This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Patch to improve PIC rtx_unstable_p & rtx_addr_can_trap
- To: law at cygnus dot com
- Subject: Re: Patch to improve PIC rtx_unstable_p & rtx_addr_can_trap
- From: John Wehle <john at feith dot com>
- Date: Mon, 21 Aug 2000 13:36:31 -0400 (EDT)
- cc: gcc-patches at gcc dot gnu dot org
> > * rtlanal.c (rtx_unstable_p): The pic register is stable
> > (within one function) and the actual rtx should be used
> > when checking the registers.
> This is not necessarily true. Consider a target where the PIC register
> is call-clobbered. Probably the thing to do is check for that explicitly.
I wondered about that, however the existing test in rtx_varies_p
doesn't check PIC_OFFSET_TABLE_REG_CALL_CLOBBERED so I made a leap
of faith and assumed that reload made sure that the pic register
has a valid value in this case. I take it that this isn't true?
Should I go ahead and install the patch changing the check in
rtx_unstable_p from:
|| x == pic_offset_table_rtx
to:
#if !defined(PIC_OFFSET_TABLE_REG_CALL_CLOBBERED)
|| x == pic_offset_table_rtx
#endif
and make the same change to rtx_varies_p?
> > (rtx_addr_can_trap_p): Pic memory addresses can't trap.
> Assuming the linker & friends all DTRT :-)
Well ... yes. A lot of things in the compiler don't work when
the linker & friends DTWT :-(
-- John
-------------------------------------------------------------------------
| Feith Systems | Voice: 1-215-646-8000 | Email: john@feith.com |
| John Wehle | Fax: 1-215-540-5495 | |
-------------------------------------------------------------------------