Re: [Axiom-developer] Project directions

[Stephen Wilson]

Gabriel Dos Reis <address@hidden> writes:
> The substantial difference is that one between a plateform specific
> assembly language (such as x86) and a portable high level language
> (such as SML or Haskell).

I am curious about the goal.  Is it to write a new
compiler/interpreter/runtime for a new Universal Assembly Language
from scratch?  Modify an existing (simple) Lisp system to provide only
the fundamental building blocks (like CCL)?  Use a combination of
existing tools to implement the foundation (new gcc front end say, or
targeting LLVM. Addressing runtime issues using common garbage
collectors like Bohem-GC plus a good chunk of custom code, etc).  Or
is the goal to identify the minimal requirements and attempt to
provide back end translators to various hosting environments (JVM,
.NET, etc) which provide the needed facilities?  Perhaps a mixture of
the above?

Clearly this has nothing to do with Lisp.  Using a full Common Lisp
system to host an assembly type language which provides nothing more
than primitives like arrays, integer arithmetic, access to system
level resources -- only to reimplemented many of Common Lisp's
features on top of -- is pure bloat.

Perhaps just picking an existing back end which was designed for such
purposes (like LLVM), and writing a boot/spad/aldor compiler on top of
it, would give the best long term results (flexibility via embedded
jit compilation, efficiency in execution, custom crafted runtime
support, native interface to C/C++ libraries, etc, etc).

I am just interested in understanding what the strategy is.
Constructing Axiom on top of a custom UAL implies a lot of work, but
certainly has its advantages.  

Of course, I am not at all confident that the end result would be any
better than just using Lisp to begin with (and in particular, choosing
one lisp, GCL, and allow ourselves to exploit all of its non-standard
features), but I am trying to keep an open mind.


Take care,
Steve