This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: i386 fix
- To: Jeffrey A Law <law at cygnus dot com>
- Subject: Re: i386 fix
- From: Jan Hubicka <jh at suse dot cz>
- Date: Thu, 27 Jul 2000 21:11:37 +0200
- Cc: gcc-patches at gcc dot gnu dot org
- References: <20000727101609.A27043@cygnus.com> <17530.964719815@upchuck>
>
> In message <20000727101609.A27043@cygnus.com>you write:
> > But the registers are in NO_REGS, so they should never be selected,
> > since they are not of the correct class.
> >
> > I seem to recall that there was some reason for the flags to be this
> > way -- non-fixed no-regs. I see that PA does the same thing; perhaps
> > Jeff remembers.
> I have no idea -- though don't some prologue expanders do things like
> "is reg X used and ! fixed and callee-saved, then save it"?
This is not the case on i386, since the expander iterate just up to
STACK_POINTER_REGNO...
Perhaps the caller-save code? Or my theory of reload needing to free
some specific register and saving the value to another register
capable to hold mode instead of spilling to stack. Then the dirflag
may look like good choice for any integral mode according to our
macros in i386.h
OK, I will try to construct testcase.
Honza
> jeff
>