[issue26219] implement per-opcode cache in ceval

[Yury Selivanov]

Yury Selivanov <yselivanov at gmail.com> added the comment:

> With the 3.7 beta deadline just around the corner, 3.8 will be the next opportunity to reconsider this idea.

Nick, Ned, Guido,

Would it be OK if we add this to 3.7beta2?

I feel kind of bad about this one... few thoughts:  

1. The speedup is about 5-10% on macro-benchmarks!  This is significant.

2. Inada-san offered me help with this, and I declined, saying that I plan to do more research/experiments and commit it before 3.7.  And I didn't have time to do it because of other work I did for CPython.

3. Even though the optimization is invasive, I feel confident about it.  It just so happens that whatever modification we do to ceval, it will be covered 100% with tests or any code you execute.  Any regression or anomaly should become apparent almost immediately.

4. We'll have quite a bit of time to see if this works or not.  This isn't a new functionality, so we can always pull it from the first RC build if necessary.

----------
nosy: +inada.naoki, ned.deily

_______________________________________
Python tracker <report at bugs.python.org>
<https://bugs.python.org/issue26219>
_______________________________________