At which time does Metal allocate and deallocate memory for textures? I've observed that the textures live for the whole time of the commandBuffer. So, if I have multiple large textures that I need in subsequent shaders, it would make sense to work with multiple commandBuffers to enable deallocation in order to reduce peak memory usage. Is that correct? Do you have any other suggestions on how to reduce peak memory usage when working with large metal textures? Hint: I am using compute shaders only.

What's the recommended pattern for sharing a complex SpriteKit game scene (and basic game code) between iOS and watchOS targets without code duplication, given the platform differences in input and rendering?

Hi! Question Could you provide some reassurance on the continued support for SpriteKit in the coming years? Trigger The deprecation of SceneKit at WWDC 2025 made me nervous. Context I use SpriteKit because: It can render any video AVFoundation can handle, which third-party engines don't do for several reasons including patents. This allows me to prototype UI that controls video with a game-like feel. It can render text and emojis through AttributedString, and text is hard. Thanks to a clear and well-exposed update loop, it can be integrated with libraries such as Core Image or Box2D. Example with Box2D here. It can draw dynamic shapes and paths with SKShapeNode, and visualize and respond to all the processing done at specific times during a frame. SKView is very handy to generate textures from nodes, and SKRenderer can be used to render a SpriteKit scene to video or sequence of images. This is possible thanks to multiple frameworks working together, including AVAssetWriter and IOSurface. Example of SpriteKit offline rendering here. Apple Alternatives When Apple announced the deprecation of SceneKit, I learned RealityKit in order to move my SpriteKit-based projects to RealityKit. RealityKit is interesting, but for 2D and 2.5D workflows I found several areas where SpriteKit remains significantly more practical: I had to lock physics to the XY plane, which could stress the engine and introduce unwanted overlap. Overlapping entities on a 2D plane was hard: ModelSortGroup did not provide the simple flexibility that SpriteKit zPosition provides. AFAIK there is no dynamic path/shape drawing in RealityKit. So drawing lines, visualizing physics and joints, rendering graph connections in the scene is not as easy as it is with SpriteKit shape nodes. SpriteKit is lighter and less power-hungry. Let me know if you need any more context or detail. Thank you!

The developer documentation for Rosetta describing its end of support (https://developer.apple.com/documentation/apple-silicon/about-the-rosetta-translation-environment) says the following: Beyond this timeframe, we will keep a subset of Rosetta functionality aimed at supporting older unmaintained gaming titles, that rely on Intel-based frameworks. Are there details about which specific functionality this is? And as a concrete example, will parts of Rosetta that are needed by programs used to run and/or evaluate Windows games, such as WINE or CrossOver, remain supported?

My app contains hundreds of games, each of which would benefit from its own set of achievements, leaderboards, etc. Is there any hope for minigame support in GameKit?

I've coded a text-adventure game in SwiftUI. (My game has no graphics or sound effects.) My app already supports keyboard navigation; I would like to add support for game controllers on iPhone. I can't figure out how to do it. I especially can't see any way to allow controller users to enter text in a TextField. I've read https://developer.apple.com/documentation/gamecontroller/supporting-game-controllers and it's all about button events. There's no reference to SwiftUI at all in that documentation, or any input-method editing at all. The only mention of "keyboard" is about treating the keyboard itself as if it were a game controller providing button events. How do I implement this?

I guess this is not an actual video Group Lab. sigh Would love to see a dedicated Reality Composer Pro 3 session - even for those of use actual Vision Pro devs that didn't get the lucky golden ticket to the Apple Park event. You need to make these interactions available to us in person during WWDC week. That all being said... How do I go about baking lighting in RCP3? I cannot find anything in the documentation. Thanks!

If you have a "launcher-type app" (a regular macOS app bundle) which is used to launch the actual game executable (a non-bundled executable, such as a Java game), Game Mode does not seem to work as desired in this case. If the launcher app has the appropriate LSSupportsGameMode and LSApplicationCategoryType set, the launcher app will activate Game Mode if fullscreened but the actual game will not (FB18026957). Is there a workaround for this? It also seems like the system hardcodes some apps to act as "game launcher apps" which do have the desired behavior - for example, if you have a sample app set to Steam's bundle identifier (or just use Steam to run the program) and have this architecture where it runs another executable which is the real game, then that game can inherit Game Mode, which is also supported by system logs in Console (note the "launcher" labelReason): default 17:11:05.008533-0400 gamepolicyd Found game GameProcess(Optional("java"), pid=33754, euid=501, labelReason=launcher, launchedBy=Steam) Is there a way to opt into or apply for this behavior for our own apps? I previously asked this on the forums but I'm also wondering if any new developments have occurred since then.

Hi, What's the best way to handle drastic changes in scene charateristics with the new MTLFXTemporalDenoisedScaler? Let's say a visible object of the scene radically changes its material properties. I can modify the albedo and roughness textures consequently. But I suspect the history will be corrupted. Blending visual information between the new frame and the previous ones might be a nonsense. I guess the problem should be the same when objects appear or disappear instantly. Is the upsacler manage these events for us (by lowering blending), or should we use the reactive or the denoise strength mask or something like that to handle them?

Please I would really appreciate a timely response to this: Providing a response and feedback to this: https://origin-devforums.apple.com/forums/thread/827311 and https://developer.apple.com/forums/thread/827043?page=1#889020022 Feedback Report: I have created a feedback report as well like recommended with this ID: FB22823733, with more elaborate images of my implementation, also see here More clarity: Yes, this is for an iOS app(iOS 18+) but testing and debugging on Xcode(MacOS 15.7.7, M4 24GB), which traps or hangs on getting to [PhotogrammetrySample] even while using the lazy sequence and the contentsOf as specified in your docs. I also tried using the PhotogrammetrySession folder run but its still failed with : CoreOC.PhotogrammetrySession.Error 6 and I understand to mean alignment failed. What can be done or do you guys expose any functions that can be used to aid, or handle these internally, can't find any internally. The ObjectMasks are actually segmentation masks from an ML algorithm . To replicate try calling PhotogrammetrySession(input: [PhotogrammetrySample]) with contentsOf as captured on your documentation, even with like 30 image set or is there something I'm missing. Please I would really appreciate a timely response to this

The documentation is unclear. I need a clarification on metal shader converter library distribution. Am I allowed to distribute the library as a part of my macos app bundle?

Hi, I integrated SpriteKit with Box2D version 3, and they work amazingly well together. I shared videos, code, and implementation notes on the project page: SpriteKit Box2D Context SpriteKit built-in physics engine is based on Box2D, likely from an older 2.x generation. This project uses Box2D 3.x directly. Box2D 3.0 was first released in 2024 as a major rewrite with a new C API. It brings improved collision handling, better performance, better stability for demanding simulations, and features that are not exposed through SpriteKit’s SKPhysicsWorld. Box2D 3 is also designed with determinism in mind. SpriteKit's physics implementation is nice to use, but it is not deterministic. See the Determinism section of my SKRenderer Demo for more information about determinism in SpriteKit. Swift & C Because Box2D 3 is written in C, this sample app also shows how to mix Swift and C in the same Xcode project. I wrote a tutorial on how to mix Swift with C in an Xcode project. I hope this is useful to anyone experimenting with SpriteKit physics or external C libraries. Let me know if you have questions or ideas.

Providing clarity and response to this: https://origin-devforums.apple.com/forums/thread/827311 and https://developer.apple.com/forums/thread/827043?page=1#889020022 Feedback Report: I have created a feedback report as well like recommended with this ID: FB22823733, with more elaborate images of my implementation, also see here More clarity: Yes, this is for an iOS app(iOS 18+) but testing and debugging on Xcode(MacOS 15.7.7, M4 24GB), which traps or hangs on getting to [PhotogrammetrySample] even while using the lazy sequence and the contentsOf as specified in your docs. I also tried using the PhotogrammetrySession folder run but its still failed with : CoreOC.PhotogrammetrySession.Error 6 and I understand to mean alignment failed. What can be done or do you guys expose any functions that can be used to aid, or handle these internally, can't find any internally. The ObjectMasks are actually segmentation masks from an ML algorithm . To replicate try calling PhotogrammetrySession(input: [PhotogrammetrySample]) with contentsOf as captured on your documentation, even with like 30 image set or is there something I'm missing.

One of my users updated his iPhone 16 Pro to iOS 26.5, and Apple TV to tvOS 26.5, both of them unable to pair a Microsoft Xbox One (Model 1708) gamepad, before updating (updated from iOS 18.x) it works normally. He also got an old iPhone on iOS 16.7 which was working fine with that gamepad. Any one met the same issue? feedback no. FB22809309

Providing a response and feedback to this: https://origin-devforums.apple.com/forums/thread/827311 and https://developer.apple.com/forums/thread/827043?page=1#889020022 I have a created a feedback/bug report with ID FB22823733 Feedback Report: I have created a feedback report as well like recommended with this ID: FB22823733, with more elaborate images of my implementation, also see here More clarity: Yes, this is for an iOS app(iOS 18+) but testing and debugging on Xcode(MacOS 15.7.7, M4 24GB), which traps or hangs on getting to [PhotogrammetrySample] even while using the lazy sequence and the contentsOf as specified in your docs. I also tried using the PhotogrammetrySession folder run but its still failed with : CoreOC.PhotogrammetrySession.Error 6 and I understand to mean alignment failed. What can be done or do you guys expose any functions that can be used to aid, or handle these internally, can't find any internally. The ObjectMasks are actually segmentation masks from an ML algorithm . To replicate try calling PhotogrammetrySession(input: [PhotogrammetrySample]) with contentsOf as captured on your documentation, even with like 30 image set or is there something I'm missing.

Our application is trying to read all resolutions of an external monitor. We have observed that, for the external monitor there is a mismatch in resolution list in our application and the resolution list in system settings. We are using the apple API "CGDisplayCopyAllDisplayModes" to read the resolutions.

Metal GPU Driver Crash on M5 Pro + macOS 26.5 — kIOGPUCommandBufferCallbackErrorOutOfMemory with <2GB working sets Summary The Metal driver AGXMetalG17X 351.2 on macOS 26.5 (25F71) for the M5 Pro chip crashes with kIOGPUCommandBufferCallbackErrorOutOfMemory (00000008) when running LLM inference workloads with working sets as small as ~1.5GB, despite 24GB of unified memory being available and Apple Diagnostics confirming the hardware is fully functional. This affects multiple tools: MLX, llama.cpp (Metal backend), and native apps using Metal for inference. System Component Value Model MacBook Pro (Mac17,9) Chip Apple M5 Pro (applegpu_g17s) GPU Cores 16 RAM 24 GB LPDDR5 macOS 26.5 (25F71) Metal Metal 4 GPU Driver AGXMetalG17X 351.2 Xcode 26.5 (17F42) Reproduction MLX (Python) pip install mlx mlx-lm python -m mlx_lm.generate \ --model mlx-community/Qwen2.5-3B-Instruct-4bit \ --max-tokens 10 \ --prompt "Hello" Expected: Normal text generation Actual: Crash with: libc++abi: terminating due to uncaught exception of type std::runtime_error: [METAL] Command buffer execution failed: Insufficient Memory (00000008:kIOGPUCommandBufferCallbackErrorOutOfMemory) llama.cpp brew install llama.cpp llama-cli --model model.gguf --prompt "Hello" --n-predict 20 --n-gpu-layers 99 Expected: Fast GPU generation Actual: Process hangs indefinitely Test Results Tool Model Peak Memory Result MLX Qwen2.5-0.5B-4bit 0.36 GB ✅ Works MLX Qwen2.5-1.5B-4bit 0.98 GB ✅ Works MLX Qwen3-1.7B-4bit 1.01 GB ✅ Works MLX Qwen2.5-3B-4bit ~1.5 GB ❌ Metal OOM crash MLX Qwen3-4B-4bit ~2.1 GB ❌ Metal OOM crash MLX Qwen3-8B-4bit ~4.5 GB ❌ Metal OOM crash llama.cpp Qwen2.5-0.5B GGUF ~0.5 GB ❌ Hangs with GPU llama.cpp Qwen2.5-0.5B GGUF ~0.5 GB ✅ Works with CPU only Key Evidence Hardware is healthy — Apple Diagnostics passed all tests Basic Metal works — matmul, array ops work fine CPU inference works — llama.cpp with -ngl 0 runs correctly The error is NOT about actual memory exhaustion — kIOGPUCommandBufferCallbackErrorOutOfMemory means the kernel rejects the Metal memory commit, not that physical memory is full. The system reports 17.76GB available for Metal working set. Crash Log Extract Thread 31 Crashed: 0 libsystem_kernel.dylib __pthread_kill + 8 1 libsystem_pthread.dylib pthread_kill + 296 2 libsystem_c.dylib abort + 148 3 Metal MTLReportFailure.cold.1 + 48 4 Metal MTLReportFailure + 576 5 Metal -[_MTLCommandBuffer addCompletedHandler:] + 104 ... Exception Type: EXC_CRASH (SIGABRT) Termination Reason: Namespace SIGNAL, Code 6, Abort trap: 6 Related Issues ml-explore/mlx#3586 — Metal compiler regression on macOS 26.5 ml-explore/mlx#3534 — M5 float32 precision issue ml-explore/mlx#3568 — M5 random divergence ml-explore/mlx#3539 — Metal residency OOM (M4 Max) Request Please investigate the AGXMetalG17X driver for M5 Pro on macOS 26.5. The driver appears to incorrectly reject Metal memory commits for LLM inference workloads, even when the working set is well within the system's reported limits (1.5GB requested vs 17.76GB available). Happy to provide full crash logs, sysdiagnose archives, or run additional tests.

Hi all, I'm working on updating my audio visualizer app. I'm adding new visualizers based on Metal 4 compute shaders. They worked in iOS 26.4 and iOS 26.5 up until beta 3. However, after that, the visualizers started crashing the phone and forcing a restart. On the latest version of iOS 26.5, the crash is still there. I submitted feedback, but haven't heard anything back just yet. I was wondering if others have faced this same issue, and if there are any workarounds. Here is my repo if you want to look at the code (forgive me if it's sloppy, I'm quite new to graphics programming and Metal): https://github.com/aabagdi/VisualMan/tree/main Thank you!

In my turn-based game, I receive GKListener event receivedTurnEventForMatch and decode the match.matchData. On occasion, the matchData is clearly stale and is from the previous turn. If I call the MatchMaker ViewController up and select that same match, the data is not stale, so it's not a matter of not calling endTurn. I have tried both loadMatchWithID and loadMatchesWithCompletionHandler after receiving the receivedTurnEventForMatch, but the data is still stale. Advice?

Hi, Can Apple provide some reassurance on the continued support for SpriteKit in the coming years? At WWDC 2025, Apple deprecated SceneKit, which was alarming for a SpriteKit fan and user. I'm aware Apple doesn't comment on future plans. However, committing to a specific framework for larger projects is a significant investment in time (and money). Developers need to know whether to continue investing in proprietary technology that could be deprecated at any time. Some reassurance regarding SpriteKit's future would go a long way. I have moved my project from SpriteKit to RealityKit. RealityKit is certainly interesting, but I'd gladly switch back to SpriteKit. SpriteKit has things going for it that no other Apple framework provides: Live drawing of shapes, paths, text, video, and images at 120 FPS across all Apple devices. RealityKit offers no procedural drawing API equivalent to shape nodes. Particles and physics fields are very easy and fun to play with. The rigid body engine has a good feel. By comparison, RealityKit sleeps too aggressively, and doesn't have spring joints. SpriteKit is lightweight and runs well on older devices. In an ideal world, I'd also love to see SpriteKit improved: A deterministic physics engine Metal shader support Soft shadows for lights One can dream. Thank you!