Store the lower single-precision (32-bit) floating-point element from a into memory. mem_addr does not need to be aligned on any particular boundary.