Pyston 0.5 Released As A Faster Python JIT
The Dropbox engineers working on their Pyston project as a high-performance JIT implementation today announced version 0.5 of the software.
Pyston 0.5 now makes use of reference counting rather than tracing garbage collectors, support for running NumPy unmodified, signal handling support, frame introspection, and many other changes.
All of the details on Pyston 0.5 -- including their motives for abandoning the tracing GC and moving to refcounting -- is described via this blog post announcement on the Pyston project site.
Pyston 0.5 now makes use of reference counting rather than tracing garbage collectors, support for running NumPy unmodified, signal handling support, frame introspection, and many other changes.
All of the details on Pyston 0.5 -- including their motives for abandoning the tracing GC and moving to refcounting -- is described via this blog post announcement on the Pyston project site.
12 Comments