On Thu, Apr 04, 2013 at 05:56:01PM -0500, Richard Henderson wrote:
> We weren't ignoring the high 32 bits during a NE comparison.
>
> Signed-off-by: Richard Henderson <r...@twiddle.net>
> ---
> tcg/ppc64/tcg-target.c | 6 ++++++
> 1 file changed, 6 insertions(+)
>
> diff --git a/tcg/ppc64/tcg-target.c b/tcg/ppc64/tcg-target.c
> index b12cbec..822eb07 100644
> --- a/tcg/ppc64/tcg-target.c
> +++ b/tcg/ppc64/tcg-target.c
> @@ -1129,6 +1129,12 @@ static void tcg_out_setcond (TCGContext *s, TCGType
> type, TCGCond cond,
> tcg_out32 (s, XOR | SAB (arg1, 0, arg2));
> }
>
> + /* Make sure and discard the high 32-bits of the input. */
> + if (type == TCG_TYPE_I32) {
> + tcg_out32(s, EXTSW | RA(TCG_REG_R0) | RS(arg));
> + arg = TCG_REG_R0;
> + }
> +
> if (arg == arg1 && arg1 == arg0) {
> tcg_out32(s, ADDIC | TAI(0, arg, -1));
> tcg_out32(s, SUBFE | TAB(arg0, 0, arg));
Given that it adds one more instruction, I do wonder if we still need a
different implementation than the EQ one and XORI 1. The latter is what
GCC choses.
That said the fix is correct so:
Reviewed-by: Aurelien Jarno <aurel...@aurel32.net>
--
Aurelien Jarno GPG: 1024D/F1BCDB73
aurel...@aurel32.net http://www.aurel32.net
