I'm curious about Cinder, the performance-focused version of Python that powers Instagram. It's been open-sourced for a while now, and one of the reasons for doing so was to enhance collaboration on performance improvements for CPython and minimize duplicated efforts. I've noticed some advancements have already made their way upstream, but I wonder why more of Cinder's features, particularly their JIT compiler approach, haven't been integrated into CPython yet.
2 Answers
Actually, some of Cinder's work has indeed been making its way into CPython. Meta is a big player pushing improvements to CPython, including the notable nogil (free-threading) changes.
True, but it's important to note that nogil wasn't originally part of Cinder's mission; that was developed separately within Meta's PyTorch team.
Honestly, it seems like there are some political dynamics at play. Companies like Microsoft have their own ideas for performance enhancements and want to push those instead of fully adopting Cinder's research. Plus, folks like Guido being at Microsoft means they hold a lot of sway over Python's direction. That said, there have been features from Cinder integrated into Python, like PEP 683 with immortalization, which did originate from Cinder.
I remember Microsoft had a similar initiative, but I heard they recently laid off that team. Crazy how things change in the tech world!