Initial PR was made by Avan for Ryubing.
This allows for Trails in the Sky 1st Chapter (and by extension Trails through Daybreak, Trails through Daybreak II, Trails Beyond the Horizon, and the future Trails in the Sky 2nd Chapter – tested via demo – to be fully playable.
Initial PR caused issues on macOS devices, specifically crashing in certain games (such as Mario Kart 8 Deluxe and the aforementioned Trails games). It was reverted on the original Ryubing project (at the time of writing this description) for causing small rendering issues on The Legend of Zelda: Breath of the Wild.
The rendering issues do not appear on macOS.
This commit improves the original PR by eliminating said macOS crash.
Trails in the Sky 1st writes its exposure/brightness value through an R32G32Float 1x1 texture and later reads the raw bits back through an R32Uint 2x1 texture mapped to the same guest memory. The two texture formats/dimensions are fully incompatible as texture views, so they end
up as separate host textures, and nothing kept their contents coherent - the reader saw stale zeroed data, making the exposure calculation (and the rendered image) too dark. (This is the initial fix for the game)
TextureGroup already had a raw-copy dependency mechanism intended to
handle this class of alias (TextureDependency, TextureGroupHandle
raw-copy plumbing), gated by CanCreateRawCopyDependency's strict same-guest-memory checks. This commit fixes three bugs that kept that mechanism from working correctly (and thus crashing Trails and other games, such as Mario Kart 8 Deluxe):
(1) TextureGroup.InitializeOverlaps() and TextureGroup.RegisterIncompatibleOverlap() only forwarded overlaps to CreateCopyDependency() when compatibility was LayoutIncompatible or better, silently excluding the fully Incompatible case the raw-copy path exists for. Widened both guards to let Incompatible overlaps through so CanCreateRawCopyDependency actually gets a chance to run.
(2) Once Incompatible overlaps were allowed through, CreateCopyDependency(TextureGroup, ...) could still fall back to a regular, non-raw textureCopy for such pairs, since ViewLayoutCompatible/CopySizeMatches only check byte size and were never meant to reason about fully incompatible pairs (e.g. a depth format aliasing a color format with the same byte size). On Vulkan running through MoltenVK on macOS, that non-raw copy path requires a pixel-format-reinterpreting texture view, and MoltenVK/Metal refuses to create any view onto a depth-format texture ("not castable"), aborting emulation. textureCopy is now forced false for Incompatible pairs, leaving raw copy (which already excludes depth/stencil formats) as the only route for that severity level.
(3) TextureGroupHandle.Inherit() copied a handle's pending DeferredCopy to the new handle when a view was recreated, but did not copy DeferredCopyRaw alongside it. A handle that inherited a pending raw copy would then execute it through the regular (non-raw) CopyTo path once acknowledged, reinterpreting the source bytes with the wrong row layout and corrupting the image - visible as vertical flickering stripes whenever a view happened to be recreated with a raw copy still pending. DeferredCopyRaw is now carried over together with DeferredCopy.
(As mentioned earlier, vertical stripes aren't present on macOS, so (3) doesn't negatively have an effect on macOS in any way).
Some fragment passes used by FFT do not have any color or depth/stencil attachments. Instead, the fragment shader writes the results directly to storage images (on OpenGL: through imageStore).
### Vulkan
When SetImage is called and FramebufferParams has no attachments, its virtual size must be updated using the width, height, and layer count of the fragment storage image. Otherwise, the attachmentless framebuffer may retain an incorrect 1x1 extent, causing backgrounds, logos, UI elements, and other rendered content to be missing.
If the first draw occurs before the storage-image descriptor is rebound, SetImage cannot yet provide the correct storage-image dimensions. Therefore, when RecreateGraphicsPipelineIfNeeded finds that FramebufferParams has no attachments and still uses the default 1x1 extent, it initializes the framebuffer dimensions from the active viewport. Otherwise, the first Vulkan draw may be restricted to a 1x1 area and render incorrectly.
The storage-image extent is therefore used as the authoritative size for FramebufferParams, while the active viewport is used as a fallback when the storage-image extent is not yet available before the first draw.
### OpenGL
A storage image is not a framebuffer attachment, so OpenGL cannot derive the framebuffer width and height from it. When the framebuffer has no attachments but a viewport with valid dimensions has already been defined, Pipeline.PreDraw must set the default framebuffer width and height from the active viewport.
Without non-zero default width and height values, the attachmentless framebuffer remains incomplete. Drawing with that framebuffer results in InvalidFramebufferOperation, so rasterization and fragment shader execution do not occur even when the storage images are bound correctly.
- The Vulkan backend now reports the driver name and version using VK_KHR_driver_properties when available, with fallbacks for raw Vulkan driverVersion parsing. Intel Windows drivers also get the full package-style version from the driver UUID when exposed by the driver.
OpenGL now carries its existing OpenGL version string through HardwareInfo.GpuDriverVersion.
Fix an "Invalid texture format 0x25A5A (sRGB: True)" error that occurs when running OCTOPATH TRAVELER 0.
The game may leave texture descriptor heap entries filled with 0x5A. Interpreting such uninitialized entries as valid texture descriptors produces an invalid format and causes texture creation to proceed with garbage descriptor data.
Validate the texture address before decoding the descriptor. Descriptors with a zero or unmapped address are marked as invalid and skipped. The invalid state is cleared when the corresponding texture pool entry is modified, allowing the descriptor to be evaluated again.
The `OpenTK.Windowing.GraphicsLibraryFramework` package is included in the Ryujinx csproj, but is never used in the project.
Checking the source code, all source files included in the package are part of the `OpenTK.Windowing.GraphicsLibraryFramework` namespace, which is never included in the Ryujinx source code. Rider also confirms that it's an unused reference.
The library includes the entirety of libglfw into the program, it clears about a half a megabyte. Not much, but it's free.
Fixes a black screen on Metal versions older than 3.1 by disabling `VK_EXT_extended_dynamic_state` (and `VK_DYNAMIC_STATE_VERTEX_INPUT_BINDING_STRIDE`). Fix imported from sunshineinabox's unpublished extended dynamic states branch.
Co-authored-by: sunshineinabox <aqemail@gmail.com>
Fixes a bug introduced in e88fd327. The problematic commit did its work directly on the updater's own arrays, calling Auto.Get() on each entry as it went, and nulling entries out along the way. The problem is that Get() can call back into Commit (via ClearMirrors -> Rebind), and when it did, that reentrant Commit would read from the same arrays the outer call was still in the middle of processing, hit one of the entries the outer had already nulled, and throw an NullReferenceException. The fix is to have Commit start by copying everything it needs into local variables and resetting _count to zero, so a reentrant call sees a clean updater and operates on its own data. The outer call then writes its snapshot back into the native arrays just before recording the Vulkan bind.
Co-authored-by: AsperTheDog <guillerman0000@gmail.com>
Co-authored-by: Max <randomgirlisweird@gmail.com>
Co-authored-by: Babib3l <gab.chevanne@gmail.com>
This PR fixes several validation errors caused by invalid Vulkan usage. These validation errors often end up invoking Undefined Behavior on the driver side, which can lead to artifacts or crashes which are driver specific and otherwise incredibly hard to track. I don't think it should have any impact on performance, but it would be good to test it with as many games as possible (maybe a bug in a game was fixed?).
Each commit fixes an error. I added to each a description with the validation error that was fixed and a small explanation on what was causing it and how I fixed it.
Co-authored-by: AsperTheDog <guillerman0000@gmail.com>
This PR marks ALL texture indexes as nonuniform to fix an issue with the paths in Tomodachi Life: Living the Dream on AMD cards. It should have a negligible impact on performance (and it should not have an impact at all on NVIDIA cards!)
It's caused by what is called 'implicit non-uniform sampler array indexing'. The idea is basically that some GPUs optimize texture lookups from indexed texture arrays, by assuming that you are never going to index different textures within a single workgroup. What this causes is that visual glitch where a subgroup is tasked with rendering a block of the screen, and in the boundaries some cores are indexing the wrong texture.
Co-authored-by: AsperTheDog <guillerman0000@gmail.com>
Fixed LoadGuestApplication hanging when cancelled.
Since startup procedure has technically changed, we should consider testing this with a variety of game formats to ensure regressions do not occur.
Fixes a few crashes:
- fixes a crash related to waking threads (priorities were wrong).
- fixes a crash from reusing the SetRenderTargets texture array (left-over data causing issues).
- fixes a mistake and an oversight in the buffer system.
- buffers were getting updated wrong causing bad data to be stored or some times cut.
- modified ranges would extend past their old buffers, crashing on syncs. Old buffers are now skipped as the new buffers already sync instead.
Introduces pooling in a few more places to increase memory efficiency.
simplified RangeList item logic.
- removed RangeItem by making all the range objects use the I(NonOverlapping)Range interface.
- BufferCache class no longer locks its RangeList, as the list is only ever accessed synchronously.
Small change to how keyboard snapshots are stored.
Increase ThreadedRenderer SpanPool size to fit slightly more data (4MB -> 8MB).
General memory improvements to decrease GC pressure and frequency.
Pool big arrays and objects that are created and deleted often.
Skip data copies when they aren't needed.
Inline flag checks to skip unneeded allocations.
From my testing the performance is about the same, but the GC frequency is much lower and collection is faster causing less and smaller spikes.
* Slightly refactors RangeLists from the last Memory Changes MR, which fixes issue 61.
* Convert as many const size array iterators to span iterators as possible. When iterating over a const size array, every iteration created a Span, now only the first iteration does in most places.
* Now using object pooling for a few object types that were rapidly deleted and recreated.
* Converted a few flag checks to binary operations to save memory allocations.
* GetSerialNumber in ISystemSettingsServer
* Bionic qualifiers
* Revert application pool size to 3285 MB for 4 GB DRAM
* Fix PPTC recompilation if mod or patch is applied