Newsgroups: comp.lang.forth,de.comp.lang.forth,comp.sys.transputer
From: Bernd Paysan <bernd.paysan@remove.muenchen.this.org.junk>
Subject: Re: TILE forth for the inmos transputer t8 family
Organization: Bernd Paysan, 81477 Muenchen, Germany
Date: Mon, 13 Apr 1998 23:47:50 +0200
Mime-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Message-ID: <35328806.3FDE86CD@remove.muenchen.this.org.junk>

Alex Stuebinger wrote:
> 
> Hi all,
> 
> I am thinking of a TILE-FORTH port to the inmos t8 family of
> transputers.
> I will not use GNU FORTH as a base for a port, because that
> uses special features of GCC, which are not available in the
> Inmos C Development Toolkit.

This is a valid concern. However, the extensions are limited to two
areas: some parts use the bracked expressions (blocks as expressions),
which is neat to write, but not necessary. The important use of GCC
extensions is that it uses computed gotos for NEXT. Since NEXT is a
macro and the rest is generaded from a file, it should be possible to
convert that to a giant switch-statement-Forth as TILE is. It is just
that nobody wanted Gforth for a machine that has C and not GCC yet*. You
are the first one.

> I also was told that GFORTH is not very FORTHish.

Gforth is as Forthish as a Forth with an engine written in C can be. And
since Gforth can use other engine implementation methods (especially the
really Forthish one: write primitives in inline assembly), this is
wrong. TILE certainly is less Forthish.

I still would recommend using tForth, as Marcel said. No amount of work
to port it, excellent code generation and all the parallel extensions
you need to use the transputer as what it is meant to be.

*) we have ported it to several CPUs that doesn't have neither C nor
GCC, and there it uses assembly primitives.

-- 
Bernd Paysan
"Late answers are wrong answers!"
http://www.jwdt.com/~paysan/