SRSaunders

> When capture is turned off in the Scheme completely, the error goes away. Thanks @cdavis5e and @WalkerWhite for figuring this out.

Probably too old and unimportant to worry about now. Most machines going forward have Apple Silicon GPUs anyways. Closing issue.

Ok that's fine. Actually my main dev machine is x86, but I test on Apple Silicon as well. I was looking at my open PRs and thought closing this one...

Thanks for fixing this old one. > I'm curious though, do any AMD devices that Apple uses have any value other than 32 for this? We have the default for...

@tomadamatkinson any updates? I have prototyped a solution to this problem, but it's the opposite to what the comments in the code imply. The existing comments seem to indicate that...

I decided to create a pull request so @tomadamatkinson can review my proposed changes. I have tested on all platforms except for Android. Seems to work properly. Happy to receive...

A quick follow-up: the following Khronos samples also fail similarly (blank output) with MoltenVK 1.2.10 on iOS 17.5.1: _16bit_arithmetic_ _layout_transitions_ _pipeline_barriers_ _subpasses_ All these run properly on macOS with MoltenVK...

After reading #2283, I decided to retest the above 5 samples with `MVK_CONFIG_USE_METAL_ARGUMENT_BUFFERS = 0` (using the environment variable method for testing purposes only). With this change the samples began...

I retested against commit ec2b23f6c9e61d8a8aa436a46a7115375d3abb11. This seems to fix the issues on iOS running on iPhone 15 for all 5 samples listed above: _timeline_semphores_, _16bit_arithmetic_, _layout_transitions_, _pipeline_barriers_, and _subpasses_. However,...

Works fine using MoltenVK main running on an AMD 6600 GPU machine + macOS Ventura (13.7.8). Re Intel GPUs unfortunately my machines are too old to test against.