https://gcc.gnu.org/bugzilla/show_bug.cgi?id=117699
--- Comment #4 from Georg-Johann Lay <gjl at gcc dot gnu.org> ---
(In reply to Andrew Pinski from comment #2)
> I think the problem is in peep2_find_free_register:
> /* Don't use registers set or clobbered by the insn. */
> FOR_EACH_INSN_DEF (def, peep2_insn_data[from].insn)
> SET_HARD_REG_BIT (live, DF_REF_REGNO (def));
>
> This only sets the live info based on the regno here.
>
>
> But it should be instead (with fixed formating):
> ```
> /* Don't use registers set or clobbered by the insn. */
> FOR_EACH_INSN_DEF (def, peep2_insn_data[from].insn)
> add_to_hard_reg_set (&live, GET_MODE (DF_REF_REAL_REG (def)), DF_REF_REGNO
> (def));
> ```
>
> This is similar to what is done in compute_regs_asm_clobbered (ira.cc).
I must admit that I don't completely understand the fix / bug.
peep2_find_free_register() currently reads:
from = peep2_buf_position (peep2_current + from);
to = peep2_buf_position (peep2_current + to);
gcc_assert (peep2_insn_data[from].insn != NULL_RTX);
REG_SET_TO_HARD_REG_SET (live, peep2_insn_data[from].live_before);
while (from != to)
{
gcc_assert (peep2_insn_data[from].insn != NULL_RTX);
/* Don't use registers set or clobbered by the insn. */
FOR_EACH_INSN_DEF (def, peep2_insn_data[from].insn)
SET_HARD_REG_BIT (live, DF_REF_REGNO (def));
from = peep2_buf_position (from + 1);
}
For the test case we have from = to = 3, so that while loop isn't even
executed.
I also tested the patch. The patch has no effect on the test case.
