For vector access to the memory it is important to have properly aligned chunks of memory. Since LLVM's malloc intrinsic will be removed and clib's malloc does not support alignment - is there a portable, space efficient (i.e. no manual padding) way of allocating memory with alignment constraints?