Store the lower double-precision (64-bit) floating-point element from a into memory. mem_addr does not need to be aligned on any particular boundary.