[plt-scheme] Performance Targets for MzScheme

From: Brent Fulgham (bfulg at pacbell.net)
Date: Thu May 13 12:15:51 EDT 2004

> Why not use regexp-split?  That's faster than
> string-tokenize.  Plus, method dispatch is slow
> in Python that the difference in this one primitive 
> should be amortized.

If my goal was to produce the fastest program in
MzScheme that could split every line in a 15 meg file,
I would be content to use regexp-split.

However, my goal was to identify places in MzScheme
where performance could be improved.  Currently, it
looks like SRFI performance is quite bad in some
cases.  This means that these constructs might as well
not exist -- what point is there in having a string
tokenization function that is 10-fold slower than the
regexp-split?  Perhaps the string-tokenize function
should be itself implemented *using* regexp-split?

I just thought it might be useful to know where we
have bottlenecks.



Posted on the users mailing list.