Ni aims to provide a module that just makes Python faster.

Ni is based on a fork of Armin Rigo's Psyco, Armin and the other Psyco developers moved onto PyPy so go check that out if you want a high performance implementation of Python (http://pypy.org).

• 64 bit port working ok but machine code is far from optimal
• Compiles against Python 3.7 and will compile simple function
• Many pysco limitation still remain, many Py2 behaviours will still remain.

Will compile very simple functions see mergepoints.c for supported opcodes.

Examples:

def test(a, b):
    return a + b

Or

def test(n):
    i = 0
    while i < n:
        i = i + 1
    return i

Like I say very simple but I suspect other opcodes could be re-enable with very little effort. Any discoveries in this area would be very interesting.

I know I should really explain now to use nianalyser and gdb to debug things.

• Improve debugging and execution performance analysis - nianalyser - DONE
• Refactor code for maintainability and to CPython coding standards - On-going
• Make Ni work on all Python code - On-going
• Improve preformance and implement better optimisation decision making - Very on-going
• Python 3 port - on-going but progress made

I really need to make Trello for all this.