The intention is to enable the extract_vector_elt load combine, and doing this for other operations interferes with more useful optimizations on vectors. Handle any type of load since in principle we should do the same combine for the various load intrinsics. llvm-svn: 341219
25 KiB
25 KiB