Speed up Python by up to 5% ?

[Tim Peters]

[Edward K. Ream]
> ...
> Among other things, this has saved me the trouble of suggesting that
> SET_LINENO opcodes be "optimized."

As was mentioned, it *would* be a worthwhile optimization to get rid of
SET_LINENO completely.  The "periodic checks" code is a heartbreaking little
drag on performance, but nobody yet has dreamed up an acceptable
(bulletproof, and portable across OSes & C compilers) way to get that test +
taken-branch off the eval loop's critical path.