MicroPython Forum (Archive)

I understand there's a tradeoff between object size's and time to complete GC but from my point of view (I integrate MP to "DSP group" chips) scanning raw data during GC seems like something I'd like to avoid, and as a new feature I can put it under compilation flag. I think MP can be improved much ...

Thanks Christian for your answer. I think the way GC works right now contradicts the concept of being "lean and efficient" and the spirit of realtime imbedded programming. My plan is to add a "GC marking" method to objects so it would not relay on excessive search, while maintaining backwards compat...

So here's another code which demonstrates the bug: import gc f = open('c:\\wavs\\tone 2kHz.wav', 'rb') x = f.read(64) x = f.read(64) gc.collect() and the partial log from running gc_mark_subtree (with additional notes): (1) 008E0590 - 4 x INT32 (2) 002B6D88 00005567 00000040 008E0540 <-- (pointer wa...

Hi, test code is: a = [x for x in range(120)] a = [x for x in range(120)] gc.collect() I've seen that gc_mark_subtree runs on the elements of 'a' looking for a pointer ("if (VERIFY_PTR(ptr)) {"). On random values of 'a' (or any other value sources) element may pass the test though its not a pointer ...