Available New release Expansion pyston-lite , which implements the JIT compiler for CPYTHON. Unlike the Pyston project, which is developed separately as a branch from the CPYTHON, Pyston-Lite, designed in the form of a universal expansion designed to connect to the Python regular interpreter (CPYTHON). The new issue is noteworthy to ensure support for the Python 3.7, 3.9 and 3.10 branches, in addition to the previously supported branch 3.8.
pyston-lite allows the use of the Pyston basic technologies without replacing the interpreter through the installation of additional expansion using the PIP or Conda package manager. Pyston -Lite is already located in Pypi and Conda repositories and to install it is enough to execute the Pip Install Pyston_lite_autoload or “Conda Install Pyston_lite_autoload -c Pyston”. Two packages were proposed: Pyston_lite (directly JIT) and Pyston_lite_autoload (performs an automatic JIT substitution when starting the Python process). It is also possible to include JIT from the application without installing the Autoload module using the Pyston_lite.enable () function.
, additional developers announced that from now on they consider Pyston-Lite as their primary product. The usual Pyston allows you to achieve higher performance, but its use is difficult to replace the interpreter, while Pyston-Lite is easily integrated into existing configurations. Specific optimization present in Pyston will be transferred to the main composition of Cpython as if possible. In a distant perspective, there is a desire to transfer to the main composition of CPYTHON and the implementation of the JIT compiler.
In the current form, compared with CPYTHON 3.8, the use of Pyston-Lite allows you to raise performance by 10% in the text of Macrobenchmarks and 28% in the Pyperformance test. For comparison, the increase in performance when using separate Pyston is 28% and 65%, respectively.
The upcoming issue of the main project of the CPYTHON 3.11 includes changes related to caching the state of bytcode, accelerating the call of functions and using fast interpreters of typical operations, and some optimization prepared by Cinder projects and hotpy . As a result, the CPYTHON 3.11RC2 in the above tests demonstrates an increase in performance of 12% and 26%, which is comparable to Pyston-Lite (the new CPYTHON and Pyston-LITE applied different optimization, porting the Pyston-Lite for CPYTHON 3.11 will still allow raising the performance of this branch ).