Multiplication: set t1 to the upper 32 bits of t2*t3 where t2 is signed and t3 is unsigned