[Bug tree-optimization/63464] compare one character to many: faster

["jakub at gcc dot gnu.org" <gcc-bugzilla@gcc.gnu.org>]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=63464

Jakub Jelinek <jakub at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2014-10-06
                 CC|                            |jakub at gcc dot gnu.org,
                   |                            |rguenth at gcc dot gnu.org,
                   |                            |steven at gcc dot gnu.org
     Ever confirmed|0                           |1

--- Comment #1 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
We have this optimization implemented for switches, if you compile
char*f3(char*s){
  do
    {
      switch (*s)
        {
        case ' ':
        case ',':
        case '\r':
        case '\n':
          ++s;
          continue;
        default:
          return s;
        }
    }
  while (1);
}

then it will do the bit test, see r189173 (and various follow-up fixes for
that).
Now, we can argue whether in this case it is beneficial to perform the
MINUS_EXPR or if maxval is small enough (e.g. when maxval is smaller than
BITS_PER_WORD), just assume minval is 0.

And then the question is, if we should teach reassoc range optimizations to
reuse emit_case_bit_tests, or convert such tests into a GIMPLE_SWITCH and
expand as such.

Richard/Steven, thoughts about this?