In the flow where VK_KHR_CREATE_RENDERPASS_2_EXTENSION_NAME does not exist
VkAttachmentReference are created inside a loop and their backing buffer is referenced in the subpass object.
the VkAttachmentReference vectors are freed once the loop exists, causing the subpass to point to freed data.
Add all the VkAttachmentReference to a vector in the scope of the entire function, to ensure they are not freed until vkCreateRenderPass is called
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".
Implements basic ASTC support:
* Only 4x4 and 8x8 block sizes.
* Other block sizes are too complex to handle for Godot image compression handling. May be implemented sometime in the future.
The need for ASTC is mostly for the following use cases:
* Implement a high quality compression option for textures on mobile and M1 Apple hardware.
* For this, the 4x4 is sufficient, since it uses the same size as BPTC.
ASTC supports a lot of block sizes, but the benefit of supporting most of them is slim, while the implementation complexity in Godot is very high.
Supporting only 4x4 (and 8x8) solves the real problem, which is lack of a BPTC alternative on hardware where it's missing.
Note: This does not yet support encoding on import, an ASTC encoder will need to be added.
- Removed empty paragraphs in XML.
- Consistently use bold style for "Example:", on a new line.
- Fix usage of `[code]` when hyperlinks could be used (`[member]`, `[constant]`).
- Fix invalid usage of backticks for inline code in BBCode.
- Fix some American/British English spelling inconsistencies.
- Other minor fixes spotted along the way, including typo fixes with codespell.
- Don't specify `@GlobalScope` for `enum` and `constant`.
For some reason AFAICT mesa reports a feature as enabled even when its
extension isn't supported. The Vulkan specification says nothing aboutd
this so this is technically more of a workaround, but it works.
Replace all TODO uses of `#warning` by proper TODO comments, and will open
matching bug reports to keep track of them.
We don't have a great track record fixing TODOs, but I'd wager we're even
worse for fixing these "TODO #warning" so we should prohibit this usage.
Mainly:
- Make `max_descriptors_per_pool` project setting Vulkan-specific.
- Use a common, render driver agnostic magic FourCC for shader binary data.
- Downgrade spirv_reflect to Vulkan-only dependency.
- Add a `RENDER_DRIVER_*` macro to GLSL shader code for per-driver customizations.
End users would get spammed with messages of varying verbosity due to the
mess that thirdparty layers/extensions and drivers seem to leave in their
wake, making the Windows registry a bottomless pit of broken layer JSON.
I'm all for helping end users clean up mess in their registry / system paths
for Vulkan ICDs, layers and extensions, but the way this is done by
VK_EXT_debug_utils is just horrible - and the way for them to fix it (manual
edit of system files) is also not a good thing to recommend.
Closes countless issues where users think Godot is broken because it reports
weird errors.
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);
```
- Validate format conservatively. (This is to have VRS images created regardless whether VRS attachments are supported, which avoids errors in places where the code assumes such images were created on low-spec GPUs.)
- Create a non-layered default VRS image, which is what Vulkan (and D3D12, by the way) expect.
Now the `linuxbsd` platform can be built headlessly (e.g. without X11
development libraries).
I also cleaned up some weird (old?) usages of the `env` variable which
seem to make no difference and are used nowhere else.
- Check block decoration in addition to type decoration to be sure to find `readonly` decorators
- Verify uniforms have same writability across all shader stages in Vulkan RD
- Include Godot version and commit hash in shader cache key
- Reject files when format doesn't match, even if it's lower, since we don't have backwards compatibility here
In far most cases it seems like it's going to message about bogus manifests
in the Windows registry which point to JSON files which have since been
uninstalled, but without clearing the registry.
This happens with bogus Vulkan overlays from Twitch, Epic Online Services,
NVIDIA Nsight Systems, OBS Studio, Rockstar Games... fix your mess folks.
Fixes#56089.
- Initialize queue indices to values meaning 'unset'
- Remove unused parameters & members
- Make texture update access flags consistent with texture copy
- Fix style and pass type of some parameters
- Synchronize setup-draw in flush with a semaphore
- Add no current list validation to draw_list_begin_splits()
- Update texture usage flags on destination of copy
- Fix misuse of Vulkan flag
Clean up and do fixes to hash functions and newly introduced murmur3 hashes in #61934
* Clean up usage of murmur3
* Fixed usages of binary murmur3 on floats (this is invalid)
* Changed DJB2 to use xor (which seems to be better)
* 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!
This method can be used to get the graphics API version currently in
use (such as Vulkan). It can be used by projects for troubleshooting
or statistical purposes.
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.
This commit removes a lot of enum values related to the color render pass in favor of a new flag-bases approach. This means instead of hard-coding all the possible option combinations into enums, we can write our logic by checking a bit-mask.
The changes in rendering_device_vulkan.cpp add support for unused attachments. That means RenderingDeviceVulkan::framebuffer_create() can take null RIDs in the attachments vector, which will result in VK_ATTACHMENT_UNUSED entries in the render pass.
This is used in this same PR to establish fixed locations for the color pass attachments (only color and separate specular so far, but TAA will add motion vectors as well). This way the attachment locations in the shader can stay the same regardless of which attachments are actually used.
Right now all the combinations of flags are generated, but we will need to add a way to limit the amount of combinations in the future.
* 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.
This updates VMA and instead of using the custom small pool approach from 4e6c9d3ae9, lazily creates pools for the relevant memory type indices, which doesn't require patching VMA.
Also, patches already merged upstream or not needed any longer are removed.
