Re: SH: use register names instead of numbers

["Zack Weinberg" <zackw at stanford dot edu>]

On Sat, Nov 18, 2000 at 06:06:27AM -0200, Alexandre Oliva wrote:
> On Nov 17, 2000, Joern Rennecke <amylaar@redhat.com> wrote:
> 
> > OTOH if we did the expansion in rtl.c, we'd get better error messages -
> > i.e. the context will be from the source file, not from some preprocessed
> > file.
> 
> Agreed.  It's definitely a better approach to the problem.  The C
> preprocessor was not designed to preprocess lisp-like languages.  I'd
> rather just drop this ill-conceived misfeature of the Makefile and do
> it right.

In this context, I'd like to point out that CGEN uses a modified RTL,
and it has a macro facility.  It'd probably be wise to copy it instead
of inventing our own incompatible mechanism.  It does not appear to
have conditional processing or file inclusion - yet.  CGEN's written
in Scheme, but I don't think it'd be that hard to implement the macro
logic in gensupport.c.

For this specific problem, I'd like to see a construct like this:

(define_constants [
   (foo 12)
   (bar 23)
   (baz 34)
   ...
])

You can have multiple instances, each one enlarges the table, it's an
error to redefine a constant with a different value.  You could do
this with CGEN's macros but it'd be a lot more verbose.

zw