Comment by danlark

In this particular case they were universal but in paper it's said the optimizations were done on x86. One of the ideas was to use LLVM IR but intuition for optimizer over optimizer was unlikely to work properly.