I compiled the x86 code using gcc 5.3 using mtune/march=native and -O3. It just turns out that the gcc compiler generates entirely scalar code for SPECInt!
That totally seems wrong. I'm certain it should fire on SPECint, since all compiler optimizations are pretty much invented just to improve the SPECint score - gcc's autovectorizer was made to do this for IBM's POWER processors, which are of course not x86.
