A benchmark has a breakdown of the following:

$39\%$ foads, $12\%$ stores, $28\%$ ALU operations other than multiplies$/$divides$,6\%$ multiplies, $2\%$ divides, $10\%$ conditional branches, $3\%$ unconditional branches

A processor has a CPI of $5$ for loads$/$stores$,3$ for all ALU operations other than multiplies$/$divides$,6$ for multiplies, $15$ for divides, and $4$ for branches.

a$)$ Improving the multiplier so that the CPI for multiplies is reduced to $3$

b$)$ Improving the ALU so non$-$multiply$/$divide operations have a CPI of $2$