On Fri, Apr 17, 2015 at 11:32:44AM -0500, Bill Schmidt wrote:
> 2015-04-17 Bill Schmidt <wschm...@linux.vnet.ibm.com>
>
> PR target/65787
> * config/rs6000/rs6000.c (rtx_is_swappable_p): Remove previous
> fix; ensure that a subsequent SH_NONE operand does not overwrite
> an existing *special value.
>
>
> Index: gcc/config/rs6000/rs6000.c
> ===================================================================
> --- gcc/config/rs6000/rs6000.c (revision 222182)
> +++ gcc/config/rs6000/rs6000.c (working copy)
> @@ -34204,17 +34204,6 @@ rtx_is_swappable_p (rtx op, unsigned int *special)
> else
> return 0;
>
> - case PARALLEL:
> - /* A vec_extract operation may be wrapped in a PARALLEL with a
> - clobber, so account for that possibility. */
> - if (XVECLEN (op, 0) != 2)
> - return 0;
> -
> - if (GET_CODE (XVECEXP (op, 0, 1)) != CLOBBER)
> - return 0;
> -
> - return rtx_is_swappable_p (XVECEXP (op, 0, 0), special);
> -
> case UNSPEC:
> {
> /* Various operations are unsafe for this optimization, at least
> @@ -34308,6 +34297,8 @@ rtx_is_swappable_p (rtx op, unsigned int *special)
> {
> unsigned int special_op = SH_NONE;
> ok &= rtx_is_swappable_p (XVECEXP (op, i, j), &special_op);
> + if (special_op == SH_NONE)
> + continue;
> /* Ensure we never have two kinds of special handling
> for the same insn. */
> if (*special != SH_NONE && special_op != SH_NONE
The " && special_op != SH_NONE" test from the second if can go then,
because it is never true. And I'd really think that you shouldn't change
just the fmt[i] == 'E' handling, but also the fmt[i] == 'e' || fmt[i] == 'u'
handling a few lines earlier (both the added
"if (special_op == SH_NONE) continue;" there and
removal of " && special_op != SH_NONE".
Jakub
