gh-121485: Always use 64-bit integers for integers bits count

[serhiy-storchaka]

Use 64-bit integers instead of platform specific size_t or Py_ssize_t to represent the number of bits in Python integer.

• Issue: Use 64-bit integers for bit counts #121485

[mdickinson]

Definitely +1 for the idea. I'm afraid I won't have bandwidth for a detailed code review in the near future.

[serhiy-storchaka]

And what do you think about imposing a hard limit on Python integers, so integers larger than $2^{2^{63}-1}-1$ or $2^{2^{64}-1}-1$ could not even be created? Then the number of bits can always be expressed as a 64-bit integer, error checks after _PyLong_NumBits and _PyLong_Frexp can be removed, and the code for shift operations and int.bit_count() significantly simplified.

[serhiy-storchaka]

I can update this PR for the later idea or implement it in a separate issue.

[mdickinson]

And what do you think about imposing a hard limit on Python integers ...

That sounds totally reasonable to me (presumably as a CPython detail, rather than an aspect of Python-the-language). Any future computing landscape that makes operating on ints of that size remotely feasible is going to look very different from today's landscape.

[serhiy-storchaka]

Thanks, @gpshead.