[Bug rtl-optimization/87763] [9 Regression] aarch64 target testcases fail after r265398

[wilco at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87763

--- Comment #22 from Wilco <wilco at gcc dot gnu.org> ---
(In reply to Steve Ellcey from comment #21)
> If I look at this specific example:
> 
> int f2 (int x, int y)
> {
>   return (x & ~0x0ff000) | ((y & 0x0ff) << 12);
> }
> 

> Is this because of x0 (a hard register) at the destination in insn 15?

Yes, the hard reg change affects the rtl shape in these tests so it fails to
match in combine. I have a simple fix for the tst_5 and tst_6 failures.

You can check this by ensuring there are no hard registers in the test:

int f2a (int x, int y)
{
   x++;
   y++;
   return (x & ~0x0ff000) | ((y & 0x0ff) << 12);
}

This also fails to match before r265398.