IntType slowness (fixed ints)

[psrok1]

IntType is implementation of a concept to make a fixed-size int that behaves similarly to real int, so it's useful for reimplementation of code written in C language. It's natural that they're slower than native implementation, but current implementation is insanely slow because each operation creates new IntType object.

Each object initialization also comes with cutting value to the expected bitness.

malduck/malduck/ints.py

Line 166 in 1f6ce66

value = int(value) & cls.mask

In the same time, base class of IntType is int which is immutable type.

Proposed solution is:

• implement wrapper based on https://docs.python.org/3/library/numbers.html#numbers.Integral instead of deriving from int
• make mutable IntType variant, so we avoid creating new objects in code that needs to work fast
• check if wrapping ctypes.c_int works faster than applying mask for each operation that could be overflowing.