This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Reducing RTL allocations
- From: law at redhat dot com
- To: Alexandre Oliva <aoliva at redhat dot com>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Sun, 09 Jun 2002 09:55:32 -0600
- Subject: Re: Reducing RTL allocations
- Reply-to: law at redhat dot com
In message <orbsaly8lj.fsf@free.redhat.lsd.ic.unicamp.br>, Alexandre Oliva writ
es:
> On Jun 7, 2002, law@redhat.com wrote:
>
> > So what I've done is initialized regno_reg_rtx for hard registers -- you
> > can use it to get a hard register of mode reg_raw_mode[regno] as long as
> > you don't mind it being shared.
>
> Erhm... It appears to me that using shared hard reg rtx is not
> appropriate in reload, since regrename modifies regs in place. What
> am I missing?
How could regrename safely modify any reg in-place since some of the hard
regs is sees come from pseudos which were turned into hard regs (which are
shared).
The cases where I see regrename making changes it generates a new REG rtx via
gen_rtx_RAW. It was a relatively quick scan, so I could have missed something.
In what function within regrename do we do an in-place modification of the
register number within a REG rtx?
jeff