3 ms·
Wow, I'm surprised that that's faster than the old byte-wise scanning of the instruction stream, and doubly surprised if the benefit isn't negated by the extra
by versteegen 10y ago
Wow, I'm surprised that that's faster than the old byte-wise scanning of the instruction stream, and doubly surprised if the benefit isn't negated by the extra cache pollution.
Actually, it's a lot simpler than I expected!
- daurnimator 10y agoIt's probably not; but it does fix a bug where bytes got misinterpreted. See http://www.freelists.org/post/luajit/Random-failures-in-compiled-code http://www.freelists.org/post/luajit/Random-failures-in-comp...
- versteegen 10y agoI should have thought of that... of course it's easy to see the code was probably wrong in hindsight. Thanks for the explanation.