The path through language design

Yesterday I went to the IronPython Tutorial at OSCON. It was pretty interesting, it reminded me the work I did on OCamlOLE (OLE/COM bindings for OCaml) which involved generating code for accessing OLE components from a TypeLibrary.

From a performances point-of-view, IronPython claims a 1.8x speed factor improvement over Python 2.4, using the pystone benchmark. This morning I wrote then the Neko version of the pystone benchmark and run it on my laptop. I got a 3x speed factor improvement over Python 2.4, so that confirms previous benchmarks using fibonacci numbers. And there is no JIT yet…

(please note that integers on Python are Objects and are automaticaly converted to Bignums when operations are causing an overflow, so a lot more checks and operations are done compared to Neko integers).