Re: memory predicates and constraints

[Richard Henderson <rth at redhat dot com>]

On Mon, Jul 26, 2004 at 02:05:34PM +0200, Roman Zippel wrote:
> Would a new (probably multi character) constraint via EXTRA_CONSTRAINT_STR 
> be an option, which behaves different depending on the target?

Absolutely this is an option.  A good one, in fact.

The only thing this relies on is that (mem (reg)) must always work,
since this is the form to which reload will fall back if it finds
that the rtl doesn't match the constraint.


r~