Re: meta-compiler

On Sun, May 29, 2005 at 02:32:18AM -0400, Mark Slicker wrote:
> I was discussing my meta-compiler ideas with Howerd, I thought I might as
> well publish them here, to see what others think. Just to note, some of
> the code below (building dictionary entries) is untested.
>
> If you don't know what a meta-compiler is, the best I can describe it is
> as an application which allows you to compile a new Forth kernel itself
> defined in Forth.
>
> If the Forth kernel is targeted to a different processor, there may be
> additional complications beyond what is described here. At the moment, I
> am interested in compiling a pentium colorForth kernel using the present
> pentium colorForth kernel. What this means is that compiling the
> colorForth kernel will be little different than compiling any other
> colorForth application. The differences however are important to note:
>
> * The kernel must be self contained, it cannot refer to words defined
> outside of itself, except for macros which compile instructions inline.
>
> * Variables must refer to locations within the kernel, not locations in
> source.
>
> * All absolute addresses which reference locations within the kernel must
> be adjusted for the kernel's target location.

I miss some premises here, that maybe are implicit.
So what is the idea about
- where is the source, presumably in colorforth blocks?
- how to handle bootcode.

But there is a problem here. The editor is in colorforth. So the
new colorforth may require tricks from the editor that are not yet
present (say e.g. a new color). Right?

Is the idea to boot into the old colorforth, then run the new
kernel, or will the boot code be meta compiled too?

<SNIP>

--
Albert van der Horst,Oranjestr 8,3511 RA UTRECHT,THE NETHERLANDS
Economic growth -- like all pyramid schemes -- ultimately falters.
[email protected] http://home.hccnet.nl/a.w.m.van.der.horst