[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Axiom-developer] new compiler, ncloopProcess

From: daly
Subject: [Axiom-developer] new compiler, ncloopProcess
Date: Wed, 17 Jan 2007 10:13:17 -0600

Scratchpad was a research project used by many people in 
different ways. As such it was always a "work in progress".

Several people worked on "the new compiler" over time.
The original plan was to have the new compiler exist in
the lisp system. Stephen Watt decided to implement it in C
as a standalone program and this became the final version.
However there were many "new compilers" in lisp both before
and during that effort. The early compilers tried new ideas
and the later compilers maintained compatibility with aldor.

The "new compiler" (remember that the word "new" basically
meant "this week's changes") was carried on in parallel in
both C (asharp,aldor) and lisp. We had to keep the syntax
and semantics as close as possible. The "old compiler" and 
the "new compiler" existed at the same time in the lisp
system. (Note that "the old compiler" refers to last week's
version and "the new compiler" was this week's version).

This was also true with the "old boot parser" and "the new
boot parser" and the interpreter, etc. Everything was always

IBM hit a rough financial patch and Scratchpad was turned
into a "product" and sold to NAG. The process was not very
clean as we had very little time. Plus we had other tasks
such as adding the fortran interface so we could talk to
the NAG library, additional code changes to handle Arthur
Norman's CCL and a huge documentation effort.

Thus the code you see has never had a "proper" cleanup.
We need to select and collect the live code, understand it,
document it, and restructure it so it can be maintained.
Thus it is hardly a surprise to find duplicate functions
with different definitions. Careful study is needed to
decide which one is used at what time (since some can be
autoloaded, overriding the existing definition).

Rather than taking each function individually I've been
walking the spanning code tree and dragging the functions
into bookvol5.pamphlet which is intended to be the new
interpreter. Someone needs to do something similar with
the compiler. This "garbage collection" approach will 
prove that any extra functions must be unused and can be

Also of note with regard to ncloopProcess.... There are
several "top level"s of the Axiom system which are not
apparent. There are functions which start the boot parser,
compiler, or interpreter in different modes to do different
things. This is not documented and not apparent. It was
intended for use by the developers of axiom so they could
debug their programs. Some of these ended up abandoned.


reply via email to

[Prev in Thread] Current Thread [Next in Thread]