One technique to help solve this problem repeatedly recomputes values that are “never killed”; that technique is rematerialization. A particular graph being perfect and chordal when the input to this problem is in SSA makes it much easier to perform. In 1981, Chaitin et al. proved this problem was NP-complete by modeling it (15[1])as coloring of the interference graph. (-5[1])In solving this problem, coalescing or splitting steps might help minimize (-5[1])“spillage,” which is insertion of expensive (*) loads and stores; (10[1])however, spilling must be (10[1])part of a solution to this problem when there are too many live variables. Calling conventions dictate portions of solutions to this late stage of compilation; for example, x86 requires (-5[1])return values to be in EAX when functions end. For 10 points, name this compilation stage that designates physical units of processor storage to hold variables and expressions. (10[1]0[9])■END■