The CPython JIT compiler in Python 3.15 alpha has achieved its performance goals ahead of schedule, running 11-12% faster than the tail calling interpreter on macOS AArch64 and 5-6% faster on x86_64 Linux. The author, a core JIT contributor, reflects on the human and technical factors behind this turnaround after the Faster CPython team lost its main sponsor. Key technical breakthroughs include a rewritten tracing JIT frontend using a 'dual dispatch' mechanism that boosted JIT code coverage by 50%, and reference count elimination that removed expensive per-instruction branches. Community stewardship, breaking complex work into approachable tasks, and daily performance benchmarking infrastructure were equally critical to the project's revival.

9m read timeFrom fidget-spinner.github.io
Post cover image
Table of contents
Ken Jin's BlogPython 3.15’s JIT is now back on trackConclusion

Sort: