Hi Apple Engineers, I am experiencing a potential memory management bug with CoreML on M1 Mac (32GB Unified Memory). When processing long video files (approx. 12,000 frames) using a CoreML execution provider, the system often completes the 'Analysing' phase but fails to transition into 'Processing'. It simply exits silently or hits an import error (scipy). However, if I split the same task into small 20-frame segments, it works perfectly at high speeds (~40 FPS). This suggests the hardware is capable, but there is an issue with memory fragmentation or resource cleanup during long-running CoreML sessions. Is there a way to force a VRAM/Unified Memory flush via CLI, or is this a known limitation for large frame indexing?

Hello, I am interested in using jax-metal to train ML models using Apple Silicon. I understand this is experimental. After installing jax-metal according to https://developer.apple.com/metal/jax/, my python code fails with the following error JaxRuntimeError: UNKNOWN: -:0:0: error: unknown attribute code: 22 -:0:0: note: in bytecode version 6 produced by: StableHLO_v1.12.1 My issue is identical to the one reported here https://github.com/jax-ml/jax/issues/26968#issuecomment-2733120325, and is fixed by pinning to jax-metal 0.1.1., jax 0.5.0 and jaxlib 0.5.0. Thank you!

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

Hi all, I'm trying to install the Xcode 26 beta on Tahoe (M4 Mac mini). I tried the latest beta 6 both the universal and the apple silicon version, but at startup the system says the app is corrupted and will be deleted. I tried this multiple times - always with a fresh download. Then I downloaded the beta 5, which told me at startup that there is no macOS SDK available and then Xcode also quits. I want to update/test my macOS apps for the upcoming Tahoe and want to adapt for the new design. On my Mac is the Xcode version 16.4 installed and this is working correctly. In the last years I started with earlier versions of Xcode betas, but this year I'm a little bit more late. Are there any steps I missed or is this a known issue yet? I had never before so much trouble installing any beta. What can I do? Best regards, Jürgen Terpe

Hi, I am curious about if hyperthreading is enabled/disabled on my macbook pro M1 or M4. Howto figure out? I am using macOS 15.5. Further, I develop a multi-threaded audio sequencer that creates threads per instrument. I use vector operations to increase performance. I recognized lowering synchronization rate from 250 Hz to 60 Hz gives additional performance advantages. Howto programmatically check if Hyperthreading is enabled/disabled and howto enable/disable it programmatically? After some research I found sysctl() and nvram SMTDisable=%01. https://support.apple.com/en-us/101870 Can anyone provide me an Objective C example? regards, Joël

The attached file bellow contains the full error error I clone this repo to my mac, change team id and group, and run it in Xcode: https://github.com/protonpass/ios-pass There's no issue when I ran it with the Debug configuration, but when I go to Product > Scheme > Edit Scheme and change the iOS target build configuration to Release then I got that error above. I have tried Archive and export the ipa, verify that the provisioning profile contains my Mac UDID, but when double clicking the ipa to install, I also got the error This app cannot be installed because its integrity could not be verified.

I am having issues deploying my iOS app, that uses ShazamKit, to get working on a Mac with Apple silicon. When uploading the archive to App Store Connect I do get ITMS-90863: Macs with Apple silicon support issue - The app links with libraries that aren’t present in macOS: /usr/lib/swift/libswiftShazamKit.dylib Is ShazamKit not supported for iOS apps that can run on Macs with Apple silicon? Or is there something I should fix in my setup / deployment?

I am worried that Rosetta 2 will eventually be removed. I rely on it to run x86-64 Docker containers as well as Windows games through the Game Porting Toolkit. I also use CrossOver very often, which relies on Rosetta. I also use an old version of MuseScore that needs it, and every now and then, I download a legacy Intel Mac app. Thus, I'm worried that Apple will no longer offer it for download in future macOS versions. Are there any plans to remove Rosetta? Since Rosetta is only installed on demand, I have a slight bit of reassurance that it might be offered for download indefinitely, since most casual users naturally won't install it in the future as most general consumer apps are compiled natively for ARM, and the on-demand install cuts down on the bloat included with macOS by default. I hope Rosetta could be offered indefinitely, since many pro-users and gamers rely on it very often, even 10 years from now when Intel Macs are completely unsupported. If Rosetta is removed, I might have to switch back to Windows for many tasks, so I really hope for the continued offering of Rosetta.

Investigating a kernel panic, I discovered that Apple Silicon Panic traces are not working with how I know to symbolicate the panic information. I have not found proper documentation that corrects this situation. Attached file is an indentity-removed panic, received from causing an intentional panic (dereferencing nullptr), so that I know what functions to expect in the call stack. This is cut-and-pasted from the "Report To Apple" dialog that appears after the reboot: panic_1_4_21_b.txt To start, I download and install the matching KDK (in this case KDK_14.6.1_23G93.kdk), identified from this line: OS version: 23G93 Kernel version: Darwin Kernel Version 23.6.0: Mon Jul 29 21:14:04 PDT 2024; root:xnu-10063.141.2~1/RELEASE_ARM64_T8122 Then start lldb from Terminal, using this command: bash_prompt % lldb -arch arm64e /Library/Developer/KDKs/KDK_14.6.1_23G93.kdk/System/Library/Kernels/kernel.release.t8122 Next I load the remaining scripts per the instructions from lldb: (lldb) settings set target.load-script-from-symbol-file true I need to know what address to load my kext symbols to, which I read from this line of the panic log, after the @ symbol: com.company.product(1.4.21d119)[92BABD94-80A4-3F6D-857A-3240E4DA8009]@0xfffffe001203bfd0->0xfffffe00120533ab I am using a debug build of my kext, so the DWARF symbols are part of the binary. I use this line to load the symbols into the lldb session: (lldb) addkext -F /Library/Extensions/KextName.kext/Contents/MacOS/KextName 0xfffffe001203bfd0 And now I should be able to use lldb image lookup to identify pointers on the stack that land within my kext. For example, the current PC at the moment of the crash lands within the kext (expected, because it was intentional): (lldb) image lookup -a 0xfffffe001203fe10 Which gives the following incorrect result: Address: KextName[0x0000000000003e40] (KextName.__TEXT.__cstring + 14456) Summary: "ffer has %d retains\n" That's not even a program instruction - that's within a cstring. No, that cstring isn't involved in anything pertaining to the intentional panic I am expecting to see. Can someone please explain what I'm doing wrong and provide instructions that will give symbol information from a panic trace on an Apple Silicon Mac? Disclaimers: Yes I know IOPCIFamily is deprecated, I am in process of transitioning to DriverKit Dext from IOKit kext. Until then I must maintain the kext. Terminal command "atos" provides similar incorrect results, and seems to not work with debug-built-binaries (only dSYM files) Yes this is an intentional panic so that I can verify the symbolicate process before I move on to investigating an unexpected panic I have set nvram boot-args to include keepsyms=1 I have tried (lldb) command script import lldb.macosx but get a result of error: no images in crash log (after the nvram settings)

Today, I submitted the following proposal to Apple through the Feedback Assistant app. I'm not confident in how I phrased it—I'd appreciate any thoughts or feedback from fellow developers. Proposal: "Apple Silicon.app" for macOS with Apple Silicon – Enhancing Performance and Swap Memory Control This suggestion has been machine translated into English, so there may be some discrepancies. If you need the actual text, please feel free to reply to this or ask at the email address below. <mail address> If the author of this proposal is to be credited, I would appreciate being listed under the nickname “DiamondGotCat,” where possible. Summary: - Currently, Apple Silicon-equipped Macs have many system-level features locked down or restricted. - This proposal suggests a new application that enables certain advanced controls for power users. - Tentatively named "Apple Silicon.app", the name may be subject to change if a more suitable alternative arises. - I propose this application be added as a pre-installed utility on compatible systems: macOS (M-series), iPadOS (A-series and M-series), and iOS (iPhones with Apple-designed A5 and newer chips, provided the latest OS is available for them). Overview: This proposal introduces "Apple Silicon.app", a new system-level utility designed to offer power users greater flexibility and control over Apple Silicon behavior, as part of a broader feature update for Apple devices. I propose that Apple Silicon.app be automatically installed as a pre-installed application on Apple Silicon devices (Mac, iPad, iPhone, and select Vision devices) that support upcoming major system updates. Suggested Features of Apple Silicon.app: 1. Performance Core Control To describe this functionality, the following terminology will be used: - P-cores: Performance cores - E-cores: Efficiency cores I understand that Apple Silicon emphasizes energy efficiency, but I believe there are users—myself included—who prioritize maximum performance regardless of power usage. Therefore, I propose that the app offer a drop-down menu with the following six modes for performance core usage: A. Automatic (Recommended) – Default macOS behavior; automatically switches between P/E cores based on workload. B. Performance Priority – Prioritizes P-cores for high-demand tasks, restricts E-cores. Ideal for developers, video editors. C. Power Saving – Uses E-cores only whenever possible; limits P-core usage. Great for battery saving. D. P-Core Exclusive Mode – User-defined processes always run on P-cores. Suitable for benchmarks or low-latency tasks. E. E-Core Exclusive Mode – Prioritizes background tasks and thermal efficiency. F. Manual Assignment (Advanced) – Users can manually assign P/E cores per application in a dedicated settings screen. Additionally, I propose the following optional checkbox settings: - Thermal Safety Mode: Automatically switches from P- to E-cores when system heat exceeds a threshold. - Restore Core Settings on Wake: Remembers P/E settings after sleep/wake. - Power Source Adaptive Mode: Switches to power-saving on battery, and performance mode when plugged in. 2. Swap Memory Configuration The app should also enable user-level control over swap memory (i.e., using part of the SSD as virtual memory). Currently, macOS manages swap space automatically with no user customization available. I propose the ability to manually configure the swap system with the following options: - Enable Manual Configuration: Checkbox to switch from automatic to manual control. - Swap Size: Adjustable in GB units, allowing users to allocate desired swap capacity. 3. Other Settings At this point, these are the core features I propose. If additional useful features exist that align with this concept, I welcome further suggestions or expansion. As users—at least speaking for myself—we look forward to such customization options becoming available.

Hello, I have an iOS app that is recording audio that is working fine on iPads/iPhones. It asks for microphone permission and after that recording works. I installed the same app on my M3 MacBook via TestFlight since iPad apps are supposed to work without a change that way. The app starts fine and everything, but it never asks for Microphone permission, so I can't record. Do I need to do something to make this happen (this is not macCatalyst, its running the arm64 iPhone binary on macOS) thanks

I'm looking for technical documentation on Apple's IEEE 754 implementation on the M1 - M4 chips. Specifically, I'd like to know how NaN payloads are handled in arithmetic operations. Does anyone know where this information is available? Thank you!

We are experiencing problems with the USB port on iPad Pro 11 inch (M4) model number MVW13NF/A. Our custom peripheral device (based on Raspberry Pi Pico + tinyUSB stack, is configured as a network adapter class and has communication with our App over UDP protocol. Our device also acts as a DHCP server, providing the IP address for iPad. The problem can be described as a “bus stall” or "bus hold" after sleep mode. To reproduce it we just send the iPad into sleep mode using a power button, the USB bus on M4 goes to the suspended state and won’t resume anymore when we wake the iPad up. The problem has occurred since the upgrade to iOS 18.2.1 and has not been observed before on the previously installed iOS 17 on the same iPad Pro M4. Also, the problem does not happen on the iPad Pro 11 inch (3rd gen with M1) model number MHW73FD/A, with the same iOS 18.2.1 installed. The problem also does not arise, if we connect our device via USB hub to the same iPad Pro M4. We have tested different versions of tinyUSB stack (either included in RPi Pico SDK or native unpatched). The problem is independent of the library version. It occurs always if our device is connected directly to the USB port of iPad Pro (M4) with iOS 18. It also stays after upgrading to the latest iOS 18.3 (beta) In the attached logs is (reduced for clarity) debug output from tinyUSB library about events on the USB bus. These logs were captured via RTT debugging output, using Segger J-Link debugger, so logging process does not affect the timings on the USB bus. There are three logs attached, for cases 1: "iPad Pro M4 + iOS18" (i.e. problematic case), 2: "iPad Pro M1 + iOS18", and 3: "iPad Pro M4 + iOS18 + external USB hub" (they are non-problematic cases). case1_usbd_log.txt case2_usbd_log.txt case3_usbd_log.txt This was already posted as feedback with id FB16366509

Issue type: Bug TensorFlow metal version: 1.1.1 TensorFlow version: 2.18 OS platform and distribution: MacOS 15.2 Python version: 3.11.11 GPU model and memory: Apple M2 Max GPU 38-cores Standalone code to reproduce the issue: import tensorflow as tf if __name__ == '__main__': gpus = tf.config.experimental.list_physical_devices('GPU') print(gpus) Current behavior Apple silicone GPU with tensorflow-metal==1.1.0 and python 3.11 works fine with tensorboard==2.17.0 This is normal output: /Users/mspanchenko/anaconda3/envs/cryptoNN_ml_core/bin/python /Users/mspanchenko/VSCode/cryptoNN/ml/core_second_window/test_tensorflow_gpus.py [PhysicalDevice(name='/physical_device:GPU:0', device_type='GPU')] Process finished with exit code 0 But if I upgrade tensorflow to 2.18 I'll have error: /Users/mspanchenko/anaconda3/envs/cryptoNN_ml_core/bin/python /Users/mspanchenko/VSCode/cryptoNN/ml/core_second_window/test_tensorflow_gpus.py Traceback (most recent call last): File "/Users/mspanchenko/VSCode/cryptoNN/ml/core_second_window/test_tensorflow_gpus.py", line 1, in <module> import tensorflow as tf File "/Users/mspanchenko/anaconda3/envs/cryptoNN_ml_core/lib/python3.11/site-packages/tensorflow/__init__.py", line 437, in <module> _ll.load_library(_plugin_dir) File "/Users/mspanchenko/anaconda3/envs/cryptoNN_ml_core/lib/python3.11/site-packages/tensorflow/python/framework/load_library.py", line 151, in load_library py_tf.TF_LoadLibrary(lib) tensorflow.python.framework.errors_impl.NotFoundError: dlopen(/Users/mspanchenko/anaconda3/envs/cryptoNN_ml_core/lib/python3.11/site-packages/tensorflow-plugins/libmetal_plugin.dylib, 0x0006): Symbol not found: __ZN3tsl8internal10LogMessageC1EPKcii Referenced from: <D2EF42E3-3A7F-39DD-9982-FB6BCDC2853C> /Users/mspanchenko/anaconda3/envs/cryptoNN_ml_core/lib/python3.11/site-packages/tensorflow-plugins/libmetal_plugin.dylib Expected in: <2814A58E-D752-317B-8040-131217E2F9AA> /Users/mspanchenko/anaconda3/envs/cryptoNN_ml_core/lib/python3.11/site-packages/tensorflow/python/_pywrap_tensorflow_internal.so Process finished with exit code 1

Currently, I am working on an iOS app with a Deployment Target set to iOS 15.0, and macOS 12.0. The app is allowed to run on Macs with Apple Silicon. A customer with a Mac running macOS Monterey (12) is complaining that in the TestFlight app, they cannot install the app since it shows "Requires OS Update", even though the deployment target is smaller than the installed version of macOS 12. Are there any specifications available on which macOS version is required in order to use iOS apps on Silicon Macs?