Languages
Mike Prince
mprince@crl.com
Wed, 14 Dec 1994 15:23:35 -0800 (PST)
On Wed, 14 Dec 1994, Francois-Rene Rideau wrote:
> That is system *execution* is founded (no chicken & egg problem), but
> system specification need not be...
I don't agree, without specifications modems, floppy disk, networks,
telephones, human language, etc would not work. All HLLs eventually
instantiate into something concrete. We need to specify that.
> Yes I'm sure the LLL should map well such manipulations. Virtual machines
> commonly used as byte-code target for Lisp or ML, etc, do that everyday !
> Why restrict the power of our LLL to manipulate only first-order objects ?
> LLL does not mean "manipulate low-level objects only" but "low-level
> transformations on object", that is, transformations that involve actual
> representations of objects in some implementation model.
Please suggest a model for the LLL that would allow rapid
compilation/execution.
> Again, at execution, we have a founded system. Founding the system is the
> usual bootstrap problem (need of original cross-compilers and
> cross-assemblers). But once you got the system running, calls to the memory
> object by the driver than manages further links to the memory object are
> already inlined. So really, this is not the difficult problem.
> As for the messages to the CPU object, these are the binary instructions,
> once properly inlined !!!
What if I modify a base class object that has been inlined everywhere.
Does the entire system get re-compiled? How do you float execution from
the old system software to the new software?
Mike