This I believe is due to overhead involved with the debug tracking of Metal commands. I would love an official Apple response on this, but I think it’s safe to assume this 0.1mb increase will not exist in a release build of your application.
Anyone from Apple have ideas about what causes this tiny memory leak even in the template apps?
I had similar memory leak. In my case I forgot to put @autoreleasepool inside of function for displayLinkCallback.
Basically, display link is creating separate thread in plain C, without autoreleasepool.
Maybe you have something similar.
I am seeing the same issue in 2 codebases I am working on, one of which is multi-threaded and the other single threaded.
It is also reproducable in the official Apple Sample: https://developer.apple.com/documentation/metal/using_a_render_pipeline_to_render_primitives?language=objc
Just run the sample and watch the memory slowly increase over time. When working on memory limited platforms where you become "Terminated due to memory issue" when not even getting near to the devices RAM capacity, a leak which we have seeming no control over is extremely frustrating.