Java vs Python Benchmarks: Java is faster

Ype Kingma ykingma at
Mon Feb 4 20:12:38 CET 2002


you wrote:
> I guess I imagine that Python's speed issues are:
>  1. Ultra-late binding of, and consequent repetitive lookups of,
>     global names, attributes, methods, module members, etc.
>     (Of course, unless names like 'int' and 'float' can be bound
>     at compile time, optional type declarations are unhelpful.)
>     Since many people like this about Python, and many have used it
>     in existing code, effort is focused on making lookups cheaper
>     (or caching the results) rather than changing the binding
>     behavior to require fewer lookups overall.
>  2. Nothing ever gets compiled to anything even close to
>     machine code.  This is largely because the Python development
>     team doesn't have the resources to invest in it.  Clearly
>     there's something to gain by moving the huge ceval.c loop/switch
>     from runtime to compile time.  No one knows how much.  Clearly
>     there's something to lose from the resulting explosion in the
>     complexity of the compiler.

Please note that jython is compiled directly to Java opcodes.
Anything that makes interpreting these faster (JIT, native compilation)
will make jython faster.


More information about the Python-list mailing list