[Python-Dev] further optimising the micro-optimisations for cache locality (fwd)

Vladimir Marangozov Vladimir.Marangozov@inrialpes.fr
Fri, 28 Jul 2000 22:20:02 +0200 (CEST)

Thomas Wouters wrote:
> On Fri, Jul 28, 2000 at 09:24:36PM +0200, Vladimir Marangozov wrote:
> > It would be interesting to collect some feedback on these ideas for
> > popular combos. Who knows...
> I like these ideas, though I think anything beyond 'further folded' requires
> a seperate switch for the non-common operators and those that do a tad more
> than call a function with a certain number of arguments and push the result
> on the stack.

I agree.

Note that all we can get though this game is some dubious percents that
we shouldn't ignore, but which persistency in the long term is quite
questionable. There's no way that, to pick a guy at random, Tim won't
confirm this!

And just like Guido, I'm not ready to trade code cleanliness for dubious

However, I take the opportunity to invite you again to "heavy test"
the object allocator's candidate -- obmalloc.c, which, in my educated
understanding of the current C implementation is the only piece of
code that is capable of increasing Python's overall performance by > 10%
assuming that your script involves object allocations.

It is my strong belief that this perfomance comes for free, by
instrumenting Python at its most inner internals, not on top of it!

Of course, improvements to this code are welcome and I'd be happy to
discuss them in this forum.

       Vladimir MARANGOZOV          | Vladimir.Marangozov@inrialpes.fr
http://sirac.inrialpes.fr/~marangoz | tel:(+33-4)76615277 fax:76615252