The developers of the Pyston project, which offers a high-performance implementation of the Python language that uses modern JIT compilation technologies, have released a new release of Pyston 2.2 and announced the return of the project to the public. The implementation aims to achieve high performance close to that of traditional system languages such as C++. The code for the Pyston 2 branch is published on GitHub under the PSFL (Python Software Foundation License), similar to the CPython license.
Recall that earlier the Pyston project was supervised by Dropbox, which stopped funding the development in 2017. The Pyston developers founded their company and released a significantly reworked Pyston 2 branch, which was declared stable and ready for widespread use. At the same time, the developers stopped publishing source codes and switched to providing only binary assemblies. Now it has been decided to make Pyston an open source project again, and to transfer the company to a business model related to the development of open source software. Moreover, the possibility of transferring optimizations from Pyston to regular CPython is being considered.
It is noted that Pyston 2.2 in performance tests that evaluate the loads typical for applications for web servers, is 30% faster than standard Python. Pyston 2.2 also boasts significant performance improvements over previous releases, primarily due to the addition of optimizations for new areas, as well as improved JIT and caching mechanisms.
In addition to performance optimizations, the new release is also interesting because of the porting of changes from the CPython 3.8.8 branch. In terms of native Python compatibility, the Pyston project is touted as the most CPython-compatible alternative implementation, as Pyston is a fork from the main CPython codebase. Pyston supports all the features of CPython, including the C API for developing C extensions. Among the main differences between Pyston and CPython is the use of DynASM JIT, inline caching and general optimizations.
Of the changes in Pyston 2.2, it is also mentioned that the codebase was cleaned from many CPython debugging features that negatively affect performance, but are almost not in demand among developers. Statistics are given, according to which the removal of debugging tools leads to a speedup of 2%, despite the fact that only about 2% of developers use these functions.
Source: opennet.ru
