspace comparison for now. We would need to create new
BM_face_calc_area and BM_face_calc_perimeter equivalent functions for that.
Not sure if it is worth the trouble.
Note: Unlike 2.79 we are not comparing face->mat_nr, but the material itself.
On top og that, if the material slot is empty we are just ignoring this face.
This was because weight paint was using the original obj to show weights.
But the mesh can be deformed by the armature and this is not visible on
the original object.
This makes the Edit Mesh display settings common to all objects. They can
also be set differently per viewport.
Modifying extra data (seams, sharp edges etc...) will no longer set them
automaticaly visible.
Bumping version because we need to force set all extra draw options for
older files.
For onion skin, the cache is set as dirty always because can be too many combinations of overlays and windows and makes impossible to control all changes of dirty flag.
This reduces performance, but it's safer and the onion is disabled while play animation, so the fps value is not affected.
This adds existing behavior from calc_weightpaint_vert_array
that was missing from the new rendering code:
- No selected Vertex Group displays as a solid pink color.
- Zero weight displays as alert color depending on Options.
- Multipaint mode correctly displays collective weight.
In order to properly implement this variety, a data structure
holding all relevant parameters is introduced.
Reviewers: fclem, campbellbarton
Subscribers: jbakker
Differential Revision: https://developer.blender.org/D3722
As the comment states, this is not really great solution, but
is good enough for now. Proper solution needs some more work,
and maybe reconsideration of what is stored where.
I think there are two possible ways to fix that.
1. Make the name a required parameter.
2. Provide a default value.
I choosed option 1 in this fix to be consistent with other .new functions.
Also I think `RNA_def_string` instead of `RNA_def_string_file_path` should be used here. Looks like a copy-paste error.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D3728
Use the 2x spin tool has 2x handles to control normal direction,
dragging either handle backwards now reverses normals.
Previously it was common for the spin result to have normals
flipped the wrong way.
Use the multiply blending mode for the weight paint overlay.
To support the opacity slider, we need a new shader. Otherwise this combination of multiplication and mixing does not seem to be supported by glBlendFunc.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D3727
Before it was only doing this for directly linked objects. This was
initially needed to solve fps issues in some of the shots. Now this
fps drop i can no longer reproduce. Was likely needed prior to
modifiers dependencies "inheriting" visibility from the "parent".
This commit makes it so objects which are linked indirectly to the
scene via collections are properly considered visible (or, rather,
evaluatable) by dependency graph.
Solves missing eyelashes and braids issues with 01_035_A.
The fields currently support a Point source and a two-dimensional
Plane source, but there is no way to create a field that pulls
toward or from a line in space other than using the Texture field
type. This adds a new simple shape option to do that.
The line is aligned along the Z axis so that it works meaningfully
with the Tube and Cone falloff modes.
Reviewers: brecht, mont29, LucaRood
Differential Revision: https://developer.blender.org/D3721
The default compositor node update function sets the need_exec flag on the node
which the Auto Render feature checks, but the custom update function that was
added as part of rB4cf7fc3b3a4d didn't do so.
Therefore, the two custom update functions that were added now also call the
default update function.
I'm not sure why we may want to sample both a sharp and an unsharp edges at the
same time, maybe to see if the selected edges all have the same values?
Either way, implemented as in 2.79. I also believe we may have a faster way to
select all the edges, but let's file this under optimizations to be done later.
I'm not sure why the original implementation was only checking for equal
comparison but I'm doing the same here. It is a one line change if we
want to support LT/GT anyways.
Also "technically" we should compare the angles in the worldspace, since
different scales will result in different angles. Added as a TODO but
honestly I think this is overkill.
I'm using kdtree here but there is nothing preventing us from using a simple
float linked list with a sorting and finding "nearest" equivalents.
At least we are benefitting from bisecting as oppose to the original solution.
Also we need epsilon for the float comparisons.
This lets us do the most common shading switching with one shortcut. We keep
alt+Z and shift+Z for quickly toggling to lookdev and rendered mode and back,
it's debatable which settings deserve a dedicated shortcut like this.
The downside is that switching X-ray is a little slower, and that there is
some risk of accidentally going to lookdev or rendered mode which can be slow
to cancel.
We now treat Xray as being the mode where Limit selection to visible is off.
If Xray is OFF, Limit selection to visible is considered ON.
To allow 'see through wires' with solid shading (not Xray shading) we still
draw solid shading if Xray is ON with Xray Alpha set to 1.0.
This is using the existing engine (workbench forward) with 0.0 xray_alpha
and forcing wireframes on all objects.
There is no workflow/shortcut changes in this commit.
Normally the time can be read from DEG_get_ctime(depsgraph), but this is a
bit more forgiving for e.g. addons that don't care too much about the details
of the COW depsgraph.
Added propagation of displacement to a higher levels.
Has the same limitation as reshape from object, which makes
grids somewhat choppy due to linear interpolation nature.
This is to be addressed next.
This brings basic implementation of displacement propagation
to a higher levels. Currently does linear interpolation between
grid points at which reshape happened. It makes results a bit
choppy, proper solution would be to use some interpolation
weights from the limit patch.
The image I get is a bit too dark, which is the same I get in the viewport
itself when there is no Display Device. So I believe for cases like this
we need to have color management on top of the buffer.
Also, on EEVEE it looks like it needs to clear some extra buffers. You can see
that, by rotating the camera view around.
That said, this is the first step to bring back fun addons
that use external offscreen buffers.
Note: When using gpu offscreen with POST_VIEW (as oppose to POST_PIXEL)
I get a crash with DST having 0xFF..
This implementation is a bit hacky but match cycles pretty close.
If pixel size is not enabled, it will use the geom shader to
compute distances between vertices. This will have a cost.
Implementation is a bit hacky in gpu_codegen to make the geom shader works
in an optional manner.
Makes it so smooth brush works properly, without causing grids
to become disconnected from each other.
This need to optimize the code for brush, to only average edges
and vertices which are adjacent to modified faces.
This information is stored for each non-loose edge.
For each of such edge we store:
- List of CCG faces it is adjacent to.
This way we can easily check whether it is adjacent to
any face which is tagged for update or so.
- List of boundary elements from adjacent grids.
This allows to traverse along the edge and average all
adjacent grids.
The main goal of this patch is to cleanup the interface of every modifier. More specifically the interface of modifiers should be DerivedMesh-free.
Internally some modifiers still use DerivedMesh. However I think it is better when the wrappers are in the modifiers so that higher level functions can use the simplified interface.
This patch removes the applyModifier_DM and applyModifierEM_DM functions. In a previous patch (rB3614d9d) the other functions that used DerivedMesh have been removed.
Reviewers: brecht
Since shape keys are stored as raw floating point data, this
unfortunately requires changes to all code that works with it.
An additional complication is that bezier and nurbs control
points have different entry size, and can be mixed in the same
object (and hence shape key buffer).
Shape key entries are changed from:
bezier: float v1[3], v2[3], v3[3], tilt, pad, pad;
nurbs: float vec[3], tilt;
To:
bezier: float v1[3], v2[3], v3[3], tilt, radius, pad;
nurbs: float vec[3], tilt, radius, pad;
The official shape key element size is changed to 3 floats,
with 4 elements for bezier nodes, and 2 for nurbs. This also
means that the element count is not equal to the vertex count
anymore.
While searching for all curve Shape Key code, I also found that
BKE_curve_transform_ex and BKE_curve_translate were broken. This
can be seen by trying to change the Origin of a Curve with keys.
Reviewers: campbellbarton, sergey
Differential Revision: https://developer.blender.org/D3676
Seems like GLEW_ARB_texture_float is not defined in core profile on OSX
even if float textures are part of the 3.3 core spec.
So removing uneeded backup cases.
Most other software expects to read indexed vertex colors, so write indices
along with the colors as we already do for UVs.
Differential Revision: https://developer.blender.org/D3704
Previously it was possible to have geometry component to
tag batch for DIRTY_ALL and shading to tag with SHADING_UPDATE.
Now there is a relation between uber eval and shading, which
solves this issue.
The idea of those flags is to avoid evaluation of operations
which are not needed to bring visible objects to an up to date
state.
Previously, dependency graph attempted to do combine those
into an ID level flag. In practice it proved to be rather
tricky, since there could be dependency cycles on ID level
which will not exist on component level.
This allows to force duplicated object to be fully evaluated for
display/draw.
Previously only transform component of duplicated object was
forced to become evaluated.
Need to ensure objects from collection which was built but
is became visible in the new "context" are poked for re-built.
This should be rather cheap, since this only will update
their visibility flag.
Can not rely on visibility flush here, since there is no
relations between collection and its objects.
Previous code would crash on whole lot of accessors in case current item
was not a real dupli instance.
And code was missing access to crucial dupli (world) matrix! Pretty
useless without that.
Also reordered a bit members in here, let's try to keep a bit of logic...
- Add Arc-Inner-Factor, to increase the inner arc radius from 0
(show as a disk).
- When showing an angle value, show the angle-end line thicker
(since thats whats being adjusted).
- Remove offset used by arc drawing which caused aliasing artifacts.
Terms get/set don't make much sense when casting values.
Name macros so the conversion is obvious,
use common prefix for easier completion.
- GET_INT_FROM_POINTER -> POINTER_AS_INT
- SET_INT_IN_POINTER -> POINTER_FROM_INT
- GET_UINT_FROM_POINTER -> POINTER_AS_UINT
- SET_UINT_IN_POINTER -> POINTER_FROM_UINT
Note: I had to add an epsilon on top of the threshould for SIMVERT_NORMAL.
Otherwise I was getting differences such as 0.000000something when comparing
supposed-to-be-identical normals.
The way I see it, the Threshold option is a user feature, where users can
control more or less what they want selected. While the epsilon is a
non-negotiable requirement for our float comparison here.
This includes support for:
* SIMVERT_NORMAL
* SIMVERT_FACE
* SIMVERT_EDGE
Not included and currently not supported/disabled:
* SIMVERT_VGROUP
* SIMEDGE_*
* SIMFACE_*
While we are working on this, we prevent users from using the non-ported modes.
Note: the bmo_similar.c file is still around, to be removed in the near future.
Everyone: Please fell free to jump in and help tackling the missing modes.
For details on the implementation discussion:
https://developer.blender.org/D3674
Also minor changes:
- Remove unused headers.
- rename SimpleExprParseState -> ExprParseState
- rename max_ops -> ops_count_alloc
Was misleading since maximums are often limits which can't be exceeded.
Simple isn't a good prefix for library names since
lots of unrelated modules could be called 'simple'.
Include 'py' in module name since this is a subset of Python,
one of the main motivations for this is to be Python like/compatible.
- Pass in the snap distance
(to avoid comparisons in the multi-object version).
- Loop directly over BMLoop's (without using the iterator).
- Use squared distance instead of manhattan.
The thing is that the quality is something a bit subjective regarding this
option value.
A more precise trace will have more noise so quality is also diminished in
a way. I believe that using "Precision" is more correct than "Quality" in
this case.
Also tested the straighten option and it is working-ish.
That said, I think straighten should take all the meshes vertices to determine
the line you want to use as reference. However we would need a different way to
determine the first and last uvs to use as reference for the line.
The first time setup screen only has the interaction preset currently, some
more work is needed to be able to set e.g. the language or compute device
here as in the mockups.
The splash screen stayed the same for now, to make room for the templates
most of the links are now in the Help menu. If there are no recent files yet
the links still show.
The splash screen buttons implementation was fully moved to Python, in the
WM_MT_splash menu.
These can now be acessed from the File > New, Ctrl+N, or the splash screen.
Since these are application templates, users can save a separate startup.blend
for each. User preferences are shared between these templates though.
This also fixes some issues in the default startup.blend (triangulated cube..).
Differential Revision: https://developer.blender.org/D3690
The goal here is to make app templates usable for default templates
that we can ship with Blender. These only have a custom startup.blend
currently and so are quite limited compared to app templates that fully
customize Blender.
But still it seems like the same kind of concept where we should be
sharing the code and UI. It is useful to be able to save a startup.blend
per template, and I can imagine some scripting being useful in the future
as well.
Changes made:
* File > New and Ctrl+N now list the templates, replacing a separate
Application Templates menu that was not as easy to discover.
* File menu now shows name of active template above Save Startup File
and Load Factory Settings to indicate these are saved/loaded per
template.
* The "Default" template was renamed to "General".
* Workspaces can now be added from any of the template startup.blend
files when clicking the (+) button in the topbar.
* User preferences are now fully shared between app templates, unless
the template includes a custom userpref.blend. I think this will be
useful in general, not all app templates need their own keymaps for
example.
* Previously Save User Preferences would save the current app template
and then Blender would start using that template by default. I've
disabled this, to me it seems it was unintentional, or at least not
clear at all that saving user preferences also makes the current
Differential Revision: https://developer.blender.org/D3690
This makes it so coordinates and normals for CCG are calculated
with mutires displacement taken into account. This solves issues
with multires displacement being lost when entering sculpt mode.
The missing part is averaging of normals along grid boundaries.
But even then sculpting shows decent results.
The plan to solve that would be to introduce function to stitch
grids, which can also be used by Smooth brush which requires
this.
This happens on NVidia GPUs, using more textures than the maximum allowed
by the gl will NOT trigger a linking issue (maybe because of bindless
texture implementation?).
So in this case we manually count the number of samplers per shader stage
and compare it against the GL limit. We discard the shader if the sampler
count is too high. This shows the user something is wrong with the shader.
Recently @sergey found that hard-coding evaluation of certain very
common driver expressions without calling the Python interpreter
produces a 30-40% performance improvement. Since hard-coding is
obviously not suitable for production, I implemented a proper
parser and interpreter for simple arithmetic expressions in C.
The evaluator supports +, -, *, /, (), ==, !=, <, <=, >, >=,
and, or, not, ternary if; driver variables, frame, pi, True, False,
and a subset of standard math functions that seem most useful.
Booleans are represented as numbers, since within the supported
operation set it seems to be impossible to distinguish True/False
from 1.0/0.0. Boolean operations properly implement lazy evaluation
with jumps, and comparisons support chaining like 'a < b < c...'.
Expressions are parsed into a very simple stack machine program
that can then be safely evaluated in multiple threads.
Reviewers: sergey, campbellbarton
Differential Revision: https://developer.blender.org/D3698
This partially solves ASAN report about unfreed memory. There is still
something in the report, need to have a closer look with debug version
of OpenEXE library.
- Now the spin tool has a persistent gizmo.
- Uses scene orientation, with additional view orientation.
- Uses the cursor center, ignoring the pivot since the
selection center is rarely useful.
- Disable most of the redo gizmo's for now since they overlap,
only allow adjusting the angle.
Note: mixing new action with adjusting previous is confusing,
we'll want to have design guidelines regarding this.
While building the AVX kernel, util_avxf.h/avxb.h were using some AVX2 intrinsics,
these were never called, so it wasn't a run-time issue, but the intrinsics headers
on centos excluded the AVX2 prototypes when building the AVX kernel causing build errors.
This commit cleans up the improper usage of the AVX2 intrinsics and provides AVX
fallback implementations for future use.
Differential Revision: https://developer.blender.org/D3696
ffi stubbornly wants to put libs in lib64 even when you tell it not to on some linux distributions.
patch based on sed fix the gentoo guys did [1]
[1] https://bugs.gentoo.org/462814
This isn't really possible to do the shuffle which was attempted to do.
While it's possible to achieve expected behavior, the function needs to
be rewritten. Since it's not used anyway, it's simpler to remove it for
now.
Fix T56789: There was issue with certain driver with glGenerateMipmap and
GPU_DEPTH_COMPONENT24.
In this case we just create a complete texture with mipmaps manually
without downsampling / initializing the data.
This reorganizes the cloth UI, and changes some of the behaviour to be
more reasonable.
Changes included here:
* Reorganized cloth panels
* Improved some tooltips
* Removed `vel_damping` option
* Removed cloth pinning checkbox
* Removed stiffness scaling checkbox
* Separated shrinking from sewing
* Separated self collisions from object collisions
Reviewed By: brecht
Differential Revision: http://developer.blender.org/D3691
Avoids invisible objects from being considered visible when they
are used as driver variables.
If those cases are actually coming from a visible object, then
deg_graph_build_flush_visibility() will ensure visibility is
properly flushed there.
Allows to go to sculpt mode, do brush strokes, get out of sculpt mode
and have deformation preserved.
The issues currently is that the current implementation of CCG
storage is created from the limit surface, without displacement
taken into account. It is trivial to get displaced coordinates,
but it is more tricky to get displaced normals. This is something
to be solved next.
Another limitation is that this only works for sculpting at a maximal
multires level. There is code to be done to support propagation
of displacement onto a higher levels.
The function is supposed to be called for original object.
Draw manager abuses this a bit, will solve later by moving
PBVH (re)creation to dependency graph.
At the time being, stop adding object evaluation to draw
manager, this is really where it does not belong to.
Flat shaded grids are now rendered as a long triangle strip with
degenerate triangles.
Smooth shaded grids are not changed.
Also add back fast navigate. The only difference with 2.79 is that fast
navigate + flat shading will show slitly "incorrect" smooth face. But that's too
much effort to fix for what it is.
Since we are not linking against OpenCV ourselves, that generated
linking errors later on (while building OSL e.g.).
Those 'open' libs link against way too many other libs... :/
Thanks to @intrah for initial report (T56785), and @LazyDodo for
suggested solution.
That bug probably did not affect 2.7x, only 2.8 with COW copying IDs in
threads... But root of the issue is that underlying boost i18n lib does
not support well multi-threaded access. So simply forbid any translation
from non-main thread. This *may* be an annoying limit at some point, but
doubt it will be any issue currently.
That pointer can be NULL, RNA default string handling does not support
that. (that whole uv_layer prop is quite nasty actually, since it does
not own that string, always borrows it from some other data :((( ).
Rebuilding depsgraph is not enough, with COW we also need to ensure COW
copies get updated accordingly.
Had to tweak the generic update system here, since it was always passed
a NULL pointer for the callback arg, this should not change existing
behavior (besides crash fixing ;) )...
Failed as in did not allocate due to possibly weight cutoff.
Tryign to allocated Extra storage for closure in such situation
will consfuse Cycles and cause crashes later one due to obscure
values in ShaderData.
Consequences of new requirement to use named keywords parameters for our
RNA API... Hopefully fixed all other cases in that file as well, but
guess we'll find more of those in comming weeks. :|
Obvious bug in recent CName cleanup commit, while BGE library
was indeed mentioned on that line it was to put Bullet libraries
after all the other libraries.
One day we should really switch to a CMake dependency graph to
take care of library dependencies.
It's more convenient to keep version patching in the same place,
this also splits out some function calls from version patching
and supports updating UserDef structs besides the 'U'l global.
Attempts to substitude CCGDM with an OpenSubdiv based structure
which has less abstraction levels. The missing part in this
substitude is a face pointers which old CCGDM/multires code was
using to stitch faces (averaging boundaries).
Another curial bit missing: "reshaping" of multires CD_MDISPS
to the state of new PBVH grids.
The new code is only available when OpenSubdiv modifier is
enabled (WITH_OPENSUBDIV_MODIFIER=ON) and with debug value of
128. This is so this WIP code is not interfering with current
production machines in the studio.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D3685
Didn't realize the index buffer is stored once in a BVH
and same pointer is reused. Surprisingly, simple files
were fixed with the previous fix.
Now disabled the optimization all together, and it was
simpler to just completely remove all residue of the
code. It is likely to be a different implementation
anyway, so no need to try to keep code in a semi-broken
state.
Nodes can have different number of grids, so can not so simply
re-use index buffer across nodes. For now allow re-using buffer
if number of grids matches.
The issue is, number of grids is probably almost never matches,
so in order to have this optimization working need more tricks.
Maybe we can "instance" index buffers?
* WITH_SYSTEM_OPENJPEG is removed and is now always on, this was already
the case for macOS and Windows.
* This should not break existing Linx builds. If there is no new enough
OpenJPEG installed, CMake will no find libopenjp2 and WITH_IMAGE_OPENJPEG
will be disabled.
* install_deps.sh was updated with new package names, since distributions
put this version in a new package.
Differential Revision: https://developer.blender.org/D3663
Mainly this is following Linux to build own xml2/lzma/ssl/sqlite and linking
them all statically. This ensures the Python ssl module uses a recent openssl
version rather than a very old one shipped with macOS.
Was using UI_BLOCK_LOOP to control draw style,
this meant we couldn't use popup theme colors for cases
where it the interface has the same purpose as a popup but happens
not to use this flag.
As mentioned in d81aeb60fe the alignment for multiple text boxes should all
respect the same overall padding.
Vertical alignment is fully implemented now.
During drawing, the depsgrah was tagged to update and this produced a full copy of the datablock. This tag was done in order to get the right data in drawing engine, but this added a great overhead while drawing and the response of the pen was not good.
Now, the depsgraph is not tagged and the drawing engine uses the original copy data of the buffer datablock. This is not a problem because only can draw in one window at time.
Now we can create new base color, roughness, metallic, specular, normal,
bump and displacement images, and linked them to the appropriate socket.
Also fixes image nodes inside groups not being visible.
Differential Revision: https://developer.blender.org/D3679
To follow Blender standards, the name of the new primitive stroke must be different of blank object.
Also changed the default orientation of Suzanne 2D to use the same as other objects (not aligned by default)
In some cases we need to use this array afterwards,
so this gives control over which objects are used for picking.
Also use an index argument as return argument
so callers that need to know the index in the array
don't need to calculate it afterwards.
- Poly build now uses a new gizmo for pre-selection
which has the same behavior as loop-cut.
This replaces hack where mouse-move set the active element
which was no longer working properly because of missing
depsgraph updates.
- Multi-object support for poly-build.
- Support for deformed cage.
- Fix error where changing active object wasn't properly
refreshing the preselect gizmo (for loopcut too).
Currently holding Alt to select non-boundary element's isn't working.
There was an issue caused by Antialiasing being done after DoF. Move TAA
after DOF and Motion Blur.
Also certain pixel with lower CoC would be spread all over the background
because the neighbooring pixel have higher CoC. So we need to apply some
bilateral filtering when downsampling. Currently we limit the influence of
neighbor pixels with a CoC inside the range [MaxCoC-2, MaxCoC].
If something is a list of data, it is to be called list, and data
is to be called data. No need to call list a data and data a weird
two letter abbreviation.
Clarity is our friend!
This is something what we need to know quite often from various places.
Added it as a cached value in Subdiv itself, so it can be queried easily
from any area.
Shouldn't be a problem from memory usage point of view, it's 4MB per
1M faces coarse mesh. This is very low percentage of mesh itself, and
even lower percentage of highres subdivided mesh.
When using the 'normal' orientation, the normal would be ignored
if the plane couldn't be calculated.
Now use only the normal if the plane is zero length,
this was already done, just not in all cases.
With the removal of blender internal, the texture preview had to be
rewritten. Now we evaluate and write the texture directly to the
preview buffer instead of going through the renderer.
Differential Revision: https://developer.blender.org/D3670
D3375 by @Al
I did a few changes before the commit:
* Initialize flot arrays with {0} instead of memset(foo, 0, sizeof(foo)).
* Use add_v4_v4 instead of for loop.
* Rename uv_map_rotation_matrix_w_offset > uv_map_rotation_matrix_ex.
bjects: orthographic support for UV project
Nothing really interesting, just starting laying down API which
seems to be a decent substitute to CCGDM, without requiring too
much work be done in sculpting area.
Tested with Debian Testing, might need some adjustements for other
distributions...
Also removed last patches we used here, we shall not need any anymore!
In a real world it is very weird to disable AA on a mask,
it will give ugly looking result. For some fast preview
passes (like in the node preview) the system can decide
to disable AA without asking user to do anything.
One thing we can consider doing is to remove Feather
option as well. If real compo becomes measurably slower
in cases when mask has no real feather, we can disable
feather internally, without user input. Disabling
feather in the interface is like making things faster
but giving a wrong result, which doesn't sound that
helpful either.
Reviewers: brecht
Reviewed By: brecht
Subscribers: hype, sebastian_k
Differential Revision: https://developer.blender.org/D3677
The pie menus were a legacy of old grease pencil and now need a full redesign or maybe a new add-on.
While we decide the way to go, the pie menus are disabled.
Displaying the labels tip immediately feels too intrusive,
make this work more like regular tooltips, displaying more quickly.
Tooltips can now uses multiple passes, each pass with it's own delay
for the next pass to show.
- Only show label-tips when tools are icon-only.
- Don't show shortcuts in label-tips.
- Position both tool label & full tips around the button bounds.
Maybe it's still early to set the new drawing api for python.
But joining these two modules is an initial step.
```
>>> gpu.
matrix
select
types
```
```
>>> gpu.types.GPU
Batch(
OffScreen(
VertBuf(
VertFormat(
```
The creation of a new offscreen object is now done by the `GPUOffscreen.__new__` method.
Reviewers: campbellbarton, dfelinto
Reviewed By: campbellbarton, dfelinto
Tags: #bf_blender_2.8
Differential Revision: https://developer.blender.org/D3667
Use kdtree for doubles (old standing TODO for this operator).
Small changes from reviewer (Dalai Felinto):
* Code style ({ is in a new line only if it is a two-line if).
* Skip objetcs loops when we know for sure nothing is selected (or all is)
https://developer.blender.org/D3441
I don't even know why this operator was ever made to work in object mode.
That said, since it does, we should have different options for it
(or rather, always do all faces for it).
Do note that we force showing the face dot if we are in face select mode
but faces and edges overlays are disable to not loose the selected faces
entirely.
The issue there was that number of layers did not include normals,
while element size counts bytes used by normals. This sounds very
fragile and dangerous to work further. Also, one value can easily
be delivered from another, so it is redundancy going on here.
Possible difference now is that multires subdivision will copy
normals to a higher levels. Shouldn't be big of a problem, since
leaving old normals and updating coordinates is not correct either.
The original code was already making a distinction between lines in the last
text box and all lines. However I removed that bit since when I tested the
values were the same (I tested with a single text box).
Bringing this distinction back.
Not addressed here: All boxes should respect the alignment. Which at the moment
they don't seem to fully do.
They way Blender handles vertical alignment is very buggy:
- Top-Base: It works perfectly.
- Bottom: It is actually bottom-baseline,
and it fails when line size is != 1.0 when working with text boxes.
- Top: Poorly implemented, it should use font's ascent
(recommended distance from baseline),
so it has room for accents,
but it's not one line distance far from the origin (as it is now).
- Center: Poorly implemented.
This is tricky since there is no silver bullet.
To clear this situation I created a new option (Bottom-Baseline),
and addressed the issues above.
I'm getting the ascent and descent from freetype2,
and use this for padding above/below the text.
Also for vertically centering the text.
Changes from reviewer (Dalai Felinto):
* pep8.
* Skip meshes that come from libraries.
* `Copy Mirrored UV coords > Copy Mirrored UV Coords`.
* Remove mesh(es) from warnings, report mesh or meshes based on number.
https://developer.blender.org/D3529
Changes from Reviewer (Dalai Felinto):
* MEM_SAFE_FREE > MEM_freeN.
* Skip loop if sync selection and no face selected.
https://developer.blender.org/D3415
Changes from reviewer (Dalai Felinto):
* MEM_SAFE_FREE > MEM_freeN.
* Skip loop if sync selection and no edge selected.
https://developer.blender.org/D3417
Annotations are not designed to edit, copy or paste, so the operators must be disabled.
By design annotations can only be added or deleted, but not edited.
Still pending clean the pie menus, but this will be done during UI cleanup.
Testing GLEW_ARB_texture_gather is not sufficient in this case. We need to
test if GL_ARB_texture_gather is defined in the shader, which is always true
on some NVIDIA drivers who does not support it...
So trying to make everything work.
We would need to at least support one more PBVH building
here, trying to squeeze everything into one function will
make it really hard to read and follow.
Simplify library remapping code to handle special collection/object links
in postprocess. Previously base contained the actual object link which
needed special handling in preprocess, now objects are linked through
collection and the base cache can be updated in postprocess.
This was introduced in the original implementation of weightpaint
in 2.8 (20f95de6ba). But this is very shortsighted, we can't assume only one
object will be edited at once, nor should we tag things during drawing.
There is a chance this introduces "bugs". If it does, we then tackle them in the
proper way (usually tagging DEG after operators changing the weight paint data).
I (Dalai) did a few changes before committing:
* Code style (mostly tabs instead of space).
* Skip loop when no vert selected.
* Keeping comment on why always return FINISHED.
Maniphest Tasks: T54643
https://developer.blender.org/D3338
Changes from reviewer (Dalai Felinto):
* Code style (replace tabs with spaces).
* Return OPERATOR_CANCELLED if no a single object get through.
Maniphest Tasks: T54643
https://developer.blender.org/D3392
"Blend from Shape" lists keys from active object and applies to all selected
objects that have keys with a matching name.
Changes from reviewer (Dalai Felinto):
* Code Style.
* Move Object **objects after sanity check to prevent leak when
returning OPERATOR_CANCELLED.
* Skip objects that have no selected vertex.
* Fixed existent comments code-style: Since we are changing the indentation of
these lines, may as well correct this as well.
Differential Revision: https://developer.blender.org/D3656
Since D3284 was inactive for a long time, I figured I could finish this.
This revision also gives warning when there are no selected faces or when
selected faces are invalid.
Changes from reviewer (Dalai Felinto) :
* Code style
* Fix checking for ngons (BM_ITER_MESH is a for loop, you need to break from it
before continuing the object iteration loop).
Differential Revision: https://developer.blender.org/D3659
This implements angular bending springs for cloth simulation. This also
adds shearing springs for n-gons.
This angular spring implementation does not include Jacobian matrices,
as the springs can exist between polygons of different vertex counts,
rendering their relationships asymmetrical, and thus impossible to solve
with the current implementation. This means that the bending component
is solved explicitly. However, this is usually not a big problem, as
bending springs contribute less to instability than structural springs.
The the old linear bending model can still be used, and is the default for
existing files, to keep compatibility. However, the new angular bending
model is the default for any new simulation.
This commit makes small breaking changes, in that shearing springs are
now created on n-gons (also in linear bending mode), while n-gons were
previously ignored.
Reviewed By: brecht
Differential Revision: http://developer.blender.org/D3662
This separates cloth stiffness and damping forces into tension,
compression, and shearing components, allowing more control over the
cloth behaviour.
This also adds a bending model selector (although the new bending model
itself is not implemented in this commit). This is because some of the
features implemented here only make sense within the new bending model,
while the old model is kept for compatibility.
This commit makes non-breaking changes, and thus maintains full
compatibility with existing simulations.
Reviewed By: brecht
Differential Revision: http://developer.blender.org/D3655
Objects in edit mode now behave as if they were joined.
Changes from reviewer (Dalai Felinto): code style (tabs, spaces, ...), missing MEM_freeN(objects);
https://developer.blender.org/D3648
MASS unit was already implemented for the C api. Only making sure it is
accessible in the python api. Also added 'CAMERA' to the documentation as a valid option.
I'm using a different shading group for the active text box and the others.
This way I can assign different colors to them.
The alternative would be to create a new `shgroup_dynlines_dashed_flat_color`,
but I find this overkill for such a simple use case.
The statement that PBVH needs to keep track of CCGDM is wrong, PBVH itself
does not care about CCGDM at all, and it's weird for it to carry on this
beast so others can access.
Even more, nobody will actually caring about CCGDM itself, all the usages
were checking whether there is CCGDM or not. This is as good as simply
checking PBVH type.
Tested with an original report T53551 and everything is still stable.
WITH_ASSERT_ABORT was not disabled for release builds. In most cases asserts
are disabled in release builds, but not always.
This also changes the buildbot to use blender_release.cmake instead of
blender_full.cmake, the only effective difference should be WITH_ASSERT_ABORT.
The preset and decorator icons were updated to be monochrome and draw in the
same color as text. Other icons are unchanged, having them as separate icon IDs
prepares for an artist to make them.
When drawing in surface mode if the stroke z-depth was not valid, the stroke was created, but it was not visible because location was something illogical.
Now, if the z-depth cannot be calculated, the stroke is calculated as it was done in other modes.
In the workspace properties a mode can now be configured that is
automatically enabled when switching to the workspace.
This is a test to validate how well it works. The weak point is
that if you don't have an appropriate object already select it will
not switch modes.
See T56475.
The assembler template was backing up and restoring ebx, which is
fair enough. However, this did not prevent compiler for putting
result variables to ebx. This was causing data corruption.
In order to prevent this easiest solution is to list ebx in clobbers
for the assembly.
This commit adds a new armature modifier for grease pencil. The deformations are done reusing the mesh deform routines.
There is also a new operator in weight paint mode to help the artist to generate weights base on armatures. This operator is required because 2D animation workflow is not equal to meshes when parent an object to armatures.
In the drawing engine has been added the option to handle the Fade object parameter used in armatures to see the strokes while move the bones.
When rename bones, all related data of grease pencil is renamed too. This not only affect new armature code, but also layers parented and hook modifiers.
Thanks @aligorith for his review and help.
Now it's possible to add the weight of the current selected vertex group to any new stroke created.
This is very handy when add drawings after the weight paint was done and it will be required with Armatures.
This small offset helps to keep the stroke over the surface and avoid the stoke is "inside" the face of the surface.
We could add this as a parameter in the UI, but after doing a lot of test, the valid values are very limited and don't worth the change, so a hardcoded value works fine.
There is chance that on a system with both versions installed this
*might* cause some issues. Such system will be pain to support out
of the box anyway.
This change allows to use precompiled libraries without extra
modifications in the config.
- Foe some reason CMake's platform and processor are not intialized there.
- Need to set variables in cache, otherwise they are not visible in the
actual CMake files.
The script clearly states:
This makes the presumption that you are include al.h like
#include "al.h"
and not
#include <AL/al.h>
The reason for this is that the latter is not entirely portable.
Windows/Creative Labs does not by default put their headers in AL/ and
OS X uses the convention <OpenAL/al.h>.
This commit makes default precompiled OpenAL to be properly detected
and also removes hack on MacOS which was finding the OpenAL package but
then was overwriting include directory.
Note, that new audaspace in 2.8 is using expected #include <al.h>.
This is an initial implementation of BVH8 optimization structure
and packated triangle intersection. The aim is to get faster ray
to scene intersection checks.
Scene BVH4 BVH8
barbershop_interior 10:24.94 10:10.74
bmw27 02:41.25 02:38.83
classroom 08:16.49 07:56.15
fishy_cat 04:24.56 04:17.29
koro 06:03.06 06:01.45
pavillon_barcelona 09:21.26 09:02.98
victor 23:39.65 22:53.71
As memory goes, peak usage raises by about 4.7% in a complex
scenes.
Note that BVH8 is disabled when using OSL, this is because OSL
kernel does not get per-microarchitecture optimizations and
hence always considers BVH3 is used.
Original BVH8 patch from Anton Gavrikov.
Batched triangles intersection from Victoria Zhislina.
Extra work and tests and fixes from Maxym Dmytrychenko.
- Added icon for To Sphere
- Added icon for Shear
- Added icons for adding mesh objects
- Added icon for Curve Draw
- Add Curve Extrude to Cursor icon
- Tweak Curve Draw icon
- Simplified Hair Cut icon. Was hard to read
- Tweaks to Hair Puff and Smooth icons
- Added icon for Extrude Along Normals
- Updated Extrude Individual icon to become clearer and more consistent
with "Extrude Along Normals". Now it's easy to see the difference.
- Rip Edge and Rip Region icon tweaks
Note, many icons are regenerated in this commit which
weren't intended to be changed, in some cases this looks to be caused
by material color. Generating again doesn't modify so it
shouldn't be changing every time in future.
There is system-wide libz development package installed by default,
needed for some other dependencies. This patch ensures Python will
use our own self-compiled Zlib.
Misleading name since it's between 0..1.
Use as a keyword argument to prepare for keyword only args.
Also document that leaving unset has special behavior.
If the brush had assigned a color not present in the object material slots, the fill crashed.
Now, the material is added to the object slot before using it.
This involved getting SSL compiled from sources first, ensuring
it is a static library placement independent code. Configuration
is based on what Debian is using. CFlags required to have own
configuration file, which i didn't find a better place that next
to the corresponding CMake file.
It is OpenSSL btw.
It is set to Python via --with-openssl= configuration argument.
This works fine in a clean chroot, but having libssl-dev installed
might make Python to prefer system wide library, This was worked
around by using libssl_pic.a name for the library and modifying
setup.py. Would be cool to ensure system wide libraries are not
a problem, but official release builder is safe against this,
since it will catch possible non-static dependencies.
There is also a new map file which shadows bunch of Python
symbols. Without this Python's shared libraries might bring
conflicting symbols to Blender namespace at runtime.
Hopefully this doesn't break other platforms.
The operator allows to copy a complete layer with all frames or only active frame to a new object.
Can be found in edit specials menu (W key) or in Layers specials menu (last button near layer list).
That is kind of mandatory with complex rigged-character groups, with
hundreds of helper objects, and a few useful ones being hidden on
specific layers (like e.g. the main rig...).
It is especially critical point with static override, which won't allow
to move objects between collections and such (that would be a nightmare
to implement and handle).
Note that this is rather basic implementation, we could go further and
move all objects in all layers they are 'active', but that would
probably be overkill.
Reviewers: brecht
Subscribers: brecht
Differential Revision: https://developer.blender.org/D3649
- Add orientation option (defaults to local, as 2.7x does)
can optionally use global, cursor, view... etc.
- Fix typo which caused select flush to fail.
- Fix for instanced objects (was only checking one instance).
- Only tag for changes if a change is made.
- Skip meshes with all vertices selected.
Before, the weight data array was created always, but now only is added when a weight value is assigned.
This change was suggested by algorithm, and both agreed it was good idea.
You cannot immediately add parent's library to newly generated hidden
child collection, since it would allow to get several of those hidden
collections with same name/library. That is strictly forbidden!
So rather loop again on collections after all hidden ones have been
generated, and assign children's library from parent one then.
there is an issue with objects destructing in a non deterministic way during process shutdown, temporary work around this until osl has a fix in place.
Only the upper triangle of the block matrix is stored, thus when
executing operations on the lower triangle, each block must be
transposed. This transposition was not ocurring in the matrix-vector
multiplication function, which is fixed by this commit.
Reviewed By: brecht
Differential Revision: http://developer.blender.org/D3619
With small tiles, the repeated allocations on GPUs can actually slow down the denoising quite a lot.
Allocating the buffer just once reduces rendertime for the default cube with 16x16 tiles and denoising on a mobile 1050 from 22.7sec to 14.0sec.
After some artist feedback the material selection was not clear.
Now, the material can be selected in the top bar or in the properties panel.
1) If the material is selected in properties panel, all Brushes except pinned will be assigned to this material.
2) If the material is selected in the brush, the properties panel is updated to set the active material.
Added a new Pin icon to keep locked the material to one brush
As the polygon drawing is handled using a buffer that is converted to stroke inmediately, the drawing engine was not drawing the buffer stroke because it was empty, and the stoke was not visible becaus eit was in fast drawing mode.
Now there is a special flag to indicate the polygon mode is enabled and force the drawing engine to draw strokes and disable fast drawing.
* Remove support for diffuse color in the pbvh buffers.
* Upload raw data to GPU.
* Only draw nodes that have mask data when drawing the overlay.
This should fix T56466
Now that this operator is working properly (in world space axis), it can
be ported for multi object support.
The issue of only running on redo is still present though, to be investigated
later.
Not, I tried using `dist_signed_squared_to_plane_v3` but profiling showed that
it is 50% slower than using regular `mul_v3_m4v3` for the verts.
I managed to get this number closer when manually inlining all the functions
called by `dist_signed_squared_to_plane_v3`. But still `mul_v3_m4v3` was better
and it makes the code simpler to understand.
Also I'm changing the default mode to positive, no idea why it was negative as
default in the first place.
Last but not least, the operator only works well on redo. This was a problem
before, not introduced by this patch.
The lock to axis is only used by grease pencil object and can produce weird results in annotations.
This lock to axis function was kept by error when the annotation code was splitted from general grease pencil code.
Was happenign for following cases:
- Deep hierarchy of non-restircted collections, which has grand-grand
parent restricted.
- Collections which are constructed from invisible object.
Available in RGB Curve node in the compositor and as modifier in the
sequencer. I reshuffled the values of the enum. But a the first commit
is just 1 day old I think that the order is more important than the file
compatibility.
While the crash is in 2.8, it's possible undo operates on data
which isn't only owned by the current scene (any object for eg).
Thanks to @mont29 for suggesting the fix.
* Move Add to Quick Favorites before Assign Shortcut, as it is more often
that we add items to favorites, than changing shortcuts.
* Move Header menu last. In some cases the Online Manual for that item
would appear after the header menu.
* No icon for "Remove from Quick Favorites", matches Remove Shortcut
Now wireframe does update in sculpt mode using the sculpt mesh VBOs.
The only thing that does not work is simplified wireframe (the wireframe
slider) when the mesh is shaded flat. In this case all edges are shown.
The title says it all actually, the idea is to speedup the following case:
- Visible duplicator of a restricted collection (reported as T56512),
One of the questionable change is that none of the view layer bases is
ignored now. This ensures corresponding objects will have copy-on-write
component evaluated, making it possible to access those pointers. The
evaluation of those objects is skipped.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D3641
For this we need to add a bias depending on the viewing angle.
But increasing the hardness of the test make float precision issues in
the mesh transformation more prominent (actual geometry is far below the
surface). So to solve this issue we use a more subdivided grid mesh
8x8 quads instead of 1 triangle.
Film-like curves for the RGB Curve node (Compositor) and Curve Modifier
(Sequencer)
Film-like curves originated from Adobe.
"It’s an RGB curve where the tone curve is applied on the largest and smallest value, and then the middle value is adapted to keep a constant hue as defined by RGB-HSL/HSV. In terms of look and saturation increase it’s very similar to a pure RGB curve, more so than a HSL-L curve or HSV-V curve, but some color shift problems are avoided."
Other tools like Natron, Krita and RawTherapee have implemented this curve tone.
Reviewers: brecht, campbellbarton
Reviewed By: brecht
Tags: #compositing, #video_sequencer
Differential Revision: https://developer.blender.org/D3638
You cannot store ID (or any other data) pointers accross undo’s/redo’s, those
are reading .blend file and hence invalidating all pointers!
Not that happy with this fix, code is rather convoluted and not elegant
at all, but cannot think of a better solution, so will do for now.
Also found and fixed another 'str ghash used for int keys' case...
Basically just concatenates ID's name (including its IDtype code) and
that library's name, if any. This must give unique string in a given
Main database, suitable for GHash keys e.g.
- Add support for key_modifiers, so grease pencil gets "D".
- Assign number keys for unmapped items
This means all tools get keyboard access,
use number/numpad to avoid confusion with other keymap items
which directly activate the tool.
Use debug_value = -1 to disable the use of the filtered depsgraph.
It's still useful to have this for benchmarking + until we're
confident the filtering works safely.
* Don't print on every frame evaluated. This was only needed
earlier to track the progress (and why things were taking so
long - answer, it was the Scene COW issue). Saves 50-100 ms
* Remove the extra calculation of the scene after evaluating
motion paths. This shouldn't be needed now with COW.
Saves about 20-30 ms
This was just randomly leaving all the Particle datablocks in the
filtered graph (and causing and extra/excess pass over the ID's
to get run). Unless we actually need those specially kept for
some reason later, it's better to leave those out for now.
This commit makes the motion path calculations use the
new Depsgraph Filtering API to obtain a more streamlined
copy of the full scene dependency graph, allowing for
faster calculations (as less data needs to be evaluated
for each frame being baked).
For example, in 01_020_A.anim.blend from the Spring production
files, the time needed to calculate paths on several bones
on Autumn went from 39.84 seconds (!) down to 9.90 seconds!
Currently, this works by just replacing the depsgraph instance
passed to the motion path calculation function. This filtered
instance contains just the ID's needed to evaluate the graph
to evaluate a specified target (i.e. the Object owning the pose).
Notes:
* By default, the filtering is not performed unless debug mode 555
is activated. Having a debug switch here allows comparing
performance and disabling it should thing it start crashing.
* It is necessary to pass in the original Scene instance (not the
COW one owned by the filtered depsgraph), otherwise changing the
current frame had no effect, due to the COW flushing from original
Scene to the new Scene overwriting the CFRA changes we make.
* The code here still needs cleaning up to debugging instrumentation, etc.
and also to optimise further (e.g. caching the filtered depsgraph
for faster updates when animating on existing paths, or fine-tuning
the exact set of nodes needed). I'm just committing this first,
since this was the quickly hacked-together test code I've been using
to check that this is all working.
* Further improvements could also be made to the time needed to
build the full graph instance (about 3.3 sec), by allowing partial
builds (e.g. by making a filtering proxy/wrapper around existing builders)
* COW data hasn't been expanded yet when we try to filter the graph
(you need to have tagged + evaluated it for this data to exist),
so all the offending nodes would just get left in
* Added more debug prints to verify whether the id_nodes vector is
getting cleared correctly
* Simplified operation-relation deletion. Now we collect the relations
to delete into a vector, then iterate through that, thus solving issues
with iterator invalidation (+ aborts arising from that)
* DEG_foreach_ancestor_ID() was assuming that all dependencies were
OperationDepsNodes, when in fact, some could be TimeSource nodes
When this works correctly, we should be able to feed in an existing
depsgraph instance, and get out a "filtered" copy of it that contains
only the subset of nodes needed to evaluate what we're interested in.
The current implementation only filters on ID blocks/nodes,
and starts by building a full new depsgraph instance first.
I'd originally intended to do it per operation instead, copying
over individual nodes as appropriate to have the smallest and least
memory intensive graph possible. However, I ended up running into
into problems with function binding + COW arguments, hence the
current slow solution.
Building the CUDA kernels takes quite a bit of memory, and when building all of
them the combined usage can be too much on some systems (especially VMs).
Therefore, this patch adds an option to force the build system to build them
sequentially by making each build step depend on the previous kernel.
Reviewers: brecht, sergey
Differential Revision: https://developer.blender.org/D3623
* Add Shortcut -> Assign Shortcut
* Add to Favorites Menu -> Add to Quick Favorites
To match the menu name and be more general for when the Quick Favorites
are accessible in places other than just menus (like panels).
Add blank icon to Remove Shortcut so it aligns with Change Shortcut
which is usually the item on top of it in the menu.
Limited to mesh type of source, not sure it ever worked for non-meshes.
While it's possible to support reshape from any object, the actual brain
of operation would need to be recoded to go away from requirement of
vertex indices matching in source and destination.
This fully works as "expected", however it seems strange when
there is no selected vertex in the non-active object.
In 2.7x if you join two monkeys, select a vertex in one of them and
try to use this operator in the other it will select the lonely newly
selected vertex.
If you split both monkeys and do this in 2.8 while multi-editing them
you won't get nothing when trying to ctrl+RMB any vertex in the monkey
object that has nothing selected yet.
I propose to have this addressed in an upcoming patch where we always
select a vertex if no vertex was previously selected.
This make the workbench draw everything in the background routine just like
eevee. This is because the workbench uses floating point buffers too and
rendering background to this buffer makes it incorrect without proper
color management.
This could be improved because in xray the background is not blended but
dithered as it's drawn after the main pass.
This supports meshes and hairs too. Matches cycles output.
This adds barycentric coords to the GPUBuiltin enum which will also be used
for the wireframe node.
This makes it more generic process to perform actions which
depend on ptex face + (u, v) and on subdivided vertex index.
Currently it is still just a subdivision calculation process,
but same foreach callbacks can easily be used to propagate
displacement from known vertex locations back to displacement
grids.
Add a general parameter to force the recalc of the triangulation data because some modifiers could change the geometry and the filling triangles would not be right.
Now, the parameter is visible in UI panel because this option reduces FPS, but maybe in the future we can keep always ON and remove the parameter.
This changes the text hinting setting to be an enum with options
Auto / None / Slight / Full. The default is Auto which currently disables
hinting.
The hinting was tested with a new FreeType version, but this is not what
is used on the buildbots an official release environment, and the fonts
look quite bad because of that. Once FreeType has been upgraded we can
change the default.
Even then the results are not ideal, perhaps due to missing subpixel
positioning and linear color blending support in BLF.
In some extreme modifications the fill triangulation cannot be right, so need a refresh.
This is done only for sculpt brushes that can change the geometry.
Supporting the ctrl+click operator for multiple objects.
When no vertex is selected we only add a new vertex for the active object.
I'm using all the selected vertices to calculate the center of transformation
regardless of the object they come from. This way we have the same effect if the
meshes are joined together or kept as individual objects.
Thanks Campbell Barton for reviewing the patch.
OpenGL rendering only implemented the deferred renderer. This commit
will add the forward renderer. The forward renderer is used when XRay
mode is enabled
The loop cut tool now works as expected for the tool-system,
where hovering previews the loop and clicking activates.
This uses a new gizmo type to pre-select the edge.
This commit adds a new method, DEG_foreach_ancestor_ID()
to accompany the existing DEG_foreach_descendent_ID().
It can be used to help print/collect all the ID's that
a given ID block depends on (i.e. all the datablocks that
need to be evaluated before the datablock of interest can
be evaluated)
The core logic of this implementation comes from the drawobject.c from 2.7x.
I'm using `ts.colorGridAxisZ` for its color because of the legacy code:
```
UI_GetThemeColor3ubv(TH_GRID, col1);
UI_make_axis_color(col1, col2, 'Z');
glColor3ubv(col2);
```
There is nothing stopping us from adding this to the theme though, as it should
have been the case to start with.
This is quite confusing in the current UI, with both startup.blend and
workspaces.blend containing a list of workspaces. In practice you'd usually
want to save workspaces to both files.
The downside of having a single file may be that you then can't disable
certain workspaces by default, but we could add a setting for that.
These are taken from HDRI Haven (https://hdrihaven.com). Many thanks to
Greg Zaal for creating and releasing them under a CC0 license!
The HDRIs are intended to be used in the Lookdev display mode. With
1K resolution and compression they are not intended for final renders,
but it keeps the file size small and the difference is not noticeable
for lighting.
Command used for compression:
oiiotool %s --ch R,G,B -d float --compression dwab:300 -o output/%s
This reuses the tool defines for the 3D view. However, for some reason,
the events don't seem to be getting fired off here, like the tool system
is getting ignored here completely.
FIXME: Resolve the event handling issues
When OpenGL renderer is selected the option for background colors are
displayed, but ignored. For now we hide this option as it makes no sense
to have a viewport color option without viewport or render a 'final'
with a theme color.
blosc embedded a copy of zlib/pthreads causing duplicate symbol linker errors. pthreads was windows specific, but zlib may apply to other platforms as well.
This avoids getting all overridden objects from said collection
instantiated in master collection, totally untidy!
Note that there is still an issue when 'active' object (i.e. the
armature usually) is in hidden sub-collection, since you have to make
that whole hidden collection visible again to see it.
We'll need to support moving objects into overridden collections am
afraid, arg. :(
Without this, there was no simple way to have
launchers for different app-templates.
Also allows force-disabling the app-template stored in the preferences.
Very dummy mistake (someone forgot to increment one of the variables in
one of the loops in that spaghetti nightmare that is nurbs shapekey
code), took half an age to spot it... :/
* 3D Markers to Mesh
* Link Empty to Track
* Set as Background
Note, the "Setup Tracking Scene" was not addressed. Pending for this script is:
* Update world for EEVEE settings
* Adjust the layer code to use collections
* Remove broken world API
The only difference from 2.7x is that we were using glLineWidth(2.0) before,
which is not supported reliably in the latest OpenGL.
Big thanks to Clément Foucault for helping with this, on the design as well as
debugging the implementation.
For the records, I'm creating and destroying batches for the path geometry at
every draw. The pipeline is as follows:
1. Create GPUBatch with all the points
2. Create new shgroup with simple shader
3. Add call with the gpubatch
4. Save it in a linked list in stl->g_data
5. Free it after all drawing is done
The function must be static. The error was the definition in header file must be removed. This error was not detected by Microsoft compiler but Linux compiler fails.
-Moved from dynamic link to static on windows
-gained lcms/tinyxml/yamlcpp deps, since we need a little more control over the build flags than the build-in options will provide.
This operator clean any stroke below a defined number of points. This is used because sometimes when use eraser some points keep missing.
Also some changes in the UI menu.
The initial design assumed that there was only one object for each unique name, but that was not the case when instances were created.
Now, instances are supported and speed has been greatly improved when repetitions are used.
As a result of this change, the option to create objects has been removed in the Instances modifier. This option was strange and was also against Blender's design rules, since a modifier should never create objects. The old functionality of the modifier can be achieved with instances.
Also, several memory leakage problems that were not previously detected have been eliminated, and especially in the grid and in the drawing process
Onion Skin is not supported in multi-user datablocks. Support this, makes incompatible with instances. We need find a solution in the long term, but now it's better keep disabled and make instances work. Anyway, the new instances makes unnecessary to use muli-user datablocks.
Adds a displacement support for OpenSubdiov based subsurf object implemented
as a callback which gives vector displacement in object space. Currently is
implemented to calculate displacement based on myltires displacement grids,
but we can support things in the future if needed.
Submitting to review to see if there is something obviously wrong in the
direction (old multires code was sharing same displacement code to both
calculate final displaced mesh and reshape an existing one, which is rather
confusing and probably can be done more cleanly?).
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D3604
This includes selections both in the clip editor as the viewport.
The selection is implemented as a synchronization function called from the eval
selection update node in depsgraph.
* Draw bundle names.
* Draw bundle shapes (but spheres).
* Draw custom colors for bundles.
* Support selecting bundles.
* Object bundles also fully tested.
Missing from this commit:
* Camera path.
* Spheres are not drawn as solid.
* Blue dashed line from camera to tracker objects.
Regenerating all to remove zero area triangles.
Previously icons included files not committed to lib/.
This is the output of 'make icons_geom' for svn rev: 62081.
Previously back-facing normal was checked which isn't ideal
because a triangle that starts out zero-area may become
visible once it's coordinates are quantized.
Add tool options to control how select operates (add/sub/set/and/xor).
Note: edit mode armature select still needs to support all options,
this is complicated by how it handles partial end-point selection.
This is a manual synchronization. If we need to re-use the same util in both
areas it would be welcomed as well. For now this at least get things back to a
working state.
Conversion code could leave object with inconsistent material data
compared to its new obdata.
Ideally, various conversion code would handle that properly, conserving
materials when possible, but for now at least ensure we get valid
result!
Related to T56363, this is not fixing the root of the bug, but ID
copying should always be a good occasion to ensure sanity of our data
(and error checking is always better than a crash!).
Currently no functional changes, just exposes all settings which we need
for OpenSubdiv, similar to what Subsurf modifier is doing already.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D3602
* http -> https
* Release Logs -> Release Notes
* Match number of items on both sides by adding a separator
* Remove Python API Reference link since it's already under Help
when Developer Extras is enabled.
Note to myself, next time, better check the fix before pushing it.
GL_ARB_texture_gather is defined if there is support for the extension
not only when the extension is enabled. Do this check ourself with
GPU_ARB_texture_gather define.
Original fix 822de6e9e1
A lot of code in codegen was used by previous GLSL system.
Now most of it is not used due to all the things being written from scratch
in 2.8 and we can clean most of it.
As a side not this make the system a bit less flexible (no support for
cubemaps, preview image, etc...) but can be extended again.
* Remove GPU_link_changed which is unused.
* Remove all GPU link function that are not used anymore.
* GPU_uniform_buffer is now GPU_uniform.
* GPU_texture_ramp is now GPU_color_band.
* GPU_uniform is now GPU_constant.
The extension GL_ARB_texture_gather is reported to be supported and does
not trigger an error when enabled but the textureGater functions are not
defined.
Workaround is to disable the use of this extension on such systems.
Currently constraints can only read the location along the
spline. This obviously limits opportunities for complex bone
interactions in rigs.
This patch exposes access to rotation and scale as well in
Copy Transforms. However, due to the way how things work,
this data cannot be smoothly interpolated, and abruptly
changes when switching to the next segment.
Reviewers: aligorith
Differential Revision: https://developer.blender.org/D3545
Scale is a multiplicative quantity, so adding it doesn't make sense.
However, for backward compatibility reasons, and in case somebody
actually desires the old additive behavior, the old way remains as
an option.
Without this change the only way to properly combine scale is via
parenting or the complicated Transformation constraint.
The new mode is turned on by a flag for file compatibility, but the
RNA option is reversed so that the new behavior feels more default.
Reviewers: aligorith
Differential Revision: https://developer.blender.org/D3558
Some constraints have an option to take the final bezier shape of
the target B-Bone into account. This shape usually depends on two
other bones in addition to the target itself, so the graph should
include the relevant dependencies.
Reviewers: sergey
Differential Revision: https://developer.blender.org/D3591
To get consistent, user-expected results here, we need to 'fake'
starting immediately after a 'skip' block (such that we start with a
full block of selected elements).
Same issue affected vertices and edges selection of course, did not
check the other usages of WM_operator_properties_checker_interval_test()
though.
seems like the openexr 2.3.0 release tarball has broken cmake support, latest from git works
we'll have to revisit this once they get a new release out.
For users it defines how accurate vertex positions are in terms
of limit surface (as in, how close the vertices locations to the
condition when they are calculated for an infinitely subdivided
mesh).
This affects things like:
- Irregular vertices (joint of 3 or more edges)
- Crease
Keep quality value low for performance.
NOTE: Going higher does not necessarily mean real improvement
in quality, ideal case might be reached well before maximum
quality of 10. Quality of 3 is a good starting point.
Internally quality is translated directly to adaptive subdivision
level.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D3599
Beautiful example of typo going unoticed and firing back up in totally
unexpected place years later. Guess nobody actually duplicated a Clip
data-block before! :P
Most likely own fault, during refactor of ID copying code.
There are following reasons to do so:
- The plan is to replace it with some sort of object or viewport option,
so we can apply OpenSubdiv subdivisions on top of modifier stack and
keep modifier stack purely CPU side.
This will solve issues when adding some relation in scene will force
modifier to be evaluated on CPU.
- With new upcoming OpenSubdiv based CPU modifier implementation we can
cache topology similar to what GPU side was doing, which will already
be reasonably faster.
- OpenSubdiv GPU does not work since the OpenGL version bump, and is
to be rewritten with all the adaptive refine options kept in mind.
Since OpenSubdiv GPU was already broken and was only causing object
to become invisible, there is no reason to keep having that option in
the modifier.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D3598
This replaces old single toggle option to subdivide UVs with
an enum which can have more options. The usecase for this is
to be compatible with other software. But we also might choose
different subdivision type as default in the future.
DNA and underlying code supports all possible options, but
only the ones which are compatible with old subdivision code
are currently exposes.
Reviewers: brecht
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D3575
C-API is way smaller than the rest of the code which uses it.
So better to conditionally compile stub implementation than
to keep adding ifdef everywhere.
Currently drop operators work mostly by specifying the name of the datablock.
However there can be datablocks with the same name in different libraries, so
this gives wrong results in some cases.
Currently only outliner drop operators have been updated to use this mechanism.
Color ramp with constant interpolation must bypass texture filtering and
use nearest neighboor sampling in order to appear correctly sharp.
This patch use a GLSL hack to use nearest sampling on thoses particular
color band.
This lower the use of texture samplers slots and let users use more real
textures in their shaders.
This patch also make the ramp texture 16 bit floating point. Meaning you
can now use value greater than one in your color ramps.
With the limit of 128 colorband per shader (a color band being either a
color ramp, a wavelength node or a curve node (and maybe wavelength node in
the future)).
Only drawback with the current implementation is that it does not remove
colorband from pruned GPUNodes but it shouldn't really matter in practice.
This should fix T56010
This is a rough (but fast) approximation that still match cycles reference
in common case.
In practice, it's just adding more of the diffuse light computed for the
diffuse contribution.
As grease pencil use multiedit frames instead of multiobject edit, this fix solves the issue.
In the future maybe will need modifications if we add multiobject support, but we need a solution now.
This is more like a static optimisation when some parameters are set to 1.0
or 0.0. In theses case we use a more optimized version of the node.
This also revisit the transmission parameter behaviour to make it closer to
cycles.
This was likely caused by some sort of race condition where the drawing
thread would request the state of the shader before the shader has been
compiled.
Thanks to Charlie Jolly (mistajolly@gmail.com) for his patch D3586 that added create materials to opacity modifier.
I had to do some more changes to get all running.
Patch porting to OpenJPEG 2.3 is by Campbell.
Once all platforms are upgraded we can remove the code for 1.5, and upgrade
or remove the openjpeg version from extern/. This intermediate step makes it
possible for platform maintainers to upgrade to 2.3 without breaking other
platforms.
This memory leak was undetected during a long time, but with new memory checking is visible.
The problem was the stroke buffer batch was realocating new batches without free the memory.
new background image/clip of Camera ID was totally wrong, down the old,
broken 'way it used to be' instead of using new, more generic system.
Those ID pointers were not even added to library_query.c file, shame! xD
The material created was not right when apply the modifiers.
These errors were related to the material modification from old palette system before the merge and for any reason this code was not changed in the right way.
Also changed the "Create Colors" to "Create Materials" to keep UI names aligned.
This is a limitation of the current operator design.
I have added a test to be sure the operator is not initializated while play animations to avoid segment fault.
In the future, we can enable this option again, but it will need a operator redesign.
This make the root flag writable using the Python API, using the
generic skin vertex flag setter function.
Reviewed By: Campbell Barton
Differential Revision: http://developer.blender.org/D3583
The object can be NULL. This was a line used in debug that it was not removed.
Also removed an old function not used.
Thanks to Charlie Jolly for catching this error.
Note that there are most certainly many other operators that’d need that
same flag... Don’t have time to hunt them down currently, will just fix
as issues are found, for now.
For now, that flag is only used in redo code, since after undo step
depsgraph is totally empty... We *may* want to add at least an assert in
op calling code too, though?
Some of them are unecessary and should be removed from the shader instead.
But for now we need a quick fix for the crashes happening on some platforms.
See T55475.
Calling code is responsible to check on NULL pointers here, or ensure
is does have a fully built and evaluated depsgraph, but this should
never crash in DEG queries themselves.
Enabled infinite sharp patches for topology refiner and evaluator,
which allows to have sharp edge at first subdivision level.
Also tweaked crease export from Blender to OpenSubdiv to have more
artistic control over the whole 0..1 range.
In some cases (e.g. using old userpref settings/keymaps)
it was possible to trigger a crash when the wrong GP/Annotation
operators were triggered in the wrong contexts (e.g. using
the old GPENCIL_OT_paint in annotation-only contexts like
all the 2D editors).
This commit resolves several issues that were caused by sloppy
code-churn + features that had been hacked on.
Since most animators find Motion Paths more useful than Armature Ghosting:
* Move Motion Paths before Ghosting settings (less scrolling)
* Collapse Ghosting panel by default
* Open Motion Paths panel by default instead
Pretty minor, from 0.6 to 0.8, but the improvement is noticeable
especially when using a stylus, without overlapping too much with
the buttons and dropdowns in headers.
Rendering OpenGL/Preview is accessible from each editor.
Render settings are accessible from the Film menu when in OpenGL/Preview engine.
It wasn't always predictable especially with Workspaces without or with many viewports.
Also reordering of items, renaming and removal of superfluous icons.
texelFetch return vec4(0.0) if the target pixel is outside the texture
rect. So we mimic the default repeate mode that we have for linear
interpolation.
Fix T56156 Mapping-Node doesn't work
The idea is simple: do not provide full topology to OpenSubdiv, leave
edges creation to OpenSubdiv itself. This solves issues with non-manifold
meshes which were known to fail, including the ones from T52059.
On a positive side we can simplify our side of converter, keeping code
shorter.
it is still possible that we'll need to ensure all loops has same
winding, but that is less things to worry about.
Applies to vertices and edges. Biggest annoyance here is that OpenSubdiv's
topology converter expects that there is no loose geometry, otherwise it
is getting confused.
For now solution is to create some sort of mapping from real Mesh vertex
and edge index to a non-loose-index. Now the annoying part is that this
is an extra step to calculate before we can compare topology, meaning FPS
will not be as great as if we knew for sure that topology didn't change.
Loose edges subdivision is different from what it used to be with old
subdivision code, but probably nice feature now is that endpoints of loose
edges are stay at the coarse vertex locations. This allows to have things
like plane with hair strands, without need to duplicate edge vertices at
endpoints.
All this required some re-work of topology refiner creation, which is now
only passing edges and vertices which are adjacent to face. This is how
topology refiner is supposed to be used, and this is how its validator
also works. Vertices which are adjacent to loose edges are marked as
infinite sharp. This seems to be good-enough approximation for now. In the
future we might tweaks things a bit and push such vertices in average
direction of loose edges, to match old subdivision code closer.
Having 'flag, flag2, flag3' is getting out of hand especially
when we support increasing the size of types.
Make flag2 into an int.
Note, this looses the 'show world' option,
but it's not such an important setting.
This also include a small optimisation (remove of a double loop and half of
the memory allocation for hit points)
This should fix T55946 Crash using knife tool on mesh with large number of
vertices.
Tried with a 500K vert suzanne and it seems fine.
instead of being ThreadLocal and leading to incorrect usage.
We still enforce no framebuffer when changing context. We can lift this
restriction later.
After clearing motion paths from objects, those objects needed to be
tagged for copy on write so that the copied data (i.e. viewport)
recieve the changes (i.e. removed paths)
Reported by Hjalti
Avoid double label for same properties in single-column.
Onion Skinning: Group custom colors together, and frame before/after together.
Small changes to tooltips.
Top bar is only for settings that apply to the next action
not a way to change existing data.
If each stroke could have a different color this would work as expected,
however it was adjusting the current layer color.
Rename "Seq. Strip" to Strip Name, "Sequence Strip" to Use Strip Metadata.
Plus re-arrange of properties and separators for sections.
Thanks @fsiddi for the feedback.
This commit merge the full development done in greasepencil-object branch and include mainly the following features.
- New grease pencil object.
- New drawing engine.
- New grease pencil modes Draw/Sculpt/Edit and Weight Paint.
- New brushes for grease pencil.
- New modifiers for grease pencil.
- New shaders FX.
- New material system (replace old palettes and colors).
- Split of annotations (old grease pencil) and new grease pencil object.
- UI adapted to blender 2.8.
You can get more info here:
https://code.blender.org/2017/12/drawing-2d-animation-in-blender-2-8/https://code.blender.org/2018/07/grease-pencil-status-update/
This is the result of nearly two years of development and I want thanks firstly the other members of the grease pencil team: Daniel M. Lara, Matias Mendiola and Joshua Leung for their support, ideas and to keep working in the project all the time, without them this project had been impossible.
Also, I want thanks other Blender developers for their help, advices and to be there always to help me, and specially to Clément Foucault, Dalai Felinto, Pablo Vázquez and Campbell Barton.
Usage of matcap image uniform had different ifdef than definition
of that uniform. Assuming the usage was correct, and the definition
needed an update.
Prevents shader from compilation failure and from aborts in debug
builds.
The idea is to create vertices along the coarse edges once, without
splitting coarse edges on separate ptex faces. This requires some
indexing magic, vertices within a patch are no longer sequential.
Not sure how to make it nicer without such a black magic looking
calculations (which are basically boiling down to mimicking order
of verts/edges creation).
In the current offsets calculation loose verts and edges are not
properly taken into account, but those are causing topology refiner
to fail anyway, so it needs a bit deeper change.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D3570
This ignores the scene color managment view settings for solid mode and
lookdev when not using scene lights and world. The scene settings are
intended for tweaking renders and should not affect studio lighting and
matcaps.
There may be cases where a simple sRGB transform is better than Filmic
and we could add configuration for this. Not sure if it really matters
and it may be better if we just assume matcaps and studiolights are all
created for one view transform.
Differential Revision: https://developer.blender.org/D3569
By popular demand, the CLean Keyframes operator will now
leave handles and other interpolation settings untouched.
Previously, it would recreate the keyframes from scratch,
keeping only the frame + value, under the assumption that
the handle information was "bad" (i.e. the source of bumps
and roughness, due to bad hand tweaking). However, since
most animators use this on hand-keyed animation instead of
motion-capture data, this assumption didn't hold, and was
actually overly destructive - wiping out lots of hand-adjusted
curve data.
This is in preparation of upgrading our library dependencies, some of which
need C++11. We already use C++11 in blender2.8 and for Windows and macOS, so
this just affects Linux.
On many distributions this will not require any changes, on some
install_deps.sh will need to be run again to rebuild libraries.
Differential Revision: https://developer.blender.org/D3568
They were added as placeholder to show something until they lazy-load.
But since the load is fast and it's hard to fit the text, and their name
is displayed already in the tooltip, we can do without for now.
In the future when we have insta-tooltips we should make them use this.
These were made by the community and curated by Pablo and William.
Thanks to everyone who contributed!
https://devtalk.blender.org/t/call-for-content-matcaps/737
Command used for compression:
oiiotool %s --ch R,G,B -d half --compression dwab -o output/%s
Code was trying to hide properties by name that may not exist.
Check if it was actually found and add 'files' to the filter,
since it is what WM_OT_studio_lights_install uses.
Just basic algebra - because all vectors have the same z coordinate, a lot of terms end up cancelling out.
Not exactly a massive improvement, but it's measurable with Branched PT and a high sample count on the lamp.
Reviewers: brecht, sergey
Reviewed By: brecht
Subscribers: swerner
Differential Revision: https://developer.blender.org/D3540
Only enforce origindex to NONE for a generated geometry. For the rest
of geometry rely on CustomData_copy() to set it to the proper value.
This will ensure origindex is set correct for cases when there is an
array modifier prior to subsurf.
Gathers information about object geometry and textures. Very basic at
this moment, but need to start somewhere.
Things which needs to be included still:
- "Runtime" information, like BVH. While it is not directly controllable
by artists, it's still important to know.
- Device array sizes. Again, not under artists control, but is added to
the overall size.
- Memory peak at different synchronization stages.
At this point it simply prints info to the stdout after F12 is done,
need better control over that too.
Reviewers: brecht
Differential Revision: https://developer.blender.org/D3566
We use a hidden window for each offscreen context we need.
On X11 (linux) it does not show any other windows in the OS task bar
but it might be the case on other operating systems (untested).
Replaced the draw world option with a shading.background_type enum.
Where the user can select Theme, World or a Custom color.
World and theme colors do not always work in workbench. We needed to
have an option what the user could control locally (per viewport).
Especially when using linked data.
I removed the world background drawing from the draw_manager. It was never used as EEVEE and Workbench both override the logic.
Not 100% sure about the naming of Theme, World, Viewport.
In other parts of blender's codebase World is sometimes called Scene.
Will stick to the names that describes its location best.
{F3990139}
Reviewers: fclem, campbellbarton
Reviewed By: fclem
Subscribers: venomgfx
Tags: #bf_blender_2.8
Differential Revision: https://developer.blender.org/D3551
There is no reason or justification to have helper functions as
class methods: they do not depend on anything in the class itself.
There are probably more cases like that.
Show "Operators Cheatsheet" and Python API link only when developers extra is enabled.
Fix URL for User Communities, Developer Community and Release Notes (which 404s
now just like the wiki anyway since the page for 2.80 is being made).
While changing the shading normal is a great way to add additional detail to a model, there are some problems with it.
One of them is that at grazing angles and/or strong changes to the normal, the reflected ray can end up pointing into the actual geometry, which results in a black spot.
This patch helps avoid this by automatically reducing the strength of the bump/normal map if the reflected direction would end up too shallow or inside the geometry.
Differential Revision: https://developer.blender.org/D2574
Also always draw the counter of elements-per-type with a dark background
regardless of the active status. It being white when active affects
readability since the icon background itself is already highlighted.
Thanks devtalk forum for feedback.
RNA API Object.ray_cast would not normalize direction vector before
doing first quick bbox intersection test, while using its returned
distance value. This could lead to wrong exclusion of object.
Thanks to @codemanx for finding that issue.
In the outliner, right click > view layer > set indirect only. This is
like clearing camera ray visibility on objects in the collection, and is
temporary until we have more general dynamic overrides.
In the outliner, right click > view layer > set holdout. This is
temporary until we have more general dynamic overrides, but helps
Spring production for now.
Color and Properties panels as sub-panels of Node. Collapse Node properties
by default, since they are already in the node itself and in material properties.
Initial work on single column layout, flow and organization of
the texture properties. More work needs to be done in the C templates
for image textures.
See D3557
But I'm not sure if the flags `BA_WAS_SEL`,` BASE_SELECTED` and `BA_SNAP_FIX_DEPS_FIASCO`(lol XD) should be added to the bases of the not evaluated `view_layer`. This needs to be discussed.
For some reason 32c5972653 broke display of
solid meshes in workbench.
After some investigation, it seems that the vertex coordinate output is
degenerated even if the input is correct and the matrix too.
Removing dead code seems to fix the problem. So maybe the GLSL preprocessor
is not doing what it should?
This reverts commit 81a93df6d2, it is not safe
to handle initialization for startup.blend differently. Instead fix the root
issue of the preview icon data structures not being initialized in time.
This puts the motion path in the same location as the transform gizmo,
which is less confusing especially if you have a custom bone shape where
the tail is not visible.
The old springs with damping 1.0 operate in a special way that
is more similar to plastic deformation than a spring. Some users
rely on that, so let the user choose which implementation to use.
This also restores full backward compatibility with 2.79.
Reviewers: sergof
Differential Revision: https://developer.blender.org/D3544
Previously it was ptex faces which were subdividing to the same
resolution. This was looking like more details for non-quad faces,
but was also causing discontinuity in the edge where quad touches
non-quad polygon.
Now ptex faces which are coming from non-quad faces are subdivided
at a half of resolution, matching old behavior and solving
discontinuity problem.
Note: Moved doversion of VSE strips uniquename to 2.8 versionning area,
and raised accordingly current file subversion, since that bug also
affected previous 2.8 .blend files...
Conflicts:
source/blender/blenkernel/BKE_blender_version.h
We actually still had cases of Meta strip duplication resulting in
non-unique strip names. Quiet surprising this went unoticed for so long. :(
Fixed that bug, and think it was last one (at least, no other case of
SEQ_DUPE_UNIQUE_NAME usage should be broken, I think...), and raised
subversion and updated doversion to run uniquename check on strips on
all previous fileversions.
Note: will have to do that again when merging in 2.8...
- Use the object referenced in `BMEditMesh` as the `ghash` key to save the bvhtrees in cache;
- Create a boundbox around edit_mesh to test the snap before creating bvhtree;
- Save the `edit_mesh`s bvhtree in the mesh bvh_cache;
This is a part of the D3504.
The issue was caused by Render Single Layer option enabled, which is
very handy for artists work, so they can hit F12 and see view layer
they are currently working in a final rendered state. This saves a lot
of time since all the "non-interesting" objects are ignored for such
iterations.
However, for the render farm we need to render view layers which are
explicitly set for render, and ignore active view layer.
Reasonable solution seems to be to ignore the Render Single Layer
option when rendering from the command line. It is really something
more like UI behavior option.
Hopefully this will fix issue with camera rig where camera properties
(like, near/far clip) are driven by custom properties from bones, and
those bones are actually belong to proxied armature.
Add new tag to bSound (runtime flags), and make read code to set a 'no
reload waveform' new tag, since it uses a mapping to get existing
waveform in undo case...
Previously it was calling `BKE_pose_rebuild()`, such thing shall never
be called from drawing code! Hopefully this now works as expected and
that horrible hack is not needed anymore.
`BKE_pose_rebuild()` should (ideally) always trigger a rebuild of the
depsgraph, since it can add or remove posechannels.
This function now takes a Main parameter to ensure that related
depsgraphes are tagged as dirty (kept it optional, for some corner cases).
We should also probably double-check calls to that function, think in
theory it should only be called from depsgraph itself? But for now...
Damped Track by specification attempts to arrive at the desired
direction via the shortest rotation. However with opposite vectors
there are infinitely many valid 180 degree rotations. Currently
it gives up and does nothing.
I think that it would be more reasonable to resolve the ambiguity
arbitrarily, so that Damped Track won't have a weird dead zone.
To make it more predictable I use a local axis.
In addition, the singularity area vicinity has some floating
point precision problems that result in significant jitter.
This applies workarounds for two causes of instability.
Differential Revision: https://developer.blender.org/D3530
GPUFrameBuffers were being free when no context was attached or in the
wrong gl context. This make sure this does not happen again.
You can now safely free any gl resource from any thread (well as long as
it's not used anymore!).
This increases stack memory usage some, and ideally we'd support a dynamic
size. But this is quite difficult on the GPU and hopefully 32 is enough even
for very complex cases.
When editing a Material the depsgraph can throw away the evaluated ID
before the preview job localized the said ID.
To fix this we localize the ID from the main thread.
Also fix WM_OT_previews_ensure crashing because of no depsgraph.
Was only visible when doing command line, since it was happening due to
cache-free policy which was aimed to bring memory usage down.
The issue is that if object with particle system is used as a nested
duplicator multiple times, it will only generate children first time,
and after that its caches are freed. After that duplication system
can not generate any instances, since the path cache is lost.
Now we delay caches free to after all objects are synchronized, which
ensures all instances are generated.
This will increase a memory peak a bit during object synchronization
time, but overall it shouldn't be that bad, since memory footprint
after synchronization will stay the same as before this change.
The ultimate thing to do here would be to drop the whole dependency
graph away, but this will require:
- API on engine side, to inform it to drop the dependency graph.
- Changes in Cycles report system to NOT use evaluated scene to get
scene name (evaluated scene will be gone with dependency graph).
The idea is to use this as a replacement of old CCG, now it is
based on OpenSubdiv. The goal is to reduce any possible overhead
which was happening with OpenSubdiv used by CCG.
Currently implemented/supported:
- Creation from mesh, including topology on OpenSubdiv side,
its refinement.
- Evaluation of limit point, first order derivatives, normal,
and face-varying data for individual coarse position.
- Evaluation of whole patches.
Currently not optimized, uses evaluation of individual coarse
positions.
- Creation of Mesh from subdiv, with all geometry being real:
all mvert, medge, mloop, and mpoly.
This includes custom data interpolation, but all faces currently
are getting separated (they are converted to ptex patches, which
we need to weld back together).
Still need to support lighter weights grids and such, but this
is already a required part to have subsurf working in the middle
of modifier stack.
Annoying part is ifdef all over the place, to keep it compilable
when OpenSubdiv is disabled. More cleaner approach would be to
have stub API for OpenSubdiv, so everything gets ifdef-ed in a
much fewer places.
This is a physically-based, easy-to-use shader for rendering hair and fur,
with controls for melanin, roughness and randomization.
Based on the paper "A Practical and Controllable Hair and Fur Model for
Production Path Tracing".
Implemented by Leonardo E. Segovia and Lukas Stockner, part of Google
Summer of Code 2018.
This patch adds a new matte node that implements the Cryptomatte specification.
It also incluces a custom eye dropper that works outside of a color picker.
Cryptomatte export for the Cycles render engine will be in a separate patch.
Reviewers: brecht
Reviewed By: brecht
Subscribers: brecht
Tags: #compositing
Differential Revision: https://developer.blender.org/D3531
This does not include all the struct and type renaming. Only files were
renamed.
gwn_batch.c/h was fusioned with GPU_batch.c/h
gwn_immediate.c/h was fusioned with GPU_immediate.c/h
gwn_imm_util.c/h was fusioned with GPU_immediate_util.c/h
This is intended for quick renders for previsualization, animation previews
or sequencer previews. It provides the same settings as found in the 3D view
Shading popover in solid display mode, but in the scene render properties.
The "Workbench" engine was removed, and this name no longer appears in the
user interface, it's purely an internal name. We might come up with a better
name for this OpenGL engine still, but it's good to be consistent with the
OpenGL Render operator name since this has a similar purpose.
metadata loading code was assuming all videos in Blender were from
FFMPEG... added empty place-holders for other types too, we probably
could load some metadata from pictures or AVI files too!
The Properties->World tab had no Viewport Display panel. The world color
itself was hidden when the 'use_node_tree' was enabled.
Also renamed the World.horizon_color to World.color as it has nothing to
do with the color of the horizon (old BI feature)
This does not fix the smokesim. It only port the drawing method.
The Object mode engine is in charge of rendering the velocity debugging.
Things left to do:
- Flame rendering.
- Color Ramp coloring of volume data.
- View facing slicing (for now it's only doing sampling starting from the
volume bounds which gives a squarish look)
- Add option to enable dithering (currently on by default.
Only fixes compilation error, the functionality will be limited.
Currently we don't care that much, since all the work is done in
the branch anyway.
Later on when we'll know which fixes we need to apply on top of
latest OpenSubdiv library we will call a library upgrade.
Main goal is to make API simpler to follow (at least ion terms what
is defined/declared where, as opposite of handful big headers which
includes all the declarations), and also avoid a big set of long and
obscure functions.
Now C-API files are split into smaller ones, following OpenSubdiv
behavior more closely, and also function pointers in structures
used a lot more, which shortens functions names,
UV integration part in GL Mesh is mainly stripped away, it needs
to be done differently. On a related topic, UV coordinates API in
converter needs to be removed as well, we do not need coordinates,
only island connectivity information there.
Additional changes:
- Varying interpolation in evaluator API are temporarily disabled,
need to extend API somewhere (probably, evaluator's API) to inform
layout information of vertex data (whether it contains varying
data, width, stride and such).
- Evaluator now can interpolate face-varying data.
Only works for adaptive refiner, since some issues in OpenSubdiv
itself.
Planned changes:
- Remove uv coordinates from TopologyConverter.
- Support evaluation of patches (as opposite to individual coordinates
as it happens currently).
- Support more flexible layout of varying and face-varying data.
It is stupid to assume varying is 3 floats and face-varying 2 floats.
- Support of second order derivatives.
- Everything else what i'm missing in this list.
Features to get the 2nd, 3rd, 4th closest point instead of the closest, and
various distance metrics. No viewport/Eevee support yet.
Patch by Michel Anders, Charlie Jolly and Brecht Van Lommel.
Differential Revision: https://developer.blender.org/D3503
Useful to store a snapshot of the current keymap state
so changes to the default keymap are ignored.
Also useful for testing keymap export works properly.
This better aligns the matcap/hdri browser, rotation/background sliders and
preferences/flip matcap button.
The remaining sub-panels are not perfeclty aligned yet, once the sub-panels
separation is more prominent (and they can be collapsed) we can align those.
Thanks devtalk forum for feedback!
This works for Cycles, Eevee, texture nodes and compositing. It helps to
reduce the number of math nodes required in various node setups.
Differential Revision: https://developer.blender.org/D3537
crashed when using filtering and no item was found ('Blender File' mode)
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D3535
- Rename "Viewport Info" to "Text Info".
Name was too vague, nearly everything is information,
this currently only controls overlay text.
- Swap text-info & 3D-cursor, making 3D-cursor less prominent.
- Logical use of fields since they define type information.
- Avoids using ordered-dict metaclass.
Properties using regular assignments will print a warning and load,
however the order is undefined.
This brings the Object Visibility dropdown closer to overlays which
also controls visibility of elements int he viewport, so it makes more
sense there. Also placing the Overlays settings closer to the viewport.
Sub-panels for shading, measurement, normals, freestyle and developer.
Less-used sub-panels will be collapsed by default once their implementation
is complete (fixing spacing between sub-panels, adding the triangle to collapse, etc).
Also added missing check for overlays on/off on existing sub-panels.
* Unify X-Ray/Shadow toggle and factor, using the whole width gives you finer control
* Shadow settings next to intensity slider
* Cavity settings next to Ridge/Valley sliders
* Collapse Cavity settings when not used
* Make MatCap preview and browser dialog smaller. MatCaps are added by the user
so she is already familiar with how they look. No need to preview them that big
while browsing. To see the MatCaps in detail we have the User Preferences/Asset Manager.
Having popovers inside popovers triggers some minor glitches that will be addressed later.
(e.g. they open on mouse over and the popover arrow is not aligned)
Revert changes from 785159e6e4
but keep 'ifdef'.
@mont29 maintains this area and prefers to keep existing logic.
Note that there was misunderstanding that '*' was intended only
to be a backup key for '=' for keyboards which require holding a
modifier.
Group Object Type Visibility with the viewport settings in the header.
The eye icon is a placeholder. A new icon is needed to better communicate
the user at a glance if all (or none) object types are enabled, or why certain types
are not visible/selectable in the viewport.
Part of design: T55863
When tweaking the shading/overlays settings through the popovers,
these get on the way of the content in the viewport. Making it hard to
see the result of our changes.
Unfortunately this breaks consistency with other headers where the snapping
tools are aligned to the right, but the benefits outweigh the costs.
Organize content in categories/sub-panels.
The result is a more organized but slightly taller popover.
Once sub-panels are implemented code-wise, the alignment issues
(like panel label) and extra space between the content and panels
headers will be reduced. Together with the ability to have certain
less-used popovers collapsed by default, this will make the
popover more compact.
Part of design: T55863
- Made OpenSubdiv_GLMesh private
Previously, it was still accessible via C-API from C++ code.
- Don't implicitly refine evaluator when updating coarse positions,
now there is an explicit call to do this.
Allows to first apply all changes to the coarse mesh and then
refine once.
- Added coarse positions update from a continuous buffer with given
starts offset and stride.
Allows to update coarse positions directly from MVert array.
- Refiner is no longer freed when CPU evaluator is created.
Allows to re-use refiner for multiple purposes.
There are now 3 categories in the overlay popover:
- Navigation
- Active (camera, lamp... etc)
- Tool (manipulator)
The user preference for mini axis now controls if the mini axis
displays minimal or a full-interactive widget.
Part of design: T55863
This reverts commit 8121010310.
After user feedback this has the downside of having no predictable
way of transforming in global space.
Since toggling between global/user is reversed when global is
the user axis.
This is no longer needed with the new offscreen draw method, so use whatever
is default and hopefully fastest. Fixes console warnings in some setups that
don't have swap copy.
Remove support for loading interlaced image sequences because
its less common now to record interlaced video,
the option to de-interlace video on load remains.
This separate probe rendering from viewport rendering, making possible to
run the baking in another thread (non blocking and faster).
The baked lighting is saved in the blend file. Nothing needs to be
recomputed on load.
There is a few missing bits / bugs:
- Cache cannot be saved to disk as a separate file, it is saved in the DNA
for now making file larger and memory usage higher.
- Auto update only cubemaps does update the grids (bug).
- Probes cannot be updated individually (considered as dynamic).
- Light Cache cannot be (re)generated during render.
- Change gl_context_mutex to a ticket mutex ensuring interactivity even
when rendering.
- Add DRW_custom_pipeline for using DRW outside of the viewport and render
pipeline.
- DRW_opengl_render_context_*** and DRW_gawain_render_context_*** to use
with DRW_custom_pipeline.
- Add possibility to bypass deferred compilation on demand (not User
demand).
- Add union to access DRWMatrixState members more easily.
- Texture creation now requires explicit data type.
- GPU_texture_add_mipmap enable explicit mipmap upload.
- GPU_texture_get_mipmap_size can be used to get the size of a mipmap level
of an existing GPUTexture
- GPU_texture_read let you read back data from a gpu texture.
Numeric input allowed mix of editing and hotkeys which were interpreted
as modifiers instead of using as numeric input.
This meant entering '1.0*3' needed to be typed as '1.0**3'
('*' to activate, and again to multiply).
Pressing '/' gave the reciprocal of the current number
which could be useful.
Test removing this feature, so only full numeric input is supported.
The approach of setting 'refresh' flags on the modifier, and performing
the associated actions when the modifier is being evaluated, is a bad
one. Instead, we use the separation of the original and the evaluated
copy to 'refresh' certain things (because they simply aren't set at all
on the original). Other actions are now done directly with BKE_ocean_xxx
functions on the original data, intead of during evaluation.
Previously CMake was raising a fatal error, which wasn't too helpful.
There is still some fatal messages about Audaspace and Game Engine,
but the latter one is on it's EOL and is removed in Blender 2.8.
The X resource database is to be explicitly destroyed. This fixes 46 bytes
leak per every window DPI query (which happens a lot on window move/resize
and even on areas resize).
Unfortunately, this does not fully fix the leak since the known leak:
https://bugs.freedesktop.org/show_bug.cgi?id=94604
Previously global was always used before the current orientation,
the order is now reversed, which assumes the user wants to first
use the value they set, with the global secondary.
A toggle for this is no longer required now that both header and background
colors are RGBA (disabling Show Header is the same setting the alpha to 0).
Thanks Brecht for reviewing!
There were a number of cases where immActivate() and immDeactivate() could
get out of sync, causing crashes due to using a freed mutex lock. Refactor
the code now to hopefully avoid this always.
Based on feedback from the 'User Feedback' devtalk forum
* More opaque panels for the viewport
* Darker state colors
* Current frame color consistency for MCE
Old behavior: tagging ID with DEG_TAG_COPY_ON_WRITE will do copy-on-write
(as requested), but will also flush changes to all operations with depends
on it. This means, for example, tagging object for copy-on-write will force
its modifier stack to be evaluated.
This was needed in the earlier days of copy-on-write when things were not
well defined and when lots of areas were not doing proper tagging.
New behavior: tagging ID with DEG_TAG_COPY_ON_WRITE will only ensure copy
of the dadatblock is up to date, without flushing updates to any dependencies.
This means following things:
- We can update parts of ID without invoking heavy computation of other
parts of same ID. For example, tagging object for COPY_ON_WRITE update
on mode change will not force modifiers stack to update.
- If some dependent datablock is dependent on pointers which are invalidated
by copy-on-write (like, evaluated mesh referencing custom data layers from
original mesh), this will either require explicit tag or explicit relation
in the dependency graph.
Currently can not find a faulty case since tagging of mesh happens with
either 0 (which means, everything) or with GEOMETRY, which also forces
all dependent modifier stacks to be re-evaluated.
This fixes missing PBVH when going into sculpt mode (sculpt mode toggle
was tagging object for COPY_ON_WRITE update, which was forcing modifier
stack to be updated, which was freeing PBVH.
Some other operations might also become faster with this change.
Instead of running code to create a keymap, store them as data.
This allows for keymaps to share content as well as running
transformations at load time.
Now we light with just a user defined HDRI by default, which is useful
for material setup and texture painting and lighting without having to
set up any scene lights.
Previously it would use the scene world without lights by default, which
in some files is just black.
Shortcut strings would be offset to the left to make space for the triangles,
breaking the alignment with other shortcut strings. Now this alignment is kept
by making menus slightly wider if there's a sub-menu triangle visible, making
room for the triangle.
Trying to have a single option for this is too likely to be
insufficient in some cases.
Instead, support object type visibility & selectability per view-port.
Remove depth testing flags from motion path pass drawing,
so that they always appear to draw in "x-ray" style on top
of everything, making it easier for animators to see what
they're doing.
This fixes an issue introduced in d192d72312.
When starting up, the UI hasn't been fully initialised yet, and calling
wm_event_do_depsgraph() can trigger icon updates when the startup file
contains an image, causing a segfault due to a not-yet-initialised ghash
for the icons.
The flag was only used in readfile.c, and resulted in a delayed call to
BKE_ocean_add(); this call is now immediately made instead as it's not
very expensive.
The previous commit only solves the problem when using the default
theme using factory settings. For previously saved themes, there could
still be problems, as the alpha values were still 0.
This commit improves the logic here so that while keyframe points on
unselected F-Curves will still get faded out (to not stick out too much
from the curves they live on), but the effect will not be as pronounced
(i.e. the points will stay visible all the time).
Restore theem setting default for Graph Editor's vertex,
with default alpha set to 1.0. The alpha value here needs
to be non-zero, as the alpha values currently get used
when drawing verts.
This makes it possible to have a single shading nodetree that contains
separate Cycles and Eevee shaders. By default the target is set to All
so shaders are shared.
* Z now goes to solid mode when in lookdev or rendered mode.
* Alt-Z was broken after removal of texture mode, now toggles lookdev mode.
* Simplify code by turning it into a single operator.
This patch will allow users to customize what object types will be drawn by the object mode overlay.
It supports: Empties, Lamps, Cameras, Speakers, Armatures and Lightprobes.
It currently does not support Physics objects due to the overlap it has with other objects types.
Also be aware that in pose mode the armature is drawn, but not by the object mode overlay
Reviewers: campbellbarton
Tags: #bf_blender_2.8
Differential Revision: https://developer.blender.org/D3524
Will help entering sculpt mode on file load by making it possible
to fully initialize sculpt session. The goal is to make sure PBVH
exists since the very beginning of file open (missing PBVH is a
reason why object is not visible before first stroke).
This is not enough yet to fully solve the issue, since entering
sculpt mode tags object for Copy-on-Write update, which frees
PBVH.
Added a option to the overlay popover that controls the visibility of
non-renderable objects like lamps, cameras, speakers, armatures, curves
empties and force fields.
After discussion we went for a single option with more detailed check in
the object_mode draw engine.
Differential Revision: https://developer.blender.org/D3524
Textures in 16 bit integer format are sometimes used for displacement, bump and normal maps and can be exported by tools like Substance Painter. Without this patch, Cycles would promote those textures to single precision floating point, causing them to take up twice as much memory as needed.
Reviewers: #cycles, brecht, sergey
Reviewed By: #cycles, brecht, sergey
Subscribers: sergey, dingto, #cycles
Tags: #cycles
Differential Revision: https://developer.blender.org/D3523
When switching the workspace in a window that does not yet have a layout
for the newly active workspace, we now duplicate the layout from the
previously active workspace. Previously it duplicated the layout from
the first window in the newly active workspace.
It was a bit odd that the scene was stored per window but not the view
layer. The reasoning was that you would use different view layers for
different tasks. This is still possible, but it's more predictable to
switch them both explicitly, and with child window support manually
syncing the view layers between multiple windows is no longer needed
as often.
* Main windows show a topbar and statusbar, and select a workspace and
scene. They are created with Window > New Main Window.
* Child windows do not show a topbar or statusbar. These follow the
workspace and scene of their parent main window. Created with Window >
New Window or View > Duplicate Area into New Window.
* The purpose of this change is to support multi monitor setups where you
just want to put more editors on the other monitors. Without multiple
topbars and statusbars, working within a single workspace and scene.
Creating multiple main windows is intended to be a concious choice to
do different tasks in different workspaces and scenes.
* Note these changes do not currently affect how the operating system
treats the windows.
* When changing the workspace, the layout in all child windows changes.
This makes sense if we consider child windows to be just a way to
extend the main window across more monitors. In some case it may be
useful to keep the same layout though, we can add an option for this
depending on user feedback.
Node link menus (like shader settings in Material properties) used a slightly
brighter variant of the menu widget. Making it hard to style and match the rest.
Make it use widget_menuiconbut, which is just the menu widget with an icon and no arrows.
Thanks Brecht for the help!
Shading is supposed to be dealing with draw manager batch cached
thingamajigs, but was causing full object update.
This was causing both flickering in sculpt mode (PBVH was removed,
why it was SOMETIMES restored before draw is a mystery), and was
also causing things to be really slow.
Workaround for now until all editors tabs share the dedicated Tab theme
settings under User Interface (which the top bar is currently using).
Thanks SimonStorl-Schulke from devtalk forum for feedback.
This deduplicates the calls for tile (un)mapping and allows to have a target buffer that is different from the source buffer (needed for baking and animation denoising).
Based on discussion with @eyecandy & @venomgfx,
we agreed that Tab drag/click, is too easy to accidentally press
while moving the cursor.
It's also not typical to activate the operator on release which
introduces a small lag switching edit-mode.
This is a shame since in some ways its a nice way to re-use the key,
overall it just feels a little too unpredictable for such an important
action.
This commit makes the following changes.
- Tab: toggles edit-mode.
- Ctrl-Tab: opens pie menu.
- Ctrl-AccentGrave: toggles manipulator.
Note, while AccentGrave isn't always available
this shortcut is not essential.
This is the same approach as 98a0bcd425
applied to soft body simulation. In short, CoW copies share the point cache,
and treat it as read-only except when the depsgraph is active.
This prevents having to wrap each call to sbFree() in an if(ob->soft)
condition and assign ob->soft = NULL after calling.
Furthermore, passing `Object *` allows us to change freeing behaviour
depending on whether the object is an evaluated copy or an original (not
done in this commit yet).
This unifies two almost-identical functions at the expense of having to
add one single 'flag' value at one call.
This makes copy_softbody() aware of the source/dest objects, allowing it
to make a distinction between doing depsgraph evaluation copies and real
object copies. This will be used in an upcoming commit to ensure that
the pointcache is shared between CoW copies, similar to the current
approach for rigidbody simulation.
This is because depth test is set before drawing anything now. There is no
case where we want to draw without depth test that is not selection and
this case is not handle by v3d->zbuf anymore.
UI assume depth test is off by default. The DRWManager assume it's on.
This should fix T55623.
In some cases we want associate use an operator for a tool
for introspection, so we can for eg, automatically
use the same binding for in the popup toolbar.
Space-G/R/S for transform now work as accelerator keys again.
Also Space-E for extrude.
Selected color is now .9 alpha so keyed/animated/driven statuses
show through when widget is active. Assign color for overriden state.
Also fix progress bar not visible.
Make room for Alt-A to be used for de-select.
Discussed with artists in studio & @venomgfx,
- Use Ctrl-Space for animation.
- Use Ctrl-Tab to toggle manipulator.
- Remove mirror transform binding.
it is possible that two threads will request same undo node, only one
of them will initialize the node. The issue is that initialization is
happening outside of a lock, which was making one thread to use non-
initialized node.
If this change is ever a bottleneck, make a lock inside of node.
When playing animations the anti-aliasing was not reset, resulting in
ghosts from previous frames to be drawn. Also when playing animations
and switching to X-Ray mode for the first time would most likely
result in a write to uninitialized memory.
The shortcut labels now use the "Item" theme color. Its contrast for hovered
items is a bit low, but not too bad.
Note that this also changes the color of the little toolbar triangle to be grey
(the same color we use for the keymap label). IMHO that looks better though.
This doesn't update any themes other than the default one.
Part of T54711.
Operators ignored edges along the screen-layout bounds. They should've ignored
those along window bounds instead.
Although the global areas can not be joined/split, the adjacent areas can. So
the menu should still be shown.
Had to change the return value of area joining operator, so that an error report
can show up when trying to join over a global area edge. Think this is fine to
do, but you never know with such stuff.
Preferably we'd gray out the "Join Area" item in the menu when clicking on the
edge of a global area. Unfotunately the operator uses coordinates passed as
operator properties to find the right edge/areas, which we cannot access from
the poll callback.
Authored by Pavel Rudko (PavelRudko) on Thu, Jun 21, 10:41 AM.
Original description:
I've encounterd into a problem with blender 2.8.
See the example screenshot (scene, shader setup). For some materials it's not enough to calculate hash based on fragment code and defines. In some cases attribute names may change. And if we change uv set name, for example, vertex shader is not recompiled, and another attribute name inside it is used.
This allows key bindings: link/append and user-preferences
to be accessed via keys - without having to assign keys to each action.
- F1: Context sensitive help.
- F2: File Context Menu.
- F3: Search.
- F4: Window Context Menu.
Other keys F5..10, remain free,
ideally they can be kept for user actions.
This reverts commit 9129319647.
This meant users needed to select everything to see relationship lines,
this isn't always easy since objects can have limit-selection set.
This could be made into an enum option if it's important.
Just like the top-bar, the status-bar can now be hidden/collapsed by dragging
its edge. We display a small line with the editor outline color then, so there
is something that can be dragged up to un-collapse the area again.
This collapsed state is not written to files yet.
For example collapsing the lower part of the topbar with 2x interface scale
would hide the top-bar header region. There were also more asserts when changing
window size and moving area edges afterwards (same assert as in T55298).
Fixes are similar to e626998a26.
With all the recent fixes I've done, area geometry handling should be stable
again. Let's hope I'm right :)
This used to be rather cryptic and it was easy to forget the `+ 1` which in fact
is needed to get the correct width/height, see e626998a26. This should
also fix some minor off-by-one errors.
This reverts commit 357b72e0a7 which caused
the issue, we need a better fix for that cosmetic issue from T50862. For
now displaying keyframes and drivers is the more important one.
Added a compile directive in order to test SH4 in stead of SH2Win.
For now I disabled SH4, it is a bit more clear, but has a small
performance impact. Will check later for a better approach
There is now a manual refresh button on the panel to update the list
of objects in case it changes, and it also gets refreshed when changing
the collection or toggling the use count option.
This is a bit more manual but the previous code of refreshing the
list while evaluating the depsgraph was unreliable.
This also fixes it to take properly take into account visibility, and
to work with linked collections for which index writing was missing.
Access to main database is actually rarely needed, but some custom
'apply' functions do need it (like Collections' overriding of objects or
children collections).
First frame of the TAA is just a regular copy of the previous buffer. so
we write directly to the final buffer and skip the taa shader. We do
init the history buffer via blit so it will be initialized for the other
iterations.
When AA was enabled a new buffer was created. As the composite pass is
already done, we can reuse earlier buffers.
For the deferred rendering we reuse the color_buffer_tx and for the
forward rendering we reuse the transparect_accum_tx.
Most of the times the materials differ due to the object_id. This was an
overhead and resulted in instabilities on Intel graphical cards. This
commit will revert the Material Data UBO and replace it with normal
uniform.
Caused by ca8f787349. This issue is in master actually. Fixing it there
would cause conflicts so won't do that unless needed ;)
Steps to recreate were:
* Split properties editor into two (default startup.blend)
* Move both properties editors to the bottom, so only their headers are visible
* Move the upper edge of the upper properties editor up and down
Although it doesn't happen everytime, you may see the edge between the timeline
and the 3D view move, while the region drawing doesn't update.
What happened is area edge snapping allowed to snap the edge outside of window
bounds (we snap to a 4px grid by default), which screen_vertices_scale
interpreted as window scaling and thus run area size updates.
Referring to ca8f787349. Thought in this case the simple `+ 1` would be
correct, but we need to make the same pixel adjustment as we do in other places.
Simply moving the properties editor of the default startup.blend all the way to
the bottom could trigger this. Calculating limits for area edge moving would use
screen bounds instead of entire window bounds (screen + global bars).
Added assert that should help debugging further issues like this.
This puts the wireframe on top of the weight in weight paint mode,
sculpt mode ...
Wireframe should be colored differently if the object is in paint mode
but it's not there yet.
Unbuffered stdout makes stdout and stderr better synchronised, and helps
when stepping through code in a debugger (prints are immediately
visible). We don't output much to stdout anyway, so this isn't likely to
cause any performance issues.
- TAA is also enabled for Forward rendering
- Uses less GPU memory (removed one history buffer)
- TAA is done after the color management
- consolidated the aa code between forward and deferred rendering
(workbench_effects_aa.c)
I.e. we can now use REPLACE operation directly on items of an RNA
collection, when those items are ID pointers.
Again, mandatory to support new Collection's objects/children override.
To prevent the pointcache from being copied-on-write too (and requiring
copying back), the cache is now shared between the original and
evaluated scenes. Reading from the cache is always allowed; running the
sim and writing to the cache is only allowed when the depsgraph is
active.
Some pointers have moved from RigidBodyWorld (RBO) to
RigidBodyWorldShared (RBOS). writefile.c copies some pointers back from
RBOS to RBO so that the file can still be opened on older Blenders
without crashing on a segfault.
The RigidBodyWorldShared struct is written to the blend file, because it
refers to the PointCache ID block.
The RigidObjectShared struct is runtime-only, and thus not saved to the
blend file.
An RNA getter-function is used to hide the new 'shared' pointer. As a
result the Python API hasn't changed.
Reviewed by: campbellbarton
Differential Revision: https://developer.blender.org/D3508
Key shortcuts and explanation about how to use the tool should go to the
status bar, but other info can in the header so it's near where the user
is working. This distinction has not been made yet for all operators.
When RNA path was for a collection item, we'd falsly get NULL final
property (and wrong final RNA pointer too).
Own stupid mistake in some previous commit, iirc.
The preview files had all the other collections excluded from the view layer.
Although this fix the problem, it probably means we ignored some existing
2.8 files in recent changes. Which is likely fine anyways given that 2.8 is
experimental.
Assert from BLI_assert by default in debug builds
(instead of just printing a warning).
Some developers ignored this, causing errors for others.
Better debug builds cause hard error so code isn't ignored.
Disabling is still useful when bisecting or testing outdated code.
When meshes has no uv layer, but has a texture assigned there was a uv
layer allocated which was corrupt. When no uv layer is available now
there won't be a vbo created. This might impact performance as the draw
cache does not cache this result.
Current implementation works with at most of 3 layers, but was not
checked during hair drawing leading to reading from uninitialized
memory.
This commit limits it to the number of layers that is used during
compilation of blender.
Based on animator feedback, accidentally not clicking on the manipulator
handle had unpredictable behavior.
This functionality will be moved into the manipulator and made optional
(add an invisible fallback manipulator).
textures were not always drawn as the material did not always include
the image in the hash. Not sure if this will solve all cases, but it
definitely works better with this change.
This translates the gl calls to the new GPU_ wrappers from D3501.
Given it's tedious and repetitive work, this patch does as much as it can with search + replace, the remainder of the gl calls will need to be manually dealt with on a case by case basis.
This fixes 13 of the 28 failing editors when building without opengl.
For the list of substitutions see D3502
Reviewers: brecht
Differential Revision: https://developer.blender.org/D3502
Added it as extension to harden. Tried out different methods to fix normals,
Though as with width and segments changes shape, orientation of new polys
a non-smooth method of fix was not possible. Current method aggregates
vertex normals into a smooth fan without affecting edge shading.
Still need to fix the crease at new vertex edges
This way we guarantee that animation component on the datablock will
totally wait for action to be fully evaluated before attempting to
evaluate f-curves.
Fixes T55642: Inconsistent result on animated DoF file
The theme is now stored in generated C code,
which can be extracted from existing user preferences.
This theme can then be used for version patching themes
instead of duplicating values
Note that theme versioning has been removed,
since minor changes from now on won't apply well on top of 2.7x theme.
Theme colors from interface_widgets.c have been removed too.
Per iteration a different AO samples are used, so the final result is
less distorted. Will improve the quality of the image a lot when the
cavity option is turned on.
Currently only attached to the Anti Aliasing of the solid mode of the
viewport. But eventually we could add other options here. Quality
setting can be found in the System tab of the userpref.
The slider goes from No Antialiasing (0.0 - 0.1) to FXAA (0.1 - 0.25) to
TAA8 (0.25 - 0.6) to TAA16 (0.6 - 0.8) to TAA32 (0.8 - 1.0)
This has wrappers for the most common gl* functions in the codebase, and is in preparation for D3502
Reviewers: brecht, fclem
Differential Revision: https://developer.blender.org/D3501
Seems to be that mix(a, b, 1.0) will not give you b. on nvidia this is
not a problem as a was initialized with 0.0, but on AMD it can be any
floating point number, resulting that mix(inf, b, 1.0) was still more to
inf than to b.
find_elem(olddata=NULL) doesn't work reliably for existence checks; it will
return NULL both when the field is found at offset 0 and when it is not
found at all.
- FXAA is now also done in the workbench_forward engine.
- User can enable TAA8 in the userpref by setting their max
viewport AA to TAA8. FXAA will still be used when navigating
Now the name/collection filters run when there's some text,
and don't run when the box is empty, thus reducing an extra
step that was needed before these options could be used.
This commit is an experiment in using popovers as a place to house some
of the filtering options for animation editors, in line with what's taking
place in the Outliner with the filtering popover there too.
Right now, the most frequently used/changed options are still available
on the headers (i.e. the "Only Selected"/"Hidden"/search fields), while
everything else (i.e. the per-datablock filters, which were already hidden
behind a collapsed-toggle button before) now live in the popover.
We have to discard the batch in smooth case, because we are modifying
the index buffer (flat shading don't need it, only changes vertex buffer
on redraw, which is safe).
Many thanks to @fclem for his help on debuging/understanding what was
wrong here!
The latest clang compiler (at least the one in Xcode 9.4.1) warns about the register keyword and macro expansions using defined().
Since these warnings come from third party code, we can't address them directly in Blender. Silencing them via #pramgas will
at least keep the warnings during a build down to the ones that are relevant to Blender code.
At the moment it's using a hardcoded gray color that conflicts with
themes using gray for the header text. Instead use a slightly brighter
background color to make it stand, yet still use a theme setting.
Force Fields and Falloff are now simpler and more compact
by removing unnecesary labels (there was a text label just for one option)
Particle Force Fields Falloff is now a sub-panel of each effector type,
rather than just as a section with label.
Move Navigation Manipulator toggle next to Mini Axis as they are related
(and in the future merged into one pulldown) and rename Manipulator
to "Transform Manipulator" to make it clear they're different kinds
of manipulators. Also move to the first column next to other viewport settings.
- Use per context menu lists to support menu editing.
- Support for different kinds of menu items since this may be needed
in the future. Only use operator types for now.
BLF' blf_font_width_to_strlen() could easily generate strings with up to
nearly two pixels length over requested limit!
Note that the fiddling between floats and ints values make things really
confusing here... :/
There is still a few limit cases where, even though computed str length
is now always below reauested limit, we still get first letter
disappearing, no idea why currently.
This method is a bit more optimized than the very first one
because it does not rely on the adjacent faces properties.
That said it's still a bit slower than the per vertex method.
createOffscreenContext left the new context bound to the calling thread
causing a race condition with the background thread doing the actual
rendering. see T55555 for a more detailed description of the problem.
this patch changes the behavior of createOffscreenContext to restore the
context to the calling context.
Reviewers: fclem
Differential Revision: https://developer.blender.org/D3499
block and layout could be NULL and checking this everywhere
wasn't practical.
Instead of lazy initializing, add UI_popup_menu_end_or_cancel
which cancels empty popup menus.
To find all effectors in the scene, we need to loop over all objects.
Doing this during depsgraph evaluation caused crashes because not all
objects are guaranteed to be evaluated yet.
To fix this, we now cache the relations as part of the dependency graph
build. As a bonus this also makes evaluation faster for big scenes,
since looping over all objects for each particle system is slow.
Fixes T55156.
In stead of a single refresh function that re-init the whole system. The
API now supports adding and removing. Which will be much faster and less
flickering of missing icons when adding/removing lights
Silences the following strict flags from external libraries:
- -Wclass-memaccess
- -Wswitch
- -Wtype-limits
- -Wint-in-bool-context
Needed to tweak macro a bit, since the old logic was wrong:
we can not use CXX flags for C compiler, need way more strict
separation between what goes where.
We draw small numbers over the icon of multiple objects of the same type.
Also, we make all inlined elements to be non-clickabled.
For modifiers we still do it the old way. In this case it is more important
to see the order then the ammount.
Reviewers: brecht, campbellbarton
Subscribers: billreynish, venomgfx
Design task and mockup: T54707
Differential Revision: https://developer.blender.org/D3497
Again, we cannot actually get rid of G_MAIN global access here, so in
most case just 'marked' them as valid, and added assert checks to ensure
we do only work with IDs in G_MAIN in those cases.
Avoids usage of uninitialized memory when inversion fails.
That uninitialized memory can cause object to become visible when
it is supposed not to or other artifacts like that.
Longer term solution would be to check every instance of invert_m#
function and to explicit fallback when needed (possibly, using
extra utility functions).
There was a Full Shading bool that was shared across the WP, VP and TP
modes. This commit makes some changes:
- Replace the bool with a factor. This gives the user more control on
the visibility.
- Also draw it on top of the Material and Rendered mode so the user can
control what he needs. In certain cases you don't want to see the final
rendered material, but the actual texture.
- Removed the skipping of objects when in paint modes. As now the paint
modes are blended.
In an effort to centralize all opengl calls in the codebase, this patch replaces
the raw opengl calls in bf_blenfont with GPUTexture so it's no longer depended
on opengl headers.
reviewer: Brecht
Differential Revision: https://developer.blender.org/D3483
This error is only in Windows and Mac when select in properties the material tab.
This commit revert dc856be839 Preview: Add own opengl context to render preview images.
Transform evaluation does depend on that matrix, but relation was never
in dependency graph. It was not even in the old dependency graph, which
makes me wonder how this thing worked reliably to begin with.
Should fix flickering issue in the render farm.
Transform evaluation doesn't not depend on objects which are being duplicated,
so should not be any regressions here.
The whole relation should probably be gone as it's kind of rooting back to
legacy times, but that would need more clear research and investigation and
regression testing.
Removing that stuff should be valid with new depsgraph/modifiers stack
evaluation system... BUT it might not be, so rather have those in
separate own well-isolated commit.
This was actually rather hairy, this code is huge and complicated, easy
to make mistakes...
Good thing is, it will allow for significant simplification and more
(name) cleanup in following commits ;)
I was originally going to just expose the filtering options fo the timeline
too (via popover), to provide full access to the filtering options here too.
However, investigating further, that would've caused problems when trying to
use the Next/Prev Keyframe operators in other editors (see comment in code).
For now, the simpler solution is to just sync the scene-level flag (used for
this option) back to the dopesheet settings (used for generating the summary
channel used for displaying keyframes), to buy some time to investigate more
carefully.
When in forward rendering and Object outline was disabled every render
call created one console warning (uniform objectId not found)
this commit removes this warning
This reverts commit fec97ec949, it's useful to be able to edit object
properties for hidden objects. The issue with mode locking that this
was meant to solve will be handled in another way.
the use of the texture drawtype is limited. so who should we have it
this prominent on the screen. By adding it as a shading.color_type
option we should save some screenspace.
The calculation of the irradiance buffer was slow. Now it is only
calculated when it is being used (background_alpha > 0.0)
this solves the freeze when switching to LookDev mode.
Also changed the default calculation to use the Spherical Harmonics
calculation. This is able to generate the irradiance buffer fast, but is
less accurate. As the irradiance buffer is only used for visual
guidance speed is more important than accuracy.
Added compile directive to switch between the implementations.
(see STUDIOLIGHT_IRRADIANCE_METHOD)
Disabled caching the irradiance buffer when
STUDIOLIGHT_IRRADIANCE_METHOD_SPHERICAL_HARMONICS is used as it is not
needed.
This is a step towards being able to just choose the "Drivers Editor"
from the editor selector, and have everything displayed, ready to use.
Currently there are a few problems to resolve still:
1) This update callback doesn't run when changing the mode via the
editors selector in the UI. (Note: This problem also affects the
Timeline vs Dopesheet, causing a lot of initialisation work to be
skipped)
2) There's some missing redraw/refresh here to cause the properties
region to actually show. You need to resize the window for that to
happen now.
This shouldn't really be part of the windowmanager code. Pulling it out
now, so that we can reuse in RNA when switching display modes,
(and perhaps other places later)
This commit fixes a number of problematic corner cases when switching between
editors after 2b5050a4cd
The root cause of these issues was that mode_prev was not being set in
many cases, resulting in mode changes to the Timeline and back (via other
editors) causing Dopesheet Editors to reset back to "Action Editor" mode.
1) Creating new Dopesheet Editors (e.g. change the default 3D view to
a Dopesheet editor) would default to displaying the "Action Editor",
since mode/mode_prev == 0 represents the "Action Editor" (for backwards
compatability reasons), while mode == 3 is for the "Dopesheet"
2) If you set the Dopesheet Editor to another mode (e.g. "Grease Pencil" mode),
change to another editor (e.g. Shaders), then come back, the mode would
get reset to "Action Editor".
Expose search like in 2.7, we could make it pop on Ctrl+F to save space if needed.
Align Filter and the less used View menu to the right.
Based on mockups and feedback session with William
- Orientation of the cubemap were not conform cycles and eevee
- one side of the cubemap was flipped, so one side was overlit and the
other side was to dark.
In the outliner there are now icons for it, Ctrl+Click isolates a
single collections.
In the 3D view, Ctrl+H key opens a menu that is more or less the
equivalent of the old layer buttons in the header. Regular Click isolates
the collection, Shift+Click toggle the collection visibility. Pressing
number keys and letters works in this menu, which can help for quickly
selecting a specific collection.
Shortcuts for quick switching by just pressing 1/2/3/.. keys are available
again. The order can be confusing with nested collections, but that seems
unavoidable. The first numbers control the top level collections, and then
sub collections if numbers are left.
Remaining design issues:
* The 3D view menu needs to be improved: support for sub collections,
staying open on shift+click, access from the 3D view header somewhere,
shortcut key display.
* Currently collection hiding just controls per-object hiding, we plan
to separate this state still so alt+H doesn't affect collection hiding.
Note this is now separate from H key hiding, and meant for more persistent
ways to define which objects are relevant to the viewport or render.
This avoids some cases where you'd have to create collection specifically
to hide objects for viewport/render.
H hides selected objects, Shift+H hides unselected objects, and Alt+H
reveals hidden objects.
This hiding state is distinct from restrict viewport and render, and
meant for temporarily hiding objects without affecting more persistent
collection hiding.
Object hiding is per view-layer, same as selection. It affects the
viewport and any preview renders in there, but not final renders.
In the outliner, different icons are now used for temporary hiding, and
restrict viewport and render. Hidden objects are greyed out.
Remaining design issues:
* For lamps we probably still want to keep their effect on the scene,
currently they are fully disabled by hiding. Arguably mesh lights or
even objects being reflected or casting shadows are not that different
but perhaps the special lamp exception from local view should remain.
* We need a feature still to disabled this hiding for specific viewports,
for render or animation preview where you want to see the entire scene
while working in another view.
* We need a new icon for restrict viewport, for now it uses a grid similar
to the 2.4 icon.
* Hiding objects does not preserve selection state as it did in 2.7,
it's probably convenient to support this again?
Validate some cases using G_MAIN instead (I don't think we want to work
on any other Main than G.main one when registering/unregistering nodes
etc.).
And when freeing, all ID not in Main shall now be tagged accordingly, so
we *should* not need to do that stupi search over all ntrees in G.main
to check wether we have to free it ourself or not!
Happened when deleting many studiolights at the same time when the
previews were still beging calculated in the background.
Added a free function callback that is filled when the preview is being
generated. This free function will then kill the preview job
This patch also removes icons that are not valid anymore so the user
cannot accidentally render an icon where the studiolight is invalid.
In the end we should use a add/remove function in the studiolight as
currently icons are recalculated too much.
`mesh_get_eval_final` and friends could call `mesh_build_data`, which in
paint/sculpt mode would call `BKE_sculpt_update_mesh_elements` which
would call `mesh_get_eval_final`... ugly!
Would compare evaluated ob pointer to original one...
Found while investigating some errors in incomming cleanup, but this was
probably generating lost of other issues in some cases...
This is a temp fix for a better system.
Currently the studiolights can be referenced by a WM_job and being freed
via the API. This can happen when removing a studiolight via the
interface.
As the studiolight has no relation with the job, it is hard to detect if
it is still being used. I tried with a Mutex and a Thread Queue but they
were failing.
So the current temp fix is to keep the studiolights in memory until you
close blender. This Must be fixed ASAP! I added this fix so normal cases
can workish.
I moved some code from ED_rigidbody_object_remove() to
BKE_rigidbody_remove_object(), so that calling the latter doesn't leave
the object in rbw->group (causing a crash later on when rebuilding the
depsgraph).
The idea is to only use pointers to particles in original object when
creating an edit structure. The derived mesh we get from evaluated
object.
The rest of the commit is just keeping pointers in sync.
This makes it possible to perform re-evaluation of the scene without
having to re-run the simulation.
The CoW → Orig copy is only performed when the depsgraph is active, so
as to not influence the current scene while rendering in the background.
Alternatively, we could have the CoW copy share the cache with the
original to prevent too much copying of cache data. This will be faster,
but I'm not sure whether we can reliably check the DEG_is_active()
status at CoW copy creation time.
- the result of the SH2-coeffiecients are stored in a bin file (3*9
floats). As the file is a local file we do not care about the
architecture.
- solved issue that also for internal lights the irradiance file were
stored. We don't want that as it could be in a not accessible location
SculptSession.mode_type wasn't initialized until painting,
making it unreliable for checks in other parts of the code.
Also remove unnecessary initialization,
matching sculpt mode more closely.
rBe0c088f8fb5a introduced offline rendering support on windows, sadly it was trying to use the desktop window for getting a context, which given SetPixelFormat can only be called once for any given HDC was an unfortunate choice.
This patch uses a temporary hidden window for getting the opengl context.
Reviewers: mano-wii
Differential Revision: https://developer.blender.org/D3481
Not sure why remap call for static override reference ID pointer was put
there, probably a mistake during some merge back in the days...
But this ended up with self_id of libquery data struct still being set
to previous ID pointer, which could randomly generate crap like falsly
reporting indirect data and such, leading to lost reference datablock at
writefile time...
Since we free collection caches when we remap one of their objects, we
also need to tell depsgraph to flush this into the COWs of that
collection, otherwise things like draw code can access freed memory...
There were two issues here, introduced by rB66aa4af836:
* Forgot to change length of some filter_glob var deep in filebrowser code.
* Truncating filter_glob in general can be dangerous, generating
unexpected patterns.
Last point was the root of the issue here, truncating to 63 chars string
left last group as 'match everything' `*` pattern.
To fix that to some extent, added a new BLI_path_extension_glob_validate
helper to BLI_path_util, which ensures we do not have last
wildcards-only group in our pattern, when there are more than one group.
By default users want AA in the viewport. For slower systems you want to
be able to turn it off. As in the future we would also like to support
TAA in the viewport we introduced it as a Max Viewport AA settings.
Also removed the drawoption to enable/disable AA per viewport
When rendering the AA is always turned on.
This mimics the behaviour of the old wireframe mode. When in Xray mode,
don't use the limit selection to visible option.
Also hide the option if Xray is enabled.
Limit to a restricted set of built-ins, as well as the math module.
Also restrict of op-codes, disallowing imports and attribute access.
This allows most math expressions to run
without any performance cost once the initial check is done.
See: D1862 for details.
This rewinds a change from own commit e3d88b021c
The only took edge overlay into account (crease, seam.. etc)
Currently active-face also uses this width.
While the difference is subtle,
this makes the active-face stand out more clearly.
In preparation for possible alternative back-ends, the dependency on glew.h needs to be confined to as few libraries as possible. This patch removes all gl types from the public gawain headers.
It'll need some more refactoring later on, but for now this will suffice.
Reviewer: brecht, fclem
Differential Revision: https://developer.blender.org/D3482
Remove all the clip cases and just pass the vertices screen position to the
fragment shader.
This does put a bit more pressure on the fragment shader but it seems to be
faster than before. And it simplify the code a lot.
It seems to fix some long standing issue on some intel GPU.
this is to highlight areas in the code that still directly do opengl calls or use
opengl types.
This is in preparation for supporting alternative rendering back-ends.
Reviewers: brecht, fclem
Differential Revision: https://developer.blender.org/D3304
This means the shader can now be used for procedural texturing. New
settings on the node are Samples, Inside, Local Only and Distance.
Original patch by Lukas with further changes by Brecht.
Differential Revision: https://developer.blender.org/D3479
- Vertex size now matches the theme setting.
- Edge width is closer to a single pixel line.
- Face dot was scaled up to be drawn as a circle,
but is currently a square.
Old implementation rendered alpha to 0.0 for background, but only during
deferred rendering. We should attach this to an DNA setting and also let
the forward shading respect this. But for now this is a better default
Currently focused on making parts which are a bottleneck for Spring,
to make things fast as possible. There are surely lots of places
where threading is not currently done, but we can keep doing this,
maybe even with help from the community :)
Without children we only need to update batch cache. This way
we don't re-evaluate the whole modifier stack just to perform
child particles redistribution.
Need to use the 'use_partial_connect' option in island connect,
so changed signatures of various functions to pass that into and
then down from BM_mesh_intersect (making true for intersect, false
for boolean).
Then fix bm_face_split_edgenet_partial_connect to work when
input edges are not necessarily wire, but at least not in the
face they are being connected in. That caused generalization
of core BM_vert_separate_hflag_wire (which is only used in
this one place in all Blender).
FXAA implementation in the deferred renderpass of the workbench.
Can be enabled per 3dview. Eventually this will also be a user setting
as it is more a system/performance setting than a scene setting.
object.base_flag was not synced back from evaluated ID to original one.
Similar thing as copying back to original transform matrix, avoids
having to get data from evaluated one in lots of cases.
Previously, render slots were hardcoded to a fixed amount.
With this change, render slots now are stored in a list. Therefore, users can add and/or remove as many slots as they want.
Credit to brecht for the UI part.
Reviewers: brecht, sergey
Differential Revision: https://developer.blender.org/D3474
I've limited it to just the RGB<->XYZ stuff for now, correct image handling is the next step.
Reviewers: brecht, sergey
Differential Revision: https://developer.blender.org/D3478
The automatic mode checks all Enviroment Texture nodes and picks the largest image's resolution.
If there are no Enviroment Textures, it just uses the old default.
Also, the sampling map now isn't limited to square shapes. The automatic detection uses the exact image size,
the manual UI option now halves the value to get the height.
A default aspect ratio of 2:1 makes sense since this is what most HDRIs use.
Reviewers: brecht, sergey
Differential Revision: https://developer.blender.org/D3477
This removes OperatorRepeatContextHandle, from 5f6c45498c
since we now only have redo in the area an operator is executed in.
These values could be invalid and crashed in some cases.
Sometimes the evaluated pose channel pointers can change during depsgraph
evaluation, so we can't store them ahead of time. The object pointer
should be ok since evaluated ID memory is reused.
This reverts commit 7529690df3.
I tried to put manually add threading protection to GHOST around
Xlib calls but that did not fix the crashes. So reverting that
for now despite the performance implication on some system.
Currently done for mesh batch cache update, and for base flags sync.
Those components do not need anything from original object, and hence
can skip CoW tag and have faster update after them used.
Currently with a debug build there is a race condition where the mutex is released before the DST structure is reset to 0xFF, meaning when multiple threads are using the DRWManager it can get cleared when the main thread is using it, causing crashes.
Reviewers: fclem, brecht
Reviewed By: fclem
Tags: #bf_blender_2.8
Differential Revision: https://developer.blender.org/D3476
this is actually adding option to add buggy behavior, but.. NPR often
expects buggy behaviors, and its one of the main targets for normal editing.
So think it's reasonable to add that option (disabled by default of
course).
Note that am not really happy with UI, but:
* Not sure where to put it, it's kind of own self-contained area option.
* Don't to make it too much visible, using this should be the exception!
This solves the problem of blocks jumping around when changing modes and
center them to the area (in case of only 2 spacers).
Which is (in my own opinion) more aestetically pleasing.
Minor changes based on user feedback.
- View aligned rotation manipulator overlapped too often
with axis aligned widgets. Increase size (matching 2.79)
- Offset arrow-heads when combining rotate & translate,
don't draw the stem since it overlaps with scale.
For grouped undo we should not skip the undo push, rather replace the
previous undo push. This way undo goes back to the state after the last
operation in the group.
Before that depsgraph tagging was done from inside notifier listener in
viewport. This had the following issues:
- If there are no viewports, selection tag was not done. Causing possible
issues when object becomes visible.
- Required special trickery to detect which data to tag for update.
- Was causing crash when transforming/selecting markers in clip editor.
This is because selecting marker needed to poke viewport to redraw, since
selected bundles will be displayed differently in viewport.
Moves the preset into a menu for the panel header, so it can be changed
without opening the panel and takes up less space. Two remaining issues:
* For long lists the add new preset button can be scrolled off screen.
* We should support showing the name of the chosen preset in the panel
header, but the current preset system does not support detecting which
preset is used.
Differential Revision: https://developer.blender.org/D3366
Note that due to RNA get/setters issue, that one may actually add some
G.main usages to the total... But at least it's not hidden anymore in a
very low-level, dark corner of BKE pointcache code!
Technically we probably need to do for the current subversion (17).
But the entirely 2.80 series is known to be experimental.
The most important part is to notify people that if they open the 2.80
file in master it will loose data.
- User preferences header at the bottom.
- Action header at the top.
Now all editors accessible from the menu have their header at the top,
default layout for the timeline remains at the bottom.
These are set in the interface and aren't useful to show to users.
Other minor changes
- Rename 'Ruler/Protractor' -> 'Ruler'
- Rename Subdivide 'Quad/Tri Mode' -> 'Create N-Gons' (and invert).
This is supposed to be handled by calling code! Henceforce, no need to
call BKE_sequencer_clear_scene_in_allseqs() here, and... no need for
that ugly G.main case. ;)
All areas are toy use "local" number generator, in order to keep
behavior predictable and threadable. This is what BLI_rng_() API
is for.
There are still lots of usages of BLI_frand(), which are to be
ported to BLI_rng_get_float(). but that is somewhat involved.
For the time being, remove unsafe API, so new areas have zero
chance using it.
Completely ignore animation for legacy zero update tag.
If one needs animation to be evaluated, tag with proper update tag, stop relying
on direct assignment of adt->>recalc with zero update tag for DEG.
This commit might cause missing updates when needed, those needs to be looked
into once they happen.
We shouldn't be using BI textures anyway. But in any case, adding noise
texture to compositor is doing proper seed updates already, so probably
was already ported to a proper RNG usage.
The ClayEngine was introduced to test the blender2.8 architecture during
development. As currently we have the wanted features implemented with
matcaps we are going to remove the clay engine as it was never intended
to be an official releasable engine
Note: The test cases are never run. But when enabled will be skipped as
they were implemented over the Clay Engine
Render depsgraph is evaluated from scratch, so it should not be needed
to do such a call.
This also prevents modifier stack from being evaluated on top of already
evaluated mesh
This is the first step to moving redo out of the top-bar, see: T55039
- Support for floating panels in the ScrArea has been added.
- A new region type RGN_TYPE_HUD is used for floating settings input.
- Only one HUD should be visible at a time.
- Currently HUD regions are only visible
for 3D view and image space types.
- There are currently some remaining glitches with hide/show
that need further investigation.
This is also a way to render an large preview icon without the selection
control. This is used by the studiolights in the userprefs that only
displays the preview of the light.
We actually get fully rid of BKE_object_free_derived_mesh_caches usages,
now always using BKE_object_free_derived_caches (bad name, btw, should
be 'evaluated_caches ;) )...
With copy-on-write we can no longer assumes the point density data is
available after Cycles synchronization with Blender data is done. So
force it to be loaded earlier, even if it's not great for interactivity.
- Added flexible separators to
Clip, Graph, Dopesheet, Image, Node, Timeline, 3D View.
- Added graying out for Proportional Editing
menus to avoid popping when right-aligned.
- Slightly re-arranged some controls,
so they can be on correct side of the separators.
Patch by @billreynish
This was causing driver editing popovers to just keep disappearing
whenever you accidentally moved the mouse away, making the popovers
there nearly useless.
After testing with the other popovers we have in other places, I don't
really see why we needed this flag enabled in the first place. You can
still switch between popovers in the topbar by mousing over the others,
and if you want to dismiss without clicking to escape, you can go over a
standard menu instead OR just move back towards the button that spawned it.
Let's just test this out for a while and see how it goes. Maybe we don't
need a more invasive solution here...
There is one legit place in the code where memcpy was used as an
optimization trick. Was needed for older version of GCC, but now
it should be re-evaluated and checked if it still helps to have
that trick.
In other places it's somewhat lazy programming to zero out all
object members. That is absolutely unsafe, at the moment when
less trivial class is used as a member in that object things
will break.
Other cases were using memcpy into an object which comes from
an external library. We don't control that object, and we can
not guarantee it will always be safe for such memory tricks
and debugging bugs caused by such low level access is far fun.
Ideally we need to use more proper C++, but needs to be done with
big care, including benchmarks of each change, For now do
annoying but simple cast to void*.
In C++ it is not really safe to memcpy objects, and newer GCC will warn
about this. However, we don't use our vector for unsafe-to-memcpy objects,
so just explicitly silence that warning.
This support layout.separator_spacer() to be used by headers as a way to
dynamically separate the ui buttons.
Right now no UI file is changed, though we can use this right away in the
timeline, and shortly after in the viewport header (moving settings from
the topbar to it).
Original design by William Reynish.
Review: Campbell Barton
D3468
This way they can be init in their owner thread. Contexts should not be
shared accross threads. Once you make a context active on a thread it is
owned by the thread.
This commit only have the GLX backend updated but should not break orther
platform.
Note: Metaballs only support the first material slot. Splicing it per
material would create empty Batches. In order to overcome this we set
the batch for other materials to NULL. We added extra checks in EEVEE
and Workbench to not draw when the geom is NULL.
thanx bblanimation (Christopher Gearhart) for spotting the issue and
providing the fix!
Reviewed By: brecht
Differential Revision: https://developer.blender.org/D3449
- Frame & object name now top left.
- Axis now top right (when view navigation disabled).
This is needed so we can use the lower area for redo popup, see T55386.
atoi usage in BLI_stringdec could overflow, use strtoll instead and
check
valid range with INT_MIN and INT_MAX
Reviewed By: campbellbarton
Differential Revision: https://developer.blender.org/D3452
This won't ever be used for this operator, as the "edit" operator should
only ever be used for dealign with a single driver - the one under the
cursor.
While we probably won't be getting a general purpose "Onion Skinning"
overlay anytime soon for meshes, etc. (at least not before the
depsgraph stabilises, and we also get geometry caching working),
for the Grease Pencil integration at least, it makes sense to move
GP objects to using a more general/future-proof solution, instead
of continuing to use a special/dedicated button in the header.
Currently the UI part of this is commented out. Also, the GP branch
doesn't need to move to this pre-merge. But, since 2.8 changes move
fast, it's better to reserve the space now to have it next to motionpaths,
than introduce it later.
The copy/paste driver variables buttons currently only work when used in the
Graph Editor as they rely on being able to have the "active F-Curve" context
info that's only present in the Graph Editor (in Drivers Editor mode).
Instead of having them constantly greyed out in the popover, it's easier
to just hide them for now. Besides, this is probably more of an "advanced"
feature that we don't need to expose in the simple UI.
use better poll and get ob with 'ED_object_active_context' (instead of
'CTX_data_active_object')
Reviewed By: campbellbarton
Differential Revision: https://developer.blender.org/D3467
This fixes the problem where the "Show in Drivers Editor" button would
not actually select and highlight the driver under the mouse.
(TODO: The channels still aren't getting selected properly in the
channels list, but at least the properties show correctly)
Previously, newly created drivers were set to "Scripted Expression"
mode and had their 'expression' field set to the pre-driver value of the
property, so that adding a driver didn't cause the property to immediately
lose its old value (and potentially causing havok in the scene).
However, this had the unintended consequence of making the driver setup
workflow more cumbersome, as you first had to replace that value with
the name of the driver variable before your driver would work.
This commit works around this issue by trying to combine the best of both
worlds (quite literally): Now, the driver expression for drivers created
using Ctrl-D will be "var + <old value>".
Thus, in the simplest case, the driver will do something as soon as you fill
out the driver variable settings (e.g. just filling out the Target Object field
will do it), meaning you get your drivers working faster. Of course, it may now
be necessary to edit out the old-value, and/or it might be a bit more confusing
what/why it's there for newbies. However, the improved ease of setup, and/or
a more tangible example of how an expression may be constructed outweigh the
downsides IMO.
This commit implements a new behaviour for the "Add Driver" functionality
(invoked from the RMB menu on a property, or by pressing Ctrl-D).
Instead of spawning a context menu asking you to pick the way
you want to create a driver, it will now just create a driver on the
property under the mouse and then show the "Edit Drivers" popover so
that you can immediately start editing the properties of this driver.
This way, the whole process is more visual and feels less blocking /
constrained, with less upfront decisions needed immediately.
Notes:
* The new behaviour is equivalent to choosing the "Manually Create (Single)"
and then doing a "Edit Driver" on the property
* Renamed the old "ANIM_OT_driver_button_add" operator to
"ANIM_OT_driver_button_add_menu". It will probably go away
in the near future, but it's better to keep it around for a
while longer still until the new workflow settles down.
The range is 0.25 to 0.75 opacity when the Xray opacity is between 0.0 to 1.0.
This is to avoid loosing completely the sense of occlusion when having no
other solid drawing than the wireframe and loosing the transparency when
xray alpha is at 1.0.
Also replace Bayer (checkerboard) method by interlieved gradient noise to
minimize the chance to loose an occluded line completely.
Other noise function could be tested in the future.
This Fix the problem when multiple objects are selected and one of them
occlude the others. You cannot see clearly what is selected.
With this option, selection is more clear when Xray mode is enabled.
By using equation 7 from the paper, we make the surfaces nearest to the
viewpoint appear more "opaque". This gives better sense of ordering than
the previous weighting function that was really not doing anything.
There was a method explained in the Weighted Blended Order-Independent
Transparency paper to support hardware that does not support per render
target blending function.
So now only 2 geometry passes are required instead of 3 (one being the
outline/depth fill pass).
This also fix how the blending is done. There was some premult confusion
in the implementation.
As in Pose Mode, the idea here it to try to reduce viewport complexity
without requiring users to turn off the overlay completely all the time.
For example, a background prop (e.g. a tree with a tyre hanging off it,
or a branch with hand-placed leaves) won't be cluttering the viewport with
its relationship lines all the time, when you're trying to do something else.
When you really do need to see these lines, you can still select the object
in question, and you'll see the lines for which objects are its children
or what its parent is. And to see all lines, you can still always select all
objects.
This patch reduce the branching in the lamp loop, improving compilation time
noticeably (2372ms to 1785ms for the default shader).
This should not change the appearance of the shader.
Performance impact is negligeable.
This mimics the 'spreadsheet' behavior. Columns and/or rows can have
equal sizes, or adapt to their content - but always in a grid way (i.e.
all items in a same column will always have same available width, and
all items in a same row will always have same available height).
Also, you can fill (order) the grid in a row- or column-major way.
Not used anywhere for now.
Differential: https://developer.blender.org/D2395
This fixes the problem that matcaps have when using a very larg FOV in
perspective view.
This was because it was stupidly using the normal direction which can not
aligned with the view vector under perspective.
So to workaround this problem we don't use the normal as is, but compute
how much it's facing the camera.
This changes how matcaps looks in perspective because they now always use
the full range of the matcap (which is expected).
Pretty straightforward changes, merely mimicking dm-related code,
which was already essentially using either Mesh or BMesh data to
build the PBVH...
Note that we "lose" the subsurf (a.k.a. grid) PBVH case here, but that
one was already dead code in current blender2.8, since final dm is
always a cddm built from evaluated mesh.
Proper fix is pending new code for subsurf/multires area.
Would try to use data from global area for the newly created fullscreen area.
This should not happen, so instead of a global area use first area from layout.
When a CacheFile datablock is copied to an in-main datablock, it means that
the new copy should be independent of the old copy. In this case certain
properties are reset/duplicated. When it is copied to a no-main datablock,
we assume this is a copy-on-write copy that's used for evaluating things,
in which case the handle and object paths are referenced instead of
reset/ duplicated. This prevents us from re-opening the Alembic file every
time a new CoW copy is made.
The same approach is taken when freeing data (only when an in-main datablock
is freed do we actually free the data).
This makes evaluated objects to point to a non-evaluated collection,
which is already really bad. What is even more worse, objects in those
collections are not evaluated either.
Proper solution would be to implement visibility flag for nodes, which
will be set to 0 for transform/geometry components, but which will be
1 for copy-on-write components.
This way we will guarantee consistency of dependency graph.
For now this change is good enough and unlocks production.
Fixes T55375: Crash when changing visibility of instanced collection
Fixes T55357: Particle geometry crash
- all known image types are supported
- BpyAPI for studiolights added
- added open user pref operator in shading menu
- possible to add multiple files in a single run
For now refreshing studio lights will free all studiolights and reinit
the whole mechanism. This can be improved by only freeing deleted, reset
updated and add new custom studiolights.
details to show currently only shows the path we perhaps want to add
other information also
After testing in the studio and extending the event system for
drag events, we've agreed on adjustments to the new keymap,
see: T55162
- Tab: Edit-mode toggle.
- Tab + Cursor Drag: mode switching pie menu.
- Accent/Grave: for 3D view pie menu.
- F3: Search
- 1..3, Shift-1..3: Edit mesh vertex/edge/face toggle.
Other minor changes were made, however they aren't part of the design.
- Ctrl-Shift-S: Image editor Save-As (was F3)
- Ctrl-Alt-R: Repeat history (was F3)
All keyboard events were sending double key events (including modifiers)
when xinput was enabled with gnome (causing much confusion!).
I cant test if XIM works,
but this isn't useful to send double events, so disabling for now.
Since we are only creating this and never updating, there is no need for
the original approach with the individual data to be updated.
Note we only populate the GPU data when binding the UBO, so we can in the
future easily create the UBOs in a separate thread than the main drawing one.
Also at the moment animated materials are not working. To fix that we need
to free/tag for free the GPUMaterials in BKE_material_eval.
This make the limited wireframe not a performance problem anymore.
However, this does change the number of edges displayed as the threshold
is now computed per vertex instead of per edges.
For this reason we extended (internaly) the range of the slider so that the
users can hide more edge.
Before that copied collection in copy-on-write were running out
of sync with original ones. This was causing crash with the
following scenario:
- Delete some objects from scene
- Add particle system to an object
- Change particle mode to Hair
Thanks Dalai for debug session! Pair programming ftw!
We now remove the shader for every update.
So at the moment the whole point of UBO (Uniform Buffer Objects) is that they
are more efficient than individual uniforms.
Next steps is a harmless refactor to stopping UBO from referring to original
data in the UBO, and simply copying it.
It would also be interesting to make the final shader more granular as far as
the library and required functions are concerned. Even if this doesn't impact
performance, it should give us smaller easy to debug shaders
(a simple shader now has > 5k lines!).
If performance for animated values is measureable slower after this commit we
can port the shader creation CPU side to the depsgraph - localizing the tree,
hashing, lookup, ...
Additionally we can stick to update the UBO when the material changes but not
its topology. This is very trick because of localized trees. So we will only
re-visit this if profiling hints at any benefit from it.
Notes:
* Really need to address RNA setters case, end up adding way too much
G.main here these days... :/
* Added Main pointer into bAnimContext, helps a lot in anim code ;)
When calling the bpy.ops.poselib.apply_pose() operator from Python, Blender
would deadlock when the rig has drivers.
Similar BPy_{BEGIN,END}_ALLOW_THREADS calls were already in place in the
rna_Scene_update_tagged() function.
This is part of the work needed to refactor the material parameters update.
Now the gpupass cache is polled before adding the gpumaterial to the
deferred compilation queue.
We store gpupasses in a single linked list grouped based on their hashes.
This is not the most efficient way but it can be improved upon later.
Not sure why exactly it is called a cleanup, the code was much more clear
and robust against possible missing return statements which are MANDATORY.
Missing return statement will:
- Cause two different BVH traversals to be run.
Not is happening currently, but if more BVH layouts are added, it will
become a problem.
- It is already causing assert() statements to fail, since functions are
no longer returning when they are supposed to.
If there is any measurable reason to keep this change, let me know.
Otherwise just stick to reliable/tested/robust code.
This reverts commit ba65f7093b.
The issue is that children drawing is done by object mode,
which operates with data from evaluated context. But that
data needs edit mode's cache to be properly updated first.
Makes ADD brush to work.
At some point children particles draw got broken, children are not
visible for until first stroke is done. Still looking into it.
Experiment: let the user be in control of the alpha channel as some rigs
are hard too see during bone selection. Especially rigs that were
designed for 2.79 wireframe mode.
That bug was also likely affecting other cases - basically it was making
auto-keying always key from evaluated ('visual') values, never base,
data values... Added a flag, in some cases we do want evaluated values
here, obviously.
This was no longer working, now this works the same way as
scene.update(), only it applied to the current view layer.
Caused crash running alembic export from Python.
A cavity shader based on SSAO. Works on all workbench deferred passes.
Per 3d viewport the cavity shader options can be set as different
shading needed different options. Some global options are in the
Viewport Display of the scene like num samples and distance.
Experimental: Naming of Ridges and Valleys
The mutex was shared between CoW copies of the CacheFile datablock, and
as a result also freed multiple times. It is now only freed when the
original datablock is freed; the CoW copies share the same mutex.
This is a beginning of series of commits which will clearly separate
building IDs which are used as object data from object building.
The goal is to be able to always build whatever ID. Required to make
driver targets to work reliably with copy-on-write concept.
DATA datablock must evaluation must never depend on object it is
used for. If there is something what depends on an object, it must
be done on object level.
At least the actual callback was empty.
When the source mesh doesn't have the primary layers (CD_VERT for vdata,
etc.) the returned mesh also didn't have those layers, even when non-zero
elements were requested (for example requesting 4 vertices would still
result in mvert = NULL).
The ObjectID pass was generating per material per dupli a specific
number for the outline what results in a GPU context switch. In spring scene
01-050 a scene with many trees (duplis) generated 28000 GPU materials.
Now only new materials are created when objectid pass is enabled. Also
added a hard limit to the number of objects for the objectid pass (255)
Basically the outline between objects will not be drawn, but it will be
very hard to detect them also.
Also fixed for XRay mode.
Test this since the popup feels disruptive/flashing when its too large
when set smaller it looks closer to a menu w/ key-accelerators which is
the intention in this case.
It's also more likely the active tool can be placed under the cursor.
The recent change also used the buildtools instead of the regular compiler, you now have to explicitly state what you want to use :
2017 - the standard msvc compiler
2017pre - the msvc compiler from the preview installation
2017b - the msvc compiler from the buildtools installation
The function definitions still reside in DerivedMesh.c. Once we're done
porting all the DerivedMesh use to Mesh, we'll move the still-relevant
functions to mesh_runtime.c. This move is now cumbersome due to shared
statically-declared utility functions in DerivedMesh.c
This helps making things clearer and cleaner. Func returning filepath of
G.main is separate, so that we can easily track its usages, and
hopefully deprecate it at some point. Though that usage of G.main is
likely the less evil one, you nearly always want current blendfile path
in those cases anyway.
When run from make.bat the environment is setup correctly and the VCToolsRedistDir environment variable exists, on later invocations of cmake this may no longer be the case and a warning was emitted about the missing runtime. we can't rely on InstallRequiredSystemLibraries.cmake here since it uses the compiler version to figure out the correct location and it doesn't know how to deal with clang.
Manipulators which aren't explicitly activated via tools are now hidden.
Tool manipulators are kept because it doesn't make sense to interact
with a tool with hidden manipulators.
Material slots are a real pain to get working, due to all the black
magic they do to hide object vs. obdata storage of the material...
Currently hitting an order problem - we need to always set 'link' (to
object or obdata) property of the slot first, before we set its
material... *super-sigh*
- users can use their own matcaps
.config/blender/2.80/datafiles/studiolights/matcap/ folder
- upto 100 matcaps can be loaded
- color of the matcap is influenced by the color of the material/single
color etc. To show the plain matcap use single color at 1.0
- chosing a matcap is at lighting level (flat/studio/matcap)
- matcap only possible in solid mode
- also works for X-Ray mode
As the old matcaps are still in used by the clay engine I didn't remove
it yet.
We are already running out of available flags in main, generic int, and
everytime I work on static override I find new special cases that will
need new specific propflag, so...
Switching the mode was being done per object,
where it should first check all pose bones from all objects.
This also removes code that changes the transform mode
from within the counting function.
This was only needed on transform initialization.
D3458 by @billreynish w/ edits.
- Context menu for dope-sheet, graph, image & node editors.
- Add type to contenxt menu header.
- Access with W-Key.
- Change UV-editor weld key binding to Shift-W.
Make it more reliable and predictable way of getting pointer to
an original mesh which came from copy-on-write engine.
Related change: made it (hopefully) more clear name for flags.
Non-manifold geom was producing inverted result when the camera was inside
the shadow volume.
When rendering non manifold geometry with the depth fail method, we need to
emit the cap as if it was the same geometry with a virtual thickness.
Another way to view it is to imagine having a set of non-manifold geometry
all facing the light.
So for every tri we emit a front cap oriented toward the light and the
back cap pointing away from it (whatever the actual orientation of the tri).
Extrusion pass stay the same as the depth pass method (double the increment
if it's a manifold edge).
In the Python API, any panel becomes a subpanel by setting bl_parent_id
to the name of the parent panel. These subpanels can contain advanced or
less commonly used settings.
Now hairs are shaded properly in workbench and support texturing.
I also added a 10% random normal direction per hair to have a bit more
variation in the shading. This is hardcoded for now.
-expanded build_deps.cmd with 2017 support, it can't locate msvc2017 so needs to be run from developer prompt.
-Newer cmake was unhappy with openal's cmakelists.txt
-collada has warning as error on and errored out on new msvc2017 warnings.
This is a dirty fix. A bit more cleaner approach would be to check if a
context is bound and delay the deletion only in this case.
Also we may want to do this orphan deletion at some other places than
wm_window_swap_buffers.
Do note that it does not match cycles implementation.
Also we could precompute the hash per strand before rendering but that would
suggest it's not per engine specific.
If we make the random value internal to blender then it won't be a matter
because other renderers will have access to the same value.
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.
(ported over from blender2.8)
By default when moving a edge of the screen it always snaps to an invisible grid with unit of 4 pixels.
This was also affecting the snap to the midpoint and adjacent.
The solution was to make the snap to areagrid optional and use values of `origmin` and `origsize` that match the transformations in screen_edit.c.
The fix bit is the split.
But since we are using col.separator() left and right we need this as well.
I still don't think the separators are the way to go, yet may as well be
consistent with the current design in place, thus the extra separator here too.
This commit restores support for Motion Path drawing in 2.8 (as it wasn't ported over
to the new draw engines earlier, and the existing space_view3d/drawanimviz.c code was
removed during the Blender Internal removal).
Notes:
* Motion Paths are now implemented as an overlay (enabled by default).
Therefore, you can turn all of them on/off from the "Overlays" popover
* By and large, we have kept the same draw style as was used in 2.7
Further changes can happen later following further design work.
* One change from 2.7 is that thicker lines are used by default (2px vs 1px)
Todo's:
* There are some bad-level calls introduced here (i.e. the actgroup_to_keylist() stuff).
These were introduced to optimise drawing performance (by avoiding full keyframes -> keylist
conversion step on each drawcall). Instead, this has been moved to the calculation step
(in blenkernel). Soon, there will be some cleanups/improvements with those functions,
so until then, we'll keep the bad level calls.
Credits:
* Clément Foucault (fclem) - Draw Engine magic + Shader Conversion/Optimisation
* Joshua Leung (Aligorith) - COW fixes, UI integration, etc.
Revision History:
See "tmp-b28-motionpath_drawing" branch (rBa12ab5b2ef49ccacae091ccb54d72de0d63f990d)
Since the viewport header now supports transparency, text on pulldowns
can be hard to read if their color matches the viewport content.
Background is drawn using the 'inner' theme color, that was unused until now.
In some heavy rigs matrix inverse can be 10% of computation time. This
reduces it to 2% by using Eigen's optimized 4x4 matrix inverse and SSE
matrix multiplication.
- Uses the roughness setting of the basic eevee material
- renamed gloss_mir to roughness
- set default of roughness to 0.25
- renamed ray_mirror to metallic
- cleaned up material rna (BI mirror struct)
- use BLINN phong model
- normalize incoming/outgoing specular light
- when using camera oriented studiolight, the SolidLight will be used
for specular highlights
- EXPERIMENT: when in world oriented studiolight only the shadow direction will be used.
- change the settings of the internal light to make scenes more
readable
Note there is some talk about changing the position of this (which would be strange in the image editor by the way,
since there we use the bottom for the result of the current pixel when dragging the mouse).
However first I wanted to fix this regardless.
This currently shows panels that were in the 2.79 3D view toolbar
which are now popovers.
In some cases it's useful for these to stay open.
This commit adds a space type to do this.
Note this is currently empty in object mode.
This will currently only work for the RelWithDebInfo configuration since asan
does not support the debug crt. for source line information in the reports,
you need a copy of llvm-symbolizer in the blender folder or set the
ASAN_SYMBOLIZER_PATH environment variable to point to it. Currently (as of
6.0.0) llvm-symbolizer does not ship with the binary clang/llvm distribution.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D3446
This overlay is showing mesh topology. It is usable with transparency
even if the mesh order can mess up with the expected result (some object
more prominent than others).
Edge thickness and alpha values are hardcoded for now but can easily be
added to theme or object settings.
Only OB_MESH is supported for now.
Creates a simple index buffer with negative indices if the edges is not a
real edge.
Also create the buffer texture representation of this buffer along with the
pos_in_order buffer texture.
This includes:
- Skip OB_RECALC_TIME tag from object duplicate operator
not sure why it is needed: even if original object was
animated, duplicating it will copy evaluated values.
- Don't tag whole ID for update when updating it after
relations rebuilt. Use the same trickery to detect
whether animation is to be re-evaluated or not as is
done for update flag=0.
- Don't tag datablocks which are expanded for update
of copy-on-write.
- Avoid flush along relation from copy-on-write
operation in action. This will not invalidate any
pointers in the copied datablock since we don't
reference anything in the action.
Will cause some bad behavior when object is shared across multiple
visible view layers with different overrides. Accept it for now,
and possibly force single view layer later.
When active dependency graph is evaluated, it will apply animation,
drivers and scalar evaluation data (such as object matrix) to an
original datablock. This way operators and tools can easily read
data from original datablock.
This will simplify porting them to copy-on-write, and solve issues
when some operator will allocate new datablock based on original one,
and will want to read data from it.
This way we allow animation system to make decisions based on which
context dependency graph is coming from, and whether it belongs to
an active edit window or not.
After discussion with Sergey and Dalai, we have decided to remove
this option completely. We're getting to the point where it is almost
impossible to really use 2.8 without COW, and keeping the old option
running ends up diverting dev resources away towards tracking down
and fixing problems with a parallel system that will be going away.
Made the highlights darker (using the defaults of blender 2.7
- sharpness = 50
- spec color = 0.025 (I bumped it to 0.1)
- added a log2 to the frontal camera light to reveal more details of the
mesh
We no longer user scissor for 3D viewport drawing, and some selection
code assumed it still. This also cleans up unnecessary scissor test
switching, we only have it temporarily enabled now.
* depsgraph.ids: all evaluated datablocks in the depsgraph
* depsgraph.objects: all evaluated objects in the depsgraph
* depsgraph.object_instances: all object instances to display or render
* depsgraph.updates: list of updates to datablocks
After more testing, I realised that bendy bone properties should also be grouped,
but probably all other per-bone settings too. Now, just group all of them, since
it's easier for everyone this way.
For many years, animators have been complaining about how keyframing a (transform)
property directly would leave them ungrouped, while keyframing them using a Keying Set
would put them into a group based on the name of the keyingset.
This commit attempts to improve (unify + make consistent) the default behaviour:
* All object transforms now get added to an "Object Transforms" group,
regardless of whether they were added individually via buttons or keyingset
* All bone transforms now get added to a group corresponding to the name of the bone
instead of only the ones added via keyingset
Saves quite a bit of CPU ticks per mesh update, giving measurable
speedup for file from T55228.
Memory usage goes up a it, most likely due to evaluated mesh having
more custom data layers than corresponding DM does.
This introduces a new depsgraph API for getting updated datablocks,
rather than getting it from bpy.data.
* depsgraph.ids_updated gives a list of all datablocks in the depsgraph
which have been updated.
* depsgraph.id_type_updated('TYPE') is true if any datablock of the given
type has been added, removed or modified.
More API updates are coming to properly handle multiple depsgraphs and
finer update granularity, but this should make Cycles work again.
Use the same key for pose & weight paint mode (instead of texture paint)
This makes more sense since pose/weight paint modes
are often used in combination.
This mean you can store data used for drawing inside the object engine
data.
Also fixes T55243 Crash in ASAN debug builds due to use-after-free memory in draw code - instances issue?
This now can shade actual poly strips that mimics cylinders.
This makes hair coverage exact compared to the line method and result in
smoother fading hair.
This does make the sampling a bit more exact but needs more samples to
converge properly.
This new system use transform feedback to compute subdivided hair points
position. For now no smoothing is done between input points.
This new system decouple the strands data (uv, mcol) with the points
position, requiring less update work if only simulation is running.
In the future, we can have compute shader do the work of the feedback
transform pass since it's really what it's meant to. Also we could generate
the child particles during this pass, releasing some CPU time.
draw_hair.c has been created to handle all of the Shading group creations
as well as subdivision shaders.
We store one final batch per settings combination because multiple viewport
or render could use the same particle system with a different subdivision
count or hair shape type.
Hair Particles shape properties are ported from cycles. Thoses properties
have the same defaults and have a do_version of their own. Cycles will use
theses properties instead of its custom ones.
Some realtime engine specific settings are also added to scene->r because
it's much easier to control as global values.
Bumping Version number so cycles can do its own do_version on top of the
default settings.
This mimics the behaviour of DRW_shgroup_empty_tri_batch_create and will
replace it eventually.
The advantage is that it's compatible with transform feedback.
Modifiers stack only get COW/evaluated IDs, so no need to go auery again
DEG for those. Further more, now unified handling of EditBMesh case (was
done on case-by-case basis in a few modifiers, not all for some reason).
We are still missing the ability to get final and cage deformed meshes
when in Edit mode though, this is to be defined/implemented in depsgraph.
This is a first step to have correspondence of legacy derivedDeform
within a new formulation. Only base ground for now to support file
reading, copy-on-write remapping and such.
Support switching non-active objects in/out of a mode from the outliner.
- This allows users to change which objects are in a mode w/o
having to exit the mode and change seleciton.
- Changing the mode of the active object applies to all other objects.
- By convention setting a mode selects, removing de-selects,
this is done for convenience so switching to a mode from object mode
maintains the set of objects in the current mode.
See: T55246
- Use Tab key for search.
- Number keys switch modes.
- The number of the current mode can open a submode menu
(currently only works for edit-mode)
- Ctrl-Tab, Ctrl-Shift-Tab - cycle workspaces.
There is a chance parts of Blender call BKE_object_workob_calc_parent
with ob->parent objects that are outside the depsgraph.
This we can tackle later since these are corner cases anyways,
and this fix fixes all parenting operators in Blender.
This commit fixes a problem the Spring team were having in the shot files 01_030/050 with
the camera rig, where when you tried grabbing/translating the bones of the camera rig,
they would only keep rotating/scaling instead.
The reason for this is that with the multi-object editing support added into the transform
system, this code would now get run for all the other objects that were being included
in Pose Mode, even if no bones from those armatures were selected. Since no valid translatable
bones were found in those other armatures, the transform mode would get reset upon
encountering one of those other bones.
Scene lights are rendered when
- v3d is not available
- or shading type is other then OB_MATERIAL
- or shading type is OB_MATERIAL and use_scene_light is true
This differential fixes a bug for resetting view (shift+c).
Bug description: When 3D cursor is placed randomly somewhere, and shift+c is pressed,
the cursor is not shown to reset at world center.
Reviewers: sergey, aligorith, brita_
Reviewed By: brita_
Subscribers: brita_
Differential Revision: https://developer.blender.org/D3423
This commit contains the minimum to make clang build/work with blender, asan and ninja build support is forthcoming
Things to note:
1) Builds and runs, and is able to pass all tests (except for the freestyle_stroke_material.blend test which was broken at that time for all platforms by the looks of it)
2) It's slightly faster than msvc when using cycles. (time in seconds, on an i7-3370)
victor_cpu
msvc:3099.51
clang:2796.43
pavillon_barcelona_cpu
msvc:1872.05
clang:1827.72
koro_cpu
msvc:1097.58
clang:1006.51
fishy_cat_cpu
msvc:815.37
clang:722.2
classroom_cpu
msvc:1705.39
clang:1575.43
bmw27_cpu
msvc:552.38
clang:561.53
barbershop_interior_cpu
msvc:2134.93
clang:1922.33
3) clang on windows uses a drop in replacement for the Microsoft cl.exe (takes some of the Microsoft parameters, but not all, and takes some of the clang parameters but not all) and uses ms headers + libraries + linker, so you still need visual studio installed and will use our existing vc14 svn libs.
4) X64 only currently, X86 builds but crashes on startup.
5) Tested with llvm/clang 6.0.0
6) Requires visual studio integration, available at https://github.com/LazyDodo/llvm-vs2017-integration
7) The Microsoft compiler spawns a few copies of cl in parallel to get faster build times, clang doesn't, so the build time is 3-4x slower than with msvc.
8) No openmp support yet. Have not looked at this much, the binary distribution of clang doesn't seem to include it on windows.
9) No ASAN support yet, some of the sanitizers can be made to work, but it was decided to leave support out of this commit.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D3304
This is a hacky fix so that animators can use this tool again with autokey enabled
(which they do all the time). The issue here is that the tool writes the new (0)
values to the original data, but insertkey now reads from evaluated data (so that
keying interpolated values works). However, the cleared values do not get re-evaluated
or flushed before insertkey gets to it (via auto keying), meaning that the wrong values
get keyed.
There may be better solutions for this, but for now, this is the simplest fix that
I can get working.
More MessageBus-related silliness to resolve issue with toggling visibility/mute/locking
status of F-Curves/Groups/etc., as well as other things like modifying active keyframe's
values, or changing properties of F-Modifiers.
I've now ended up just whitelisting in the region subscribers all the animation-related
structs in RNA. But still, that may not be enough to deal with potential issues later
with the property sliders (shown per channel, optionally), which can come from anywhere.
(Spring bug)
This patch adds support for IES files, a file format that is commonly used to store the directional intensity distribution of light sources.
The new IES node is supposed to be plugged into the Strength input of the Emission node of the lamp.
Since people generating IES files do not really seem to care about the standard, the parser is flexible enough to accept all test files I have tried.
Some common weirdnesses are distributing values over multiple lines that should go into one line, using commas instead of spaces as delimiters and adding various useless stuff at the end of the file.
The user interface of the node is similar to the script node, the user can either select an internal Text or load a file.
Internally, IES files are handled similar to Image textures: They are stored in slots by the LightManager and each unique IES is assigned to one slot.
The local coordinate system of the lamp is used, so that the direction of the light can be changed. For UI reasons, it's usually best to add an area light,
rotate it and then change its type, since especially the point light does not immediately show its local coordinate system in the viewport.
Reviewers: #cycles, dingto, sergey, brecht
Reviewed By: #cycles, dingto, brecht
Subscribers: OgDEV, crazyrobinhood, secundar, cardboard, pisuke, intrah, swerner, micah_denn, harvester, gottfried, disnel, campbellbarton, duarteframos, Lapineige, brecht, juicyfruit, dingto, marek, rickyblender, bliblubli, lockal, sergey
Differential Revision: https://developer.blender.org/D1543
If the object is manifold and the camera is in the shadow side, we can
use the depth fail method to fix the inverted shadow glitch.
Unfortunately this does not really work for non-manifold.
Implementation details:
We try to be as efficient as we can, we precompute camera near plane
projected into 2D shadow space so we can test for intersection with the
shadow boundbox easily.
As the intersection test is done in 2D it's pretty fast.
Unfortunately, this means the shadow bounds are all aligned to the same
space and are not the smallest bound we could extract.
This new API aim to provide simple function that can be called by the draw
engines during any phase of the draw pipeline. All calls are saved and
issued after all engines have finished rendering.
This removes the need of setuping special passes and shading groups for some
simple debug drawing.
ninja is an alternative to msbuild designed for fast rebuilds. However there is no IDE support, builds only from the command line.
Comparison between msbuild and ninja for a full build, build time in seconds.
Full Clean Build
msbuild 867.5
Ninja 801.2
Difference -66.3 (-7.6%)
Minor Change
msbuild 43.0
Ninja 14.9
Difference -28.1 (-64.4%)
No Changes
msbuild 23.0
Ninja 6.1
Difference -16.9 (-73.5%)
make.bat was starting to become hard to maintain, this refactors it into separate batch files for each stage of the process.
-Improved detection of msvc2013/2015
-Improved failure handling.
-Added check for working msbuild and C++ compiler
-Added verbose switch to ease trouble shooting.
-Added Check if svn/cmake/git are in the path before using them
-Display the build configuration before asking to download the libraries
-Offer an option to recover an interrupted checkout of the libraries.
-Automatically check out sub-modules in-case they are missing.
The problem was that the particle system modifier was reading ob->derivedDeform
during modifier stack evaluation. Due to the mesh -> DM conversion this was no
longer set leading to wrong results.
In fact we don't really need the deformed mesh, just the original mesh topology
for face/poly index remapping. So the solution is to use that instead.
This commit adds number formatting (thousands separator) to the baking panel. It also adds a new function to format memory sizes (KB/GB/etc) and applies it to the baking panel and scene stats. The new function is unit tested.
Reviewers: Severin
Tags: #user_interface
Differential Revision: https://developer.blender.org/D1248
The registry hack we were using wasn't very reliable, the recommended way to locating visual studio is using vswhere (15.2 and up), using it also allows to switch between the regular and pre-release versions.
Just as with my earlier fix for bones (0492e56fec),
here we're reading evaluated object values into the "i" versions of each transform
property, and using the original/non-evaluated data for the pointers.
XXX:
What's not clear though is what we should be doing with the "BKE_object_where_is_calc()"
calls here. They currently use ob, but that probably shouldn't happen... and ob_eval
should in theory have a more up-to-date version of what it would need to evaluate!
* Added BKE_object_get_evaluated_mesh, which can also be used outside of
Modifier context.
* BKE_modifier_get_evaluated_mesh_from_object is now a dummy wrapper
around BKE_object_get_evaluated_mesh, we do not need anything special
anymore for RENDER quality option, since this is supposed to be handled
at depsgraph level... Maybe we can get rid of it at some point, but kind
of like the idea of keeping it for now, sounds more consitent.
* Always use BKE_id_new, unless you have a very good reason to use
lower-level code!
* Prefer to pass actual ID user pointer to functions like id_us_plus & co,
rather than 'floating' ID pointer, when possible. It makes it more clear
who is the user we increase count for!
Don't store pointers to ViewLayer in the workspace, only names. Add specific
relation type since the generic mechanism makes the code hard to follow.
Integrate with pointer restore for undo and library remapping code to avoid
data going out of sync.
Also add relation automatically if there doesn't exists one yet in
BKE_workspace_view_layer_get, because in general it's really hard to ensure
it will exist when making arbitrary scene changes.
Differential Revision: https://developer.blender.org/D3432
This caused a glitch with COW, where forcing edge selection
caused the evaluated scene to enable this afterwards.
Now pass the selection mode as an argument to the draw function.
Note from revisor: Edit the error message to match the original one.
Maniphest Tasks: T54643
Differential Revision: https://developer.blender.org/D3351
These needed to be using the COW evaluated data, instead of the raw bone
positions.
All other datatypes still need converting to work with this though.
The GPU kernel needs to use atomics for accumulation since all offsets are processed in
parallel, but on CPUs that's not the case, so we can disable them there for a considerable speedup.
This commit adds a new menu entry - "Edit Driver" - the RMB menu that
will show a popover panel displaying the settings for the driver you
activated the menu item on. This shows the popover panel defined in
yesterday's commit (GRAPH_PT_drivers_popover).
It is possible to edit the driver settings from this panel now.
However, do be warned that the functionality presented is highly
WIP still. There are some unresolved issues, such as:
- The popover disappears too easily on any mouse movements/clicks
on anything, making the panel less useful right now than it should.
- The layout still needs refining. Currently the layout that's there
is a bit of a placeholder until we can play around with it a bit
more to see/feel what feels good/right or what is too much.
- The "Open Drivers Editor" on the bottom of the panel doesn't work.
There are some tricky context tricky things that need to happen here
to make that case work, since the operator button won't have the necessary
context info.
The Math node currently has the normal atan() function, but for
actual angles this is fairly useless without additional nodes to handle the signs.
Since the node has two inputs anyways, it only makes sense to add an arctan2 option.
Reviewers: sergey, brecht
Differential Revision: https://developer.blender.org/D3430
The implementation is pretty straightforward.
In Cycles, sampling the shapes is currently done w.r.t. area instead of solid angle.
There is a paper on solid angle sampling for disks [1], but the described algorithm is based on
simply sampling the enclosing square and rejecting samples outside of the disk, which is not exactly
great for Cycles' RNG (we'd need to setup a LCG for the repeated sampling) and for GPU divergence.
Even worse, the algorithm is only defined for disks. For ellipses, the basic idea still works, but a
way to analytically calculate the solid angle is required. This is technically possible [2], but the
calculation is extremely complex and still requires a lookup table for the Heuman Lambda function.
Therefore, I've decided to not implement that for now, we could still look into it later on.
In Eevee, the code uses the existing ltc_evaluate_disk to implement the lighting calculations.
[1]: "Solid Angle Sampling of Disk and Cylinder Lights"
[2]: "Analytical solution for the solid angle subtended at any point by an ellipse via a point source radiation vector potential"
Reviewers: sergey, brecht, fclem
Differential Revision: https://developer.blender.org/D3171
This fix applying the following modifiers:
* Boolean (working already)
* Array
* Mesh Deform
* Surface Deform
* Vertex Weight Proximity
This function was to return evaluated mesh. So it should get the evaluated
object at all times. So in this case it makes more sense to simply pass the
depsgraph (or in this case the ModifierEvalContext that contains both the
depsgraph and the flag.
Solution discussed with Bastien Montagne.
- Adjusted order and names for better consistency.
- Group add/edit/delete into menu sections.
- Move UV/Vertex color into 'Face Data' menu, matching 'Edge Data'.
- De-duplicate items between vertex/edge/face menu.
- Remove undo/redo menu (to be moved into 'Edit' menu).
- Remove Auto-Merge & Proportional Editing
(already available from the 3D view).
Patch from @billreynish w/ edits
Was breaking static override chaining (static override of a linked
datablock that is already a static override in it original lib file).
Dummy mistake, thanks to @dfelinto for finding the bug.
574c0fe38f caused complications
restoring the cursor.
Use WM_cursor_modal_set/restore which is intended for operator use,
calling ED_region_cursor_set after an operator runs isn't reliable.
Fixes bug with changes to properties not being flushed to the COW data.
It fixes T55144.
This is the part of rBb4b745b72064 that is required although slow.
Which was partially addressed but the rest of the commit, which in
turn broke things. So for now let's get RNA to flush slow cow, and
deal with the consequences.
affects one item
UI editing multiple selected items missed the case of PROP_POINTER
properties
Reviewed By: campbellbarton
Differential Revision: https://developer.blender.org/D3373
Some conversion helper functions were (most likely by accident) contained
inside an ifdef for SSE2 support, so on e.g. ARM they would be undefined
and therefore cause compilation to fail.
* Add horizontal bar at bottom of all non-temp windows, similar to the Top-bar.
* Status-bar is hidden in UI-less fullscreen mode
* Current contents are preliminary and based on T54861:
** Left: Current file-path if needed. "(Modified)" note if file was changed.
** Center: Scene statistics (like in 2.7 Info Editor).
** Right: Progress-bars and reports
* Internally managed as own "STATUSBAR" editor-type (hidden in UI).
* Like with the Top-bar, Status-bar data and SDNA writing is disabled.
* Most changes in low-level screen/area code are to support layout bounds that differ from window bounds.
Design task: T54861
Main changes approved by @brecht.
This reverts commit b4b745b720.
This was causing a problem in 01_025_A.anim.blend from the Spring
production files, where selecting one of Autumn's bones would
result in character jumping back to the origin.
Note the relationship we need is from the geometry/object to the curve data,
not the curve object.
Related to T55167, though when copy-on-write is enabled, we still get a crash.
Otherwise, it's not clear where things are coming from.
FIXME: The icons for datablocks may not always be correct. It uses the
innermost struct's icon instead of the datablock's icon - e.g. Bone
vs Object/Armature. But, that may make more sense for users?
The idea is that we may be able to just take this code and put it into a
new popover panel that gets displayed when called from the UI. This should
at least work well for a first pass test of what we want this UI to look like.
In theory, this should mean that we can get rid of the "Update Dependencies"
button. In practice, there may still be cases where it's still needed
(as somehow, it did end up being needed in the past, even though the RNA
calls should in theory be doing everything needed already).
There are multiple other ways to do it. Leaving this here just made it
easy (and dangerous) to accidentally remove the driver, and was causing
other problems with other layouts.
* Remove "Show Debug Info" option. Everyone has it turned on all the time,
since it's just useful to have
* Make the "Remove Driver" button less prominent. It doesn't happen that much,
so it shouldn't take up as much room
* Make "expressions" textbox wider (i.e. taking up the whole column width)
by separating the label and textbox on separate lines.
* Rename "Add Variable" button to make it clearer to users what "variables"
may be (i.e. they serve as a way to specify Inputs, just like adding a
"Source Data" node in a nodetree)
* Regroup buttons
This is intended to help developers to know how and when to use each shader.
There are plenty of undocumented shaders, but it's a matter of filling them in.
The script I used to quickly find the related shaders for a const is: P700
Original patch: D2318
Unmultiplied the final color during compositing. Same as the revealage
buffer would do
Also use the DRW_STATE_ADDITION_FULL as it is aware of premultiplied
colors
The modifier is still quite slow; this could be due to caches being written
to a CoW datablock instead of the original one. More investigation is
needed.
Logic here was a bit broken.
1) We need to send updates even when deselecting bones, not only when we
actually end up select then.
2) Also, when bones are locked for "selectability", they should not be able
to be deselected.
To bring the UI more in line with the proposed design in T54653 for the "Add Drivers"
popup panel (NOTE: this is separate from the "Drivers Editor", in previous commit!),
this commit adds a new panel - "Driven Property" to the Drivers Editor UI.
This basically duplicates the "Active F-Curve" panel (with less options)
to make it easier to see at a glance which property the Drivers Editor is
showing you.
This commit adds an operator, "Show Drivers Editor", to the RMB menu when
clicking on properties.
As per T54653, this will open a new Graph Editor instance in a new/separate
window (much like how the User Preferences show up in a popup window now),
and will configure all the relevant panels so that you can see and edit the
driver settings immediately without doing a lot of the view configuration steps
that were previously needed.
When doing so on a property that is driven, the driver/fcurve for that property
will be made active in the editor, ready for you to start editing its settings
without having to hunt it down again first.
Allows more than one snap mode to be enabled. So different combinations are possible.
Reviewers: campbellbarton
Reviewed By: campbellbarton
Subscribers: Christopher_Anderssarian, duarteframos
Tags: #bf_blender_2.8
Differential Revision: D3400
Only show hierarchy relationship lines when bone or its parent is selected.
This cuts down the clutter visible in general when relationship lines are
enabled (currently they can't be disabled), which should make it more useful
to keep them on (e.g. constraint lines/hints can still be drawn this way).
As was decided at today's dev kickoff, we're now moving to having
Copy-on-Write enabled by default, as 2.8 is barely functional with
it off.
To run Blender *without* COW (e.g. for testing), use:
--disable-copy-on-write
You need the whole chain of pointers... This was breaking static
overrides of any rig basically.
Note that this kills performances again (adding several hundreds of
thousands more stuff on a blendrig...), need a better way to handle RNA
override walking.
Fixes camera view locking, camera related transitions and movements with the active object as pivot point
Note there can still be problems if the active object is not selectable
Seriously... Not all edges have faces!!!!!!!!
Quick fix to make it possible to load/use customshaped bones again, not
100% sure it is correct, but... At least it does not crash anymore!
- This allows quick, consistent toolbar access
w/o conflicting w/ the keymap.
Where pressing space before a key activates that operator as a tool
instead of running immediately.
- Search can still be accessed by pressing spacebar again.
- When there is no toolbar for a space, operator search still opens.
If a mesh is known to be manifold, then it's not necessary to increment the
stencil buffer 2 times anymore. But we still need to account properly for
degenerate triangles.
In this case, only generate a quad if the tri is facing the lamp. If there
is a degenerate loop, the other edge will either cancel the increment (if
it is also facing the light) or not produce a quad (if not facing).
This will always give the correct count.
Seems that degenerate tris are somewhat widely used in modeling. So we need
to fix this. Test the edges in the geom shader since the adjacency info is
only dependant on topology, not actual vertex placement.
This fixes most of the remaining noise issues. Only a few artifacts appear
on really weird models. So if you want to get rid of the artifacts, fix
your model!
This makes the shadows ~10 times faster in the general case.
This only create extrusion geometry on the outline edges. Also we increment
or decrement the stencil buffer by 2 for each manifold edge and only by 1
for non manifold. This make the algorithm robust yet less heavy than creating
one prism for each triangles.
Operators did not suffer from this problem as they were still just using notifiers
directly. The "proper" fix is to use the new message bus system. But, we've
got enough problems dealing with COW already as it is now... moving on.
The transform code needed to use data from the "_eval" copy of the posebone
instead of the main-db version, otherwise the initial/reset value would be
wrong, causing the jumping.
Without the exception, adding new poses to pose libraries took several seconds
with only <= 4 bones selected. While we may still need this for other cases too,
since bones are such a common use case, it makes sense to provide some level
of optimisation for them.
This now works by getting the RNA Path from the given PointerRNA to go from the
ID block to the data it points to, then uses this path to find the new data
relative to the COW ID.
Note: This currently still has all the debug prints left in - As can be seen,
I was testing this against the earlier PoseBone hack/special case. We may still
need to bring such special cases back in future, since looking up RNA Paths
like this can be slow.
When using copy on write, insert keyframe operators were reading from old
bmain data instead of COW data. This meant that inserting keyframes would
often read old/stale data, resulting in invalid keyframes getting created
(e.g. from last transform operation, instead of actual current state).
This commit makes it so that keyframing operators will ask depsgraph for
the evaluated copy of the data, so that it can read values from that. It
introduces a new function - `DEG_get_evaluated_rna_pointer()`, which when
working correctly/fully, should work just like the other `DEG_get_evaluated_*()`
functions, except it lets you pass in an RNA Pointer.
However, currently, this is only done for Pose Bones (as a dirty hack, since this
is an important/pivotal requirement for production) and/or datablock
properties directly (since we can just use the DEG_get_evaluated_id() directly).
on the datablock.
Committing to a branch for now as this all needs more testing. More work to come
later at a more sane time of day!
Experimental support for using popovers like menus,
use this when the user hold the mouse down
(previously this did nothing).
This means turning frequently accessed menu items into
popovers doesn't add more clicks to the existing use case.
There are still probably other massive problems to solve (i.e. which copies
of data cache/sim gets written/read from for COW eval) that need to be solved
before we can get the sims actually running though.
This adds Eevee render tests using the Cycles files. Currently it must
be enabled by setting WITH_OPENGL_RENDER_TESTS=ON. Once we have reference
images we can enable it by default.
Some of the Cycles and Eevee tests are also currently broken due to
modifier and particle changes.
Differential Revision: https://developer.blender.org/D3182
OVERVIEW
* In 2.7 terminology, all layers and groups are now collection datablocks.
* These collections are nestable, linkable, instanceable, overrideable, ..
which opens up new ways to set up scenes and link + override data.
* Viewport/render visibility and selectability are now a part of the collection
and shared across all view layers and linkable.
* View layers define which subset of the scene collection hierarchy is excluded
for each. For many workflows one view layer can be used, these are more of an
advanced feature now.
OUTLINER
* The outliner now has a "View Layer" display mode instead of "Collections",
which can display the collections and/or objects in the view layer.
* In this display mode, collections can be excluded with the right click menu.
These will then be greyed out and their objects will be excluded.
* To view collections not linked to any scene, the "Blender File" display mode
can be used, with the new filtering option to just see Colleciton datablocks.
* The outliner right click menus for collections and objects were reorganized.
* Drag and drop still needs to be improved. Like before, dragging the icon or
text gives different results, we'll unify this later.
LINKING AND OVERRIDES
* Collections can now be linked into the scene without creating an instance,
with the link/append operator or from the collections view in the outliner.
* Collections can get static overrides with the right click menu in the outliner,
but this is rather unreliable and not clearly communicated at the moment.
* We still need to improve the make override operator to turn collection instances
into collections with overrides directly in the scene.
PERFORMANCE
* We tried to make performance not worse than before and improve it in some
cases. The main thing that's still a bit slower is multiple scenes, we have to
change the layer syncing to only updated affected scenes.
* Collections keep a list of their parent collections for faster incremental
updates in syncing and caching.
* View layer bases are now in a object -> base hash to avoid quadratic time
lookups internally and in API functions like visible_get().
VERSIONING
* Compatibility with 2.7 files should be improved due to the new visibility
controls. Of course users may not want to set up their scenes differently
now to avoid having separate layers and groups.
* Compatibility with 2.8 is mostly there, and was tested on Eevee demo and Hero
files. There's a few things which are know to be not quite compatible, like
nested layer collections inside groups.
* The versioning code for 2.8 files is quite complicated, and isolated behind
#ifdef so it can be removed at the end of the release cycle.
KNOWN ISSUES
* The G-key group operators in the 3D viewport were left mostly as is, they
need to be modified still to fit better.
* Same for the groups panel in the object properties. This needs to be updated
still, or perhaps replaced by something better.
* Collections must all have a unique name. Less restrictive namespacing is to
be done later, we'll have to see how important this is as all objects within
the collections must also have a unique name anyway.
* Full scene copy and delete scene are exactly doing the right thing yet.
Differential Revision: https://developer.blender.org/D3383https://code.blender.org/2018/05/collections-and-groups/
Without this we need to have the context to get the
(space_type, mode) args for an active tool lookup.
For event handling & poll its more convenient to have direct access.
This piece of code remains annoyingly verbose, but at least now we do
not have anymore twice the same logic duplicated between single array
element handling, and non-array properties.
This patch adds support for:
- Per space-type tools (3D view and edit).
- Per mode tools (object, edit, weight-paint .. etc).
The top-bar shows the last activated tools options, this is a design
issue with using a global topbar to show per-space settings.
See D3395
We handle doversion for the scene properties, but not for the
view layer overrides.
Overrides will be implemented in a different way via dynamic overrides.
For now this data is completely lost.
For now, do not override/instantiate objects used as boneshapes.
Note that this is a rather poor/dirty hack, it seems Spring char groups
still have a lot of other (unused???) shape objects.
Ideally lib groups should be designed more cleanly for the new Static
Override system, putting only really usable objects into 'main' group,
and placing the others in sub-groups, helpers groups, or so...
*Always* use BKE_library API to handle IDs allocation and freeing,
unless you have a very, very, very good reason not to do so - and
perfectly know what you are doing.
No idea why that was working this morning, now we do not get CoW tag
anymore (which kind of makes sense since it's set *after* copying
happens...), so just resorting to using the NO_MAIN flag instead.
This is a usefull feature that can be used to do a lot of precomputation on
the GPU instead of the CPU.
Implementation is simple and only covers the most usefull case.
How to use:
- Create shader with transform feedback.
- Create a pass with DRW_STATE_TRANS_FEEDBACK.
- Create a target Gwn_VertBuf (make sure it's big enough).
- Create a shading group with DRW_shgroup_transform_feedback_create().
- Add your draw calls to the shading group.
- Render your pass normaly.
Current limitation:
- Only one output buffer.
- Cannot pause/resume tfb rendering to interleave with normal drawcalls.
- Cannot get the number of verts drawn.
Patch D3205 by Kanzaki Wataru
Only implemented in Eevee for now. Collapse a closure to RGBA so we can
do NPR stuff on the resulting color.
Use an emission shader to convert the color back to a closure.
Doing this will break PBR and will kill any SSR and SSS effects the shader
the shader rely on. That said screen space refraction and ambient occlusion
are supported due to the way they are implemented.
There are a few places where DerivedMesh is still used, most notably
when calling the (not yet ported) cloth simulation. There is also still
the use of Object.derivedDeform and Object.derivedFinal. Those places are
marked with a TODO.
Some functions in the editors module were copied to accept Mesh. Those
already had 'mesh' in the name; the copies are suffixed with '__real_mesh'
for easy renaming later when the DM-based functionality is removed.
Locked bones of proxies should not be editable, at all. But lack of
update from linked rest pose in 2.7 allows to pose and animate locked
bones (not to pose them without animation though, or you'd lose your
pose on next file save & reload).
this is used by artists to always lock all their bones in a rig, so that
proxies fully update when lib rig is modified...
For now, restore that broken behavior in 2.8 by not updating proxies
against lib armature in CoW context (makes sense anyway, we are
currently doing a lot of useless thing when copying data for depsgraph
evaluation!).
This patch adds support for clip_planes (ie ignore what is behind a face)...
The idea is to first execute a raycast to get the polygon to which the mouse cursor points.
Then a snap test is done on the vertices or edges of the polygon.
Then with the normal and location obtained in raycast a new clip_plane is created and the snap over the whole scene is processed ignoring the elements behind the clip_plane.
Here 2 gif of how the previous patch would work on blender2.79:
{F497176}
{F497177}
Reviewers: mont29, campbellbarton
Reviewed By: campbellbarton
Subscribers: bliblubli
Tags: #bf_blender_2.8
Differential Revision: https://developer.blender.org/D2527
Doing that copy is the right thing, but only if the mesh has its own
deformed_only flag set correctly. This isn't generally the case, so
keeping dm->deformedOnly = 1 is better.
This flag is copied when converting between DM and Mesh.
This flag is set to true in get_mesh(), to mimick the behaviour of
CDDM_from_mesh_ex. This is necessary for the particle system to work
correctly.
Reviewers: dfelinto (changed test from totvert to totfacesel before committing)
Maniphest Tasks: T54643
Differential Revision: https://developer.blender.org/D3306
This ports the main modifier eval loop.
A few things haven't been ported yet: tessface and shapekey related code. As
far as I can tell, shapekey code was never entered into.
https://developer.blender.org/D3342
Non modifying version of `BKE_mesh_validate`, mirrors `DM_is_valid` more
closely. Will be used in port of `mesh_calc_modifiers`
from `DerivedMesh` to `Mesh`.
This relations is only to force geometry evaluation to happen
after scene's CoW is done. it was never meant to update geometry
when scene is being tagged.
Don't try to bend existing design, just follow it. It's not nice, but is
working. Just bring it back, then repeal and replace system as a whole,
making sure every bit is working according to a design.
Such quick patches only trying to make local sense of a system, defeating
it's design.
- Default Lamp increased strength (10x stronger)
- 3D View & Camera Lens = 50mm
- Camera film size = 36x24mm Full Frame
- Render Size Percentage = 100%
- Render Display = New Window
- Scene Units = Metric
- Color Management View = Filmic
- Workbench Object Overlap = ON
- Headers on top for all editors, except the Timeline at the bottom
- Default Properties tab = Object Properties
- Generate UV's = ON
See T47064
The outliner can redraw quicker without rebuilding the tree, for example when
just moving the mouse and highlighting different items. The way this worked is
that the outliner would be tagged to avoid rebuilding, however if another
operation that does require rebuilding happens in the meantime we could go out
of sync and crash.
Currently all are stacked together, but perhaps if we don't get any other
tools in the meantime, it would make sense to have them separate.
Notes:
* No icons for now...
* There are some 2.8 bugs with the underlying operators (notably multi object)
For fun, I tried adding some more edit armature tools to the toolbar to
check how this is all working. Icons are missing currently, but it's a
nice test.
Regression in recent undo system changes,
This caused T55048.
When each mode had its own undo stack it was important
to initialize it when entering edit-mode.
This patch does not make any difference for a user's POV. But it is a step for adding the occlusion test for snapping functions.
This new function finds the node(aabb) whose projection is closest to a screen coordinate.
Reviewers: campbellbarton
Reviewed By: campbellbarton
Tags: #bf_blender_2.8
Differential Revision: https://developer.blender.org/D3180
With a few changes:
* Use space after if.
* Skip for loop when no selected vertices.
Maniphest Tasks: T54643
Differential Revision: https://developer.blender.org/D3349
With changes by Dalai Felinto:
* Indentation
* Skip for loop when no vert/edge are selected
Maniphest Tasks: T54643
Differential Revision: https://developer.blender.org/D3233
Previously, the "layers_used" value was getting updated by the drawing code.
However, when using copy on write, the drawing code gets evaluated copies of
the armature data instead of the original data, so any updates here fail to
get flushed to the original data, hence the lack of updates in the UI.
Fixed by moving the calculation to RNA when setting bone layers, as it should
have been done originally. (The one downside to this is if we set individual
layer memberships one by one - this could be slower as the recalc would have to
happen each time this changes).
With changes by Dalai Felinto:
* Move WM_operator.* outside for loop.
* Update error message to handle Mesh and Meshes.
* Skip main functionality when no vert/edge/face is selected.
Maniphest Tasks: T54643
Differential Revision: https://developer.blender.org/D3371
* Move Shading modes out of the popover
* Move Show Overlays out of the popover
* Test moving the Mode to the topbar again
* Move submode (select vertex, edge, face) to the topbar
* Remove icon from show_manipulator toggle
With changes by Dalai Felinto (skip for loop when no vert/edgeface selected).
Maniphest Tasks: T54643
Differential Revision: https://developer.blender.org/D3302
SSR does not work with hair strands. Basically, the rays are too much
random to ever converge properly and just result in a soup of self
inter reflections.
So forcing it to not produce any SSR. Could potentially save some bandwidth
by not rendering hair to the SSR buffers.
This is a hack to properly shade wire hairs. Use stochastic sampling and
let TAA solve the noise.
At least it's way more correct than the previous hack.
Linking/appending in edit mode currently isn't supported. For workspaces it
should probably be, but we can look into supporting this later.
For now gray out buttons in "Add Workspace" menu while in edit mode.
This avoids having to store the tool definition in the operator
properties (on redraw).
Now just set the name, note this means we can't show the keymap in the
tool-tip anymore. This will eventually be shown in the status bar.
This means only one texture to draw to and only one sprite per pixel.
The texture is twice as large and near and far planes are side by side.
The sprite choose the biggest coc to expand to and is redirected to the
area (layer) it belongs to.
The fragment shader discard every pixel that does not belong to the correct
layer.
Due to the scatter operation being done at half resolution, undersampling
is visible at bokeh shape edges (because of the hard cut).
This commit adds a smoothing function to minimize the problem.
Also optimize the bokeh shape parametrization by precomputing a lot of
constants.
This new blending allows background to fill the gaps left by forground
objects. However this has a drawback, background objects that should be
partially occluded in this case can be seen through the blurred objects.
This does not fix the problem of blurred foreground over sharp background.
Also cleanup code to be simpler and remove unused geometry shader.
The Studio lights are now loaded from disk. The location is
`datafiles/studiolights` they need to be JPG for now. JPG cannot store
HDRI color range but they are clamped inside the Workbench
engine for speed reason. I didn't select JP2K as it might not be
enabled.
Users can add upto 20 HDRI files. This limitation is inside the
RNA_space.c Currently the icons are calculated when you first open the
selection box for the HDRI's. We could add them to a background
rendering later.
I added 2 test files a sky texture rendered in Cycles and an HDRI from
cloud.blender.org.
We can not rely on edit->psys, it is not set for particle edit,
and there is some logic deeper inside which does different things
dependent on that.
We need to replace those checks with some some HAIR vs. PARTICLES
flag and always set psys pointer.
The idea is that edit mode structure is owned by original object,
and used for drawing. This is a bit confusing, especially since
path cache is also in that structure and needs evaluated object
to calculate cache.
In the future we should split edit data from visualization data,
but that's bigger refactor.
Freeing sequencer would always do usercount, which is now forbidden when
called from main ID freeing code.
Annoying in 2.7x, much more critical issue in 2.8!
Also, moved RNA sequencer API functions to proper rna_scene_api.c file.
Freeing sequencer would always do usercount, which is now forbidden when
called from main ID freeing code.
Annoying in 2.7x, much more critical issue in 2.8!
Also, moved RNA sequencer API functions to proper rna_scene_api.c file.
Two things to indicate (which als apply before the DM → M port):
- Face count in the UI is not updated and stays at 0
- When planar mode is used, the result is inverted (mirrored along X)
compared to 2.79
Reviewers: sybren
Differential Revision: https://developer.blender.org/D3303
Bone selection overlay is only available in pose mode.
and when active overrules the selection buffer.
This is currently `tricked` by switching the draw engines, but this is
an exception. Not sure how to solve this in a better way.
After this is solved we can look at how to localize the dim effect to only the objects connected to the active armatures. Currently it dims the whole screen (including background).
@campbellbarton I added you as reviewer as it you have done a lot in the DRW_draw_select_loop
Reviewers: campbellbarton, fclem
Reviewed By: fclem
Subscribers: campbellbarton
Tags: #bf_blender_2.8, #code_quest
Maniphest Tasks: T54983
Differential Revision: https://developer.blender.org/D3241
Initial add-cube tool which uses the scale cage to resize.
The 3D cursor is currently used to project the cursor onto.
We'll likely have more orientation options in the future.
NOTE: This is all still heavily WIP, and still requires some additional
layout engine magic to make it all nice.
* Moved current frame, and start/end frame buttons to the end
* Put playback controls in the center
* Remove other more obscure buttons (e.g. Frame menu, lock frame to frame range)
More to follow in next commit...
Now all Keying Set and Auto Keyframing settings can be found the in NKEY
Properties Region (right-hand-side hidden tab/drawer).
This way, we have more room to display the full names of the settings,
without clogging up the timeline header.
Bvhtrees take up a lot of memory space, reusing the common bvhtree of looptris to snap to vertices and edges is a good way to save memory.
Unfortunately we have a worsening performance in the snapping operation around 63% (addition to the original time).
But as we often do not need to build a bvhtree of loose verts and loose edges, we have an improvement in cache time :)
Since the CPU time of snapping operations (no matter how higth poly the object is) corresponds to less than 0.01% of all CPU time of a blender frame, that change is not really significant.
Snapping operations on a mesh in edit mode have not changed significantly.
Signed-off-by: Germano <germano.costa@ig.com.br>
Everything seems to be working without this, which makes since there is no
evaluation-time edit code in master. The only tricky part here would be that
this means we might want to ensure edit structure exists from inside particle
mode in draw manager.
Other point for not having this code in evaluation is that it's possible to
have same object evaluated in different contexts at the same time, and since
edit structures are supposed to be in original object we might run into
threading conflict.
Readcode always set relative paths of indirectly linked libs relative to
*current* .blend file, not to the library using it.
But BKE_library_filepath_set was then setting them relative to their
parent library, breaking checking code (and saved files even :((( ).
in Scene DNA is available in scene.display New DNA Struct DisplayData
Added RNA (scene.display).
We already have scene.display_settings which contains non viewpoert
specific color management settings. I did not merge those two.
Patch should be in line with the ideas that @brecht wrote in T55008.
As I am not in detail aware of the decisions that have been made for the override feature I want a short review if it is in line with the plans.
Reviewers: brecht, dfelinto
Reviewed By: brecht
Subscribers: brecht
Tags: #code_quest, #bf_blender_2.8
Differential Revision: https://developer.blender.org/D3290
This feature is limited (only byte PPM output, no multi-view),
only works with specific configurations.
This also causes some maintenance overhead
when testing changes to the render pipeline.
There is still an issue with the modified mesh not being updated until you
disable and re-enable the modifier. However, after that it'll now also work
while editing the deformation mesh.
The modifier performed the 'bind' operation not in the bind operator,
but delayed in the mesh evaluation. This saved the result in a CoW copy
instead of in the actual modifier data. The binding operator now follows
the same approach as Mesh Deform: it forces the modifiers to run on the
real (non-CoW) data, making it possible for the modifier to store the
binding data.
This commit also ports the usage of DerivedMesh to Mesh.
Patch description:
- Reports "Nothing selected" only when all objects has no selection
and it is in Separate type(Mode)
- Tested all types (Seperate/By Material/By loose parts)
- Instead of using
```
BKE_view_layer_array_from_objects_in_edit_mode_unique_data
```
I used
```
BKE_view_layer_array_from_bases_in_edit_mode_unique_data
```
Because it needs the "Base" not the "Object" itself.
Maniphest Tasks: T54643
Differential Revision: https://developer.blender.org/D3267
With minor changes:
* Have BMOperator declaration close to its initialization.
* Move const use_prepare and const use_interp_simple outside the for loop.
Maniphest Tasks: T54643
Differential Revision: https://developer.blender.org/D3222
Gets edit more from the current object and displays it as a path.
this is how both hair and particle edit modes are supposed to work.
This only covers path itself, it doesn't do anything like keys
visualization or selection. However, it's already possible to
comb and such.
Only implements particle mode. There are also some settings to
do soft body and cloth. No idea yet what that all is about.
Copy-on-write is not supported either, this is due to some
edit mode ownership problems which are to be addressed from
dependency graph side.
Shading is dead-simple: uses tangent as a color. This is where
i hope to get some help from Clément.
BLI_bvhkdop functions were not made to work with zero-leaf trees.
Perhaps a better solution would be to modify BLI_bvhkdop to work with zero leaf trees.
But this solution of returning NULL was already used for bvhtrees of looptris.
Some years-old deprecated stuff has now been removed.
Correct solution is probably to use valid defines etc. in own code, but
this is more FFMEPG maintainer task (since it also may change how old
FFMPEG we do support...).
When entering edit mode the texture coordinates is not calculated by
drawcachemesh. For now fallback to solid mode. This gives wrong
artifacts when disabling overlays when in edit mode. We should make it
possible to calculate the UVLOOP from bmesh. This will be done in a next
patch
The actual code is a bit convoluted but allows good and "pseudo efficient"
drawing. (pseudo efficient because rendering instances with that amount of
vertices is really inneficient. We should go full procedural but need to
have bufferTexture implemented first) But drawing speed is not a bottleneck
here and it's already a million time less crappy than the old (2.79) immediate
mode method.
Instead of drawing actual wires with different width we render a triangle
fan batch (containing 3 fans: bone, head, tail) which is then oriented in
screen space to the bone direction. We then interpolate a float value
accross vertices giving us a nice blend factor to blend the colors and
gives us really smooth interpolation inside the bone.
The outside edge still being geometry will be antialiased by MSAA if enabled.
Now we use vcos when provided, and fall back to mesh vertices' co
otherwise.
Deform modifiers usually do not have up-to-date coordinates in Mesh
itself, only in given vcos array!
Aside from identation, seed should change for each object. Otherwise to objects
with a copied mesh would get the exactly same "random" selection.
And we do so in a consistent way so it doesn't vary depending on
object order.
Seed bump using hash suggested by Campbell Barton and Sergey Sharybin.
Maniphest Tasks: T54643
Differential Revision: https://developer.blender.org/D3231
This modifier still has issues that are not related to this port:
- While editing the deformation mesh, the deformed mesh doesn't update.
This update only happens after exiting edit mode, making editing
cumbersome.
- Binding doesn't work yet. It works fine when binding in master and
loading pre-bound in 2.8. This was also an issue before this port, and
will be investigated separately.
This function creates a Mesh struct with a number of vertices/edges/etc.
It allocates the minimal number of CD layers needed.
Currently not yet used, but will be soon in the upcoming
BKE_new_mesh_from_curve_displist().
With changes by me:
* Indentation (tabs)
* Removing RNA_boolean_get tests out of the for loop.
* Continue iterator if no selected vertice.
Maniphest Tasks: T54643
Differential Revision: https://developer.blender.org/D3210
Some notes here:
* Proximity with non-mesh objects (like curve, see TEST_2 scene in
weightvg testfile) are not working currently. This is known TODO of COW
depsgraph project.
* Proximity modifier is slower, due to some other TODO pending on
BVHTree creation/caching for Mesh.
This commit brings back motion path calculations working on same level
as 2.7x. Caveats:
* You cannot actually see anything, as all the drawing code for motion
paths was removed and not reimplmented in 2.8 draw engines yet
* This doesn't actually use a separate depsgraph instance for fast
background calculations yet. That's a separate topic, and will require
further investigation (probably after code quest, according to original
planning).
* The work here *does* however pave the path forwards towards fully separate
COW evaluation though. We only need to stub in a different depsgraph instance
here to get things working.
frames
caused division by zero if both dupli_frames_on and dupli_frames_off are
zero. doing this doesnt seem useful, dupliframes can be disabled in
other ways.
Reviewed By: campbellbarton
Differential Revision: https://developer.blender.org/D3132
frames
caused division by zero if both dupli_frames_on and dupli_frames_off are
zero. doing this doesnt seem useful, dupliframes can be disabled in
other ways.
Reviewed By: campbellbarton
Differential Revision: https://developer.blender.org/D3132
Note that custom normals drawing seems to be broken, and there also are
some refresh issues in some cases... But this is same with old DM-based
code, so not related to modifiers themselves probably.
Steps to recreate were:
* Go into fullscreen mode (Alt+F10)
* Open debug menu (Ctrl+Alt+D)
* Set value to 1
* Confirming should crash (at least in debug builds)
Was hidden behind debug values 1, 2 and 3.
Also cleaned up logic in region_azone_edge_initialize. It was using a variable
called is_hidden that actually should've been called is_visible.
Since modifier stack is still primarily based on DM in blender2.8, we
need to copy over 'dirty normals' flag from temp Mesh to DM after
modifier has been evaluated.
Scroll-bars are now hidden unless the cursor approaches them, in which case they
smoothly grow and become more & more visible. Note that since 0d309144020168e55,
scroll-bars are drawn on top of editor contents. There's no more jumping of
buttons when scroll-bars appear.
Technical notes:
* AZones are used to adjust scrollbars based on mouse movements
We may want to support screen level AZones if we want scrollbars to also
smoothly appear when approaching them from a different area.
I also plan to make further changes to AZones to clean up stuff a bit.
* Had to move AZone handling to a post ARegion init stage, since we need the
updated View2D data from there.
* View2D masks and scroller rectangles are now updated on every redraw. It's
cheap to do that though.
Scroll-bars used to draw in a little extra space in the editor, causing
buttons to jump a bit when they appeared/disappeared. Now they draw on
top of the buttons, just small enough to avoid bigger overlaps. Followup commits
will do further adjustments.
With this we can get rid of a hack that was calling the (often Python
defined) panel definition - the panel 'draw' callback - twice.
Note: This rely on the property having a pre-defined default.
Also, be aware that trying to multi-drag and multi-ui-edit is not working
at the moment (T54976).
With changes by Campbell Barton.
Differential Revision: https://developer.blender.org/D3207
* Ensure popover does not change direction or location.
* Open popover towards the relevant editor, like pulldown menus.
* Use a bigger maximum assumed size to deal with some corner cases.
* Do proper 3D viewport header refresh on shading mode changes.
In object mode, the axes are drawn like any other wire objects with
depth test and depth write. Thus enabling MSAA to work but not their xray
behaviour.
In edit armature/pose mode, draw smooth line without depth testing. This
produces wrong draw ordering problem but still gives the desired xray
behaviour. We do it outside of the MSAA pass since the xray behaviour is not
compatible with it. But we are drawing smoothed lines so no need for MSAA.
The lines are 2px thick and improve readability.
Now the axes are displayed correctly at the tip of the bone and with the
axes names.
I've made some modifications though:
- Axes are colored. (should not be in object mode but that's TODO)
- Axes ends are not flat arrows anymore. Replaced with a small diamond.
- Axes names are now scale by their respective axes instead of being
affected by other axes.
- Changed axes names "font" to be a bit more sexy.
This is for the 3D viewport and image editor. A remaining issue is
that pulldown menu buttons only show text, which is not visible on
all background colors, so making the header entirely transparent
does not work well at the moment.
There is no object to update anyways, and there is no valid basact when the
nified_findnearest() test fails.
How to reproduce the bug: Try to select linked (L) with selected faces but
withotu mouse hovering any mesh.
Bug introduced on: rBbfc9d426bb95 (original multi-object edit commit).
This makes a few changes:
- Remove the old "overlay" wires.
- Add constraints colors to bones.
- Specify a a new "hint" color per bone. Making selection/Active state
more obvious.
- Unify Octahedral/B-Bones/Envelope shading and colors.
- Change outline size depending on the selection/active state of the bone.
Note that thoses changes are not final and needs review.
This will enable us to do more nice stuff in future commits.
This commit is a temporary commit, it will compile but will crash if
trying to display any armature. Next commit does work.
For some we may add per object overrides, but for most we plan to keep them
strictly per viewport settings. Display settings from the mesh still need to
be moved here, only collections were done to remove that code.
While testing UI_OT_reset_default_button, discovered that using
Reset to Default Values on bone locations didn't work. Turns out
to be missing update on this property. So, this is probably
dependent on the property used.
This was the otherway around before. But since we can have a different size*
for cube texture now, we can compute the correct-ish texture size.
This will give us on average the same texture appearance when we will add
support for real cubemap shadows.
As much as I want to give freedom to the user, 1.5G of vram for a
single shadow is a big of a stability issue.
So limiting to 4096 for now, we may remove this limit in the future.
This mean we can now have different shadow resolutions for both.
However each shadow type keep the same size accross all lamps because of
future "real" Cube Shadowmaps limitation and to save texture sampler slots.
That said the cascade shadow resolution could (in the future) still be
changed to be adjustable per sun lamp.
Existing code checked pose/edit mode to check for transforming in an
objects local space.
This added many similar checks all over the code,
which leads to confusion.
Multi-edit caused a regression in UV transform since where UV's
had the object matrix applied by accident.
Now there is a boolean to use a local matrix,
this allows for any mode to have a 4x4 matrix
applied w/o adding mode specific checks everywhere.
Initial review of the shard shadows in the workbench engine.
Speed optimizations like transform feedback are not implemented yet. I first want this part to be reviewed and merged.
@fclem please check the note in drw_stencil_set it was holding back nequal == 0 as by default DST.stencil_mask was set to 0. questioin is should we remove the whole check or not.
Also I am still looking for a better name (or split the enum) for DRW_STATE_STENCIL_DEPTH_FAIL_INCR_DECR_WRAP
Reviewers: fclem
Reviewed By: fclem
Tags: #code_quest
Differential Revision: https://developer.blender.org/D3198
Those should be used in priority when you need to create either a new
datablock in Main, or a new temp one. Lower level BKE_libblock_...
should only be used when you need a very specific, uncommon behavior.
The value of epsilon was never used to create this bvhtree because whenever we activate this constraint, a bvhtree with parameter epsilon 0.0 was created and cached.
Seems to be only related on linked nature of particles.
This is caused by some conflicting optimization done for viewport, which
does not do particles re-calculation if they do not depend on time
(which is crucial for big layout scene grass fields) and particle render
setting switch which was relying on fact that render pipeline will do
particle update via time dependency.
Now we extent an old workaround for invisible objects, which now also
deals with particles in the same way as old dependency graph was dealing
with this: tag object data for update if there is particle system.
There shouldn't be any speed difference between old and new depsgraph,
since tagging was already needed and was happening.
In Blender 2.8 such things should be easier to deal with since the whole
depsgraph is to be evaluated for render engine anyway.
`MREMAP_RAYCAST_APPROXIMATE_BVHEPSILON(ray_radius)` greatly increased the radius making for example that 0.1 becoming 1.5
Now the result is much more predictable.
* Remove the "sync_mode" dropdown from timeline header, and move it into
the Playback menu instead.
* Remove the confusing "Frame Dropping" and "AV Sync" entries from the
Playback menu. These were supposed to be mutually exclusive (or else,
the "sync_mode" menu would break).
* Turn AV Sync on by default
* Commented out the Frame Dropping and Av Sync RNA Properties too.
THere doesn't seem to be any valid reason for these to exist?
Note: This still needs some copy-on-write magic to get the bone pointers
to get remapped properly (e.g. evaluated pose still refers to original
arm->bones, and arm->act_bone isn't getting remapped).
After copying animation data we need to re-map action, otherwise
our original DNA values will go out of symc, causing flickering
when tweaking values on animated node tree.
The idea is to re-use evaluated datablock from within current context.
This relies on the following factors:
- It expects scene to be fully evaluated before preview jobs starts.
- It expects datablocks to be localized.
The workaround removal was wrong, the whole id->adt of a local
copy is to be NULL.
But now instead of modifying original datablock, we tell library
manager to not copy animation data.
This was due to the background being drawn by a batch that had its VAO
generated in the windows "UI" context.
Since we use the DRW ogl context to draw the entire area, we have to
regenerate the VAO for thoses UI batches to be drawn correctly.
The actual weighting calculation is not smooth as the bone display.
The bone itself can be smooth for esthetic purpose but the distance display
should match the underlying weighting formula.
Past shader was too slow and had bad artifacts. This method is much simpler
and eficient and only exhibit some popping when the raidus of the head/tail
is changed.
We now use a more pleasant and efficient way to display enveloppe bones
and their radius.
For this we use a capsule geometry that is displaced (in the vertex shader)
to a signed distance field that represents the bone shape.
The bone distance radius are now drawn in 3D using a "pseudo-fresnel" effect.
This gives a better understanding of what is inside the radius of influence.
When capsules are not needed, we switch to default raytraced points.
The capsules are not distorded by the bone's matrix (same as their actual
influence radius) and are correctly displayed even with complex scaled
parents hierarchy.
Here is how it works:
We render a high poly disc that we orient & scale towards the camera so that
it covers the same pixel of the sphere it's supposed to represent.
Then the pixel shader raytrace the sphere (effectively starting from
the poly disc depth) and outputs the depth to gl_FragDepth.
This approach has many benefit:
- high quality obviously: per pixel accurate depth!
- compatible with MSAA: since the sphere horizon is delimited by polygons,
we get the coverage computed by the rasterizer. However we still gets
aliasing if the sphere intersect directly other meshes.
- virtually no overdraw: there is no backface to shade but we still get
overdraw because by little triangle [gpus rasterize pixel by groups of 4].
- allows early depth test: since the poly disc is set at the nearest depth
we can output, we can use GL_ARB_conservative_depth to enable early depth
test and discard pixels that are already behind geometry.
- can draw outline pretty easily without geometry shader.
This manually resolve the content of a multisample FB to a single sample FB.
It resolves color (combine the 2 framebuffers in a logical maner keeping
depth buffer occlusion etc..) instead of a plain glBlitFramebuffer copy.
Group all fetches together without interleived alu to let compiler optimize.
Also do the color samples only if needed.
Went from 3.86ms to [1.11-2.22]ms [min-max] for the 16samples resolve pass
on my nvidia card.
This shader is used instead of blitting back and forth to a single sample
buffer.
This means it resolves the color and depth samples and outputs a fragment
which can be depth tested and blended on top of an existing framebuffer.
We do static shader variation with manual loop unrolling for performance
reason. In my test I get 25% more perf with intel integrated gpu and 75%
performance gain with dedicated nvidia card compared to a single shader
with a uniform for sample count.
This fix the issue with the zfighting we were getting at bones edges.
Moreover, this enables us to render arbitrarly large outline with
varying thickness.
It's usefull in some scenario to tweak the specular intensity of a light
without modifying the diffuse contribution.
Cycles allows it via lamps material which we currently not support in Eevee.
This is a good workaround for now.
The asserts were introduced on rB5f6c45498c92 (top-bar).
Although the asserts are technically correct, they would fail even in master.
And the commit simply added the asserts without fixing the situation itself
(as you can see in the report, it is really simple to reproduce this issue).
I propose we remove the asserts and bring them back only when the situation
itself is fixed. It doesn't make sense to introduce asserts that would fail
with the current state of the code.
This was a hard to reproduce bug, but it happens often enough.
Basically the tooltip of the active tool was been invoked when
context had no valid ScrArea which would lead to a crash.
We now just return no tooltip in these cases.
There are various values which depends on context in there, for example
current driver value and original DNA value f-curve is applied for.
This partially fixes issue with not being able to tweak keyed values
when material preview is open.
The material preview is not being currently updated against non-keyed
changes since every tweak of material property does full preview scene
depsgraph evaluation.
Now they are properly converted to Linear space before interpolation.
Since the only way to get vertex color in eevee and cycles is via the
attribute node with the CD_AUTO_FROM_NAME flag, we have to know at binding
time which type of buffer will be connected to this auto input.
We store this information inside the batch cache (together with the according
uniform name) and pass it as uniform to the shader which does conversion if
needed. The same shader can then be reused to draw another mesh with
different auto layers configuration.
This introduces `BKE_mesh_to_bmesh_ex()`, which exposes all of the
`BMeshFromMeshParams` parameters to the caller. This is required to enable
the `calc_face_normal` flag, which is required for the Bevel modifier.
This also introduces `BKE_bmesh_to_mesh()`, which allocates a new `Mesh`,
converts the `BMesh` to it, and returns it. The returned mesh is owned by
the caller.
- ability to switch between Single Color, Object Color, Material Color
and Random Color
- fixed Shading and Lighting popover
- Renamed Solid -> Single Color
Note: As already commented in the code, the ideal solution would
be to pass select mode as parameter to ED_view3D_backbuf_validate.
Without that we have to resort to the hack solution of writing to
evaluation data.
Does not yet support applying those operations, only detecting
insertions and generating matching rules was already rather complicated.
Also got rid of ;ost rna_path string allocation in collection diffing
code, could help a bit with speed too.
This matches the new convention for left-handed mode selectors,
however we're still undecided on exactly how this should work.
For now test this out as a convention for all space types.
The contents of the ModifierEvalContext struct are constant while iterating
over the modifier stack. The struct thus should be only created once, outside
any loop over the modifiers.
This commit introduces `EditMeshData`. The fields in this struct are
extracted from `EditDerivedBMesh` into their own struct `EditMeshData`,
which can then also be used by the `Mesh` struct. This allows passing
deformed vertices efficiently to the draw routines.
The modifier code constructs a new Mesh instead of writing to ob->data;
even when ob->data is a CoW copy, it can still be used by different
objects and thus shouldn't be modified by a modifier.
Makes the follow changes:
- Add new `deform*` and `apply*` function pointers to `ModifierTypeInfo` that take `Mesh`, and rename the old functions to indicate that they take `DerivedMesh`. These new functions are currently set to `NULL` for all modifiers.
- Add wrapper `modifier_deform*` and `modifier_apply*` functions in two variants: one that works with `Mesh` and the other which works with `DerivedMesh` that is named with `*_DM_depercated`. These functions check which type of data the modifier supports and converts if necessary
- Update the rest of Blender to be aware and make use of these new functions
The goal of these changes is to make it possible to port to using `Mesh` incrementally without ever needing to enter into a state where modifiers don't work. After everything has been ported over the old functions and wrappers could be removed.
Reviewers: campbellbarton, sergey, mont29
Subscribers: sybren
Tags: #bf_blender_2.8
Differential Revision: https://developer.blender.org/D3155
We use 'reference' to designate the linked ID which is being overridden
by the local one, so using 'reference' to designate the stored local ID
in apply RNA code was... not a good idea. ;)
Actions shouldn't be copying by new library management code.
Or at least should be possible to make it to not copy actions
or do unneeded user management.
This way we avoid modification of original data which could
and does cause threading conflict with copy-on-write which
could be happening for viewport.
Skip access to any evaluated data when operator is run on file load,
we don't have depsgraph evaluated yet. In this case we skip part of
sculpt session initialization, since it will be done during depsgraph
evaluation which happens after DEG_on_visible_update().
We can not skip sculpt session initialization since during normal
operation we want all the data to be initialized on mode change,
and not on initial brush stroke.
Since area lights are affected by scaling them, it only makes sense to support applying the scale to the lamp size.
Of course, applying location or rotation does not work.
If a scaling that changes the aspect ratio is applied to a square lamp, the mode is automatically changed to Rectangle.
This reduces the number of headers we need to import (and/or dependencies
on the order of them), as these depend on PBONE_VISIBLE(), which is already
in BKE_armature.h
Like the select/deselect operators, we only support doing this for the
active object for now. From the UI panels, it doesn't make sense to be
operating on bone groups which may not appear/line up with those in the panel
that we're currently dealing with.
This can be reviewed later, but for now, it's easier and makes more sense
this way.
This operator was only partially converted to multi-object editing,
as on one hand, it was using the new "objects in mode" iterator,
while on the other hand, it was also using the context iterator inside
that, making all selected bones across armatures get included.
We only want these to operate on the "active" armature only at a time
(where the "active" one is whichever the groups from the UI came from).
The fix therefore is to make it not use the context functions
(which were changed to always take bones from all selected armatures
instead).
Because:
- Less redundancy.
- Better suffixes.
Also a few modification to GPU_texture_create_* to simplify the API:
- make the format explicit to the texture creation process.
- remove the component count as it's specified in the GPUTextureFormat.
- Disable VertexPaint and WeightPaint for OB_MATERIAL and OB_RENDER. Users want to see the final result
- When in OB_SOLID, the active object should be rendered without any color. The lighting information is multiplied with the VertexPaint/WeightPaint color
- Removed the use_shading flag from VertexPaint and WeightPaint
- add method to check if render engine should draw without color (DRW_object_in_only_lighting_mode)
Reviewers: fclem
Tags: #code_quest
Maniphest Tasks: T54894
Differential Revision: https://developer.blender.org/D3191
A recent change in the UI layout code probably broke how the scale_x
for layouts was getting handled. This was leaving a large gap
(and causing layouts to pop) when trying to scrub the timeline.
This commit fixes this with a manually-found value that largely seems
to get rid of the popping problem. There's still a little jumping (1-2 px)
but it's less distracting now.
Otherwise, trying to change color/material properties of EEVEE materials
that have been keyframed doesn't work until a new keyframe has been inserted.
NOTE: There's still some wonkiness if you edit more than one keyframed value
before keyframing (i.e. the other unkeyed value temporarily gets reset). But
that's more of a general COW+animation issue.
Was causing temporal sampling artifacts when scene is set to Cycles
and viewport is set to Eevee. Visually was looking like ghosting or
motion blur when moving objects.
Was using classes to define tools, however this makes it awkward to
dynamically generate them (we can do it, but its not very "Pythonic").
Move to a named tuple.
This commit does two things:
- Adds an option to do the calculation in different color spaces (BT601
or BT709).
- Changes the default caluclation from legacy BT601 to BT709.
This affects several areas:
- UI areas (mainly scopes)
- ViewLevelsNode
- Several other nodes that use `COM_ConvertOperation.h`
Dynamically sized regions in the topbar were flickering due to only updating
their size after redraws. Now there is an optional layout() callback for
all regions in an area to do UI layout first, then refresh the region layout,
and then do the actual drawing for each region.
Task T54753
* This is just moving buttons to get a bit closer to the intended design,
better naming and layout is needed.
* Popovers currently work best when the 3D view header is at the top, with
the most important settings nearest to the mouse. Open design question is
if we should flip (part of) the buttons if header is at the bottom.
* Another question is if selecintg a shading mode enum should immediately
close the popover since those are changed often, unlike the other settings
for which it seems more convenient to keep the popover open.
This is to address things like a percentage slider with a fixed soft mininum.
For example, the render resolution ranges from 1% to 100% and it is really
strange to have the slider showing nothing filled when the ui shows 1%.
Now we always fill the slider with a vertical boundary. A bit hard to explain,
but very easy to see the difference.
I split the widget in three parts and used fragment shader discard to remove the
undesired bits. That means all the widget program is doing a bit extra
calculation.
Reviewers: fclem
Subscribers: billreynish
Differential Revision: https://developer.blender.org/D3186
U.ui_scale is the setting from the user preferences and should never be used
for drawing. UI_DPI_FAC is the final scale after DPI from the operating system
is taken into account.
Having single & multi-column checks inline were becoming unmanageable.
Use a generator to define the layout,
allows for easily adding different layouts in the future.
Not really happy with the fix, but it works. With the new window draw method
we are no longer storing the 3D viewport in 4 buffers, by having the GPU
viewport function directly as the 3rd buffer. This means we need to draw the
action zones into it, and so we need to keep the framebuffer bound a little
longer.
Note: Quad-View region overlay drawing is not working because of e01cadd657.
However I tested this patch after reverting that commit locally and this
patch's code seems fine.
Partial revert of 3d62230ed6.
The problem is that some local path is always returned by
bpy_resource_path. The function does not enforce a check for existence
of the path in the low level code.
Since client code may indeed not care about existence of the directory,
I leave bpy_resource_path alone and patch the python code to check for
existence. An extra argument that enforces the check might be a better
solution, but I will be leaving this to the UI team.
Originally it was nice to have a small list of definitions
with tools inline.
However we need to be able to define drawing functions for tools
which Python can't easily inline.
Use function for keymap definition,
support creating a function from a tuple as well
(handy for simple key-maps).
For Blender 2.8 we had to be compatible with very old OpenGL versions, and
triple buffer was designed to work without offscreen rendering, by copying
the the backbuffer to a texture right before swapping. This way we could
avoid redrawing unchanged regions by copying them from this texture on the
next redraws. Triple buffer used to suffer from poor performance and driver
bugs on specific cards, so alternative draw methods remained available.
Now that we require newer OpenGL, we can have just a single draw method
that draw each region into an offscreen buffer, and then draws those to
the screen. This has some advantages:
* Poor 3D view performance when using Region Overlap should be solved now,
since we can also cache overlapping regions in offscreen buffers.
* Page flip, anaglyph and interlace stereo drawing can be a little faster
by avoiding a copy to an intermediate texture.
* The new 3D view drawing already writes to an offscreen buffer, which we
can draw from directly instead of duplicating it to another buffer.
* Eventually we will be able to remove depth and stencil buffers from the
window and save memory, though at the moment there are still some tools
using it so it's not possible yet.
* This also fixes a bug with Eevee sampling not progressing with stereo
drawing in the 3D viewport.
Differential Revision: https://developer.blender.org/D3061
The issue was that every object tweak was doing a full copy
of original datablock onto evaluated version, and was updating
animation. This made it impossible to tweak properties which
has keyframes.
Proposed solution is to:
- Always apply animation on frame change, and when object
is explicitly tagged for animation update.
This will store original DNA value of animated property
in the f-curve.
- When applying animation in other cases, we check original
DNA value stored in f-curve with the actual original DNA
property. If they differ, it means user started to tweak
animated property, and we should skip animation.
If the value is the same, we apply animation.
This is just a first step towards proper final implementation,
but seems to be the direction we want to take.
Both the scene and workspace had an active view layer, and it was confusing
which settings were being used or displayed where. Now we always have one,
so there is no mismatch.
The "View Layers" tab in the properties editor is now "View Layer", no longer
showing a list of layers. Instead view layers can be added and removed with
the workspace view layer selector. They are also listed and selectable in the
outliner.
Single layer rendering uses the active view layer from the workspace.
This fixes bugs where the wrong active view layer was used, but more places
remain that are wrong and are now using the first view layer in the scene.
These are all marked with BKE_view_layer_context_active_PLACEHOLDER.
Use render settings and active view layer will be handled elsewhere.
Also change icon to not be confusing with render layers.
Probably we should get rid of the workspace tab entirely and do it in
the user preferences, but that's for later.
When creating an offscreen context we need wglCreatePbufferARB to create
a drawable. In non-background mode wglCreatePbufferARB would have been set
by the main window creation code. But in background mode this is not the
case so we need to create a dummy context using the screen window to init
wglew properly.
Active tool name needed to be unique from all other tools.
This caused problems since different modes have different tools with the
same name (Armature/Mesh click to extrude for eg).
Tool names now only need to be unique per mode.
Normal packing. The sign is stored in the A of the color buffer.
if the A == 1.0 the normal should be inverted. before use.
The reason is that packing has more precision for frontfaces, than for
backfaces
This allows for background rendering with EEVEE and other opengl render
engine.
I've only tested it on Linux for the moment so I can't say about other
platforms.
We do lazy init because we cannot assume we will need Ghost for rendering
before having parsed all arguments and we cannot know if a script will
trigger rendering. This is also because it currently does not work without
any display server (blender will crash).
Previously it could have happened on every request to evaluated scene
or view layer.
This commit also removes expansion of view layer and scene from iterator.
Iterator is not to be used before depsgraph is evaluated.
This isn't working well when zoomed out in the properties editor,
we should be taking into account properties editor zoom level.
This reverts commit 1ba91ae2c0.
This caused too much trouble, also it's possible users run with
'release' in their CWD causing issues.
Developers can symlink "release/" to "bin/2.79".
Technically this is the following operator:
bpy.ops.view3d.edit_mesh_extrude_move_normal
But this is a Python operator that in turns calls:
MESH_OT_extrude_region_move
Which in turns calls:
* MESH_OT_extrude_region
* TRANSFORM_OT_translate
Multi-View was never a per-viewlayer option. And now that we have viewlayer
it is better to name it in a more recognizable way:
"Stereoscopy" instad of plain "Views".
The Blender Internal removal (51b796ff15)
removed support for texture slots from Materials/Lamps/Worlds,
but didn't remove the relevant parts from the animation editor filtering
code that were responsible for adding the texture slots for those channels.
Added Object Overlap Overlay
- Added R32UI support to GPU_framebuffer
- Added R32U support to draw manager
- The overlay mode has a object data pass that will render 'needed' data to specific buffers so we can mix them together via a deferred rendering. In future will also add UV's and other data
- Overlap is implemented as an overlay so it could be used on top of the Scene lighted Solid mode (that will be rendered by Eevee.
Reviewers: fclem, brecht
Reviewed By: fclem
Subscribers: sergey
Tags: #code_quest
Maniphest Tasks: T54726
Differential Revision: https://developer.blender.org/D3174
There were two issues here actually:
* The hack to allow running Blender directly from the source directory
would just check for a 'release' directory, without actually ensuring it
is release dir from blender source tree, and not some other random
folder.
* GHOST_getSystemDir returns nothing for portable installations, now
we'll then check directly in the blender binary dir in that case.
This fix is more critical in 2.8 branch, where that system path is used
to retrieve new '3D' icons...
* Draw the frame/time number box over the scrollbar instead of above it,
to reduce the clutter/clashes with markers.
* Draw the box centered around the line instead of off to one side,
making it clearer that the frame shown is the one being affected.
* Make the box larger than the scrollbar + use white text to make it
stand out from the neighbouring frame numbers (otherwise, it's easy
to misread that it's just another one of those)
Based on discussions here in the office, this commit introduces a number of
changes to make the "Bind Camera to Markers" feature (introduced during Sintel
to facilitate camera switching, and now an integral part of the UI for doing this)
nicer to use.
Main changes:
* "Camera Markers" are now drawn using Camera icons, making it easier to
distinguish between them and other ("normal") markers
* "Camera Markers" will display the name of the bound marker, making it
easier to see what camera each marker uses. This will then also stay
in sync with the camera being used (though a manual refresh is needed
after renaming objects to make the timeline update), fixing the problem
where the marker's camera and the name get out of sync.
* Behaviour of Ctrl-B has been modified to make it easier to quickly add
these markers. Now, it will directly add a new marker on the current frame,
bound to the currently selected camera object. If there's an existing marker
on that frame, the existing marker's camera will be replaced instead of making
a new marker.
With the clear separation between data and a state we need to make sure
operators and other areas are readingstate from evaluated datablocks.
Code-wise it means that all evaluated values are to be read from dataglock
which is owned by dependency graph, using DEG_get_evaluated_id() or similar
helper.
Reviewers: brecht, mont29, campbellbarton, dfelinto
Differential Revision: https://developer.blender.org/D3036
Each parameter of the function is copied into the memory stack.
This also brought an improvement in peformance of snapping functions between 5% and 12% in my tests.
* "Scenes" now shows for each scene lists of all view layers, collections and
objects contained in it. This is the place to see all collections and objects
in the scene even if they are not used in any view layer. Objects are nested
according to parenting here.
* "Collections" now shows all collections in the view layer, and the objects
in those collections. Objects are not nested by parenting, only collections
since it would be too confusing if the children are in a different collection.
* "Groups" is unchanged.
* "View Layer" was removed, replaced by "Collections".
Part of T54790.
Works on every probe type.
The function to see is EEVEE_lightprobes_obj_visibility_cb.
Set pinfo->vis_data.cached to true makes the computation faster for multiple
views using the same group.
We could even sort the probes by group for that mater to speed things up
even more (only applies to dynamic probes like the planar reflections because
other probes are only rendered one at a time).
This add a callback function that runs after frustum culling test.
This callback returns the final visibility for this object.
Be aware that it's called for EVERY drawcalls that use this callback even
if their visibility has been cached.
Some widgets would have rounding relative to the button size, others
absolute. Now it's always absolute. Note changing Display Scale in the
user prefs is not zooming, the rounding still scales with that.
This is actually rather slow process, commeting it out gives us another
10% speedup... On a non-animated char! Don't even want to know how much
this would have costed on a rig with hundreds of fcurves!
And checking this is not really critical for us anyway, once animated
you do not really care whether props are also statically overridden or
not.
- Added UInt R support to framebuffers
- Added the overlap as an overlay so should be reusable by other engines
(Scene lighted Solid mode)
Differential Revision: https://developer.blender.org/D3175
Applied similar fix to T54233 to get the "Record with NLA" feature working with
active action blending + influence settings. Extrapolation is explicitly ignored
though, as it shouldn't be used with this feature (i.e. it is already disabled
with the new strips and also on the animdata by default)
While window manager is supposed to exist after file was fully read
and do-versioned, we can not rely on window manager to exist while
reading file and setting up an environment.
* Fixes crashes in versioning code for the removed editor types (see d8c719d8d8)
* Removes 'Game Logic' workspace.
It still only contains workspaces converted from old screen-layouts, proper
defaults are on the TODO.
Timelines and Logic Editors are gone. So far they were simply replaced by broken
Info Editors, now they are replaced by Dopesheets in the new Timeline mode.
We reuse ScrArea.butspacetype to temporarily store the space-type identifier of
the deprecated editor (see 9db492de6d). That way we can identify it in
versioning code and replace it nicely.
Action editor creation needs a scene to set the scrolling based on frame range.
Active screen-layouts use the active scene of the window they are displayed in.
Inactive screens simply use the first scene in the main data base.
Note that inactive editors don't need version patching, readfile.c converts them
to SPACE_EMPTY already, so users can't activate them.
Files saved since the editors were removed will still be broken.
Workspace config files saved before this will also crash (will update default
one in followup commit).
The only real reason we need `butspacetype` is while switching areas, where we
need to delay the actual switch to the RNA _update callback since only there we
can access context.
So instead of trying to sync it with `spacetype`, only set while needed and
unset it afterwards (as in set to `SPACE_EMPTY`).
This should also allow us to re-use `butspacetype` in versioning code when
trying to read removed editors. It'll store the space type value of the removed
editor which we can then use on versioning.
For backwards compatibility, we store `butspacetype` with the value of
`spacetype`.
This makes easier to distinguish between different editors
(as oppose to an editor and its regions).
Note action zones look a bit strange with this. I recommend we do next:
* Make sure all 4 corners can be used as action zones.
* Remove their drawing code (or show them only on mouse hover).
- UILayout.popover(.. panel_type ..)
A single panel
- UILayout.popover_group(.. panel categories ..)
Expands all panels matching args.
Currently used in the topbar for redo and paint options.
Requiring context means we can't easily create new editors to replace deprecated
ones in versioning code.
Think it's reasonable to give editors access to scene and area data for their
initial setup though. They mostly need it for setting "the view", as in,
scrolling values.
Also did minor cleanup in top-bar creation function.
Steps to recreate were:
* Create custom transform orientation.
* Change properties editor into 3D View.
* Trigger refresh of 3D view header by mouse hovering it.
Mistake in rB7d055da327b9555f.
Dithering the output color for 8bit precision framebuffer with bayer matrix.
On my tests the bayer matrux patterns are not noticeable at all.
Note that it also does that in opengl rendered mode which can be in a much
higher bitdepth. We can fix that if that's a problem in the future but I
doubt it will.
* Make bottom half of topbar a bit higher
* Make tabs higher and put them closer together
* Remove screen layouts dropdown, we'll have one layout per window
* Hide action zones from topbar
* Don't change topbar background color when activating
This will never get run, because direct_link_area() already flags/resets
every space type isn't registered, meaning that we don't have any opportunity
to apply our patching.
This commit removes all references to the old timeline editor.
Unfortuantely, the removal of the Timeline spacetype defining
functions has ended up breaking the version patching code I'd
been working on earlier (as now, the editor gets marked as
"unknown/info" before we get a chance to patch it!)
After a lot of failed attempts and head banging working trying to find a way to reuse
the standard editor-switching/creation code, I've just hacked in a temporary solution
here so that users can load old files and see the old timeline instances replaced
with Dopesheet-Timelines.
Note: This is not nice code, and copies a lot of the standard initialisation code,
but it works well enough for now. We can revisit this later when the other mode changes
come along.
These were getting set in the init() callback instead of the new(). As a result,
the settings would get reset everytime you resized the window/area - not quite
something you'd really want happening everything the size changes!
instead of using "black" curtains
With most editors now showing the start/end range by default, we need a way of
easily distinguishing when preview range is now enabled. By using a different color
(the exact color used is something we can change/adjust later), there is a more distinct
visual difference between them, making it easier to see what's happening.
This uses the global scene range, with styling matching the sequencer's start/end
frame drawing.
(The graph editor's "drivers" mode is exempt, as that doesn't really display time
in a linear way, so the start/end frames don't apply)
Eventually the idea is that they'll get remapped to some more global/generic hotkey
that can get used across all animation editors (see T54728). However, to facilitate
the removal of the timeline editor, it's better we do this now.
These were runtime only data, used in pre 2.8 Blender to make use of GL vertex arrays
to draw these more efficiently. Maybe we might restore these sometime as an optimisation
step, but for now, they're not needing and were confusing.
These now live in the action editor/dopesheet related files.
Apart from these, the timeline didn't actually have other settings
of its own that were of any interest to anyone.
For many years, animators have been requesting the ability to edit keyframes in the
timeline. However, implementing such tools in the timeline quickly becomes a slippery
slope, where we'll eventually end up having to duplicate all the functionality from the
dopesheet editor.
Discussing with William and Pablo this morning, we realised that perhaps it might be possible
to just make the Timeline a mode of the Dopesheet Editor (and kill off the old standalone
Timeline), meaning that we essentially get all the Dopesheet Editor goodness for free!
Also, with some proposed UI updates (i.e. allowing "submodes" of editors to be part of the
the main editors selector), it might not even matter that there isn't an "actual" timeline
editor anymore.
This commit implements the following changes (which are actually sufficient for supporting
most basic workflows):
* Timeline mode in Dopesheet Editor
* Tweaks to UI code to make the Timeline header/menus show up in Dopesheet editor
TODO:
* Hide channels list when switching to timeline mode
* Port over cache-file indicators
* Add missing timeline-only settings that need a new home in the dopesheet
* Go through fixing all timeline editor operators (e.g. Bind to camera)
* Port over start/end frame shading (and adjust preview range rendering to make the
distinction between these clear)
* Remove old timeline editor, and transfer over any leftover code
This "improve" the viewport experience by reducing the noise from random
sampling effects (SSAO, Contact Shadows, SSR) when moving the viewport or
during playback.
This does not do Anti Aliasing because this would conflict with the outline
pass. We could enable AA jittering in "only render" mode though.
There are many things to improve but this is a solid basis to build upon.
This pass create a velocity buffer which is basically a 2D motion vector
texture. This is not yet used for rendering but will be usefull for motion
blur and temporal reprojection.
== Main Features/Changes for Users
* Add horizontal bar at top of all non-temp windows, consisting out of two horizontal sub-bars.
* Upper sub-bar contains global menus (File, Render, etc.), tabs for workspaces and scene selector.
* Lower sub-bar contains object mode selector, screen-layout and render-layer selector. Later operator and/or tool settings will be placed here.
* Individual sections of the topbar are individually scrollable.
* Workspace tabs can be double- or ctrl-clicked for renaming and contain 'x' icon for deleting.
* Top-bar should scale nicely with DPI.
* The lower half of the top-bar can be hided by dragging the lower top-bar edge up. Better hiding options are planned (e.g. hide in fullscreen modes).
* Info editors at the top of the window and using the full window width with be replaced by the top-bar.
* In fullscreen modes, no more info editor is added on top, the top-bar replaces it.
== Technical Features/Changes
* Adds initial support for global areas
A global area is part of the window, not part of the regular screen-layout.
I've added a macro iterator to iterate over both, global and screen-layout level areas. When iterating over areas, from now on developers should always consider if they have to include global areas.
* Adds a TOPBAR editor type
The editor type is hidden in the UI editor type menu.
* Adds a variation of the ID template to display IDs as tab buttons (template_ID_tabs in BPY)
* Does various changes to RNA button creation code to improve their appearance in the horizontal top-bar.
* Adds support for dynamically sized regions. That is, regions that scale automatically to the layout bounds.
The code for this is currently a big hack (it's based on drawing the UI multiple times). This should definitely be improved.
* Adds a template for displaying operator properties optimized for the top-bar. This will probably change a lot still and is in fact disabled in code.
Since the final top-bar design depends a lot on other 2.8 designs (mainly tool-system and workspaces), we decided to not show the operator or tool settings in the top-bar for now. That means most of the lower sub-bar is empty for the time being.
NOTE: Top-bar or global area data is not written to files or SDNA. They are simply added to the window when opening Blender or reading a file. This allows us doing changes to the top-bar without having to care for compatibility.
== ToDo's
It's a bit hard to predict all the ToDo's here are the known main ones:
* Add options for the new active-tool system and for operator redo to the topbar.
* Automatically hide the top-bar in fullscreen modes.
* General visual polish.
* Top-bar drag & drop support (WIP in temp-tab_drag_drop).
* Improve dynamic regions (should also fix some layout glitches).
* Make internal terminology consistent.
* Enable topbar file writing once design is more advanced.
* Address TODO's and XXX's in code :)
Thanks @brecht for the review! And @sergey for the complaining ;)
Differential Revision: D2758
- need a draw mode in workbench engine.
- reorganized render engine retrieval in 3d view. There are 2 places
where this happenes 1. 3d view draw code and 2. draw manager.
the draw manager code is not used for external engines, currently added
an exception in for cycles. will need to have a better solution in
place.
issue was the factoring of the workspace engine that was removed. the logic implied that the clay could not be rendered as clay will be a draw mode we placed it already there so it is accessible in any engine. Should eventually fix the clay engine by migrating it to the workbench engine.
- Removed the depth pass as it will reuse the depth pass of the render
engine
- Used gl_FrontFacing to determine the facing
- Blender the result with the render engine result
use stack instead of always allocating memory for RNA paths of checked
properties! From average 167ms to 118ms here with Autumn rig... Still a
lot to improve, but that's already much better.
It is hidden behind the --debug-io flag for now.
Idea is to try to catch broken libraries state in current Main before we
actually write the file on disk, should help catching and understanding
what happens in Spring corruption cases.
Implemented the face orientation overlay for testing.
Overlay mode is only drawn when there are overlays to be rendered.
The overlay mode is rendered before the object mode.
While the feature is interesting, it's not much from what we can tell.
Retargeting is an important feature but needs
to fit in better with typical animation work-flows.
See: T52809
This is rather uncommon when operator will operate on a non-active view layer,
so there is no need to do full scene update.
This change solves lag first time using Extrude operator in edit mode.
This is still broken I cant tell if it is the fact that the in_band
funtion does not work properally or an issue in the box algorithm, or
both.
It seems like the calculation of the size of the box while roatated
needs to be fixed also.
@campbellbarton: This operator works (as intended) on an object level, wich means that it won't remove doubles for vertices that are close to each other but contained in different objects - is that really helpful?
Brecht authored this commit, but he gave me the honours to actually
do it. Here it goes; Blender Internal. Bye bye, you did great!
* Point density, voxel data, ocean, environment map textures were removed,
as these only worked within BI rendering. Note that the ocean modifier
and the Cycles point density shader node continue to work.
* Dynamic paint using material shading was removed, as this only worked
with BI. If we ever wanted to support this again probably it should go
through the baking API.
* GPU shader export through the Python API was removed. This only worked
for the old BI GLSL shaders, which no longer exists. Doing something
similar for Eevee would be significantly more complicated because it
uses a lot of multiplass rendering and logic outside the shader, it's
probably impractical.
* Collada material import / export code is mostly gone, as it only worked
for BI materials. We need to add Cycles / Eevee material support at some
point.
* The mesh noise operator was removed since it only worked with BI
material texture slots. A displacement modifier can be used instead.
* The delete texture paint slot operator was removed since it only worked
for BI material texture slots. Could be added back with node support.
* Not all legacy viewport features are supported in the new viewport, but
their code was removed. If we need to bring anything back we can look at
older git revisions.
* There is some legacy viewport code that I could not remove yet, and some
that I probably missed.
* Shader node execution code was left mostly intact, even though it is not
used anywhere now. We may eventually use this to replace the texture
nodes with Cycles / Eevee shader nodes.
* The Cycles Bake panel now includes settings for baking multires normal
and displacement maps. The underlying code needs to be merged properly,
and we plan to add back support for multires AO baking and add support
to Cycles baking for features like vertex color, displacement, and other
missing baking features.
* This commit removes DNA and the Python API for BI material, lamp, world
and scene settings. This breaks a lot of addons.
* There is more DNA that can be removed or renamed, where Cycles or Eevee
are reusing some old BI properties but the names are not really correct
anymore.
* Texture slots for materials, lamps and world were removed. They remain
for brushes, particles and freestyle linestyles.
* 'BLENDER_RENDER' remains in the COMPAT_ENGINES of UI panels. Cycles and
other renderers use this to find all panels to show, minus a few panels
that they have their own replacement for.
Will be part of the collection manager where per collection the
ob->col can be set. This currently depends on DepsGraph +
CollectionManager.
I removed it for now so the code won't influence development
Was caused by ec0756af6c, once again, we can't pass view layer,
need to pass index.
The sad part is that currently we don't have quick way to look up
view layer by index. Can do similar thing as we do for bones and
bases.
The work is mainly from Lukas Toenne, with some modifications from myself.
Includes following obvious changes:
- Particle system selection is now name-based, with lookup menu.
- Lots of new options to control varieties.
Changes comparing to the Gooseberry branch:
- Default values and versioning code ensures same behavior as the
old modifier.
- Custom data layers are coming from vertex color, the modifier
does not create arbitrary layers now. The hope is to keep data
more manageable, and maybe make it easier to select in the shader
later on.
This means, values are quantized to 256 values, but it should be
enough to get varieties in practice.
Reviewers: brecht, campbellbarton
Reviewed By: brecht
Subscribers: eyecandy
Differential Revision: https://developer.blender.org/D3157
- added `object_color_type` where the user can set if the collection
determines the color, or the object will be used for the color.
Implemented it as an enum as later this can have a random color option.
- moved OB_LIGHTING_* to DNA_view3d_types and renamed it.
- Fixed some DRY in workbench_materials.c. Can remove more DRY's but
will need to discuss the responsibility of the workbench engine as it
might become part of the eevee renderer.
ViewRender was removed, which means we can't get the render engine for files
saved in 2.8. We assume that any files saved in 2.8 were intended to use Eevee
and set the engine to that.
A fix included with this is that .blend thumbails now draw with Clay mode,
and never Eevee or Cycles. These were drawn with solid mode in 2.7, and should
be very fast and not e.g. load heavy image textures.
Differential Revision: https://developer.blender.org/D3156
Assume files saved in 2.8 were intended for Eevee and set them to material
viewport shading. In Eevee this is equal to rendered draw mode, in Cycles
this will draw with Eevee. This way Eevee demo files still show something
interesting when opened.
Similar materials will reuse the same shadergroup. Currently using
a custom hash function that might select too similar colors into the
same material.
Reintroduced the workbench_materials.c this file will be responsible for
material lookup/creation and shader compilation
Fixed a GPUShader mem leak
This statement is only relevant in 2.8, but causes confusion in master.
I kept the 'default' label to prevent compiler warnings about unhandled
cases. The break is needed because there should be at least one statement
after 'default'.
I.E. only enable auto-override for 'active' selected object when making
an override of a linked group. This will ease on auto-override creation,
and you typically do not want to auto-override most objects in the group
anyway (in proxy system, you could only proxyfy one object of the group
anyaway!).
This is by default. We can still enable the thicker outlines for high dpi
screens or personnal preference but it's not used atm. This also improve
the performance removing 1/3 of the outline cost.
This finally allows us to use Random factor to add variations to the
interpolated children. This feature never worked since 2007L there was
random factor slider in the interface, but it was only used by simple
children. Now it has affect on interpolated children as well.
Technically, this will break compatibility if older file had random
factor set to something else than 0 (default value is 0 though). But
we are leaving 2.7 series, so can accept such breackage in the name
of supported features.
In more complex models the object color uniform data is freed before rendered.
We should copy it to local data. but for now we redirected it to a
constant.
Currently uses static lighting. Will become HDRI lighting.
Added do_versions to set default drawtype_solid and drawtype_texture to
OB_LIGHTING_STUDIO. When View3D space is created drawtype_solid and
drawtype_texture are also set to OB_LIGHTING_STUDIO.
Current studio lighting uses a dot product to simulate static lighting.
Will need to be changed in the future with different lighting models.
Tab after C++ comment broke parsing and didn't remove the line at all.
Was there since 2002 at least, probably confused some peeps.
This means commented out code was actually written to SDNA.
- added the drawtype_solid, drawtype_wireframe, drawtype_texture to
View3D
- enabled workbench panels for important render engines
- merged workbench_materials to solid_flat_mode. All draw modes will get
its own fast implementation in the workbench
Folders removed entirely:
* //extern/recastnavigation
* //intern/decklink
* //intern/moto
* //source/blender/editors/space_logic
* //source/blenderplayer
* //source/gameengine
This includes DNA data and any reference to the BGE code in Blender itself.
We are bumping the subversion.
Pending tasks:
* Tile/clamp code in image editor draw code.
* Viewport drawing code (so much of this will go away because of BI removal
that we can wait until then to remove this.
This is a way to deal with animated properties in evaluated version
off datablock. Previously, running blender with copy-on-write enabled
will show original values. Now we can see proper properties, while
typing values in still goes to the original datablock.
Thanks Brecht for the review!
Was caused by two things from the past:
- Tagging would set id->recalc to COW update flag.
This one is to be ignored.
- Particle tagging will use psys recalc flags on id->recalc,
but we only need to use flags from particles. Otherwise
it will be some collisions in bit masks.
Why exactly this happens remains unclear, found that in the
autumn.blenrig file of Spring production while working on static
overrides... Tons of ugly IDProps in that rig. xD
This is just to have hair rendering and editing mostly working as in
master. A better fix is probably needed, there seems to be some
missing depsgraph relations for particle edit settings, and particle
edit code doesn't rebuild caches after applying edits. But at least
you can see and interact with hair now until those things can be
sorted out.
The depsgraph was always created within a fixed evaluation context. Passing
both risks the depsgraph and evaluation context not matching, and it
complicates the Python API where we'd have to expose both which is not so
easy to understand.
This also removes the global evaluation context in main, which assumed there
to be a single active scene and view layer.
Differential Revision: https://developer.blender.org/D3152
This changes quite a few things.
- Outline is now per object.
- No more outline at object intersection (fix hairs problem).
- Simplify the code quite a bit.
We use a R16UI buffer to save one id per object outline. We convert this id
to color when detecting the outline.
Added textureGatherOffsets to the code but could not test on current hardware
so leaving it commented for now.
This uniforms can be used to have a unique id for each drawcall of a shgrp.
This only works for standard shgroups and is an exception for the outline
drawing.
This adds initial multi-object editing support.
- Selected objects are used when entering edit & pose modes.
- Selection & tools work on all objects however many tools need porting
See: T54641 for remaining tasks.
Indentation will be done separately.
See patch: D3101
Blender allows this.
The Cube in the file in the report would always disappear with the non camera view.
The clip_end was too small.
The correction here is only on the assert.
Note: tried with a complex production file characters, this is currently
totally non-functionnal (crashing even in a infinite recursion loop),
IDProps and override need some love :(
Although somewhat less micro efficient, I decided to separate the `viewinv` matrix to calculate the world position separately.
This makes it easier to understand the code.
While 3d cursor is mainly an UI thing and isn't needed for scene evaluation,
it is stored in scene DNA. This means, operator is inform depsgraph that data
has changed, so all copies of that scene can copy new values.
Fixes lack of visual feedback when changing cursor location in viewport
with copy-on-write enabled.
Added a lock-free deferred queue for deletion. Now if ID icon
is requested to be freed from non-main thread, it will be added
to the deferred list. Actual deletion will happen later from main
thread.
Currently actual deletion only happens next time BKE_icon_id_delete()
is called, which might not be enough. But it's easy to enforce
deferred deletion.
Icons for preview images are not covered by deferred deletion yet.
Reviewers: mont29
Differential Revision: https://developer.blender.org/D3146
The tests as they are now make string comparisons. This only works
on Windows because the reference files look different for different
operating systems because of different number formatting.
The collada tests need a complete rework (wip)
This gets rid of the need of a geom shader and instancing.
Both are pretty slow compared to the new method.
The only moment the old method could be better is when scene is filled
with lots of objects and most of the objects in the shadow map appear
on every layer.
But even then, we could optimize the culling and minimize the overhead.
E.g. the vertices created for each of the defines would require a
certain offset. If you don't know what to look for, finding out about
this is pretty difficult. Make them easily searchable instead.
There is quite some mess going on in that most of the old triangle
drawing code is still there, but does almost nothing effectively.
Instead values are hardcoded in the shader, however it doesn't support
the drawing options the triangle functions expose.
E.g. the 'where' variable to set triangle direction doesn't work.
- Wasn't clear which functions handle edit-bones.
- Mixed both ebone and edit_bone in names.
- Didn't use ED_armature_* prefix for public API.
See P655 to apply to branches.
- Move static undo variable into 'WriteData',
'memfile_chunk_add' used arguments in a confusing way,
sometimes to set/clear static var.
- Replace checks for 'wd->current' with 'wd->use_memfile'
move memfile vars into 'wd->mem' struct.
For correct results these must have been set already when the depsgraph was
created and evaluated, so all dependencies have appropriate resolutions too.
For particle we no longer backup and restore the viewport particles to avoid
overwriting them during render, as copy-on-write solves this for us. Even
without COW particles seem to work ok.
This also removes the particle simplification options based on camera. This
was never used much and only available in Blender Internal.
Differential Revision: https://developer.blender.org/D3148
This was only used for viewport rendering, where we can just pass the engine
type directly. There is no technical reason why we can't draw the same depsgrpah
with different render engines.
It also led to some weird things like requiring a render engine for snapping
and raycast API functions.
Differential Revision: https://developer.blender.org/D3145
Scene, view layer and mode are now set in the constructor and never changed.
Time is updated on frame changes to indicate which frame is being or has been
evaluated last.
This is a step towards making EvaluationContext obsolete.
Differential Revision: https://developer.blender.org/D3144
Cycles is no longer using this. There are still addons using it but for
correct results with the new depsgraph this API should not be used.
Differential Revision: https://developer.blender.org/D3143
Having that one when opening a file or loading some lib makes absolutely
no sense, and switching that 'temp' editor to some other type can
trigger all kind of funny bugs...
Note that using the shortcuts keys (Shift-F5 etc.) is still possible,
removing those seems a bit more involved. :/
We now only look into dupli groups to find point caches to edit. This
feature is a leftover from the old proxy system, and evaluating the
full dupli list and all transforms was overkill. With static overrides
we may want to get rid of using duplis entirely, and just let users
select the objects directly.
I can see how it's slowing things down: glFinish make sure that every query
are finished but the first query may have been finished a long time ago.
This might create bubbles because of the PIL_sleep_ms.
It's not just the Graph Editor that needed this - the NLA also uses similar code
and thus suffers from a similar problem.
(My first commit from the Blender Institute v2.0 - Just testing that everything works)
Unless there is an external action from an user, there should not
be need in re-copying original datablock to a copied one.
This brings performance up from 5fps to 11fps with Spring runcycle
(performance in master is 14fps).
Quite straightforward implementation, allows us to remove all the cherry-picking
update of specified scene/view layer/collection fields. Makes it possible to use
generic function to update scene.
The tricky part is that we need to know view layer pointer before the whole
evaluation starts. So we actually expand scene at initialization of evaluation.
context. This is still a bit of an exceptional case, but at least we still avoid
dangerous cherry-picking update.
For the performance we convert object bases list to an array
during view layer evaluation. This makes it possible to have
very cheap index-based base lookup.
The goal of this change is to get rid of base used for function
binding, and avoid scene datablock expansion at the depsgraph
construction time.
Use single function to evaluate all the collections for the given view layer.
This way we avoid need to get scene ID sub-data. Similar to pchan index, this
allows us to avoid build-time scene expansion, which also simplifies update of
the scene datablock.
Well, sort of. There is still work to be done to get rid of build-time scene
datablock expansion, which includes:
- Need to pass view layer by index.
Annoying part would be to get actual view layer for that index. In practice
doing list lookup might not be such a bad idea, since such lookup will not
happen very often, and it is unlikely to have more than handful of view
layer anyway.
Other idea could be to use view layer from evaluation context.
Or maybe from depsgraph, which is supposed to be in the context. Can have
some assert statements to make sure everything is good.
- Need to get id of base binding for flags flush.
We can replace that with index-based lookup from an array created by view
layer evaluation.
Reviewers: dfelinto
Differential Revision: https://developer.blender.org/D3141
In preparation of the removal of blender internal render we
moved the vectorblur code that was placed in the render package
(legacy) to the compositor. The compositor is only using this
code even the blender internal renderer did not use the code at
all.
Got lost in big undo refactor.
Note that this is probably (maybe) not how we want to have it in the
end, things like EditMode undo should probably not trigger this check?
The previous assert assumed '..' is always there, which isn't necessarily
true (for example when in the root of an Asset Engine repository).
The new code asserts that if '..' is present it should be the first entry
(rather than forcing the first entry to be '..').
the Vector Transform node was added to the "Vector" category in
nodeitems_builtins.py
but was using the "NODE_CLASS_CONVERTOR" internally (thus using e.g. the
'wrong' theme color)
thanx @dingto for review
Differential Revision: https://developer.blender.org/D3138
Only recreate ogl context if we cannot reuse the one of the previous thread.
Adding lots of shaders were recreating as many ogl context which was very
slow.
This happened when creating a window with the cursor over the timeline area.
I still don't know exactly what happened but for a reason batches were not
reset in this case.
This means fewer indices to store. That being said, it seems to be a little
slower because of the restart index. But that's in the case we would be
vertex bound, which is mostly never going to happen.
Fixes the "emtpy scrolling" glitch by clamping the scroller offset to
the boundary of the view when it's smaller than the previous.
Fixes T45197. Patch by @januz.
Differential Revision: D1580
The new constraint is slower and not backward compatible, but should
be better, especially in the damping side. The new constraint also
has a different valid range of the damping coefficient, and a limit
implementation that bounces instead of making the object stationary.
Reviewers: sergof
Differential Revision: https://developer.blender.org/D3125
WEBM is the codec name, and VP9 is the encoder (the older encoder "VP8"
is less efficient than VP9).
WEBM/VP9 and h.264 both have options to control the file size versus
compression time (e.g. fast but big, or slow and small, for the same
output quality). Since WEBM/VP9 only has three choices, I've chosen to
map those to 3 of the 9 possible choices of h.264:
- BEST → SLOWER
- GOOD → MEDIUM
- REALTIME → SUPERFAST
The VERYSLOW and ULTRAFAST options give very little extra benefit.
Reviewed by: @Severin
Some of the code is simpler because we use Blender's triangulation directly
instead of dealing with quads. Also some progress printing code was removed
because the depsgraph can not tell us the number of objects ahead of time.
Differential Revision: https://developer.blender.org/D3127
- Disable scissor test for fast clear. This could lead to some issues but
I cannot think of one and could not find one either.
- Manually wait for queries to be available instead of making the driver
wait and issue warnings.
The encoding panel mentions "None" in a few places, which is confusing.
- "Codec: None" now reads "No Video"
- "Audio Codec: None" now reads "No Audio"
- "Output Quality: None; ..." now reads "Constant Bitrate"
When selecting "No Video" the remaining video encoding options are
hidden, making it even more explicit that there will not be video in the
output file.
The label "Codec" now reads "Video Codec" for symmetry with "Audio
Codec".
Previous code was assuming that the glyph texture would remain bound to
GL_TEXTURE0 until the cache would be drawn. This is not always the case,
so better save the texture and rebind it before drawing.
This port the Blurring of blf fonts to the final drawing shader.
We add a bit of extra padding to each glyph so that jittering the texture
coord does not sample the neighbor glyphs.
Overall 10% more performance on general UI drawing time.
This commit can introduce ordering problem on some elements.
In this case you need to flush the widget cache to ensure the element that
is going to be drawn is drawn on top of any widget base.
To flush the cache use UI_widgetbase_draw_cache_flush.
This is already done for BLF and Icons.
When importing multiple materials for one object,
the imported material animation curves have all been
assigned to the first material in the object.
This fix also improves the console logging whenever the importer
finds a consistency problem with the imported animation data.
Replace the 12 iterations of UI_draw_roundbox_4fv with only one batch.
This mean less overdraw and less drawcalls.
I had to hack the opacity falloff curve manually to get approximatly the
same result as previous technique. I'm sure with a bit more brain power
somebody could find the perfect function.
The MovieSequence and MovieClip classes now have a metadata() function
that exposes the `IDProperty *` holding the video metadata.
Part of: https://developer.blender.org/D2273
Reviewed by: @campbellbarton
This is useful to create a mapping from the frame range in the video to
frame index in the blend file.
Part of: https://developer.blender.org/D2273
Reviewed by: @campbellbarton
This is currently only supported by FFmpeg (so not frameserver, AVI RAW,
or AVI JPEG), and only seems to work when using Matroska or Ogg Theora
containers.
Only metadata that doesn't change from frame to frame is written to
video files. This distinction is visible in the UI by looking at the
stamp checkbox tooltips (they either mention "image" or "image/video").
Part of: https://developer.blender.org/D2273
Reviewed by: @campbellbarton
- Metadata handling is now separate from `ImBuf *`, allowing it to be
used with a generic `IDProperty *`.
- Merged `IMB_metadata_add_field()` and `IMB_metadata_change_field()`
into a more robust `IMB_metadata_set_field()`. This new function
doesn't return any status (it now always succeeds, and the previously
existing return value was never checked anyway).
- Removed `IMB_metadata_del_field()` as it was never actually used
anywhere.
- Use `IMB_metadata_ensure()` instead of having
`IMB_metadata_set_field()` create the containing `IDProperty` for
you.
- Deduplicated function declarations, moved `intern/IMB_metadata.h` out
of `intern/`. Note that this does mean that we have some extra
`#include "IMB_metadata.h"` lines now, as the metadata functions are
no longer declared in `IMB_imbuf.h`.
- Deduplicated function declarations, all metadata-related declarations
are now in imbuf/IMB_metadata.h.
Part of: https://developer.blender.org/D2273
Reviewed by: @campbellbarton
Use the new GPU_SHADER_2D_NODELINK and GPU_SHADER_2D_NODELINK_INST to
accelerate nodelink drawing.
This commit does not include the batching functionnality. So this should
not make a lot of difference.
Now use a list of preset batches with a function to add new ones to this
list.
This removes the need of new functions all over the place to reset/exit.
Special shader to draw nodelinks for the node editor.
We only pass bezier points to the GPU and vertex position is handled inside
the vertex shader.
The arrow is also part of the batch to avoid separate drawcalls for it.
We still draw 2 pass one for shadow and one for the link color on top.
One variation to draw instances of theses links so that we only do one
drawcall.
The issue was caused by component tag forcing CoW component to be run,
without actually flushing changes down the road from the CoW operation.
In a way, this is what we want: we do want CoW to run on changes, but
we don't want tiny change forcing full datablock update.
This commit makes it so order of updates is all correct, but the bigger
issue is still open: what parts of datablock CoW should be updating?
Now it's possible to open spring file and play around.
The issue was mainly visible when copy-on-write was enabled. This was forcing
lots of meshes to be freed from multiple thread, causing all sorts of race
conditions in Gawain's VAO code.
OpenGL resources seems already to be doing deferred deletion, need to do the
same for CPU side arrays.
Free code should not handle ID refcounting at all. This has to be done
at higher level, since in some case we want to free (temp) data that
actually did not refcount at all its IDs.
This change seems to be working OK, but as usual in that area, only
lots of testing in real-case situation will say whether there are some
hidden bugs or not.
Issue was, *some* IDs (like infamous nodetrees from materials etc.)
would not go through the 'main' read_libblock() func, so their tags were
never reset.
So now, we ensure direct_link_id() always clear the tags, and move
setting them in read_libblock() after the call to direct_link_id().
Needed for depsgraph, but general healthier fix actually.
The problem was that textures were assigned to different slots on different draw
calls, which caused shader specialization/patching by the driver. So the shader
would be compiled over and over until all possible assignments were used.
This is a part of copy-on-write sanitization, to avoid all the checks
which were attempting to keep sub-data pointers intact.
Point is: ID pointers never change for CoW datablocks, but nested
data pointers might change when updating existing copy.
Solution: Only bind ID data pointers and index of sub-data.
This will make CoW datablock 7update function was easier in 2.8.
In master we were only using pose channel pointers in callbacks,
this is exactly what this commit addresses. A linear lookup array
is created on pose evaluation init and is thrown away afterwards.
One thing we might consider doing is to keep indexed array of
poses, similar to chanhash.
Reviewers: campbellbarton
Reviewed By: campbellbarton
Subscribers: dfelinto
Differential Revision: https://developer.blender.org/D3124
Back in the days (2.4x and before), it was rather easy to get some
invalid utf-8 strings in Blender. This is totally breaking modern code,
so this commit adds a simple 'check & fix strings' operator, available
from the main File menu.
E.g. typing `bpy.data.bl_rna.properties[8].<tab>` in console would hard-crash
trying to dereference NULL pointer. Was a missing check in rna_Property_tags_itemf().
This is really minor but anyways, now it will only leak if you cancel the menu.
And that only if htis is the last time you called this operator before closing
Blender.
Better fix for T54457. It seems Debian compiles OpenVDB without ABI 3
compatibility, while Arch does enable it as is the default in the OpeVDB
CMake build system.
So now there's an option that the distribution can set depending on how
they compile their OpenVDB package.
Some functions always returned the input argument
which was never used.
This made code read as if there might be a leak.
Now return a boolean (true the imbuf is modified).
- Undo that changes modes currently asserts,
since undo is now screen data.
Most likely we will change how object mode and workspaces work
since it's not practical/maintainable at the moment.
- Removed view_layer from particle settings
(wasn't needed and complicated undo).
- Use a single undo history for all operations.
- UndoType's are registered and poll the context to check if they
should be used when performing an undo push.
- Mode switching is used to ensure the state is correct before
undo data is restored.
- Some undo types accumulate changes (image & text editing)
others store the state multiple times (with de-duplication).
This is supported by checking UndoStack.mode `ACCUMULATE` / `STORE`.
- Each undo step stores ID datablocks they use with utilities to help
manage restoring correct ID's.
Needed since global undo is now mixed with other modes undo.
- Currently performs each undo step when going up/down history
Previously this wasn't done, making history fail in some cases.
This can be optimized to skip some combinations of undo steps.
grease-pencil is an exception which has not been updated
since it integrates undo into the draw-session.
See D3113
For this we use a new shader that gets it's data from a uniform array.
Vertex shader position the vertices using these data.
Using glUniform is way faster than using imm for that matter.
Like BLF rendering, UI icons are always (as far as I know) non occluded and
displayed above everything else. They also does not overlap with texts so
they can be batched at the same time.
This adds less than a megabyte of mem usage.
FT_Get_Kerning was the 2nd hotspot when profilling. This commit completly
remove this cost.
One concern though: I don't know if the kerning data is constant for every
sizes but it seems to be the case. I tested different fonts at different
dpi scalling and saw no differences.
I left a flag to quickly debug if something is wrong.
But now that everything uses shader, it seems to be alright since a shader
is always set active before drawing.
I've made a separate version of the geom shader that works with full
3D modelviewmat.
This commit also includes some fixup inside blf_batching_start().
You can now use BLF_batching_start and BLF_batching_end to batch every
drawcall to BLF together minimizing the overhead introduced by BLF and the
opengl driver.
These calls cannot be nested (for now).
If the modelview matrix changes, previously batched calls are issued and a
the process resume with the new matrix.
However the projection matrix MUST not change and gl scissors as well.
This is not a perfect win just yet. It's now calling glBufferSubData for
every call (instead of using glMapBufferRange which is almost faster), but
with this system we will be able to batch drawcalls together.
See next commit.
This allows us to specify a the number of vertices to upload to the gpu.
This is to keep the same allocation on the System Memory but send the least
amount of data to the GPU/Driver.
- See `--log` help message for usage.
- Supports enabling categories.
- Color severity.
- Optionally logs to a file.
- Currently use to replace printf calls in wm module.
See D3120 for details.
Two issues are fixed with this commit:
1) When we build OIIO (on unixoid build environments) and no /src/doc/oiiotool was present we had no build target for it (which led to a build error). As we don't need docs for OIIO, we disable it now.
2) We specified a var that OIIO does not recognize (was removed upstream a long time ago): ILMBASE_VERSION.
Selecting not only the objetcs directly linked to the selected collection.
So we also do it for the objetcs in the nested collections, just as we can
do from the outliner.
Use Shift+G > Collection. If there is only one collection, it just selects it,
if there are multiple ones user get to pick which one to select.
This is the same behaviour we have for groups. Note, we only select objects
directly in the collection, not the ones in any nested collection.
Feature suggested by Pablo Vazquez (venomgfx)
This means smaller imm buffer usage.
This does not reduce the number of drawcalls.
This uses geometry shader which is slow for the GPU but given we are really
CPU bound on this case, it should not matter.
A perfect implementation would:
- Set the glyph coord in a bufferTexture and just send the glyph ID to the
GPU to read the bufferTexture.
- Use GWN_draw_primitive and draw 2*strllen triangle and just retrieve the
glyph ID and color based on gl_VertexID / 6.
- Stream fixed size buffer that the Driver can discard quickly but this is
the same as improving IMM directly.
This update adds a link to the Wikipedia article "Euler angles" to the description of the mathutils.Euler class.
I initially was not sure what a "Euler" represented in Blender API, but found the Wikipedia article helpful. I believe others will find the link helpful too if it appears in the class documentation.
This is similar to the Wikipedia links that appear in the mathutils.Matrix class, e.g: https://docs.blender.org/api/blender_python_api_current/mathutils.html?highlight=euler#mathutils.Matrix.adjugate
Author: @justasb
Reviewers: campbellbarton, trumanblending, Blendify
Reviewed By: Blendify
Subscribers: Blendify
Tags: #bf_blender
Differential Revision: https://developer.blender.org/D3077
Fix for T54437: Sequencer preview uses last updated scene
The fix started in master, moving EvaluationContext initialization
before we leave `deg_evaluate_on_refresh()`.
Upon merging master we can fix the actual issue which was to set
the EvaluationContext depsgraph even if the depsgraph was already updated.
This is required to T54437 (sequencer preview uses last updated scene).
Although the fix itself needs to be in 2.8, for the 2.8 specific
initialization code.
Introduce a UI batch cache. For the moment it's only used by widgetbase so
leaving it interface_widgets.c. If it grows, it can have its own file.
Like all preset batches (batches used by UI context), vaos must be refreshed
each time a new window context is binded.
This still does 3 GWN_batch_draw in the worst cases but at least it does
not use the IMM api.
I will continue and batch the 3 calls together since we are really CPU
bound, so shader complexity does not really matters.
I cannot spot any difference on all the widgets I could test. I did not use
any unit tests so I cannot tell if there is really any defects.
This is not a complete rewrite but it adresses the top bottleneck found
after a profilling session.
Use more generic id->recalc flag.
Also sanitize flag flush from settings to particle system.
Need to do such flush before triggering point cache reset, since
point cache reset will do some logic based on what flags are set.
This will solve crash caused by threaded update which will set
some bitflags while point cache reset is in progress.
The way how particle state is to be accessed or used did not change
in Blender 2.8, so the drawing code should follow old design.
This code is somewhat duplicated from drawobject.c, but old draw
code is on the way to be removed anyway.
This fixes issue with disappearing particles when tweaking number
of particles.
Tagging based on components might not be granular enough.
For example, for particles we would want to know what part
of particles was changed exactly. For the flushing we wouldn't
worry too much, because we will want less granular updates
there anyway.
Roughness baking previously defaulted to 1.0 for all diffuse materials,
now we also bake roughness values of Oren-Nayer and Principled Diffuse.
Differential Revision: https://developer.blender.org/D3115
How to use: Select a few objects, and press "M" in the viewport.
If you hold ctrl the objects will be added to the selected collection.
Otherwise they are removed from all their original collections and moved
to the selected one instead.
Development Notes
=================
The ideal solution would be to implement an elegant generic multi-level
menu system similar to toolbox_generic() in 2.49.
Instead I used `uiItemMenuF` to acchieve the required nesting of the menus.
The downside is that `uiItemMenuF` requires the data its callback uses to be
always valid until the menu is discarded. But since there is no callback we
can call when the menu is discarded for operators that exited with
`OPERATOR_INTERFACE`.
That means we are using static allocated data, that is only freed next time
the operator is called. Which also means there will always be some
memory leakage.
Reviewers: campbellbarton
Differential Revision: https://developer.blender.org/D3117
When importing an xfov curve, we must transformed the data to
Lens opening angles in degrees. While the curve value itself is
correctly transformed, the transformation of the tangents has been
forgotten. this is fixed now.
Use C++11 threads when available, and native critical section on Windows.
Later on we can remove pthread code when C+11 becomes required.
Differential Revision: https://developer.blender.org/D3116
Drawcall per window redraw on default layout:
- 4100+ without patch
- 1270 with patch
Theses drawcalls meant a lot of driver overhead since they each correspond
to one glMapBuffer which is slow.
- Get memory usage from MemFile instead of MEM API
avoids possible invalid when threads alloc memory.
- Use size_t instead of uint and uintptr_t to store size.
- Rename UndoElem.str -> filename
- Rename MemFileChunk.ident -> is_identical
Random numbers for step offset were correlated, now use stratified samples
which reduces noise as well for some types of volumes, mainly procedural
ones where the step size is bigger than the volume features.
Remain convinced that it should not be possible for undo code to run in
parallel with DEG eval... But for now, this whould prevent static
override code to dive into this collection.
This refactor modernise the use of framebuffers.
It also touches a lot of files so breaking down changes we have:
- GPUTexture: Allow textures to be attached to more than one GPUFrameBuffer.
This allows to create and configure more FBO without the need to attach
and detach texture at drawing time.
- GPUFrameBuffer: The wrapper starts to mimic opengl a bit closer. This
allows to configure the framebuffer inside a context other than the one
that will be rendering the framebuffer. We do the actual configuration
when binding the FBO. We also Keep track of config validity and save
drawbuffers state in the FBO. We remove the different bind/unbind
functions. These make little sense now that we have separate contexts.
- DRWFrameBuffer: We replace DRW_framebuffer functions by GPU_framebuffer
ones to avoid another layer of abstraction. We move the DRW convenience
functions to GPUFramebuffer instead and even add new ones. The MACRO
GPU_framebuffer_ensure_config is pretty much all you need to create and
config a GPUFramebuffer.
- DRWTexture: Due to the removal of DRWFrameBuffer, we needed to create
functions to create textures for thoses framebuffers. Pool textures are
now using default texture parameters for the texture type asked.
- DRWManager: Make sure no framebuffer object is bound when doing cache
filling.
- GPUViewport: Add new color_only_fb and depth_only_fb along with FB API
usage update. This let draw engines render to color/depth only target
and without the need to attach/detach textures.
- WM_window: Assert when a framebuffer is bound when changing context.
This balance the fact we are not track ogl context inside GPUFramebuffer.
- Eevee, Clay, Mode engines: Update to new API. This comes with a lot of
code simplification.
This also come with some cleanups in some engine codes.
This is a bit useless because gpu lamps are only used by the game engine
and it is planned to be "remove" in some way.
Doing this to clean gpu_framebuffer.c.
This includes a few modification:
- The biggest one is call glActiveTexture before doing any call to
glBindTexture for rendering purpose (uniform value depends on it).
This is also better to know what's going on when rendering UI. So if
there is missing UI elements because of this commit look for this first.
This allows us to have "less calls" to glActiveTexture (I did not
measure the final count) and less checks inside GPU_texture.
- Remove use of GL_TEXTURE0 as a uniform value in a few places.
- Be more strict and use BLI_assert for bad usage of GPU_texture functions.
- Disable filtering for integer and stencil textures (not supported by
OGL specs).
- Replace bools inside GPUTexture by a bitflag supporting more options to
identify texture types.
Undo sometimes reserved too much space in the buffer,
now assert when this happens and allocate the exact size needed.
Note prepares for moving text editor undo out of the text block (D3113)
which will split the undo buffer into a list of undo steps.
This commit essentially introduces a new RNA property flag, which when
set prevents affected property from being processed at all in comparison
code (also used to automatically generate static override rules).
The idea is to use it on very low-level data in RNA, like e.g. mesh's
geometry or psys' particles collections.
For now only applied to psys' particle collections, on the main mesh of
Agent327 pigeon, it goes from 100ms to 0.5ms on a full
auto-override-generating comparison...
Also added some new RNA property helper funcs to check on comparable and
overridable status.
With better directory layout and more proper include
statements we can avoid several local modifications,
such as changing config.h for Windows Glog and the
ones related on pass-through statements in logging
headers in Glog.
This commit also makes unused functions not-a-warning
for external code.
The list of editor-types is rather long by now, so better to arrange them into
sections.
Original patch by @jeske with updates by @Blendify and myself.
Design Task: T36028
Patch: D3112
* Pressing "OK" wouldn't close Blender anymore
* Using File -> Quit would use popup version, not OS native window
Cleaned up code a bit to avoid duplicated logic.
Steps to reproduce were:
* Open Blender (no need for factory settings, "Promt Quit" needs to be enabled)
* Edit the file (e.g. translate some object)
* Quit Blender but don't skip quit promt
* Press "Save & Quit"
* Save the file
Not sure if Windows supports the "Save & Quit" behavior, so this may not have
applied to Windows.
You only had to close Blender through File -> Quit.
Leaks happened because WM_exit() was called from within operator, UI wasn't able
to free some of it's heap data then. This data was the handler added in
uiTemplateRunningJobs() and the IDProperty group added in uiItemFullO_ptr_ex().
There was obviously a general design issue which only became visible in this
specific case.
We now delay the WM_exit call by wrapping it into a handler that gets registered
as usual. I didn't see a better way to do this, all tricks done in
ui_apply_but_funcs_after() to prevent leaks didn't work here. In fact they may
be redundant now, but am not brave enough to try ;)
Seems we can not use include directories order trick, since
files are included form inside ".." string, which forces current
directory to be checked first.
This module has no use now with the new DrawManager and DrawEngines and it
is using deprecated paths.
Moving gpu_shader_fullscreen_vert.glsl
to draw/modes/shaders/common_fullscreen_vert.glsl
Steps to reproduce were:
* Append a workspace (via '+' icon) - make sure its from the default workspaces.blend
* Activate it
* Should crash
Was accessing data from view-layer which wasn't updated yet (and thus could be
NULL). Crash occured after rB8153f89518b4a.
@campbellbarton, you may want to check if all object-mode stuff still works as
expected, not sure what's the state of it.
It is possible that datablock will not be re-used for the new
dependency graph building. Freeing function was freeing all
the nested pointers of databnlock, but not datablock memory
itself.
MSVC still defines __cplusplus as 199711L until it's in full conformance with the newer c++ standards, however the things we need from the standard are fully supported, hence a check for the msvc version was needed.
With the API recently added to gawain, it is now possible to update the vbos linked to the batch.
So the batch does not have to be destroyed.
The optimization is more sensitive when sculpt is made on low poly meshs
Without this a "Clearcoat" link could be moved to "Clearcoat Normal"
for example, which doesn't make much sense.
Differential Revision: https://developer.blender.org/D3105
Increasing the samplig dimensions like this is not optimal, I'm looking
into some deeper changes to reuse the random number and change the RR
probabilities, but this should fix the bug for now.
We revert to the malloc/realloc and manually manage the upload.
There seems to be a performance penalty from using glMapBuffer on some
hardware, prefering way is glBufferData(NULL) with glBufferSubData.
PointCache was having a collection of items of PointCache type, having a
collection of items of PointCache type, having...
Nuff said.
For now, chose the 'ugly' way to fix it, that is, the one that changes
nothing to API and scripts using it: we define another 'PointCacheItem'
RNA type for items of our point cache collection, which has exact same
interface as PointCache except for the collection.
This is doomed to be rewritten at some point anyway, not worth spending
time trying to define a really correct data layout for now.
- Upload the data to the GPU directly when creating the element buffer in
GWN_indexbuf_build_in_place().
- Convert data in place when squeezing the indices and removing the need
for another allocation.
- GWN_indexbuf_build_in_place() can be used with already used element
buffers and reupload their data without changing vbo id (keeping vaos
up to date).
We now alloc a vbo id on creation and let OpenGL manage its memory directly.
We use glMapBuffer to get this memory location.
This enables us to reuse and modify any vertex buffer directly without
destroying it with its associated Batches.
This commit does not really improve performance but will let us implement
more optimizations in the future.
We can also resize the buffer even if this can be slow if we need to keep
the existing data.
The addition of the usage hint makes dynamic buffers not a special case
anymore, simplifying things a bit.
* In the Collada Module parameters are typically ordered
in a similar way. I changed this to:
extern std::string get_joint_id(Object *ob, Bone *bone);
* The Object parameter was not used in get_joint_sid().
I changed this to:
extern std::string get_joint_sid(Bone *bone);
When a name property is defined for collection's struct, but no name is
actually set, we want to also fallback to index case. We cannot handle
empty names to address items of a collection!
We had a mix of two issues here actually:
* First, Brush are currently using their own sauce for custom previews,
this is not great, but moving them to use common ImagePreview system of
IDs is a low-priority TODO. For now, they should totally ignore their
own ImagePreview.
* Second, BKE_icon_changed() would systematically create a PreviewImage
for ID types supporting it, which does not really makes sense, this
function is merely here to 'tag' previews as outdated. Actual creation
of previews is deferred to later, when we actually need them.
They are used to start and end colored output in console.
Use with care, it is up to you to check that console actually
supports Truecolor ANSII.
In thew future we can extend this to other consoles and platforms.
Previous approach was not clear enough and caused problems.
UBOs were taking slots and not release them after a shading group even
if this UBO was only for this Shading Group (notably the nodetree ubo,
since we now share the same GPUShader for identical trees).
So I choose to have a better defined approach:
- Standard texture and ubo calls are assured to be valid for the shgrp
they are called from.
- (new) Persistent texture and ubo calls are assured to be valid accross
shgrps unless the shader changes.
The standards calls are still valids for the next shgrp but are not assured
to be so if this new shgrp binds a new texture.
This enables some optimisations by not adding redundant texture and ubo
binds.
Unity itself is deprecated, but the API is also supported by KDE and the GNOME Dock extension,
which means that it will be useful for a wide variety of distributions.
To get a progress bar, the system must have a blender.desktop file and libunity installed.
The need for libunity is annoying, but the only alternative would be to integrate a DBus library...
Reviewers: campbellbarton, brecht
Differential Revision: https://developer.blender.org/D3106
Basically, don't do full in-place copy of node tree datablock if it's
already expanded. Current way how node tree is evaluated is fully
built around the idea that evaluation copies values from original
to copied datablocks.
Changing links is handled on another level.
Old solution was to create a new vbo and copy it to the location of
the old vbo hoping for the batches to update their vaos before drawing.
The issue is that the new VAO caching is not updating the VAOs at all
unless the shader interface changes.
So unless we expand gawain to support updating of vertex buffers (in a
better way than the current "dynamic" buffer) we need to delete every batch
linked to the vbo we want to recreate.
This solution might have performance implications.
Requires BLI_utildefines.h to be included first,
(already noted in other inline code).
Possible alternative could be to move BLI_assert into own header.
Even if they are for safety they are not free to use !
On my system (Mesa + AMD Vega GPU) calling:
glBindVertexArray(1);
glDrawArrays(GL_TRIANGLES, 0, 3);
glBindVertexArray(0);
in a loop, shows the same overhead as a full vao switching (which is more
or less 10 times slower than just calling glDrawArrays)
Moreover, now that we use OpenGL 3.3 binding a VAO is REQUIRED to issue a
drawcall so it is garanted to be overwritten before the next drawcall.
Problem can only happen if someone draws directly with opengl commands.
This allows to draw multiple primitive of the type
GWN_PRIM_LINE_STRIP
GWN_PRIM_LINE_LOOP
GWN_PRIM_TRI_STRIP
GWN_PRIM_TRI_FAN
GWN_PRIM_LINE_STRIP_ADJ
with only one drawcall. This should speed up some areas that are really
sensitive to drawcall counts : UV drawing, Hair drawing...
For IDProps IDarray, IDP_EqualsProperties was called for each item,
instead of IDP_EqualsProperties_ex, discarding value of `is_strict`
option.
Probably not an issue with current code, though.
The issue was only visible with copy-on-write enabled, and related to the
fact, that dependency graph builder binds original FCurves.
For now use smallest patch possible to make things to work and to make
draguu happy.
Need to think of a smarter way to deal with drivers, bones and view layers.
I tried to use real multisampling but the main problem is the outline detection that needs to have matching depth samples.
So adding FXAA instead. Always on for now, may add a parameter for it later.
One thing to note is that we need to copy the final output once again to the main color buffer because we cannot swap the dtxl textures (they can be referenced elsewhere like GPUOffscreen).
We could improve upon this and add TAA on top if viewport is still.
This means that rendering clay with AO only needs 1 geometry pass.
Thus greatly improving performance of poly heavy scene.
This also fix a self shadow issue in the AO, making low sample count
way better.
We also do not need to blit the depth anymore since we
are doing a fullscreen shading pass.
The constant cost of running the a deferred shading pass is negligeable.
This include quite a bit of code cleanup inside clay_engine.c.
The deferred pipeline is only enabled if at least one material needs it.
Multisampling is not supported yet.
Small hacks when doing deferred:
- We invert the normal before encoding it for precision.
- We put the facing direction into the sign of the mat_id.
- We dither the normal to fight the low bitdepth artifacts of the normal
buffer (which is 8bits per channel to reduce bandwidth usage).
- The common name in computer science are 'getters' and 'setters', so by
adding these names to the documentation (while 'get' and 'set are still
also mentioned) we improve findability. Having 'Getters/Setters' as a
title also makes it clearer that this example is not just about
getting or setting the property value.
- Added a little prefix to each printed value, so that print statement,
expected output, and real output can be matched easier.
The old example had two downsides:
- It promoted a blocking UI design, where the user is shown a popup
before actually executing the operator.
- It didn't show how to actually use the property values.
The new code avoids these mistakes. The properties are also shown in the
redo panel in the 3D view.
Note that I also changed the bl_idname, as this is an example about
properties, not about dialogue boxes, and changed the class name to use
the standard operator naming convention.
I also extended the example to include a panel that sets multiple
properties of the operator, since I see questions about this relatively
frequently.
Sequencer rendering can use multisample render targets. Be sure to sync
thoses after rendering.
Also disable the sample loop when not needed.
Do note that currently the color correction is broken with the sequencer.
Looks like someone changed the signature of some RNA update callback,
and for some reason that 'change skey' update function was not updated
(or later got merged from master)...
We'll need RNA to check for its func signatures, some day...
When adding scene strips to the sequencer, the wrong scenes were
getting getting added if some were skipped. For example:
Given 4 scenes (A, B, C, D) if you're trying to add the last 3 scenes
(B, C, D) as strips to the first scene (A), it would ended up adding
"A, B, C" instead of "B, C, D" as expected.
Fix provided by Andrew (signal9).
Calling the rendering operator seems to kill any other WM_job running, leaving
uncompiled materials into a GPU_MAT_QUEUED state. This then made the probe update
looping indefinitely (all_materials_updated remaining to false).
To fix this, we resume compilation for materials that are in this state.
Cancelling Render before all material compilation could make certain material
remain uncompiled. Fortunately, this is not allowed as of now.
Need Clear ID recalc flag on load. Otherwise it's possible to have
some IDs considered always updated by Cycles, when they were saved
in a tagged-for-update state.
Thanks Bastien for feedback and review!
Nothing user visible, only things needed for multi-object support,
making picking functions more flexible too.
- Support passing in an initialized hit-struct,
so it's possible to do multiple nearest calls on the same hit data.
- Replace manhattan distance w/ squared distance
so they can be compared.
- Return success to detect changes to a hit-data
which might already be initialized (also more readable).
This is mostly to avoid re-compilation when using undo/redo operators.
This also has the benefit to reuse the same GPUShader for multiple materials using the same nodetree configuration.
The cache stores GPUPasses that already contains the shader code and a hash to test for matches.
We use refcounts to know when a GPUPass is not used anymore.
I had to move the GPUInput list from GPUPass to GPUMaterial because it's containing references to the material nodetree and cannot be reused.
A garbage collection is hardcoded to run every 60 seconds to free every unused GPUPass.
* Suspicious usage of pointer:
short *type = 0; // this creates a null pointer
When this is later used for anything then blender would crash.
After following the code and check what happens i strongly believe
the author wanted to use a short and not a pointer to a short here.
* local variable where reused later in same function
While this did no harm, i still felt it was better to use a different
name here to make things more separated:
- moved variable declaraiotns into loop (for int a=0; ...)
- renamed uv_images to uv_image_set
- renamed index variable from i to j in inner loop that
reused same index name from outer loop
The iterator was redeclared 3 times. I fixed this to avoid future issues.
I commit separately because so the changes are less cluttered all over
the place.
The variable child was redeclared multiple times in the same function.
While this has not created any issues i still changed this to avoid
confusion and keep the usage of the variables more local.
The function validateConstraints() potentially causes a null pointer
exception. I changed this so that the function returns a failure as soon
as the validation fails. This avoids falling into the null pointer trap.
The 2 methods add_bezt() and create_bezt() do almost the same.
I combined them both into add_bezt() and added the optional parameter
eBezTriple_Interpolation ipo
Don't write the multichannel metadata when there is only a single layer,
and don't unnecessarily consider single layer images with Blender metadata
as multi layer.
This save a little memory and copying in the kernel by storing only a 4x3
matrix instead of a 4x4 matrix. We already did this in a few places, and
those don't need to be special exceptions anymore now.
This is in preparation of making Transform affine only, and also gives us
a little extra type safety so we don't accidentally treat it as a regular
4x4 matrix.
The purpose of the previous code refactoring is to make the code more readable,
but combined with this change benchmarks also render about 2-3% faster with an
NVIDIA Titan Xp.
This leads to less lookups to the GWNShaderInterface and less uniform upload.
We still keep a legacy path so that Builtin uniforms can still work. We might restrict this path to Builtin shader only in the future.
You can now cancel your renders that are too long. This will still output the current status of the render. For example if you cancel at 50% rendering progress, you will have a render result with only half the render samples.
This also fix a bug with the probe debug display when there was more than 2 probes. ped->probe_id was equal to 0 for all planar probes until the next frame. Resulting in all planar data debug to show probe 0.
If no custom URL was set, add-ons would get a "Report a Bug" button opening
the default developer.blender.org bug tracker. Now we only add this default
button if the add-on is bundled and not installed by the user.
Premise: When pose bones are selected, applying a pose library should
only affect the selected bones.
This commit fixes a bug where the pose was also applied when there was
no overlap between the selected bones and the bones in the pose. For
example, applying a pose which contains only keyframes for the left
hand, while only right-hand bones are selected, would apply the pose
to the left hand anyway.
The code is now also slightly more efficient; the removed 'selcount'
counter was only used as a binary (i.e. zero or non-zero). It's now
stored as a bitflag instead.
Currently only covering handful of files from reports about wrong fps detected.
It will need D3083 applied first to get tests passed, also tests themselves
are to be committed to svn.
But there are some python code which needs to be reviewed, like blendfile
passed to run_blender().
Reviewers: sybren, mont29
Reviewed By: sybren, mont29
Subscribers: mont29
Differential Revision: https://developer.blender.org/D3096
This is an issue with which value to trust: fps vs. tbr. They both cam be
somewhat broken. Currently the idea is:
- If file was saved with FFmpeg AND we are decoding with FFmpeg we trust tbr.
- If we are decoding with Libav we use fps (there does not seem to be tbr in
Libav, unless i'm missing something).
- All other cases we use fps.
Seems to work all good for files from T53857, T54148 and T51153. Ideally we
would need to collect some amount of regression files to make further tweaks
more scientific.
Reviewers: mont29
Reviewed By: mont29
Differential Revision: https://developer.blender.org/D3083
- put render iterator in own scope
(would shadow it's own variable if used multiple times).
- enforce semicolon at end of iterator macros.
- no need to typedef one-off macro structs.
Each AnimData block has a set of Blend/Extrapolation/Influence settings
that can be used to control how the active action is blended with the
NLA stack. However, these settings were not getting copied over to the
newly created strips (as the push-down code existed long before these
settings were added).
This commit solves this in several ways:
* Active Action Blend/Extrapolation/Influence settings now get copied
to the new strips when adding them to the NLA stack via Push Down.
Note: This doesn't happen when there are no existing NLA tracks,
as these settings don't get used in that case.
* Strip Influence will be copied across when inf < 1.0 (i.e. when a
non-default value is used), to maintain the effect. To make this work,
the influence value will get added as a keyframe to the strip's
"Influence" Control FCurve.
- See code comments for an alternative approach and why that was not chosen
- Strip Time still doesn't get keyframes added automatically yet.
* To ensure the "extrapolation mode" settings don't get always overwritten,
I've put in place a compromise: the extrapolation will only get changed
if the chosen setting will cause problmes (i.e. hold forward & back -> hold forward
if there are other tracks before it already).
Not safe for backporting to 2.79[x] stable releases.
Now repeating the operator will use the previously chosen offset, either with
the modal operator or typed in. The modal operator will still start at zero.
Reverts rBb9ae517794765d6a1660 and fixes the issue properly. Old fix could cause
NULL to be passed to functions that expect all arguments to be non-NULL.
The idea is to separate the most common case from symmetrical frustum. And to make a simple but efficient calculation.
The new radius is usually 98% the size of the radius size of the asymmetric solution.
Thanks to @fclem for reviewing the patch on IRC
Also get rid of the static var and initialization.
This enables the user to see the progress on the info header.
Closing blender or reading a file also kill the job which is good.
Unfortunatly, this job cannot be interrupt by users directly. We could make it interruptible but we need a way to resume the compilation.
In the gpus like `AMD Radeon HD 7570M` and `Intel(R) HD Graphics 4000` this solution improves performance a hundreds or even thousands of times depending on the resolution.
Reviewed By: @brecht and @fclem
Differential Revision: https://developer.blender.org/D3095
Vertex group remapping utility function,
now shared between object join and array modifier cap-ends.
Weights which don't exist are removed.
D3092 by @Foaly
Main purpose is to make it possible to cover FPS detection with regression test.
But it might also be handy for some other scripters.
Thanks Campbell for review!
The issue was happening with fast Gaussian blur, and caused by NaN value pixels
in the input buffer.
Now made it so Map Range output does not produce NaN, by returning arbitrary
value of 0. Still better than NaN!
Previous fix for T53430 caused T54200.
The edge case for soft & hard cuts weren't working,
where the strip used start/end-still & the frame was placed exactly on
the start/end of of the sequence content.
T54200 fixed the end-still case but broke hard-cuts for all other cases.
This fixes the case for soft/hard cuts with/without start/end-still.
I keep reading that texture painting is not working yet. However it is fully
working. We even have a "Full Shading" option in the viewport display panel.
Clay/EEVEE still need their UI figured out. But the context itself is doing
its part after this patch, and at least for Cycles it's working like 2.79.
Instead of creating a new instancing shading group without attrib, we now have instancing calls. The benefits is that they can be culled.
They can be used in conjuction with the standard and generate calls but shader must support it (which is generally not the case).
We store a pointer to the actual count so that the number can be tweaked between redraw.
This will makes multi layer rendering more efficient.
Introduced explicit ID property node for driers in depsgraph,
so it is clear what is the input for driver, and what is the
output.
This also solved relations builder throwing lots of errors
due to ID property not being found.
It was too tricky to know ahead of time if an object would still
be visible in the new window/workspace/scene/layer combination,
especially since other windows may share some of these data-blocks.
So store the context, make the change, then check if the object is
still visible, freeing mode data of it's not.
It was possible to have relations like A -> B -> C -> A (import thing is
that no other operations points into this cluster) which were not detected
or reported by dependency cycle solver.
Now this is solved by ensuring we don't leave unvisited nodes behind.
This is probably a better way to handle it: instead of totally
discarding scaling of non-free axes, keep the ratio between them.
Basically the logic of the constraint is now that it rescales the
object uniformly in the non-free axis plane in order to force the
total volume change to the desired value.
Note that this code will likely be generalized,
currently each new case is a little different though
so it's too early to move them into general functions.
This (now removed code) calls gl_Vertex deprecated draws. It was doing
background drawing (color gradient, flat background) which is not used
by any engine.
It seems the reason the old version of the constraint overcompensates
as reported in T48079 is to allow the constraint to work with uniform
scaling on all axes. However the way it did that actually _requires_
uniform scaling for the constraint to work correctly, and breaks if
only the free scaling axis is used to avoid redundant channels.
This version attempts to allow both by discarding scaling in the non-
free directions instead of applying the correction on top of it.
This merges changes in internals, runtime-only of existing custom
normals code, which make sense as of themselves, and will make diff of
soc branch easier/lighter to review.
In the details, it mostly changes two things:
* Now, smooth fans (aka MLoopNorSpaceArray) can store either loop
indices, or pointers to BMLoop themselves. This makes sense since in
BMesh, it's relatively easy to get index from a BMElement, but nearly
impracticable to go the other way around.
* First change enforces another, now we cannot rely anymore on `loops`
being NULL in MLoopNorSpace to detect single-loop fans, so we instead
store that info in a new flag.
Again, these are expected to be totally non-functional changes.
around the volume.
We generate a tight mesh around the active voxels of the volume in order
to effectively skip empty space, and start volume ray marching as close
to interesting volume data as possible. See code comments for details on
how the mesh generation algorithm works.
This gives up to 2x speedups in some scenes.
Reviewed by: brecht, dingto
Reviewers: #cycles
Subscribers: lvxejay, jtheninja, brecht
Differential Revision: https://developer.blender.org/D3038
Selection code relies on being able to set the depth functions
however passes have their own depth settings.
Add DRW_state_lock to ignore passes settings for particular flags.
This fixes occlusion queries cycling through objects under the cursor.
By default select wasn't picking the nearest object,
this could have been fixed by not clearing the depth buffer,
but calling GPU_select_(begin/end) without the binded frame-buffer
caused issues for depth-picking. So move GPU_select begin/end to a
callback.
This also has the advantage that only needs to populate the engines once
to draw two passes.
Note that cycling through objects fails with occlusion queries still,
will fix shortly.
This is very efficient and add a pretty low overhead (0.1ms of drawing time for 10K objects passing through all tests, on my i3-4100M).
The like the rest of the DRWCallState, test is "cached" until the view matrices changes.
In some cases it doesn't make sense for add-ons to be listed for hiding.
Especially for import/export which use minimal UI space.
This adds `bl_info["use_owner"]` to add-ons,
currently defaulting to True for all non Import-Export add-ons.
This is more important now that we will have tigther volume bounds that
we hit multiple times. It also avoids some noise due to RR previously
affecting these surfaces, which shouldn't have been the case and should
eventually be fixed for transparent BSDFs as well.
For non-volume scenes I found no performance impact on NVIDIA or AMD.
For volume scenes the noise decrease and fixed artifacts are worth the
little extra render time, when there is any.
* Use a subsurface color equal to the base color, and give the subsurface
radius skin like values by default. This is how the parameter should
typically be used.
* Use GGX by default, multiscatter GGX is still quite noisy and has some
fireflies so let's keep it optional for now.
Now the only missing bit seems to be in Cycles to pass depsgraph to
builtin_image_float_pixels().
Ideally we could get evaluation context instead of using depsgraph + settings.
But for the other rna EvaluationContext functions this is how we are doing.
Reviewers: sergey, brecht
Differential Revision: https://developer.blender.org/D3087
While doing so with Bone_R.001, Bone_R.003, Bone_R.003 etc. is doomed to
issues, doing that on duplicates of actually correctly named bones can
be handy, and safe.
So adding back as an option (was removed in rB702bc5ba26d5).
Flip names operator changed in rB702bc5ba26d5, to some sensible
behavior. But this breaks common workflow of 'duplicate part of the
bones, scale-mirror new ones, and flip their names'.
So now, instead of doing this in two steps, trying to guesstimate which
bones should get which name, just add option to flip names to duplicate
operator itself. Simpler, safer, and much, much more consitent behavior
and predictable results.
This solves issue with tweaking brush size when interleaving particle edit
and texture paint modes. The issue was caused by texture paing setting more
operator properties then it's done for particle edit mode, which made window
manager to use saved proeprties for the "missing" ones.
Don't see any reason why we would want to save any of those properties.
This is a regression since rB83b60dac57a1.
Allows for each workspace to have it's own add-ons on display.
Filtering for: Panels, Menus, Keymaps & Manipulators.
Automatically applies to add-ons at the moment.
Access from workspace, toggled off by default
once enabled, add-ons can be white-listed.
See D3076
When changing the mode of an object, apply this to all other
workspaces that share the same active object.
Also use copy the object-mode when duplicating workspaces.
Note that setting `glDepthFunc` isn't important,
since 2.8 branch changes this value it might seem like an error
however it's harmless in this case - so better make note of this.
Refactor include:
- Removal of DRWInterface. (was useless)
- Split DRWCallHeader into a new struct DRWCallState that will be reused in the future.
- Use BLI_link_utils for APPEND/PREPEND.
- Creation of the new DRWManager struct type. This will enable us to create more than one manager in the future.
- Removal of some dead code.
This still does not make point density to work in Cycles, but at least it pass
the depsgraph down the line.
Note this was working fine before the depsgraph/render refactor to pass
evaluated depsgraph to the engines.
User notes
----------
Compositing, rendering of multi-layers in Eevee should be fully working now.
Development notes
-----------------
Up until now we were still using the same depsgraph for rendering and viewport
evaluation. And we had to go out of our ways to be sure the depsgraphs were
updated.
Now we iterate over the (to be rendered) view layers and create a depsgraph to
each one, fully evaluated and call the render engines (Cycles, Eevee, ...) with
this viewlayer/depsgraph/evaluation context.
At this time we are not handling data persistency, Depsgraph is created from
scratch prior to rendering each frame. So I got rid of most of the partial
update calls we had during the render pipeline.
Cycles: Brecht Van Lommel did a patch to tackle some of the required Cycles
changes but this commit mark these changes as TODOs. Basically Cycles needs to
render one layer at a time.
Reviewers: sergey, brecht
Differential Revision: https://developer.blender.org/D3073
-Make the view and object dependant matrices calculation isolated and separated, avoiding non-needed calculation.
-Adding a per drawcall matrix cache so that we can precompute these in advance in the future.
-Replaced integer uniform location of only view dependant builtins by DRWUniforms that are only updated once per shgroup.
Now that Eevee has support for offline rendering (F12) we can use it for
the Material previews.
Note: This makes the duplicated UI issue one panel worse. That happens when
Cycles if your scene engine, and Eevee is your workspace engine.
Depth picking needs to read the depth buffer after drawing
since GPU_select_end runs in a different OpenGL context
reading the depth buffer wasn't working.
This caused the last object to be unelectable.
This patch adds F12 offline Freestyle rendering support to Eevee.
Most functionalities are identical with those found in Cycles.
The only major difference is that the per-view layer "use Freestyle" toggle
option is currently placed in the "Passes" panel of the "View Layers"
properties window instead of a "Layer" panel as in Cycles. Since Freestyle
is a post-processed overlay and not a pass, the present option location is
a compromise. To describe this fact, the per-layer "use Freestyle" option
is in a subsection labeled as "Layer".
Reviewers: fclem, brecht, campbellbarton
Reviewed By: fclem, brecht
Subscribers: dfelinto
Differential Revision: https://developer.blender.org/D3084
This separate context allows two things:
- It allows viewports in multi-windows configuration.
- F12 render can use this context in a separate thread and do a non-blocking render.
The downside is that the context cannot be used while rendering so a request to refresh a viewport will lock the UI. This is something that will be adressed in the future.
Under the hood what does that mean:
- Not adding more mess with VAOs management in gawain.
- Doing depth only draw for operators / selection needs to be done in an offscreen buffer.
- The 3D cursor "autodis" operator is still reading the backbuffer so we need to copy the result to it.
- All FBOs needed by the drawmanager must to be created/destroyed with its context active.
- We cannot use batches created for UI in the DRW context and vice-versa. There is a clear separation of resources that enables the use of safe multi-threading.
Offscreen contexts are not attached to a window and can only be used for rendering to frambuffer objects.
CGL implementation : Brecht Van Lommel (brecht)
GLX implementation : Clément Foucault (fclem)
WGL implementation : Germano Cavalcante (mano-wii)
Other implementation are just place holder for now.
The exporter does export matrix data (4*4 Transformation matrix) only for Skeletal animation. For object animation only exporting to trans/rot/loc is implemented.
This task implements Matrix export also for simple Object animation.
Differential Revision: https://developer.blender.org/D3082
Turns out to be the call that was destroying performance.
I get 18ms->6ms improvement of drawing time with 10 000 unique objects.
And we can still improve upon this!
This started with a fix for an animated Object Hierarchy. Then i decided to cleanup and optimize a bit. But at the end this has become a more or less full rewrite of the Animation Exporter. All of this happened in a separate local branch and i have retained all my local commits to better see what i have done.
Brief description:
* I fixed a few issues with exporting keyframed animations of object hierarchies where the objects have parent inverse matrices which differ from the Identity matrix.
* I added the option to export sampled animations with a user defined sampling rate (new user interface option)
* I briefly tested Object Animations and Rig Animations.
What is still needed:
* Cleanup the code
* Optimize the user interface
* Do the Documentation
Reviewers: mont29
Reviewed By: mont29
Differential Revision: https://developer.blender.org/D3070
This is used to determine which voxels are to be considered empty space.
Previously it was hardcoded for converting dense grids to OpenVDB grids
to reduce disk space usage.
This value is also useful for rendering engines to know, i.e. to
optimize ray marching.
Some other software cannot handle grid names with spaces in them. We still check for names with spaces so as to not break old
files.
This fixes T53802.
This replaces the blackbody to RGB code with the simpler and faster one from
Cycles. It's a little different but the other placing using this is the legacy
volume drawing, so no need to stay compatible with that.
Similar to the Principled BSDF, this should make it easier to set up volume
materials. Smoke and fire can be rendererd with just a single principled
volume node, the appropriate attributes will be used when available. The node
also works for simpler homogeneous volumes like water or mist.
Differential Revision: https://developer.blender.org/D3033
Technically the original issue is that xof/yof in render result is calculated
for drawing border render. So a simpler patch could be:
```
- rr->xof = re->disprect.xmin;
+ rr->xof = re->disprect.xmin + BLI_rcti_cent_x(&re->disprect) - (re->winx / 2);
```
However everywhere in the code we are getting border directly from re->disprect
which we may as well do here too.
Besides I'm taking this as a chance to get rid of RenderResult in the internal
loop of eevee, to help prepare the code to the upcoming rendering pipeline
changes.
When you were using autosmooth to generate some custom normals, and
created empty custom loop normal data, you would go back to an 'all
smooth' shading, cancelling some sharp edges generated by the mesh's
smooth threshold.
Now we will first tag such edges as sharp, such that shading remains the
same. This is not crucial in current master, but it is for clnors
editing gsoc branch!
Regression caused by earlier commits to improve the automerge behaviour.
In this case, the problems only occurred when moving a selected keyframe
forwards in time to overlap an unselected keyframe.
While it is necessary to ignore duplicates when doing Deselect/Column Select
(where double-updates may result in nothing being selected), for borderselect,
not including the duplicates meant that sometimes, nothing would happen
if you were trying to borderselect keyframes originating from hidden channels.
This was first noticed in the greasepencil-object branch, but affects all
animation channel types.
This is not an ideal solution but blender freeing system is already well tangled.
So tracking and clearing vao caches when destroying contexts does prevent bad behaviour.
This allows us to:
- Not mock around with tags stored in a global space,
and not to iterate over all datablocks in the database
to clear the tags.
- Properly deal with datablocks which might not be in main database.
While it sounds crazy, it might be handy when dealing with preview,
or some partial scene updates, such as motion paths.
- Avoids majority of places where depsgraph construction needed bmain.
This is something what could help in blender2.8 branch.
From tests with production file here did not see any measurable slowdown.
Hopefully, there is no functional changes :)
Was temporarily removed when moving object mode to workspace.
Note: there is an issue where eval_ctx->view_layer is NULL on load,
for now pass a view layer argument, we might wan't to set the value
instead.
This reverts commit 87c72a7d27.
Caused T54121 which breaks blend file saving.
For now crash on exit is preferable.
Possible solution is to free screen-manipulator batches in a separate
loop.
We now continue transparent paths after diffuse/glossy/transmission/volume
bounces are exceeded. This avoids unexpected boundaries in volumes with
transparent boundaries. It is also required for MIS to work correctly with
transparent surfaces, as we also continue through these in shadow rays.
The main visible changes is that volumes will now be lit by the background
even at volume bounces 0, same as surfaces.
Fixes T53914 and T54103.
A major bottleneck of current implementation is the call to create_bindings() for basically every drawcalls.
This is due to the VAO being tagged dirty when assigning a new shader to the Batch, defeating the purpose of the Batch (reuse it for drawing).
Since managing hundreds of batches in DrawManager and DrawCache seems not fun enough to me, I prefered rewritting the batches itself.
--- Batch changes ---
For this to happen I needed to change the Instancing to be part of the Batch rather than being another batch supplied at drawtime.
The Gwn_VertBuffers are copied from the batch to be instanciated and a new Gwn_VertBuffer is supplied for instancing attribs.
This mean a VAO can be generated and cached for this instancing case.
A Batch can be rendered with instancing, without instancing attribs and without the need for a new VAO using the GWN_batch_draw_range_ex with the force_instance parameter set to true.
--- Draw manager changes ---
The downside with this approach is that we must track the validity of the instanced batch (the original one). For this the only way (I could think of) is to set a callback for when the batch is getting free.
This means a bit of refactor in the DrawManager with the separation of batching and instancing Batches.
--- VAO cache ---
Each VAO is generated for a given ShaderInterface. This means we can keep it alive as long as the shader interface lives.
If a ShaderInterface is discarded, it needs to destroy every VAO associated to it. Otherwise, a new ShaderInterface with the same adress could be generated and reuse the same VAO with incorrect bindings.
The VAO cache itself is using a mix between a static array of VAO and a dynamic array if the is not enough space in the static.
Using this hybrid approach is a bit more performant than the dynamic array alone.
The array will not resize down but empty entries will be filled up again. It's unlikely we get a buffer overflow from this. Resizing could be done on next allocation if needed.
--- Results ---
Using Cached VAOs means that we are not querying each vertex attrib for each vbo for each drawcall, every redraw!
In a CPU limited test scene (10000 cubes in Clay engine) I get a reduction of CPU drawing time from ~20ms to 13ms.
The only area that is not caching VAOs is the instancing from particles (see comment DRW_shgroup_instance_batch).
This allows allocation of VAOs from different opengl contexts and thread as long as the drawing happens in the same context.
Allocation is thread safe as long as we abide by the "one opengl context per thread" rule.
We can still free from any thread and actual freeing will occur at new vao allocation or next context binding.
The other approach was causing too much error in some cases (e.g. favouring
the lower-valued keyframes). This fix should make the resulting curves less
bumpy/jagged.
This commit removes an earlier attempt at optimising the lookups
for duplicates of a particular tRetainedKeyframe once we'd already
deleted all the selected copies. The problem was that now, instead
of getting rid of the unselected keys (i.e. the basic function here),
we were only getting rid of the selected duplicates.
With this fix, unselected keyframes will now get removed (as expected)
again. However, we currently don't take their values into account
when merging keyframes, since it is assumed that we don't care so much
about their values when overriding.
that end up on the same frame
Currently, when scaling keyframes in the Dopesheet, if multiple
selected keyframes end up on the same frame post-scaling, they
would not get removed by the "Automerge" setting that normally
removes duplicates on the same frame.
This commit changes the behaviour so that when multiple selected
keyframes end up on the same frame, instead of keeping all these
around on the same frame (e.g. resulting in a column of keyframes
on different values), we will instead merge them into a single
keyframe (by averaging the values). This should result in a
smoother F-Curve with fewer "stair-steps" that need to be carefully
cleaned out afterwards.
Requested by @hjalti
This bug took a while to track down. In the test file with this report,
the Nla-Strip Control Curve for strip time would get disabled if you
changed the NLA Editor to a second Graph Editor instance.
It turns out that because this second Graph Editor would have the
"filter fcurves by name" option enabled, this would trigger a lookup
of the referenced property's name (in order to test whether it matched
the filtering criteria). However, since that filtering code was written
before the introduction of these curves, it still assumed that the names
for these Control Curves should be handled the same as for standard FCurves.
Unfortunately, that doesn't work, as the property lookups fail if the standard
method is used - when the lookups fail, the F-Curves get tagged as being
invalid/disabled (and need to be reset using the "Revive Disabled FCurves"
operator).
Note: The changes in this patch look complicated, as I've had to shuffle
a bit of code around so that the name-filtering check can have access to
the additional info it needs. In the process, I've also removed the earlier
(hacky) approach where the control curves were getting added to a temp
buffer to get changed from normal FCurves to special ANIMTYPE_NLACURVES.
Not sure why we need a relation from solver to a tip local transform, this
will be handled via parent relation.
Fixes remaining dependency cycles reported in T54083.
It is not possible to address transform at particular position of constraint
stack, and when constraint is being addressed is usually from driver variable.
This fixes some of dependency cycles reported in T54083.
We need to move the render result logic outside the render engine code.
It makes no sense for Eevee/Clay/... to have to re-implement the render resilt
creation logic. Beside the original implementation really got it wrong, by
ignoring the different render layers needed for the final render.
Finally, there is no need to re-create the logic for views. So this was also
fixed.
Note 1: This will break still if the depsgraph of the needed view layers is not
updated / created. We need to address this separately. For now if users want
to test this, just show each view layer in the viewport at least once.
Note 2: We are still getting depsgraph from scene and creating if needed.
`BKE_scene_get_depsgraph(scene, view_layer, true);` according to Sergey we need
to move the render depsgraph for the Render struct instead. I will do it
separately as well.
This is a regression in rB4f1c0a1 which only allowed cutting haior at the
second segment only, while there is nothing wrong with cutting hair at the
first segmewnt.
Don't use dm->get*Array for DM you don't own. This call can allocate temporary
CD layer, which is not thread safe at all.
Also removed hard-coded logic around CDDM check. new functions will do same
logic, but are mode DM-type-=independent.
We shouldn't mix image pool acuisition with and without user provided,
the fact that internally image.c uses last frame from Image datablock
confuses the logic.
Optionally don't remap indices for objects.
Checking all objects parent's would reference a freed pointer
while freeing all objects.
In the case of dynamic topology there is no use in keeping track
of hook/vertex-parent indices.
Also disable this when creating meshes for undo storage
since adding an undo step shouldn't be modifying other objects.
This reuses the Cycles regression test code to also work for OpenGL UI drawing.
We launch Blender with a bunch of .blend files, take a screenshot and compare
it with a reference screenshot, and generate a HMTL report showing the failed
tests and their differences.
For Cycles we keep small reference renders to compare to in svn, but for OpenGL
developers currently have to generate the references manually. How to use:
* WITH_OPENGL_DRAW_TESTS=ON in CMake
* BLENDER_TEST_UPDATE=1 ctest -R opengl_draw
* .. make code changes ..
* ctest -R opengl_draw
* open build_dir/tests/opengl_draw/report.html
Differential Revision: https://developer.blender.org/D3064
Once 'losing lib' issue is fixed (in previous commit), we have new issue
that this could lead to several copies of the same linked data-block in
.blend file. Which is not good. At all.
So had to add a GHash-based check in libraries reading code to ensure we
only load a same ID from a same lib once.
The issue was that when a same lib was found several times in loaded
.blend, we'd only keep the first occurence. But since Blender expects
next data-blocks to belong to last found library, we could actually
be adding data-blocks assigned to copies of the duplicated lib to
another, totally unrelated lib.
Those data-blocks were then obviously not found when actually loading
libs content, and lost.
Note that this only fix one part of the issue, current code can
generate several copies of same linked data-block now, will fix in
another commit.
While the script should be using INVOKE_PREVIEW for operators in clip view,
window manager was lacking some switch statements.
Thanks Brecht fore review!
- Use BLI_threadpool_ prefix for (deprecated)
thread/listbase API.
- Use BLI_thread as prefix for other functions.
See P614 to apply instead of manually resolving conflicts.
- When returning the number of items in a collection use BLI_*_len()
- Keep _size() for size in bytes.
- Keep _count() for data structures that don't store length
(hint this isn't a simple getter).
See P611 to apply instead of manually resolving conflicts.
This is kind of doesn't matter where macro itself is defined.
We should stick to the following:
- If some macro is actually more an inline function, follow regular
function name conventions.
- If macro is a macro, type it in capitals. Use module prefix if that
helps readability or it if helps avoiding accidents.
This completes twist feature, which is now possible to also control by
texture. Since textures can not easily contain negative values as well,
same trick with 0.5 neutral as vertex groups is used.
All in all, this twist features allows to do following things.
Original hair:
{F2287535}
Hair with scientifically calculated twist value of 0.5:
{F2287540}
And we can also twist braids in opposite directions dependent on left/right
side:
{F2287548}
The idea is to give a control over direction of twist, and maybe amount of
twist as well. More concrete example: make braids on left and right side of
character head to be twisting opposite directions.
Now, tricky part: we need some negative values to flip direction, but weights
can not be negative. So we use same trick as displacement map and tangent normal
maps, where 0.5 is neutral, values below 0.5 are considered negative and values
above 0.5 are considered positive.
It allows to have children hair to be twisted around parent curve, which is
quite an essential feature when creating hair braids.
There are currently two controls:
- Number of turns around parent children.
- Influence curve, which allows to modify "twistness" along the strand.
This isn't supported since there are subsequent reads to all point coordinates
after modification started.
Probably we need to create a temp copy of point, but that's like extra CPU
ticks.
view_layer is NULL when the render engine is created, this gets passed
around and ends up in this code causing a crash. This should be reverted
after the render engine api is updated to set view_layer.
The conditionals in particle code are... some sort of madness... I'm not
even sure what the correct behavior should be from looking at it.
In this case the path cache generation was being skipped in edit mode.
Due to changes in draw code particles from old files that may have had a
default draw size of 0 will not be visible. Old draw code would check
for this and adjust the size, however the unit for draw size has changed
from pixels to BU, and it no longer makes sense to have such checks.
This patch is to ensure particles from such files remain visible.
It seems to be useful still in cases where the particle are distributed in
a particular order or pattern, to colorize them along with that. This isn't
really well defined, but might as well avoid breaking backwards compatibility
for now.
This removes the need of custom attribs for instancing.
Instancing works fully with dynamic batches & Gwn_VertFormat now.
This is in prevision of the VAO manager patch.
This manager allows to distribute existing batches for instancing
attributes. This reduce the number of batches creation.
Querying a batch is done with a vertex format. This format should
be static so that it's pointer never changes (because we are using
this pointer as identifier [we don't want to check the full format
that would be too slow]).
This might make the original Instance Data manager useless but it's currently used by DRW_object_engine_data_ensure().
Theses batches keeps their memory chuck allocated after transfer to be reused and updated very often.
NOTE: This commit break instancing in DRW. (it's fixed in the next commit)
Now that the new 3D viewport draws to a multisample offscreen buffer, there is
no good reason anymore to create an entire multisample window and pay the
performance/memory cost for other regions that don't need it.
GL_MULTISAMPLE now only gets enabled for offscreen buffers, so we don't need
to check for it throughout the UI code anymore.
Differential Revision: https://developer.blender.org/D3062
This is like the only way to add variety to hair which is created
using simple children. Used here for the hair.
Maybe not ideal, but the time will show.
Burley SSS uses a bit of strange thing where the albedo and closure weight are
different, which makes the subsurface color act a bit like a subsurface radius
indirectly by the way the Burley SSS profile works.
This can't work for random walk SSS though, and it's not clear to me that this
is actually a good idea since it's really the subsurface radius that is supposed
to control this. For now I'll leave Burley SSS working the same to not break
backwards compatibility.
This can be very slow if it contains a big texture, and it's not
necessarily setup in a useful way anyway, and materials can be used
in multiple scenes.
Instead of cloning the window to create dummyHWNDs and dummyHDCs to avoid calling the SetPixelFormat more than once in the same window, use the original window and HDC and do not call the SetPixelFormat again.
In addition to avoiding a lot of unnecessary calls, it simplifies the code and makes it match the others OS
Cycles already uses 1.5 as default. BI original 1.0 filter doesn't look good for
Eevee. The ideal scenario would be for both Cycles AND Eevee to use the same DNA
setting.
But for now it is nice to at least have Eevee renders to look better by default.
Note: This handles doversion for 2.7x files only. Files previously created in
2.8 need to be manually corrected.
It is basically brute force volume scattering within the mesh, but part
of the SSS code for faster performance. The main difference with actual
volume scattering is that we assume the boundaries are diffuse and that
all lighting is coming through this boundary from outside the volume.
This gives much more accurate results for thin features and low density.
Some challenges remain however:
* Significantly more noisy than BSSRDF. Adding Dwivedi sampling may help
here, but it's unclear still how much it helps in real world cases.
* Due to this being a volumetric method, geometry like eyes or mouth can
darken the skin on the outside. We may be able to reduce this effect,
or users can compensate for it by reducing the scattering radius in
such areas.
* Sharp corners are quite bright. This matches actual volume rendering
and results in some other renderers, but maybe not so much real world
objects.
Differential Revision: https://developer.blender.org/D3054
Looks like there was no way to avoid that so far, since
WM_event_add_timer_notifier can set mere int-in-pointer there, this can
cause issues. So added mere flags system to wmTimer to allow
controlling this.
Instead of calling an operator I just call `collection.new()`. Moving the
code into a separate function also simplifies it. In its new form there is
also no undefined behaviour when me.vertex_colors is non-empty but without
active layer.
We were not passing a scene collection parent to the BKE_collection_add
function, which in turn made syncing not work.
Right now we:
* Explicitly pass the master collection in this case
* Fallback to the master collection in other cases
With unittest.
- normalize → average the vector: the vector isn't normalized here, because
it doesn't necessarily becomes unit length. Instead, the sum is converted
to an average vector.
- angle is the acos()…: the dot product between the vertex normal and the
average direction of the connected vertices is computed, and not the
opposite.
- The initial `con` list was discarded immediately and replaced by a new
list.
- File didn't end with a newline.
We've got quite comprehensive BMesh based implementation, which is way easier
for maintenance than abandoned Carve library.
After all the time BMesh implementation was working on the same level of
limitations about manifold meshes and touching edges than Carve. Is better
to focus on maintaining one boolean implementation now.
Reviewers: campbellbarton
Reviewed By: campbellbarton
Differential Revision: https://developer.blender.org/D3050
Previously quads always split along first-third vertices.
This is still the default, to avoid flickering with animated deformation
however concave quads that would create two opposing triangles now use
second-fourth split.
Reported as T53999 although this issue has been known limitation
for a long time.
- Read-only access can often use EvaluationContext.object_mode
- Write access to go to WorkSpace.object_mode.
- Some TODO's remain (marked as "TODO/OBMODE")
- Add-ons will need updating
(context.active_object.mode -> context.workspace.object_mode)
- There will be small/medium issues that still need resolving
this does work on a basic level though.
See D3037
option(WITH_INTERNATIONAL"Enable I18N (International fonts and text)"ON)
option(WITH_PYTHON"Enable Embedded Python API (only disable for development)"ON)
option(WITH_PYTHON_SECURITY"Disables execution of scripts within blend files by default"ON)
mark_as_advanced(WITH_PYTHON)# dont want people disabling this unless they really know what they are doing.
option(WITH_PYTHON_SECURITY"Disables execution of scripts within blend files by default"ON)
mark_as_advanced(WITH_PYTHON)# don't want people disabling this unless they really know what they are doing.
mark_as_advanced(WITH_PYTHON_SECURITY)# some distributions see this as a security issue, rather than have them patch it, make a build option.
option(WITH_PYTHON_SAFETY"Enable internal API error checking to track invalid data to prevent crash on access (at the expense of some effeciency, only enable for development)."OFF)
option(WITH_PYTHON_SAFETY"Enable internal API error checking to track invalid data to prevent crash on access (at the expense of some efficiency, only enable for development)."OFF)
mark_as_advanced(WITH_PYTHON_SAFETY)
option(WITH_PYTHON_MODULE"Enable building as a python module which runs without a user interface, like running regular blender in background mode (experimental, only enable for development), installs to PYTHON_SITE_PACKAGES (or CMAKE_INSTALL_PREFIX if WITH_INSTALL_PORTABLE is enabled)."OFF)
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.
