Investigate memoryleaks

[almarklein]

In #306 we explored the memoryleak of Lavapipe, and found out that Device objects leak memory on Windows too. To be fair, in practice you'd likely only use one device per process, but there are use cases for using multiple devices. Let's also do some testing on other objects, and report/fix errors upstream.

The purpose of this issue is to track known leaks. We want wgpu-py to be a reliable system without leaks.

Status

• Adapter
  • All systems: Memory does grow, albeit very slowly. Could be a small leak?
• Device
  • LavaPipe with Vulkan: Leaks significantly.
  • Windows with Vulkan/D3D12: Leaks significantly. Crashes at the 64th iter!
  • MacOS with Metal: Does not seem to leak.
• Buffer
• Texture
• ...

---

Code to track (updated):

import gc
import psutil

import wgpu.backends.rs
import wgpu


p = psutil.Process()


def print_mem_usage(i):
    megs = p.memory_info().rss / 1024**2
    print(f"memory usage (round: {i}): {megs:.3f} MB")


if __name__ == "__main__":
    print_mem_usage(0)
    for i in range(500):
        adapter = wgpu.request_adapter(canvas=None, power_preference="high-performance")
        device = adapter.request_device()
        device.destroy()
        del device
        wgpu.backends.rs.delayed_releaser.release_all_pending()
        gc.collect()
        print_mem_usage(i + 1)

    wgpu.print_report()

[almarklein]

Closing this because we have #403. Following up in #405 and #406.