Subj : Re: Compiler and an interpreter
To   : comp.programming
From : Jon Harrop
Date : Mon Aug 08 2005 07:10 pm

Paul Dietz wrote:
> Rob Thorpe wrote:
>> Jon Harrop wrote:
>>>Incidentally, someone has written a Lisp implementation that is 500x
>>>slower than the OCaml and almost twice as long. :-)
>> 
>> 
>> Really, I'm not surprised it's slower, but I'm surprised it's longer.
> 
> I'm surprised it's that much slower (or, indeed, slower at all).  It must
> be using a different algorithm.  And maybe he forgot to compile it?

No, same algorithm. The difference was a slow Lisp implementation. IIRC,
nothing was specialised by the compiler and the program was only compiled
to interpreted byte-code, not to native-code. I was still surprised to see
such a huge difference though.

Scheme people are currently developing a new port of my ray tracer on
comp.lang.scheme. The program is 2-3x as long as the OCaml but is now as
fast as the first OCaml version on our site.

Note that it took a compiler writer (one of the authors of Stalin) to
optimise the Scheme though. :-)

Also, and I'm not sure about this, it seems that optimising a lisp/scheme
program is very compiler-dependent.

-- 
Dr Jon D Harrop, Flying Frog Consultancy
http://www.ffconsultancy.com

.