[Bug tree-optimization/94779] Bad optimization of simple switch

[marxin at gcc dot gnu.org via Gcc-bugs]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=94779

--- Comment #19 from Martin Liška <marxin at gcc dot gnu.org> ---
> 
> One of the work items for the next release is to multi-range enable all
> these consumers that can make use of the information.

I would really appreciate that. I'm don't like integrating VRP into the CSWITCH
pass mainly because it's duplicate work and EVRP runs right before that pass.
And I would prefer a more canonical form of switch statements:

f1(unsigned x) {
  if (x >= 3)
    __builtin_unreachable();

  switch (x) {
    case 0:
      return 1;
    case 1:
      return 2;
    case 2:
      return 3;
  }
}

right now we generate:

switch (x)
    default:
      return 1;
    case 1:
      return 2;
    case 2:
      return 3;

I would rather prefer:
switch (x)
    case 0:
      return 1;
    case 1:
      return 2;
    case 2:
      return 3;
    default:
      __builtin_unreachable ();

Similarly:

int y;

f1(unsigned x) {
  if (x == 2)
    __builtin_unreachable();

  switch (x) {
    case 0:
      return 1;
    case 1:
      return 2;
    case 2:
      return 3;
    default:
      y = 123;
      return 2;
  }
}

is transformed into:
 switch (x) {
    case 0:
      return 1;
    case 1:
      return 2;
    default:
      y = 123;
      return 2;

but I would prefer:

 switch (x) {
    case 0:
      return 1;
    case 1:
      return 2;
    case 2:
      __builtin_unreachable ();
    default:
      y = 123;
      return 2;

having that, we could be able to do a CSWITCH transformation.
I'm planning to return to these situations in the next stage1.