On Mon, Feb 25, 2013 at 10:51 PM, Eric Botcazou <ebotca...@adacore.com> wrote:
>> That's what I believe, too. Still, combine appears to add REG_UNUSED
>> notes for clobbers intentionally. Do you know why it does that?
>
> Hmm, right, it apparently clearly wants to immediately close the "live" ranges
> it creates by adding new clobbers. And, indeed, regrename.c for example uses
> these CLOBBER/REG_UNUSED pairs to compute live ranges for non-operands (the
> pass does its own forward scan of the insn stream). So DF's choice might have
> been made explicitly after all...
You mean this piece of nice code:
/* Step 4: Close chains for registers that die here, unless
the register is mentioned in a REG_UNUSED note. In that
case we keep the chain open until step #7 below to ensure
it conflicts with other output operands of this insn.
See PR 52573. Arguably the insn should not have both
notes; it has proven difficult to fix that without
other undesirable side effects. */
for (note = REG_NOTES (insn); note; note = XEXP (note, 1))
if (REG_NOTE_KIND (note) == REG_DEAD
&& !find_regno_note (insn, REG_UNUSED, REGNO (XEXP (note, 0))))
{
remove_from_hard_reg_set (&live_hard_regs,
GET_MODE (XEXP (note, 0)),
REGNO (XEXP (note, 0)));
scan_rtx (insn, &XEXP (note, 0), NO_REGS, terminate_dead,
OP_IN);
}
?
Good pointer, because it lead me to:
http://gcc.gnu.org/PR52573#c6
which is the same patch as what I posted the other day.
The code in regrename.c is a different "fix" for the same PR, and it
looks to me like it's not the right fix...
On top of the REG_EQ* work, I see more REG_* notes work for GCC 4.9 :-)
Ciao!
Steven
