Failure to optimize 128-bit division

|  |  |
| --- | --- |
| Bugzilla Link | [52168](https://llvm.org/bz52168) |
| Version | trunk |
| OS | Linux |
| CC | @ecnelises,@RKSimon |

## Extended Description 
```cpp
__int128 b(__int128 j)
{
    return j / 5;
}
```

This can be inlined into a series of operations avoiding the division, but LLVM just outputs a call to __divti3. GCC does this optimization.