Consider an integer benchmark with a breakdown of instructions of $50\%$ load$/$store$,40\%$ ALU and

$10\%$ branch and a CPI of $5$ for loads and stores and $4$ for ALU and branch. Given that the

benchmark uses only int registers, the FP registers are not used at all. In such a case, imagine that

the compiler is able to freely use the FP registers to store integer values, thus reducing the number

of loads and stores in the program. The CPU time does not change but the IC $*$ CPI does as some

instructions go from load$/$store to ALU. Assuming the compiler can successfully reduce $20\%$ of

the loads$/$stores$,$ what is the speedup? Use both the CPU Time formula and Amdahl$$s Law.