Re: RFC: arm_legitimate_address_p change for minipool constants

[Daniel Jacobowitz <drow at mvista dot com>]

On Fri, Jan 23, 2004 at 10:48:05AM +0000, Richard Earnshaw wrote:
> Yep.  Hmm, I've just remembered.  The point of looking up the value was to 
> solve the reverse case: when we are accessing a subreg of a larger object, 
> we want to have the value being accessed in the minipool, not the wider 
> constant.

Ah, this makes sense.

> >  Ideal would be to, at
> > this point, replace the MEM with a const_int instead of fixing up its
> > minipool reference, if CONST_OK_FOR_LETTER_P.
> 
> I really think an earlier pass should have resolved that case.  If you can 
> find a test case which demonstrates this, I'll have a look at it.

I will try.  I need to sit down with my 3.3 tree and figure out how
that expression is produced, first.

> I'm going to approve this for HEAD only for now, unless you can produce a 
> test case that fails on the 3.4 branch.

OK, I'll check it in for HEAD, and work on a testcase.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer