[Bug target/83210] New: __builtin_mul_overflow() generates suboptimal code when exactly one argument is the constant 2

[lh_mouse at 126 dot com]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83210

            Bug ID: 83210
           Summary: __builtin_mul_overflow() generates suboptimal code
                    when exactly one argument is the constant 2
           Product: gcc
           Version: 7.2.0
            Status: UNCONFIRMED
          Keywords: missed-optimization
          Severity: normal
          Priority: P3
         Component: target
          Assignee: unassigned at gcc dot gnu.org
          Reporter: lh_mouse at 126 dot com
  Target Milestone: ---

This snippet illustrates the problem:

<https://godbolt.org/g/TssrYv>

```
bool mul_and_check(unsigned *dst, unsigned src){
    return __builtin_mul_overflow(src, 2, dst);
}
```

With `g++ -O3`, this compiles to:

```
mul_and_check(unsigned int*, unsigned int):
  mov eax, esi
  mov edx, 2
  xor ecx, ecx
  mul edx
  mov esi, eax
  jo .L5
.L2:
  mov eax, ecx
  mov DWORD PTR [rdi], esi
  and eax, 1
  ret
.L5:
  mov ecx, 1
  jmp .L2
```

This is very suboptimal. GCC could have used a bitwise shift operation instead,
as follows:

```
mul_and_check(unsigned int*, unsigned int):
  xor eax, eax             # EAX  = 0
  shl esi, 1               # CF   = MSB(src)
                           # src  = src * 2;
  rol eax, 1               # EAX  = CF
  mov dword ptr[rdi], esi  # *dst = src
  ret
```