Context I’m deploying large language models on iPhone using llama.cpp. A new iPhone Air (12 GB RAM) reports a Metal MTLDevice.recommendedMaxWorkingSetSize of 8,192 MB, and my attempt to load Llama-2-13B Q4_K (~7.32 GB weights) fails during model initialization. Environment Device: iPhone Air (12 GB RAM) iOS: 26 Xcode: 26.0.1 Build: Metal backend enabled llama.cpp App runs on device (not Simulator) What I’m seeing MTLCreateSystemDefaultDevice().recommendedMaxWorkingSetSize == 8192 MiB Loading Llama-2-13B Q4_K (7.32 GB) fails to complete. Logs indicate memory pressure / allocation issues consistent with the 8 GB working-set guidance. Smaller models (e.g., 7B/8B with similar quantization) load and run (8B Q4_K provide around 9 tokens/second decoding speed). Questions Is 8,192 MB an expected recommendedMaxWorkingSetSize on a 12 GB iPhone? What values should I expect on other 2025 devices including iPhone 17 (8 GB RAM) and iPhone 17 Pro (12 GB RAM) Is it strictly enforced by Metal allocations (heaps/buffers), or advisory for best performance/eviction behavior? Can a process practically exceed this for long-lived buffers without immediate Jetsam risk? Any guidance for LLM scenarios near the limit?

Hello! I'm a developer working on a plugin for the Elgato Stream Deck, called GPU Metrics. The plugin currently only works on Windows but I'd like to bring it to macOS. However, based on forum posts I've read (and StackOverflow) there isn't a very clear path to query GPU metrics like usage, temperature, used GPU memory, and power consumption. There are some tools out there that do similar things, but I wanted to see what would be the recommendation from Apple's engineering team to get this data via a public API. Requirements: Access GPU utilization, temperature, memory usage, power usage C/C++ based API for querying the metrics so I can expose the data to JavaScript via Node Addon No need to compatibile with Intel-based Macs, as Apple silicon will be fine for now Plugin GitHub Thank you! Noah

After watching WWDC 2025 session "Combine Metal 4 machine learning and graphics", I have decided to give it a shot to integrate the latest MTL4MachineLearningCommandEncoder to my existing render pipeline. After a lot of trial and errors, I managed to set up the pipeline and have the app compiled. However, I am now stuck on creating a MTLLibrary with .mtlpackage. Here is the code I have to create a MTLLibrary according the WWDC session https://developer.apple.com/videos/play/wwdc2025/262/?time=550: let coreMLFilePath = bundle.path(forResource: "my_model", ofType: "mtlpackage")! let coreMLURL = URL(string: coreMLFilePath)! do { metalDevice.makeLibrary(URL: coreMLURL) } catch { print("error: \(error)") } With the above code, I am getting error: Error Domain=MTLLibraryErrorDomain Code=1 "Invalid metal package" UserInfo={NSLocalizedDescription=Invalid metal package} What is the correct way to create a MTLLibrary with .mtlpackage? Do I see this error because the .mtlpackage I am using is incorrect? How should I go with debugging this? I'd really appreciate if I could get some help on this as I have been stuck with it for some time now. Thanks in advance!

Hello Apple team, I'm working on an iOS AR app using SwiftUI and RealityKit, and I was wondering if the Cinematic API can be used with a RealityKit scene. I’d like to achieve a shallow depth of field while keeping the 3D asset in focus, and vice versa. Thanks!

Hi, I have an Unity game. I need to have multiple App Icons for my game for it to be able to be recognized in different countries. In other words, is it possible to have an iOS app in which the App Icon changes based on device locale/language? On Android this is possible using Unity Localization package "com.unity.localization"

after launching a nearby exoerience on quick look or inside our app, all the user in the group watch sometimes teh model blinking abd becoming transparet... ... just one user hasnt the issue, either the one who launched shareplay or the user who force align the immersive space in front weird

RealityKit spatial audio crackles and pops on iOS 26.0 beta 5. It works correctly on iOS 18.6 and visionOS 26.0 beta 5. The APIs used are AudioPlaybackController, Entity.prepareAudio, Entity.play Videos of the expected and observed behavior are attached to the feedback FB19423059. The audio should be a consistent, repeating sound, but it seems oddly abbreviated and the volume varies unexpectedly. Thank you for investigating this issue.

My IOS app generates pdf files. Every time my users open the generated pdf files, the autofill popup jumps out, but my pdf file is NOT for interacting. I'm here to ask if there's a way to mark my pdf files as "not a form", like in metadata or anywhere else?

我们想在游戏类 App 内接入 Game Center。用户可以在游戏内创建多个角色，若用户在游戏内创建了2个角色：角色1、角色2，请问： 当用户将角色1与 Game Center 绑定后，数据将上报至 Game Center。此时玩家想要将角色1与 Game Center 解除绑定，解绑后，再将角色2与 Game Center 绑定。那么这时角色1的数据是留存在 Game Center 中，还是将被移除？

I do not understand how offline leaderboard submission is supposed to work in Game Kit: While the documentation briefly states that offline submission is supported, how is that even possible when you first have to fetch a leaderboard object in order to then call its submitScore function? How can I get the leaderboard object in the first place when offline? Can anyone enlighten me how this works? Or maybe point me to some relevant documentation?

Hi all, I've developed some code that enables an arcball camera interaction with my scene. I've done this using components and systems. The implementation feels a bit messy as I've got gesture code on my realityView, and then a bunch of other code that uses those gesture inputs in my component and system. Is there a demo app, or some example code that shows a nice way to encapsulate these things in to one item for custom cameras, something like Apple's .realityViewCameraControls(.orbit) If not can anyone recommend an approach to take?

How can one match the walls and floor of a given CapturedRoom ? The transform.eulerAngles of a floor z & y are always 0 ! And the polygons seems to have a different orientation than the walls. So how to figure out the rotation and match the one from the walls ?

I am puzzled by the setAddress(_:attributeStride:index:) of MTL4ArgumentTable. Can anyone please explain what the attributeStride parameter is for? The doc says that it is "The stride between attributes in the buffer." but why? Who uses this for what? On the C++ side in the shaders the stride is determined by the C++ type, as far as I know. What am I missing here? Thanks!

Hello, I am currently working on a research project under ENINCA Consulting, focused on advanced diagnostic tools for pseudorandom number generators (structural metrics, multi-seed stability, cross-architecture reproducibility, and complementary indicators to TestU01). To validate this diagnostic framework, I prototyped a small non-linear 64-bit PRNG (not as a goal in itself, but simply as a vehicle to test the methodology). During these evaluations, I observed something interesting on Apple Silicon (Mac M1): • bit-exact reproducibility between M1 ARM CPU and M1 Metal GPU, • full BigCrush pass on both CPU and Metal backends, • excellent p-values, • stable behaviour across multiple seeds and runs. This was not the intended objective, the goal was mainly to validate the diagnostic concepts, but these results raised some questions about deterministic compute behaviour in Metal. My question: Is there any official guidance on achieving (or expecting) deterministic RNG or compute behaviour across CPU ↔ Metal GPU on Apple Silicon? More specifically: • Are deterministic compute kernels expected or guaranteed on Metal for scientific workloads? • Are there recommended patterns or best practices to ensure reproducibility across GPU generations (M1 → M2 → M3 → M4)? • Are there known Metal features that can introduce non-determinism? I am not sharing the internal recurrence (this work is proprietary), but I can discuss the high-level diagnostic observations if helpful. Thank you for any insight, very interested in how the Metal engineering team views deterministic compute patterns on Apple Silicon. Pascal ENINCA Consulting

The “explore spatial accessory input on visionOS” presentation from WDC25 interests me. I bought both the MUSE Logitech stylus and the PS VR2 sense controllers to try out with the sculpting app presented by the author, engineer Amanda Han. Unfortunately the app itself was not included. Could the app be made available for downloading as well as the Xcode project? I appreciate any assistance the author and your team could provide. Thank you.

Create the QRCode CIFilter<CIBlendWithMask> *f = CIFilter.QRCodeGenerator; f.message = [@"Message" dataUsingEncoding:NSASCIIStringEncoding]; f.correctionLevel = @"Q"; // increase level CIImage *qrcode = f.outputImage; Overlay the icon CIImage *icon = [CIImage imageWithURL:url]; CGAffineTransform *t = CGAffineTransformMakeTranslation( (qrcode.extent.width-icon.extent.width)/2.0, (qrcode.extent.height-icon.extent.height)/2.0); icon = [icon imageByApplyingTransform:t]; qrcode = [icon imageByCompositingOver:qrcode]; Round off the corners static dispatch_once_t onceToken; static CIWarpKernel *k; dispatch_once(&onceToken, ^ { k = [CIWarpKernel kernelWithFunctionName:name fromMetalLibraryData:metalLibData() error:nil]; }); CGRect iExtent = image.extent; qrcode = [k applyWithExtent:qrcode.extent roiCallback:^CGRect(int i, CGRect r) { return CGRectInset(r, -radius, -radius); } inputImage:qrcode arguments:@[[CIVector vectorWithCGRect:qrcode.extent], @(radius)]]; …and this code for the kernel should go in a separate .ci.metal source file: float2 bend_corners (float4 extent, float s, destination dest) { float2 p, dc = dest.coord(); float ratio = 1.0; // Round lower left corner p = float2(extent.x+s,extent.y+s); if (dc.x < p.x && dc.y < p.y) { float2 d = abs(dc - p); ratio = min(d.x,d.y)/max(d.x,d.y); ratio = sqrt(1.0 + ratio*ratio); return (dc - p)*ratio + p; } // Round lower right corner p = float2(extent.x+extent.z-s, extent.y+s); if (dc.x > p.x && dc.y < p.y) { float2 d = abs(dc - p); ratio = min(d.x,d.y)/max(d.x,d.y); ratio = sqrt(1.0 + ratio*ratio); return (dc - p)*ratio + p; } // Round upper left corner p = float2(extent.x+s,extent.y+extent.w-s); if (dc.x < p.x && dc.y > p.y) { float2 d = abs(dc - p); ratio = min(d.x,d.y)/max(d.x,d.y); ratio = sqrt(1.0 + ratio*ratio); return (dc - p)*ratio + p; } // Round upper right corner p = float2(extent.x+extent.z-s, extent.y+extent.w-s); if (dc.x > p.x && dc.y > p.y) { float2 d = abs(dc - p); ratio = min(d.x,d.y)/max(d.x,d.y); ratio = sqrt(1.0 + ratio*ratio); return (dc - p)*ratio + p; } return dc; }

Description: I'm developing an AR effect using SceneKit and applying a transparent material to a face mesh. However, I'm facing an issue where the front faces of the mesh overlap each other, causing incorrect rendering. Problem: The front faces of the mesh overlap with each other when transparency is applied. This causes areas like the cheeks to be visible through the nose, even though they should be occluded. Expected Behavior: The material should behave as if it were opaque to itself—that is, overlapping front faces should be occluded properly, while still allowing transparency for background elements. Actual Behavior: The mesh renders its own front faces incorrectly, making parts of the face visible through others when they should be blocked. What I Have Tried: testMaterial.writesToDepthBuffer = true testMaterial.readsFromDepthBuffer = true Question: 👉 How can I prevent SceneKit's transparent material from rendering overlapping front faces? 👉 Is there a way to force SceneKit to treat its own mesh as opaque for itself while still being transparent to the background? 👉 Does SceneKit support a proper depth pre-pass or an equivalent to Unity’s ZWrite shaders to solve this issue? Attached screenshots demonstrate the problem visually. Any help would be greatly appreciated! 🚀

Hi, I am working with a large project. We are compiling each material to its own .metallib. They all include many common files full of inline functions. Finally we link it all together at the end with a single big pathtrace kernel. Everything works as expected, however the compile times have gotten completely out of hand and it takes multiple minutes to compile at runtime (to native code). I have gathered that I can do this offline by using metal-tt however if I am wondering if there is a way to reduce the compile times in such a scenario, and how to investigate what the root cause of the problem is. I suspect it could have to do with the fact that every materials metallib contains duplications of all the inline functions. Any ideas on how to profile and debug this? Thanks, Rasmus

Hi, I'm working with a very simple app that tries to read a coordinates card and past the data into diferent fields. The card's layout is COLUMNS from 1-10, ROWs from A-J and a two digit number for each cell. In my app, I have field for each of those cells (A1, A2...). I want that OCR to read that card and paste the info but I just cant. I have two problems. The camera won't close. It remains open until I press the button SAVE (this is not good because a user could take 3, 4, 5... pictures of the same card with, maybe, different results, and then? Which is the good one?). Then, after I press save, I can see the OCR kinda works ( the console prints all the date read) but the info is not pasted at all. Any idea? I know is hard to know what's wrong but I've tried chatgpt and all it does... just doesn't work This is the code from the scanview import SwiftUI import Vision import VisionKit struct ScanCardView: UIViewControllerRepresentable { @Binding var scannedCoordinates: [String: String] var useLettersForColumns: Bool var numberOfColumns: Int var numberOfRows: Int @Environment(.presentationMode) var presentationMode func makeUIViewController(context: Context) -&gt; VNDocumentCameraViewController { let scannerVC = VNDocumentCameraViewController() scannerVC.delegate = context.coordinator return scannerVC } func updateUIViewController(_ uiViewController: VNDocumentCameraViewController, context: Context) {} func makeCoordinator() -&gt; Coordinator { return Coordinator(self) } class Coordinator: NSObject, VNDocumentCameraViewControllerDelegate { let parent: ScanCardView init(_ parent: ScanCardView) { self.parent = parent } func documentCameraViewController(_ controller: VNDocumentCameraViewController, didFinishWith scan: VNDocumentCameraScan) { print("Escaneo completado, procesando imagen...") guard scan.pageCount &gt; 0, let image = scan.imageOfPage(at: 0).cgImage else { print("No se pudo obtener la imagen del escaneo.") controller.dismiss(animated: true, completion: nil) return } recognizeText(from: image) DispatchQueue.main.async { print("Finalizando proceso OCR y cerrando la cámara.") controller.dismiss(animated: true, completion: nil) } } func documentCameraViewControllerDidCancel(_ controller: VNDocumentCameraViewController) { print("Escaneo cancelado por el usuario.") controller.dismiss(animated: true, completion: nil) } func documentCameraViewController(_ controller: VNDocumentCameraViewController, didFailWithError error: Error) { print("Error en el escaneo: \(error.localizedDescription)") controller.dismiss(animated: true, completion: nil) } private func recognizeText(from image: CGImage) { let request = VNRecognizeTextRequest { (request, error) in guard let observations = request.results as? [VNRecognizedTextObservation], error == nil else { print("Error en el reconocimiento de texto: \(String(describing: error?.localizedDescription))") DispatchQueue.main.async { self.parent.presentationMode.wrappedValue.dismiss() } return } let recognizedStrings = observations.compactMap { observation in observation.topCandidates(1).first?.string } print("Texto reconocido: \(recognizedStrings)") let filteredCoordinates = self.filterValidCoordinates(from: recognizedStrings) DispatchQueue.main.async { print("Coordenadas detectadas después de filtrar: \(filteredCoordinates)") self.parent.scannedCoordinates = filteredCoordinates } } request.recognitionLevel = .accurate let handler = VNImageRequestHandler(cgImage: image, options: [:]) DispatchQueue.global(qos: .userInitiated).async { do { try handler.perform([request]) print("OCR completado y datos procesados.") } catch { print("Error al realizar la solicitud de OCR: \(error.localizedDescription)") } } } private func filterValidCoordinates(from strings: [String]) -&gt; [String: String] { var result: [String: String] = [:] print("Texto antes de filtrar: \(strings)") for string in strings { let trimmedString = string.replacingOccurrences(of: " ", with: "") if parent.useLettersForColumns { let pattern = "^[A-J]\\d{1,2}$" // Letras de A-J seguidas de 1 o 2 dígitos if trimmedString.range(of: pattern, options: .regularExpression) != nil { print("Coordenada válida detectada (letras): \(trimmedString)") result[trimmedString] = "Valor" // Asignación de prueba } } else { let pattern = "^[1-9]\\d{0,1}$" // Solo números, de 1 a 99 if trimmedString.range(of: pattern, options: .regularExpression) != nil { print("Coordenada válida detectada (números): \(trimmedString)") result[trimmedString] = "Valor" } } } print("Coordenadas finales después de filtrar: \(result)") return result } } }

Problem Description I'm encountering an issue with SCNTechnique where the clearColor setting is being ignored when multiple passes share the same depth buffer. The clear color always appears as the scene background, regardless of what value I set. The minimal project for reproducing the issue: https://www.dropbox.com/scl/fi/30mx06xunh75wgl3t4sbd/SCNTechniqueCustomSymbols.zip?rlkey=yuehjtk7xh2pmdbetv2r8t2lx&st=b9uobpkp&dl=0 Problem Details In my SCNTechnique configuration, I have two passes that need to share the same depth buffer for proper occlusion handling: "passes": [ "box1_pass": [ "draw": "DRAW_SCENE", "includeCategoryMask": 1, "colorStates": [ "clear": true, "clearColor": "0 0 0 0" // Expecting transparent black ], "depthStates": [ "clear": true, "enableWrite": true ], "outputs": [ "depth": "box1_depth", "color": "box1_color" ], ], "box2_pass": [ "draw": "DRAW_SCENE", "includeCategoryMask": 2, "colorStates": [ "clear": true, "clearColor": "0 0 0 0" // Also expecting transparent black ], "depthStates": [ "clear": false, "enableWrite": false ], "outputs": [ "depth": "box1_depth", // Sharing the same depth buffer "color": "box2_color", ], ], "final_quad": [ "draw": "DRAW_QUAD", "metalVertexShader": "myVertexShader", "metalFragmentShader": "myFragmentShader", "inputs": [ "box1_color": "box1_color", "box2_color": "box2_color", ], "outputs": [ "color": "COLOR" ] ] ] And the metal shader used to display box1_color and box2_color with splitting: fragment half4 myFragmentShader(VertexOut in [[stage_in]], texture2d<half, access::sample> box1_color [[texture(0)]], texture2d<half, access::sample> box2_color [[texture(1)]]) { half4 color1 = box1_color.sample(s, in.texcoord); half4 color2 = box2_color.sample(s, in.texcoord); if (in.texcoord.x < 0.5) { return color1; } return color2; }; Expected Behavior Both passes should clear their color targets to transparent black (0, 0, 0, 0) The depth buffer should be shared between passes for proper occlusion Actual Behavior Both box1_color and box2_color targets contain the scene background instead of being cleared to transparent (see attached image) This happens even when I explicitly set clearColor: "0 0 0 0" for both passes Setting scene.background.contents = UIColor.clear makes the clearColor work as expected, but I need to keep the scene background for other purposes What I've Tried Setting different clearColor values - all are ignored when sharing depth buffer Using DRAW_NODE instead of DRAW_SCENE - didn't solve the issue Creating a separate pass to capture the background - the background still appears in the other passes Various combinations of clear flags and render orders Environment iOS/macOS, running with "My Mac (Designed for iPad)" Xcode 16.2 Question Is this a known limitation of SceneKit when passes share a depth buffer? Is there a workaround to achieve truly transparent clear colors while maintaining a shared depth buffer for occlusion testing? The core issue seems to be that SceneKit automatically renders the scene background in every DRAW_SCENE pass when a shared depth buffer is detected, overriding any clearColor settings. Any insights or workarounds would be greatly appreciated. Thank you!