[PATCH] Fix combine.c (count_rtxs) (PR middle-end/54879)
Richard Biener
richard.guenther@gmail.com
Wed Oct 10 08:37:00 GMT 2012
On Wed, Oct 10, 2012 at 8:46 AM, Jakub Jelinek <jakub@redhat.com> wrote:
> Hi!
>
> Seems I wrote a patch 7.5 years ago for and on the 3.4 branch
> (supposedly for a testcase that went latent on the trunk at that time,
> due to SSA merge?), and missed the fact that GET_RTX_CLASS changed
> on the trunk from characters to enum.
>
> Fixed thusly, bootstrapped/regtested on x86_64-linux and i686-linux, ok for
> trunk?
Yes, that's obvious.
Thanks,
Richard.
> 2012-10-10 Jakub Jelinek <jakub@redhat.com>
>
> PR middle-end/54879
> * combine.c (count_rtxs): Use RTX_BIN_ARITH resp. RTX_COMM_ARITH
> instead of '2' resp. 'c' for GET_RTX_CLASS comparisons.
>
> --- gcc/combine.c.jj 2012-09-27 12:45:55.000000000 +0200
> +++ gcc/combine.c 2012-10-09 23:35:17.577659296 +0200
> @@ -12015,8 +12015,8 @@ count_rtxs (rtx x)
> const char *fmt;
> int i, j, ret = 1;
>
> - if (GET_RTX_CLASS (code) == '2'
> - || GET_RTX_CLASS (code) == 'c')
> + if (GET_RTX_CLASS (code) == RTX_BIN_ARITH
> + || GET_RTX_CLASS (code) == RTX_COMM_ARITH)
> {
> rtx x0 = XEXP (x, 0);
> rtx x1 = XEXP (x, 1);
> @@ -12024,15 +12024,15 @@ count_rtxs (rtx x)
> if (x0 == x1)
> return 1 + 2 * count_rtxs (x0);
>
> - if ((GET_RTX_CLASS (GET_CODE (x1)) == '2'
> - || GET_RTX_CLASS (GET_CODE (x1)) == 'c')
> + if ((GET_RTX_CLASS (GET_CODE (x1)) == RTX_BIN_ARITH
> + || GET_RTX_CLASS (GET_CODE (x1)) == RTX_COMM_ARITH)
> && (x0 == XEXP (x1, 0) || x0 == XEXP (x1, 1)))
> return 2 + 2 * count_rtxs (x0)
> + count_rtxs (x == XEXP (x1, 0)
> ? XEXP (x1, 1) : XEXP (x1, 0));
>
> - if ((GET_RTX_CLASS (GET_CODE (x0)) == '2'
> - || GET_RTX_CLASS (GET_CODE (x0)) == 'c')
> + if ((GET_RTX_CLASS (GET_CODE (x0)) == RTX_BIN_ARITH
> + || GET_RTX_CLASS (GET_CODE (x0)) == RTX_COMM_ARITH)
> && (x1 == XEXP (x0, 0) || x1 == XEXP (x0, 1)))
> return 2 + 2 * count_rtxs (x1)
> + count_rtxs (x == XEXP (x0, 0)
>
> Jakub
More information about the Gcc-patches
mailing list