Move misplaced assignment in num_sign_bit_copies1

[rsandifo]

[ This patch is part of the SVE series posted here:
  https://gcc.gnu.org/ml/gcc/2016-11/msg00030.html ]

The old assignment to bitwidth was before we handled VOIDmode with:

  if (mode == VOIDmode)
    mode = GET_MODE (x);

so when VOIDmode was specified we would always use:

  if (bitwidth < GET_MODE_PRECISION (GET_MODE (x)))
    {
      num0 = cached_num_sign_bit_copies (x, GET_MODE (x),
                                         known_x, known_mode, known_ret);
      return MAX (1,
                  num0 - (int) (GET_MODE_PRECISION (GET_MODE (x)) - bitwidth));
    }

For a zero bitwidth this always returns 1 (which is the most
pessimistic result).

Tested on aarch64-linux-gnu and x86_64-linux-gnu.  OK to install?

Thanks,
Richard


gcc/
2016-11-15  Richard Sandiford  <richard.sandif...@arm.com>
            Alan Hayward  <alan.hayw...@arm.com>
            David Sherwood  <david.sherw...@arm.com>

        * rtlanal.c (num_sign_bit_copies1): Calculate bitwidth after
        handling VOIDmode.

diff --git a/gcc/rtlanal.c b/gcc/rtlanal.c
index 4617e8e..35e95f2 100644
--- a/gcc/rtlanal.c
+++ b/gcc/rtlanal.c
@@ -4795,7 +4795,6 @@ num_sign_bit_copies1 (const_rtx x, machine_mode mode, 
const_rtx known_x,
                      unsigned int known_ret)
 {
   enum rtx_code code = GET_CODE (x);
-  unsigned int bitwidth = GET_MODE_PRECISION (mode);
   machine_mode inner_mode;
   int num0, num1, result;
   unsigned HOST_WIDE_INT nonzero;
@@ -4812,6 +4811,7 @@ num_sign_bit_copies1 (const_rtx x, machine_mode mode, 
const_rtx known_x,
     return 1;
 
   /* For a smaller object, just ignore the high bits.  */
+  unsigned int bitwidth = GET_MODE_PRECISION (mode);
   if (bitwidth < GET_MODE_PRECISION (GET_MODE (x)))
     {
       num0 = cached_num_sign_bit_copies (x, GET_MODE (x),