Thanks to reports from some of our users we are happy to announce that the memory leak that was introduced in version 0.2 has now been patched.

For technical users, the leak was coming from our switch to the metal api of which generated autorelease pool allocated command buffers. While command buffers are extremely lightweight single use objects which are released in the autoreleasepool lifecycle, mining calculations on the GPUs can require over 100 buffers a second, which adds up fast. The solution here was to force handle the command buffer memory deallocation.

For further reading see Command Organization and Execution Model.

Happy new year