As anticipated, the LLVM Clang compiler has now enabled the use of its SLP Vectorizer by default.
LLVM developers have been anticipating turning on SLP vectorization by default
and with today's code activity they finally have turned it on when using the -O3
compiler optimization level.
The SLP Vectorizer is able to vectorize certain straight-line code and will then be in place for LLVM/Clang 3.4, after LLVM/Clang 3.3 had enabled use at -O3 of the Loop Vectorizer. The SLP Vectorizer change happened with this commit
and this commit
For those out of the loop on this change that will impact the LLVM 3.4 release, read LLVM 3.3 To Introduce SLP Vectorizer
and Benchmarking The SLP Vectorizer On LLVM Clang 3.4