Python 3.11 - Hackdot

Re:Python 3.11 by azinman2 on Monday October 04, 2021 @03:10AM (#28743973)

> bpo-44590: All necessary data for executing a Python function (local variables, stack, etc) is now kept in a per-thread stack. Frame objects are lazily allocated on demand. This increases performance by about 7% on the standard benchmark suite. Introspection and debugging are unaffected as frame objects are always available when needed. Patch by Mark Shannon

That’s pretty massive! My impression of CPython was there were few performance improvements globally since 3. I’m having trouble finding benchmarks for all the minor versions.

[ Parent | on HN ]

Re:Python 3.11 by vland on Monday October 04, 2021 @06:30AM (#28745046)

7% is massive for C, nothing for Python, and the benchmark suite is unreliable.

I'd wait for independent measurements in real world applications. Python core developers are very good at writing bloated benchmark suites that are so convoluted that no one knows what is going on.

Since more than two decades, big projects like "The need for speed", pythonspeed.com pop up periodically without anything happening. But apparently that's enough to get folks excited.

[ Parent | on HN ]