Fixes#90017Fixes#90030Fixes#98044
This PR makes the following changes:
# Force processing of GPU commands for frame_count frames
The variable `frames_pending_resources_for_processing` is added to track
this.
The ticket #98044 suggested to use `_flush_and_stall_for_all_frames()`
while minimized.
Technically this works and is a viable solution.
However I noticed that this issue was happening because Logic/Physics
continue to work "business as usual" while minimized(\*). Only Graphics
was being deactivated (which caused commands to accumulate until window
is restored).
To continue this behavior of "business as usual", I decided that GPU
work should also "continue as usual" by buffering commands in a double
or triple buffer scheme until all commands are done processing (if they
ever stop coming). This is specially important if the app specifically
intends to keep processing while minimized.
Calling `_flush_and_stall_for_all_frames()` would fix the leak, but it
would make Godot's behavior different while minimized vs while the
window is presenting.
\* `OS::add_frame_delay` _does_ consider being minimized, but it just
throttles CPU usage. Some platforms such as Android completely disable
processing because the higher level code stops being called when the app
goes into background. But this seems like an implementation-detail that
diverges from the rest of the platforms (e.g. Windows, Linux & macOS
continue to process while minimized).
# Rename p_swap_buffers for p_present
**This is potentially a breaking change** (if it actually breaks
anything, I ignore. But I strongly suspect it doesn't break anything).
"Swap Buffers" is a concept carried from OpenGL, where a frame is "done"
when `glSwapBuffers()` is called, which basically means "present to the
screen".
However it _also_ means that OpenGL internally swaps its internal
buffers in a double/triple buffer scheme (in Vulkan, we do that
ourselves and is tracked by `RenderingDevice::frame`).
Modern APIs like Vulkan differentiate between "submitting GPU work" and
"presenting".
Before this PR, calling `RendererCompositorRD::end_frame(false)` would
literally do nothing. This is often undesired and the cause of the leak.
After this PR, calling `RendererCompositorRD::end_frame(false)` will now
process commands, swap our internal buffers in a double/triple buffer
scheme **but avoid presenting to the screen**.
Hence the rename of the variable from `p_swap_buffers` to `p_present`
(which slightly alters its behavior).
If we want `RendererCompositorRD::end_frame(false)` to do nothing, then
we should not call it at all.
This PR reflects such change: When we're minimized **_and_**
`has_pending_resources_for_processing()` returns false, we don't call
`RendererCompositorRD::end_frame()` at all.
But if `has_pending_resources_for_processing()` returns true, we will
call it, but with `p_present = false` because we're minimized.
There's still the issue that Godot keeps processing work (logic,
scripts, physics) while minimized, which we shouldn't do by default. But
that's work for follow up PR.
- Adds Swappy for Android for stable frame pacing
- Implements pre-transformed Swapchain so that Godot's compositor is in
charge of rotating the screen instead of Android's compositor
(performance optimization for phones that don't have HW rotator)
============================
The work was performed by collaboration of TheForge and Google. I am
merely splitting it up into smaller PRs and cleaning it up.
Changes from original PR:
- Removed "display/window/frame_pacing/android/target_frame_rate" option
to use Engine::get_max_fps instead.
- Target framerate can be changed at runtime using Engine::set_max_fps.
- Swappy is enabled by default.
- Added documentation.
- enable_auto_swap setting is replaced with swappy_mode.
`core/os/os.h` doesn't use `core/io/image.h`. It just brings
transitive dependencies. Lots of dependencies because `core/os/os.h`
is transitively included in almost every file of godot
Also added `core/io/image.h` into files^1 where `Ref<Image>` and `core/os/os.h`
were used to prevent obscure errors involving `Ref<Image>`
^1 except those which include `core/io/image_loader.h` or `core/io/image.h` by
corresponding .h file with the same name
Signed-off-by: Yevhen Babiichuk (DustDFG) <dfgdust@gmail.com>
Co-authored-by: A Thousand Ships <96648715+AThousandShips@users.noreply.github.com>
- Implements asynchronous transfer queues from PR #87590.
- Adds ubershaders that can run with specialization constants specified as push constants.
- Pipelines with specialization constants can compile in the background.
- Added monitoring for pipeline compilations.
- Materials and shaders can now be created asynchronously on background threads.
- Meshes that are loaded on background threads can also compile pipelines as part of the loading process.
This is needed to allow 2D to fully make use of 3D effects (e.g. glow), and can be used to substantially improve quality of 2D rendering at the cost of performance
Additionally, the 2D rendering pipeline is done in linear space (we skip linear_to_srgb conversion in 3D tonemapping) so the entire Viewport can be kept linear.
This is necessary for proper HDR screen support in the future.
As many open source projects have started doing it, we're removing the
current year from the copyright notice, so that we don't need to bump
it every year.
It seems like only the first year of publication is technically
relevant for copyright notices, and even that seems to be something
that many companies stopped listing altogether (in a version controlled
codebase, the commits are a much better source of date of publication
than a hardcoded copyright statement).
We also now list Godot Engine contributors first as we're collectively
the current maintainers of the project, and we clarify that the
"exclusive" copyright of the co-founders covers the timespan before
opensourcing (their further contributions are included as part of Godot
Engine contributors).
Also fixed "cf." Frenchism - it's meant as "refer to / see".
Adds a FramebufferCache singletion that operates the same way as UniformSetCache.
Allows creating framebuffers on the fly (and keep them cached if re-requested) such as:
```C++
RID fb = FramebufferCache::get_singleton()->get_cache(texture1,texture2);
```
* Map is unnecessary and inefficient in almost every case.
* Replaced by the new HashMap.
* Renamed Map to RBMap and Set to RBSet for cases that still make sense
(order matters) but use is discouraged.
There were very few cases where replacing by HashMap was undesired because
keeping the key order was intended.
I tried to keep those (as RBMap) as much as possible, but might have missed
some. Review appreciated!
Didn't commit all the changes where it wants to initialize a struct
with `{}`. Should be reviewed in a separate PR.
Option `IgnoreArrays` enabled for now to be conservative, can be
disabled to see if it proposes more useful changes.
Also fixed manually a handful of other missing initializations / moved
some from constructors.
* Changed syntax usage for RD::Uniform to create faster with a single RID
* Converted render pass setup to use this in clustered renderer to test.
This is the first step into creating a proper uniform set cache system to simplify large parts of the codebase.
