Age | Commit message (Collapse) | Author |
|
When rendering using the command line the curvature wasn't rendered. The reason
was that the ui_scale wasn't initialized and therefore the same pixels where
sampled to detect the curvature. This is fixed by setting the ui_scale to 1 for any
image render.
|
|
This was visible outside of camera view and was not respecting the
"Depth of Field" checkbox on the Camera properties.
Now return early if DoF should not be visible.
Maniphest Tasks: T94262
Differential Revision: https://developer.blender.org/D13631
|
|
Caused by own {rB5aa3167e48b2}.
Related commit: {rBebaa3fcedd23}.
For stereo renders, `BKE_image_is_multilayer` is true, however we seem
to get to down to `space_image_gpu_texture_get` [where this is called]
from `IMAGE_cache_init` with a NULL Image->RenderResult.
So what then happens is that `BKE_image_multilayer_index` is called and
even though it has an appropriate codepath for stereo, it earlies out
and does not set multi_index correctly.
Still a bit puzzled why RenderResult is NULL for a render, but since
other places also check for a valid RenderResult before going down the
_multilayer_ route (and doing _multiview_ instead), now do the same
thing, BKE_image_multiview_index is now called in these cases (and seems
to behave correctly, checked with layers and passes and all seems to
display correctly, either in stereo or choosing individual eyes).
thx @jbakker & @brecht for double-checking.
Maniphest Tasks: T92608
Differential Revision: https://developer.blender.org/D13063
|
|
The planar reflections being rendered at the same resolution as the HiZ max
buffer, do not need any uv correction during raytracing.
However, the GTAO horizon buffer being at output resolution do need the
uv factors in order to match the pixels visible on screen. To avoid many
complication, we increase the size of the GTAO texture up to the hiz buffer
size. This way, if planar reflections need GTAO the texture is big enough.
We change the viewport of the GTAO framebuffer for the main view in order
to not have to modify Uvs in many places.
|
|
This was caused by camera background being rendered in world space, causing
floating point imprecision issues when camera was far from origin.
Adding a uniform to change vertex shader to process everything in viewspace
to fix the problem.
|
|
Up lower clamp on spec_angle to prevent NaN from being generated on intel GPUs at low roughness.
Fixes T88754
Reviewed By: fclem
Maniphest Tasks: T88754
Differential Revision: https://developer.blender.org/D12508
|
|
The issue was caused by `textureSize()` returning the size of the level 0
even when the min texture level is higher than 0.
Using a uniform to pass the correct size fixes the issue.
This issue also affected the downsampling of radiance for reflections and
refractions.
This does not affect anything other than the recusive downsampling shaders.
|
|
The simplify was hardcode to be disabled in render.
|
|
selected pass
Caused by {rBebaa3fcedd23}.
Seems this above commit assumed an ImageUser's multi_index is only used
for Multiview/Stereo? This is not the case, multi_index also stores the
index for layer/pass combination.
If we call both BKE_image_multilayer_index and BKE_image_multiview_index
(even though this is not appropriate/needed for multilayer images?), we
might end up overwriting multi_index again.
note: looking at this I was also wondering why we update the ImageUser
in image-buffer-aquiring funnctions [and not from the UI, e.g.
template_image_layers, but that is a whole different story I guess, see
comment in T90772 as well]
note2: this could also use a utility function (this is not the only
place where this is done), this is fo a cleanup commit.
Maniphest Tasks: T90772
Differential Revision: https://developer.blender.org/D12267
|
|
This was working differently in 2.79, tried tracking this down and it
seems this was wrong since the 2.8 beginning in {rB7907dfc40018}.
This would not only crash without an active scene camera, but would also
result in different tracks from different camera's constraints could not
be selected.
So select id depends on corresponding camera, remove the dependency on
scene camera completely.
Maniphest Tasks: T90651
Differential Revision: https://developer.blender.org/D12230
|
|
While trying to get Blender 2.93.x LTS to build fine on all release architectures in Debian, I noticed that the misleading use of "mips" as integer variable caused problems when compiling on mips64el. The patch should fix the issue.
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D12194
|
|
Channel packed images should not have their RGB affected by alpha.
rendering in Cycles and Eevee was fine already, but displaying these was
not right in the Image and Node editors.
Not 100% sure what to do for the "Color and Alpha" mode, but I guess
this should stay like it was before (applying the alpha).
"Color", "R", "G", and "B" modes were changed to not have color be
affected by alpha though.
ref. T89034
Maniphest Tasks: T89034
Differential Revision: https://developer.blender.org/D11871
|
|
Issue is that due to the strange definition of render in grease pencil
(meaning should be rendered similar to rendering). This included normal
viewport rendering in OB_RENDER and OpenGL render in OB_RENDER.
For other rendering modes the overlay vertex opacity would be used. This
patch sets this value to 1 when rendering via a scene strip override.
NOTE: that this isn't a good solution as I expect that users want to use
the opacity of the Grease pencil object. Perhaps the GPencil team has a
better solution for it.
|
|
EEVEE uses hashing to sync aov names and types with the gpu.
For the type a hashed value was overridden making `decalA`
and `decalB` choose the same hash. This patches fixes this
by removing the most significant bit.
|
|
When using ``Path`` alignment, if the stroke has one point the texture rotates randomly when move the viewport. This was because with one point is impossible to calculate a path.
Now, if the stroke has only one point, the texture for this stroke is aligned to Object.
|
|
This was caused by unsafe sqrt calls.
Fixes T86578 white artifacts in EEVEE
Reviewed By: brecht, dfelinto
Differential Revision: https://developer.blender.org/D11428
|
|
Both before and after can have artifacts with some normal maps, but this seems to give
worse artifacts on average which are not worth the minor performance increase.
This reverts commit 5c4d24e1fd752a8a89d44d05e8e3f9b31f2d7db0.
Ref T88368, D10084
|
|
When filtering the render using view layer, could be possible the mask is NULL and just need to be ignored.
|
|
|
|
This was a bug uncovered by rB50782df42586.
Previously, the lightcache was always discarded between redraw and forced
to be updated again.
Now we check for update inside the render loop making it compatible with
accumulation motion blur and long exposure.
|
|
This was caused by the slight focus gather not being wide enough
for small radii. Now the cast to int will properly round the radius to
the nearest integer.
This is related to T86244 Black Artefacts in EEVEE on Transparent BSDF
|
|
This was caused by the bokeh LUT being sampled outside the valid range.
But `texelFetch` is only valid if the sample actually exists. This lead to
undefined behavior.
The fix is to increase `DOF_MAX_SLIGHT_FOCUS_RADIUS` (which just offsets the
LUT along the X=Y axis) to avoid any sample outside the defined range.
|
|
This was caused by the drivers not optimizing the shader enough to remove
the samplers and data used by closure eval. Removing the lighting loops
from the depth shader fixes the perf regression.
|
|
This enables the overlay for instanced geometry.
After this change, objects that are an instance of the current active
object (which are also being modified in the current active mode) won't
fade, which is different from the previous behavior.
Reviewed By: fclem
Maniphest Tasks: T82155
Differential Revision: https://developer.blender.org/D9362
|
|
When a scene uses cryptomatte the viewport rendering would lead to a
memory leak. The reason was that all image renders (viewport+final)
activated cryptomatte. But is only used for final rendering.
This patch only activates cryptomatte when doing final rendering.
|
|
|
|
This artifact was already present in previous version but was hidden
by the faulty SSS scale.
The issue comes from the translucence using the geometric normal
(computed using fragment shader derivative) leading to poor precision at
depth discontinuity.
Replacing using the same geometric normal reconstruction as the ambient
occlusion pass removes most of the issue.
|
|
This was caused by the material grouping system which was missing
a per material uniform update in the opaque case.
|
|
This was caused by the SSR option resetting the accumulation. But the
render passes were only cleared in the init phase. This means that
when SSR was resetting the taa_render_sample the actual renderpasses
would still contains 1 sample. This means the renderpasses were always
divided by the wrong number of samples.
The fix is to clear just before accumulation if the sample is 1.
The fact that it works for motion blur is kind of a blessing. This is
because we check stl->effects->ssr_was_valid_double_buffer before
resetting the sampling. So this only happens on the first motion step
and does not affect the rest of the rendering.
Reviewed by: jbakker
Differential Revision: https://developer.blender.org/D11033
|
|
Some discard of vertbuf were not correctly followed by discards of
the GPUBatches that were using them. This lead to a use-after-free
situation where GPUBatches would reuse old VBO information. This did
not crash immediatly because the VBO indices were cached by our VAO
caching system. It kept working on some implementation because VBO
reference in the VAO (probably) preventing the VBO from being freed
by reference counting.
This fixes T85977 NVidia: Random crashes in 'DrvPresentBuffers'
|
|
Logical error let this function to write one past the buffer bounds.
|
|
This was caused by the new depsgraph persistence.
The GPUbatches we got from the cache being the same for each frame
means that we need to be more careful about cleanning the additional
VBOs references.
Moving the `EEVEE_motion_blur_swap_data` function call at the end of
the loop makes sure the references are cleaned.
|
|
In cases where the same node tree is used in different materials with
small changes, the wrong material could be selected.
Cause: Hair shaders GPU resources weren't updated and used the previous bound data.
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D11036
|
|
This was caused by the closure refactor. The radiance being masked
for SSR, we need to not enable SSR when trying to render the specular
color pass.
Reviewed By: jbakker
Differential Revision: https://developer.blender.org/D11028
|
|
The `filtered_radiance` is not a texture form the texture pool.
It needs manual freeing if not used.
|
|
This was caused by some sort of degenerated normals.
|
|
It seems the pow result is unstable on some implementations.
Also avoid undefined behavior by clamping aoFactor to strict positive values.
|
|
Now, instead to offset the stroke color to make it visible over fill, the stroke keeps the original color and the fill is offset.
Related to the issue in T87406.
|
|
The color of the strokes was shifted when using `Single Color` or
`Object Color`in Solid Display Mode. This was originally done so
that it was still possible to differentiate strokes and fills.
The fix allows for a completly flat shading by checking if the
Lighting Mode is set to `Flat`. If it is, then the colors are not
shifted for the strokes and everything will use the same color.
Reviewed By: antoniov, fclem
Maniphest Tasks: T87406
Differential Revision: https://developer.blender.org/D10957
|
|
|
|
Points are drawn as half octahedron (aligned to the camera).
Getting the appropriate matrix for facing the camera would fail in in
orthographic view, points were not facing the camera (revealing their
missing other half octahedron)
Maniphest Tasks: T87150
Differential Revision: https://developer.blender.org/D10923
|
|
The problem was that you could getting write access to a grid from a
`const Volume *` without breaking const correctness. I encountered this
when working on support for volumes in the bounding box node. For
geometry nodes there is an important distinction between getting data
"for read" and "for write", with the former returning a `const` version
of the data.
Also, for volumes it was necessary to cast away const, since all of
the relevant functions in `volume.cc` didn't have const versions. This
patch adds `const` in these places, distinguising between "for read"
and "for write" versions of functions where necessary.
The downside is that loading and unloading in the global volume cache
needs const write-access to some member variables. I see that as an
inherent problem that comes up with caching that never has a beautiful
solution anyway.
Some of the const-ness could probably be propogated futher in EEVEE
code, but I'll leave that out, since there is another level of caching.
Differential Revision: https://developer.blender.org/D10916
|
|
|
|
|
|
To check if an "is_mesh_verts_only" mesh, the overlay engine checks if the
mesh has no "totedge" and has "totvert".
However, sometimes this engine can check the wrong mesh since editmesh
works on `embm->mesh_eval_final`.
Reviewed By: fclem
Differential Revision: https://developer.blender.org/D10917
|
|
Use same implementation as Cycles. Is also a bit faster.
Should also fix T86578 at the same time.
|
|
The issue was that where_on_path uses a resampled curve to get the data
from the curve. This leads to disconnects between the curve the user
sees and the evaluated location data.
To fix this we simply use the actual curve data the user can see.
The older code needed a cleanup either way as there were hacks in other
parts of the code trying to work around some brokenness. This is now
fixed and we no longer need to clamp the evaluation range to 0-1 or make
helper functions to make it do what we actually want.
Reviewed By: Campbell, Sybren
Differential Revision: http://developer.blender.org/D10898
|
|
Before this change, you could have the new sculpt symmetry code and the
older weight paint symmetry code active at the same time. This would
lead to users easily trashing their weigh paint data if they were not
careful when switching between modes.
Now the specific weight paint symmetry code is an exclusive toggle so
the user can't accidentally mirror strokes and vertex groups at the same
time. This also paves the way of supporting Y and Z symmetry in the
future for weight groups mirroring if we decide to add it in the future.
Reviewed By: Sybren
Differential Revision: http://developer.blender.org/D10426
|
|
Now `ED_view3d_backbuf_depth_validate`, `ED_view3d_draw_depth` and
`ED_view3d_draw_depth_gpencil` are unified in `ED_view3d_depth_override`.
This new function replaces `ED_view3d_autodist_init`.
Also, since `ED_view3d_depth_update` depends on the render context, and
changing the context is a slow operation, that function also was removed,
and the depth buffer cached is now updated inside the new unified drawing
function when the "bool update_cache" parameter is true.
Finally `V3D_INVALID_BACKBUF` flag has been renamed and moved to
`runtime.flag`.
Differential revision: https://developer.blender.org/D10678
|
|
Display the bone axes at the head (root) of the bone by default, instead
of the tail (tip), and add a slider so that it's possible to adjust this
position.
Versioning code is in place to ensure existing files behave the same
(axes shown at tail), whereas new Armatures will be using the new
default (axes shown at head).
Reviewed By: #animation_rigging, #user_interface, Severin, Sybren
Differential Revision: https://developer.blender.org/D7685
|