This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: Sharing hard registers
- From: law at redhat dot com
- To: "David S. Miller" <davem at redhat dot com>
- Cc: jh at suse dot cz, apl at alum dot mit dot edu, echristo at redhat dot com, gcc at gcc dot gnu dot org, gcc-patches at gcc dot gnu dot org
- Date: Fri, 14 Jun 2002 04:23:59 -0600
- Subject: Re: Sharing hard registers
- Reply-to: law at redhat dot com
In message <20020614.030603.82192502.davem@redhat.com>, "David S. Miller" write
s:
> From: Jan Hubicka <jh@suse.cz>
> Date: Fri, 14 Jun 2002 11:59:16 +0200
>
> Pseudo registers are always shared.
>
> I thought reload rewrites individual pseudo REG rtx expressions as it
> performs spills/reloads, and therefore each pseudo instance must be
> unique and not shared.
Pseudos are modified in-place and this requirement implies that pseudos must be
shared. ie, when we change a pseudo to a mem because it did not get a hard
register, we modify the rtx for the reg in-place to make it a mem and we
expect that doing so will fix all references to the pseudo. Similarly when
we assign a hard register to a pseudo -- we modify REGNO (pseudo) and expect
that all references will be updated with the new pseudo.
The latter is the reason why I think we can make shared hard regs work -- the
compiler already has to deal with shared hard regs after reload. The only
code that should cause us problems is stuff generated in the prologue/epilogue
expanders by the backend.
jeff