Serhiy Storshaka wrote a kind of cache for PyArg_ParseTupleAndKeywords(): _PyArg_Parser structure used by _PyArg_ParseTupleAndKeywordsFast(). It makes argument parser much faster.

Would it be possible to modify it to be able to use it in _PyArg_ParseStack(). (Would it be faster?)

See also issue #29419 "Argument Clinic: inline PyArg_UnpackTuple and PyArg_ParseStack(AndKeyword)?" for a different kind of optimization.