This is the mail archive of the
guile@cygnus.com
mailing list for the Guile project.
Re: Well, that's interesting.
Mikael> :) As I've reported previously (long ago) I've seen
Mikael> execution times change by 30% as a direct effect of
Mikael> *moving* one function from one place in eval.c to
Mikael> another [...]
Mikael> (I think those hardware guys must be completely nutty. :)
Ulrich once talked to me about using a stochastic search technique to
optimize the *order* of the functions in glibc. Various programs
would be run, and re-run with a different order of the functions,
trying to optimize short jumps. A simulated annealing search would
then guide you to the optimal configuration.
I don't think he's implemented it yet, but it seems like a fun idea.
Many commercial compilers try to optimize this kind of thing.