Store packed single-precision (32-bit) floating-point elements from a into memory using mask. See: "Note about mask load/store" to know why you must address valid memory only.