> And no one does inline assembly for that, they use intrinsics Depends on your architecture. With ARM, compilers will often not set the alignment bits in vect

> And no one does inline assembly for that, they use intrinsics Depends on your architecture. With ARM, compilers will often not set the alignment bits in vector loads and stores, and this can be a big performance hit depending on the microarchitecture. In general, compilers sometimes deal poorly with instructions that have particularly strange register constraints, or loads/stores with address writeback.