[RFA]: Fix reload of paradoxical subreg on big-endian platforms when reg is in memory

[Alexandre Oliva]

On Apr 13, 2003, Jan Hubicka <jh@suse.cz> wrote:

>       else if (regno >= FIRST_PSEUDO_REGISTER
> #ifdef LOAD_EXTEND_OP
>                && (GET_MODE_SIZE (GET_MODE (x))
>                    <= GET_MODE_SIZE (GET_MODE (SUBREG_REG (x))))
> #endif

> ^^^^ This will disqualify the paradoxical subregs entirely

> I believe that condition is right.  I don't follow the logic on why
> there is bypass on LOAD_EXTEND_OP.

I think it's because when you have LOAD_EXTEND_OP, you can obtain the
extended value by simply loading the inner subreg, and the implicit
extension will take care of the paradoxical subreg.

-- 
Alexandre Oliva   Enjoy Guarana', see http://www.ic.unicamp.br/~oliva/
Red Hat GCC Developer                 aoliva@{redhat.com, gcc.gnu.org}
CS PhD student at IC-Unicamp        oliva@{lsd.ic.unicamp.br, gnu.org}
Free Software Evangelist                Professional serial bug killer