[Python-Dev] Idea for a fast calling convention (original) (raw)

Christian Tismer tismer at stackless.com
Mon Mar 1 08:04:37 EST 2004

Michael Hudson wrote:

Christian Tismer <tismer at stackless.com> writes:

p.s.: I believe some automatic source analysis and rewrite might pay off in other areas as well. Grepping through the sources, there are still very many similar patterns of PyArgParseTupleXXX calls, which could be replaced by less general, optimized versions. This would even not cause code bloat, since all those calling sequences would be smaller than now. Well, yes. C sucks seriously for things like this, though. It's frankly embarassing that every time, say, ''.split() is called, some silly string is being parsed. Unclear what to do about this (excpet PyPy, I guess).

Why not a switch over the arg tuple size? Or do you mean, this happens in so many different places that it would be tedious to change that by hand. Sure, we should not try to morph Pytho into PyPy in the first place. But maybe some runtime analysis with a modified version of PyArg_Parse... things, we could see how often which format string appears, and how often which optional paths are used, as a hint for hand optimisation.

I could do that with Stackless, btw. For Win32, I have a few introspection tools which can do C stack analysis from a debug build and find out what is being called. They were meant for later C stack synthesis, which never worked, but this would give them some use, again.

ciao - chris

-- Christian Tismer :^) <mailto:tismer at stackless.com> Mission Impossible 5oftware : Have a break! Take a ride on Python's Johannes-Niemeyer-Weg 9a : Starship http://starship.python.net/ 14109 Berlin : PGP key -> http://wwwkeys.pgp.net/ work +49 30 89 09 53 34 home +49 30 802 86 56 mobile +49 173 24 18 776 PGP 0x57F3BF04 9064 F4E1 D754 C2FF 1619 305B C09C 5A3B 57F3 BF04 whom do you want to sponsor today? http://www.stackless.com/

More information about the Python-Dev mailing list