Joy to the world

...or at least to me. I've translated my Python atmega CPU emulator and I can now officially say: Python is unbearably slow.

Pure Python naive implementation is 16 times slower than the real CPU. Pure Python with improved assembly command lookup is 4 times slower. C++ with each command invoked from Python is two times slower. And finally entire emulation loop in C++ is ... 10 times faster!

I did not expect that last difference. Just writing a Python loop around C++ slows it down by the factor of 20? This insanity!

I still intend to write extensions (such as timer subsystem) in Python, but the rule is clear: no Python in the tight loop for me.