[Bug target/93722] New: rorq is not produced for rotate on some cases

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=93722

            Bug ID: 93722
           Summary: rorq is not produced for rotate on some cases
           Product: gcc
           Version: 10.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: enhancement
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: pinskia at gcc dot gnu.org
  Target Milestone: ---
            Target: x86_64-linux-gnu

Take:

void f0(unsigned long *a)
{
  __uint128_t t0 = ((__uint128_t *)a)[0];
  __uint128_t t1 = t0>>sizeof(unsigned long)*8;
  __uint128_t  t2 = t0<<sizeof(unsigned long)*8;
  ((__uint128_t*)a)[0] = t1 | t2;
}

--- CUT ---
We should just produce:
        rolq    $32, (%rdi)
        ret

--- CUT ---
But we produce:
        movq    8(%rdi), %rdx
        movq    (%rdi), %rax
        movq    %rdx, (%rdi)
        movq    %rax, 8(%rdi)
        ret

Note it gets worse when using rotate to half-way