Re: [Chicken-users] help please

[Linh Dang]

On 13 Apr 2004, address@hidden wrote:

> On Tue, 2004-04-13 at 23:21, felix wrote:
>>> - the new script takes 10 secs to run
>>> - the old script took 6 min to run
>>> - the Perl script took 5-6 secs to run
>>
>> Hey, 10 secs vs. 5 secs is actually pretty good!
>
> Yep, that's excellent.  I'm impressed.  I wonder how it goes with
> the new(ish) option to use PCRE as the Chicken regexp engine?

It was with PCRE.

>
>
>>> I guess Perl's IO is much more efficient than chicken's IO
>>
>> Indeed, there are worlds between them. But that is to be expected,
>> since Perl is an I/O and regexp engine.

In this example, it's pipe of 3 cpu-intensive processes:

        cleartool | ctags | scheme-program

I believe that:

        - the bottleneck now is in chicken's IO subsystem.
        - the bigger the pipe's buffer, the better it will be because
          it's not efficient to preempt hi-cpu processes too often.


>>
>> Chicken will never be in the same ballpark as Perl for these kind
>> of things.
>
> Perl is good, but it's not unbeatable.  I've managed to tune bits of
> Gwydion Dylan (IO and hash tables, mostly) to get very close to (or
> even beat) Perl for many of the "challenge" snippets that float
> around the net from time to time.  And that's despite the *far*
> higher generality of Dylan's I/O system and collection classes.
>
> There a regexp implementation for CL (CL-PCRE) that compiles regexps
> to linked closures and claims to be twice as fast as Perl.  I wonder
> how hard that would be to port to Scheme (or Dylan)?

I'd be REALLY interested in this. I hope it'd also has some kind of
s-expression notation.


--
Linh Dang