Re: librep's indirect threaded bytecode interpretter

[Per Bothner <per at bothner dot com>]

Mikael Djurfeldt <mdj@mdj.nada.kth.se> writes:

> Personally, I now suspects that a byte-code interpreter can be faster
> than a tree interpreter ever can be.

And if you use Java bytecodes you don't even have to write the
interpreter.  Of course a simple interpreter is easy to write, and
Java bytecodes are admittedly far from optimal for Scheme.  The point
about using Java bytecodes is to benefit from all the engineering and
research that people are putting into *fast* interpreters, such as
Just-in-Time compilers, as well as all the other Java-related tools
and engineering.

Clearly compiling Scheme into efficient Java bytecodes is non-trivial,
but that task has been more-or-less solved in Kawa.  (Not to say that
Kawa can't be improved, of course.)  And if you don't like some of my
design decisions, you are free to design another language front-end
sharing the Kawa compiler and utility classes.  (By the way, the
Emacs Lisp language is now supported quite well.  And Kawa now has an
efficient and flexible module system.  I also have a design for
implementing GOOPS/CLOS-like classes in Java.)
-- 
	--Per Bothner
per@bothner.com   http://www.bothner.com/~per/