[LispM] World Files

[Robert Swindells]

ams at gnu.org (Alfred M. Szmidt) wrote:

>   For an odd bit of entertainment, I have been attempting to decypher
>   world files.  (I have an Ivory/XL1200.) Based on information
>   available in i-sys/sysdef.lisp and netboot/world-substrate.lisp I
>   have made some progress.  I have, however, hit a wall and am
>   looking for some inspiration.
>
>   When the FEP loads and boots a world, it must read the world file
>   into memory, build an environment for it and, eventually, vector to
>   it.  It appears that the definition of this "environment" is not
>   readily obvious from the source code.  It appears, for example,
>   that the function QLD is the initial entry point inside the LISP
>   world (although how we get to that function is unclear to me).
>   Also, the world file describes both wired and un-wired regions of
>   memory.  How the unwired page table is communicated to the code
>   within the world and where it should be located is equally not
>   obvious.
>
>   Has anyone tinkered around at this level, and would be in a
>   position to provide some hints?  I am trying to understand the FEP
>   -> World protocol.
>
>Ten years later, we get an answer -- but not exactly what you might be
>looking for.  Did you get any further in your digging? :-)

>Here is a fun document on cold loading:

Your document describes how LMI bootstrapped a new Lisp image, a
process that had evolved from the way the MIT CADR did things, their
final source tree is in Joe Marshall's repository.

I have made changes to the MIT sources to allow QLD to pull in the
extra object files needed using TFTP.

I have no idea how much Symbolics had changed their bootstrap process,
it is interesting though that the entry function still has the same
name.

Robert Swindells