Speed of Python vs. Perl

[Mike Coffin]

pj at sgi.com (Paul Jackson) writes:

> Tim wrote
> |> Talk someone into writing a patch; I've got no bandwidth to spare.
> 
> See another branch of this news thread for a patch (both 1.5.2
> and 2.0 variants).
> 
> Me thinks this (getting startup time competitive with Perl) is
> a hard problem, and that the two tweaks in my patch help, but
> not greatly.
> 
> The combination of the dynamic nature of Python, with having
> much of its run time library support loaded from Python code,
> not compiled in, would seem to force Python to do upwards of
> 10 million instructions on startup.
> 
> I don't see anyway around this.

One way is to provide some sort of "dump" facility that allows the
program to initialize, dump the state, and then make an executable
from the dumped state.  Emacs does (or at least, did) this to reduce
startup time.

-mike