[LispM-Hackers] Interrupts work (mostly), and Yet Another Question

[Tim Moore]

On Mon, 11 Mar 2002 jm@mak.com wrote:

> I know this is kind of, er, twisted, but...
> 
> If we get the Linux version running, can't we just put the code in the
> band (probably as the contents of an editor buffer?), and then save the
> band, and then boot that band on the emulator, and then compile the
> buffer contents?
> 
> I realize this presupposes having a fully functional Linux version, and
> enough of the bare iron stuff to fire up the compiler.

If you get the Linux emulator version fully running, you shouldn't need to
go through such contortions, although you would go through minor
contortions.  I think the procedure would be:

* Tweak the compiler, if needed, for the new target architecture i.e., the
bare iron emulator.  The fasl files produced by this compiler might
not be loadable into the running image;

* Write Lisp device drivers to your heart's content, and compile them;

* Write out fake load band that is intended to be booted by the bare iron
system.  This is not a memory dump of the running system, but is
synthesized from the target system fasl files.  In the CMUCL world this
process is called "genesis"; in the TI world I think it's spelled
"GENASYS".  There are some references to it in the SSDN2, but I haven't
seen the source code for it in the selection on line.

Moving between different versions of the Lisp Machine and Explorer must
have involved something like this, no?

Tim