[Bug tree-optimization/127504] New: `-cmp * b` -> `cmp ? -b : 0`

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

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127504

            Bug ID: 127504
           Summary: `-cmp * b` -> `cmp ? -b : 0`
           Product: gcc
           Version: 17.0
            Status: UNCONFIRMED
          Keywords: easyhack, missed-optimization
          Severity: enhancement
          Priority: P3
         Component: tree-optimization
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org
  Target Milestone: ---
            Target: riscv aarch64

Take:
```
int f(int a, int b)
{
  int c = b != 10;
  c = -c;
  return a * c;
}

int f1(int a, int b)
{
  int c = b != 10;
  return -(a * c);
}
```

These both should produce the same code.

Both aarch64 and riscv (with zicond) produce better code for f1.

There is a pattern to handle (CMP * a) but none to handle ((-CMP) * a).

The pattern for CMP*a:
```
/* (m1 CMP m2) * d -> (m1 CMP m2) ? d : 0  */
(if (fold_before_rtl_expansion_p ())
 (for cmp (tcc_comparison)
  (simplify
   (mult:c (convert (cmp@0 @1 @2)) @3)
   (if (INTEGRAL_TYPE_P (type)
        && INTEGRAL_TYPE_P (TREE_TYPE (@0)))
     (cond @0 @3 { build_zero_cst (type); })))
/* (-(m1 CMP m2)) & d -> (m1 CMP m2) ? d : 0  */
  (simplify
   (bit_and:c (negate (convert (cmp@0 @1 @2))) @3)
   (if (INTEGRAL_TYPE_P (type)
        && INTEGRAL_TYPE_P (TREE_TYPE (@0)))
     (cond @0 @3 { build_zero_cst (type); })))
 )
)
```

So adding one for this case here would be useful.  This could show up if you
are using -1 as a mask to emulate vector masks.