Problem $8$:

Consider a program that consists of a purely sequential part which takes $500$ cycles to execute and a $40-$iteration

loop which takes $35$ cycles per iteration on a multicore processor. Assume the loop iterations are independent and

cannot be further parallelized, i$.$e$.,$ each loop iteration must fully run on a single core, but different iterations can

run on different cores. Using Amdahl's law, calculate the program speedups $($compared to a single core$)$ in case the

processor has $2,4,8,10,$ and an infinite number of cores.Problem $7$:

Suppose we enhance a computer to make all floating$-$point instructions run $5$ times faster. Answer the following

questions:

a$.$ If the execution time of some benchmark before the floating$-$point enhancement is $80$ seconds, what will

the speedup be if $70\%$ of the $80$ seconds is spent executing floating$-$point instructions?

b$.$ How much of the execution time would floating$-$point instructions have to account for in a benchmark

that runs for $120$ seconds with the old floating$-$point hardware in order to yield a speedup of $4?$

c$.$ Assuming that we enhance the computer as described above, plot the speedup obtained versus the

fraction of time in the original benchmark spent doing floating$-$point operations