This option is used when the user wants to avoid emitting double precision FP
loads and stores. Double precision FP loads and stores are expanded to single
precision instructions after register allocation.

llvm-svn: 181718