Compare packed double-precision (32-bit) floating-point elements in a and b based on the comparison operand specified by imm8.