[Python-Dev] Re: a serious threat to 2.3's speed?

[Fredrik Lundh]

Tim Delaney wrote:

> From the microbenchmarks, it looks like function calls are the big
> bottlenecks.

in my pytte1 experiment from last year, I was able to run pystone about 25
times faster than a stock 2.1 interpreter.

(that would be, say, about 20 times faster than 2.3)

key features contributing to this is true gc, call-site caching, and "traditional"
(C-style) argument passing for common cases.

not sure if any of that can be applied any of that to today's C python, though
(but that doesn't mean it cannot be done ;-).

> Another option for a masters thesis Brett - speeding up function calls :)

that would be most excellent.

</F>