🚀 go-pugleaf

In article <39bjl5$bi0@csnews.cs.Colorado.EDU>,
Tom Christiansen  <tchrist@mox.perl.com> wrote:
>:-> In gnu.misc.discuss, rsanders@mindspring.com (Robert Sanders) writes:

>:>Translating ksh93, perl5, or python into scheme will only slow them
>:>down.  Why ever would you want to do this?

>:Why do you think that?  Almost every Scheme interpreter I've tried outruns
>:python, ksh, and perl on simple benchmarks like arithmetic and function 
>:calling. 

>If you write a fibonacci program, full of deeply nested subroutine
>calls, than scheme will run faster -- because perl hasn't optimized
>the argument call stack yet.  But for many other things it's plenty
>fast.  Don't listen to this "Perl is 10x slower than Scheme" drivel,
>because it's simply not true.

Scheme outran perl in simple iteration and arithmetic -- there were no
subroutine calls in the perl version.  Remember Ilya Zakharevich's
tiny benchmark on perl5-porters that showed that %OVERLOAD wasn't very
expensive?  Well, taking this as an example, uncompiled SCM, Scheme48,
OakScheme and VSCM all outran perl4.  SCM and Scheme48 both outran perl5
by a few percent.

Compiled SCM and Scheme->C versions are around 30x faster than perl5 on this 
particular `benchmark', and Scheme compiled with bigloo is around 90x faster.
So you're right, you shouldn't listen to that "10x slower" drivel.  It's
a bit conservative :-)

Seriously, most of the things you do with Perl or GEL aren't (going to be)
limited by the speed of the interpreter itself.  I do most of my day-to-day
programming in Perl, and profiling would probably show that the majority of
the time is spent in I/O and regexp matching.

Perl is usually `plenty fast', but I'm still not sure why you think 
translating ksh93, Perl, or python into Scheme will slow them down.  Could
you clarify your point, please?

  -- Robert