According to the ARM documentation for the CPU models available in Apple Silicon, the CoreSight implementation includes an Embedded Trace Macrocell which can perform a complete "Instruction Trace" (https://developer.arm.com/documentation/102119/0200/What-is-trace-). Although other operating systems such as Linux make this easy, we have not been able to find any tools or even a system-level API for accessing this feature of the ETM. In the "Instruments" window of Xcode 16+, there is a "Processor Trace" instrument, but this performs sampling and is totally unrelated to the Instruction Trace we need for debugging and analysis purposes. Because it produces a complete, contiguous sequence of branch instructions, the Instruction Trace is essential for identifying precise execution behaviors that are otherwise invisible to the developer. On other platforms, an alternative is debugger scripting, but we have found far too many bugs and reliability issues with the macOS implementation of lldb. Any suggestions would be greatly appreciated!

Hello, I am trying to investigate my apps memory usage using the instruments tool. Whenever I attach instruments my app crashes, when I have it connected to xcode as well I can see that the crash occurs in liboainject: ___lldb_unamed_symbol117 with SIGBUS. Here's what the stack looks like. Is this a known issue? Are there any workarounds or things I can do to figure out what is causing the crash?

Hi there, How to let instruments - app launch template wait for process to launch , rather than actively launch by instruments? I need to profile the app launch performance when clicking a push notification or cold launch via a url link. How to let instruments to wait for the process and collect the data? Currently, I tried the command xctrace record --template "App Launch" --attach MyApp --device-name 'Phone-Dev' --output mytrace.trace But it soon failed with 'Cannot find process matching name: MyApp'. How to make it work?

Hello, I’m encountering an issue with the Instruments app while running a benchmark on an M2 Ultra Mac Studio. Despite being certain that GPU activities involving memory read and write operations are occurring, all related performance counters consistently return 0. Interestingly, this problem does not occur when using the same code on an M1 MacBook Air, where the counters behave as expected. What could be causing this discrepancy? Any insights or suggestions would be greatly appreciated. Thank you!

Hey folks, We are looking for a way to increase the sampling frequency beyong what's currently called "high frequency sampling" for CPU profiler (or time profiler -- doesn't really matter for us). We are aware that this is not offered through the UI but wondering if we can somehow experimentally enable this via the .tracetemplate (plist). Basically, we see that samplingRate exists (in the plist) but don't see it having an effect on the actual runs. The resulting trace file always lists sample-rate-micro-seconds="1000" for the data table. E.g., <table trigger="time" pmc-events="Cycles" target-pid="SINGLE" schema="counters-profile" needs-kernel-callstack="0" sample-rate-micro-seconds="1000"/> Cheers

It looks like, for some reason, our apps are using a bunch of power sometimes. sysdiagnose has this in the power log: Never mind. Including the output of sysdiagnose has "sensitive language," and it won't tell me what is sensitive, making this a waste of my time. ETA: Ok, I I can attach the file: power.log I've gone through the energy documentation, but it seems geared towards embedded, not macOS, so I'm not sure how I can figure this out more. The extra problem, of course, is that we have a network extension, two daemons, and a GUI app. 😄

I am using XCode16 and macOS 14.7.2. Previously, using the instruments on an iPhone with iOS 14.3 was normal, but when I upgraded to iOS 18, the instruments often couldn't find the library. I have to restart the instruments to restore normal operation, but the problem will occur again after using it for a period of time

I am trying to record the requests and responses in a WKWebView, but instruments does not seem to record them. Is this to be expected? The webView is set to inspectable, and I am using the HTTP Traffic instrument. All the requests the app is doing are recorded, but neither the original request for the webView, nor subsequent traffic is recorded. When I use Safari to inspect the webView, all I see is the last page (even when I start the inspector before the first request is made). How can I see these requests?

I made a box with MDLMesh.newBox(). I added normals. let mdlMesh = MDLMesh.newBox(withDimensions: SIMD3<Float>(1, 1, 1), segments: SIMD3<UInt32>(2, 2, 2), geometryType: MDLGeometryType.triangles, inwardNormals:false, allocator: allocator) mdlMesh.addNormals(withAttributeNamed: MDLVertexAttributeNormal, creaseThreshold: 0.25) After I convert to MTKMesh the normals are (0,0,0) for a group of vertices. I can only inspect the geometry after I convert to MTKMesh. Is there a way you can use Geometry Viewer on a MDLMesh?

hello, I got a question about coreml. I loaded the coreml model in the project and set the computing unit to CPU+GPU. When I used instruments to analyze the performance, I found that there was an overhead of prepare gpu request before each inference. I also checked the freezing point graph and found that memory was frequently allocated. Is this as expected? Is there any way to avoid frequent prepares? I have tried some methods, such as memory sharing of predict interface input parameters, but it seems to be ineffective.

I could use the Metal System Trace before the most recent update, but now whenever I try to profile using the Metal Counter instrument, I get the [Warning] GPU Service reported error: Selected counter profile is not supported on target device. What is the issue here?

It seems as though using any initializer of SubscriptionOfferView or StoreView will create a memory leak. This can be simply reproduced by adding this to your SwiftUI view: SubscriptionOfferView(groupID: "yourgroupID", visibleRelationship: .all, useAppIcon: true) or StoreView(ids: ["monthly", "yearly"]) Tested on iOS 26 beta 2

Dear Apple Developer Support team, I would like to request an official confirmation regarding the handling of transaction status in the App Store Server API, specifically for the GET /inApps/v1/transactions/{transactionId} endpoint. As per our current understanding from the official documentation (Get Transaction Info), the API’s behavior appears to be: If a transaction is finalized and successfully processed by App Store, querying this API will return HTTP 200 OK along with transaction details. If a transaction is still in a pending or deferred state (such as awaiting Ask to Buy approval or pending authorization), the API will not return a 200, and instead respond with HTTP 404 Not Found or an appropriate error. Could you please confirm if this behavior is accurate and officially supported? Specifically: Does a 200 OK response guarantee that a transaction is finalized and successfully recorded on App Store servers? In cases where a transaction is pending approval (e.g. Ask to Buy), is it correct that GET /transactions/{transactionId} would return 404 Not Found until the transaction is finalized? We would greatly appreciate your confirmation to align our server-side logic for transaction validation accordingly. Thank you very much for your support! Kind regards, cuongnx

Hello We use Datadog Mobile Vitals in our app and I'm trying to run some tools in Instruments for comparison. I'm not sure what tool should I use for some of those metrics: Slow Renders Description: With slow renders data, you can monitor which views are taking longer than 16ms or 60Hz to render. Instruments equivalent: Hangs including microhangs (?) CPU ticks per second Description: RUM tracks CPU ticks per second for each view and the CPU utilization over the course of a session. The recommended range is <40 for good and <80 for moderate. Instruments equivalent: CPU Profiler (?) Frozen Frames - Description: Frames that take longer than 700ms to render appear as stuck and unresponsive in your application. These are classified as frozen frames. Instruments equivalent: Hangs with > 500ms (?) Memory Utilization Description: The amount of physical memory used by your application in bytes for each view, over the course of a session. The recommended range is <200MB for good and <400MB for moderate. Instruments equivalent: Allocation (?)

I tried using Create ML of Xcode 26.0 beta 7 to generate a model using the "Word Tagging" template, and I received the error: Training progress unavailable - Unexpected error. Using Create ML of XCode 16.4 with the same documentation, I was able to build the model and use it in a test app. I'd like to understand why Create ML of Xcode 26 no longer works.

Hello Apple team, I am using xctrace to record an Allocations trace on iOS. For example: xctrace record --template "Allocations" --launch com.example.myapp --time-limit 30s --output alloc.trace After recording, I can export the results in Allocations List format (flat list of allocations) using: xcrun xctrace export --input ./alloc.trace --xpath '/trace-toc/run/tracks/track[@name="Allocations"]/details/detail[@name="Allocations List"]' --output ./alloc.xml This works fine and produces an XML output. However, what I really need is to export the data in Call Tree format (as shown in Instruments GUI). I checked xctrace export --help, but it seems that the Allocations template only supports the List view for export, not the Call Tree breakdown. My question is: 👉 Is there a way to export an Allocations trace in XML with Call Tree details using xctrace? 👉 If not, is there an API or recommended workflow to automate this instead of exporting manually from Instruments GUI? Thanks in advance for your help!

I'm working on a custom instrument that displays intervals from os_signpost data. I'd like to color the intervals in the graph based on data from an accompanying aggregate. For example, color the interval red if its duration is greater than 3 standard deviations from the mean. Is this possible?

Hello, I'm no macOS 26.1 I'm running through the example shown in the WWDC video "Optimizing CPU Performance with Instruments." After right-clicking a test, clicking "Profile...", and trying to run a processor trace I get the error in the subject. I have processor trace enabled on the CPU via system settings. I confirmed this by disabling it and re-enabling it and noting the error that appears when this is disabled is different from the one I'm now getting. This did previously work but I haven't tried since macOS 26 is released. Is there something new I need to be doing to my Xcode project settings in order for this to be working?