Many of the CPython core team members are looking forward to a new dispatch technique called tail calling interpreter, which shows up to 40% performance improvement and 10% geometric means improvement. https://github.com/python/cpython/issues/128563
Comments
Log in with your Bluesky account to leave a comment
The downside of this technique is that it is limited to the latest compiler, which should support the __attribute__((musttail)) and __attribute__((preserve_none)). But the computed-goto had the same issue at that time, so I think that this would be just a time issue.
Comments
Interesting results are out for the macOS Apple Silicon build, too :)