Python3.13 memory usage is more

[Pooja-ganiger]

Bug description:

After upgrading to Python3.13.11 it is seen that most of the python processes are taking more memory compared to previous version 3.11.14.
Below is an example script run with redis and iproute, where ~0.7MB memory increase is seen.

And also each for python process running has shown a memory increase of 1.5 to 2MB.

CPython versions tested on:

3.13

Operating systems tested on:

Linux

[picnixz]

We need a reproducer without redis and iproute. The increase may have come from them. Since 3.11 is security-only and 3.13/3.14 were quite different at the interpreter level, we may not help you more unless what caused the exact memory increase. If you're able to reproduce this with something that doesn't rely on those libs (or at least only on redis maybe), then we could try to do soemthing, but there is nothing actionable on our side with this reproducer.

I would also consider that 0.7MB out of 42 MB of increase is fine (we may have used more memory for caching for instance, or because a structure got larger for some reason, or because of some additional requirements from the free-threaded build). If a memory regression for this snippet happens between 3.13 and 3.14, this may also be because of the incremental GC which has different thresholds (TL;DR: the GC runs less frequently and thus improves the runtime, at the cost of memory, and if you have isolated reference cycles, those cycles are collected less frequently).