This is a complete refactor over the old system. The goal was to increase quality
first and then have something more flexible and optimised.
|{F9603145} | {F9603142}|{F9603147}|
This fixes issues we had with the old system which were:
- Too much overdraw (low performance).
- Not enough precision in render targets (hugly color banding/drifting).
- Poor resolution near in-focus regions.
- Wrong support of orthographic views.
- Missing alpha support in viewport.
- Missing bokeh shape inversion on foreground field.
- Issues on some GPUs. (see T72489) (But I'm sure this one will have other issues as well heh...)
- Fix T81092
I chose Unreal's Diaphragm DOF as a reference / goal implementation.
It is well described in the presentation "A Life of a Bokeh" by Guillaume Abadie.
You can check about it here https://epicgames.ent.box.com/s/s86j70iamxvsuu6j35pilypficznec04
Along side the main implementation we provide a way to increase the quality by jittering the
camera position for each sample (the ones specified under the Sampling tab).
The jittering is dividing the actual post processing dof radius so that it fills the undersampling.
The user can still add more overblur to have a noiseless image, but reducing bokeh shape sharpness.
Effect of overblur (left without, right with):
| {F9603122} | {F9603123}|
The actual implementation differs a bit:
- Foreground gather implementation uses the same "ring binning" accumulator as background
but uses a custom occlusion method. This gives the problem of inflating the foreground elements
when they are over background or in-focus regions.
This is was a hard decision but this was preferable to the other method that was giving poor
opacity masks for foreground and had other more noticeable issues. Do note it is possible
to improve this part in the future if a better alternative is found.
- Use occlusion texture for foreground. Presentation says it wasn't really needed for them.
- The TAA stabilisation pass is replace by a simple neighborhood clamping at the reduce copy
stage for simplicity.
- We don't do a brute-force in-focus separate gather pass. Instead we just do the brute force
pass during resolve. Using the separate pass could be a future optimization if needed but
might give less precise results.
- We don't use compute shaders at all so shader branching might not be optimal. But performance
is still way better than our previous implementation.
- We mainly rely on density change to fix all undersampling issues even for foreground (which
is something the reference implementation is not doing strangely).
Remaining issues (not considered blocking for me):
- Slight defocus stability: Due to slight defocus bruteforce gather using the bare scene color,
highlights are dilated and make convergence quite slow or imposible when using jittered DOF
(or gives )
- ~~Slight defocus inflating: There seems to be a 1px inflation discontinuity of the slight focus
convolution compared to the half resolution. This is not really noticeable if using jittered
camera.~~ Fixed
- Foreground occlusion approximation is a bit glitchy and gives incorrect result if the
a defocus foreground element overlaps a farther foreground element. Note that this is easily
mitigated using the jittered camera position.
|{F9603114}|{F9603115}|{F9603116}|
- Foreground is inflating, not revealing background. However this avoids some other bugs too
as discussed previously. Also mitigated with jittered camera position.
|{F9603130}|{F9603129}|
- Sensor vertical fit is still broken (does not match cycles).
- Scattred bokeh shapes can be a bit strange at polygon vertices. This is due to the distance field
stored in the Bokeh LUT which is not rounded at the edges. This is barely noticeable if the
shape does not rotate.
- ~~Sampling pattern of the jittered camera position is suboptimal. Could try something like hammersley
or poisson disc distribution.~~Used hexaweb sampling pattern which is not random but has better
stability and overall coverage.
- Very large bokeh (> 300 px) can exhibit undersampling artifact in gather pass and quite a bit of
bleeding. But at this size it is preferable to use jittered camera position.
Codewise the changes are pretty much self contained and each pass are well documented.
However the whole pipeline is quite complex to understand from bird's-eye view.
Notes:
- There is the possibility of using arbitrary bokeh texture with this implementation.
However implementation is a bit involved.
- Gathering max sample count is hardcoded to avoid to deal with shader variations. The actual
max sample count is already quite high but samples are not evenly distributed due to the
ring binning method.
- While this implementation does not need 32bit/channel textures to render correctly it does use
many other textures so actual VRAM usage is higher than previous method for viewport but less
for render. Textures are reused to avoid many allocations.
- Bokeh LUT computation is fast and done for each redraw because it can be animated. Also the
texture can be shared with other viewport with different camera settings.
This patch moves the EEVEE depth of field shaders to eevee_shaders.c and
adds them to the eevee shaders test suite.
Reviewed By: Clément Foucault
Differential Revision: https://developer.blender.org/D8771
- add the use of DRWShaderLibrary to EEVEE's glsl codebase to reduce code
complexity and duplication.
- split bsdf_common_lib.glsl into multiple sub library which are now shared
with other engines.
- the surface shader code is now more organised and have its own files.
- change default world to use a material nodetree and make lookdev shader
more clear.
Reviewed By: jbakker
Differential Revision: https://developer.blender.org/D8306
This cleans up a bit of duplicated code and some confusion about
what was culled and what wasn't.
Now everything is culled based on the given object pointer.
If the object pointer is NULL there is no culling performed.
The unit system is designed for displaying and editing and not for
rendering. Eevee, Workbench and GPencil used these settings to convert
the focal length and sensor size to world units. Making depth of field
render differently with Cycles.
For now we will remove the scale in the draw engines to match cycles,
until we implemented a camera parameters specific scale.
Reviewed By: brecht, fclem
Maniphest Tasks: T64988
Differential Revision: https://developer.blender.org/D4925
There is now a checkbox to enable/disable depth of field per camera. For Eevee
this replace the scene level setting. For Cycles there is now only an F-Stop
value, no longer a Radius.
Existing files are converted based on Cycles or Eevee being set in the scene.
Differential Revision: https://developer.blender.org/D4882
Viewport now displays alpha checkerboard pattern like Cycles does when
film alpha is set to "Transparent".
Some small workarounds were necessary for Depth of Field and correct TAA
support.
BF-admins agree to remove header information that isn't useful,
to reduce noise.
- BEGIN/END license blocks
Developers should add non license comments as separate comment blocks.
No need for separator text.
- Contributors
This is often invalid, outdated or misleading
especially when splitting files.
It's more useful to git-blame to find out who has developed the code.
See P901 for script to perform these edits.
Even though the fragment shader was already discarding all members of
dof_bokeh_sides when blades was zero, the C code was still trying to
use this for a few divisions leading to runtime asserts.
Those are harmless yet can lead some to waste time while pursuiting
other bugs (namely a near freeze when blades aspect ratio is too low).
NOTE: There is a float imprecision near the focus plane
due to the current technique used for DOF. This makes the alpha channel
transparent on nearly in focus objects even when they should not.
This artifact should be fixed when the DOF will use scatter as gather for
low brightness areas.
Fix T57042 : Eevee does not render alpha when DOF is turned on
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.
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.
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.
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.
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.
