The article states that the compilation was tailored for this specific task, which means it doesn't represent how the CPU would normally perform it.
So it's definitely cheating.
A benchmark is not supposed to be a compiler optimization competition. If they showed both, and revealed it was optimized, it would be another matter.