Given combinational block can be divided into 3 partitions so each partition can be executed parallely using pipeline processing with 3 stages.
and in pipelining having K stages we know first output comes after K cycles and then after each cycle one output will come,
so if there are n tasks to be done then in a K stage pipeline,
number of cycles($N_{c}$) = K+n-1 as for first task K cycles needed and for remaining n-1 tasks 1 cycles needed as they are being processed parallely.
one cycle time in pipelining ($T_{p}$) = Max(stage delays)+ buffer delay
Complete operation execution time($C_{n}$) =($N_{c}$) x ($T_{p}$)
so here,
$N_{c}$ = 3 + n-1 = n + 2
$T_{p}$= max(70ps, 40ps, 65ps) + 20ps
= 70ps + 20ps = 90ps
$C_{n}$= (n+2) x 90 ps
Throughput ($T_{n}$) is “number of operations performed per unit time”
In $C_{n}$ time, n operations are performed
hence in 1 unit time number of operations performed = $_{\frac{n}{C_{n}}}$
But here number of operations(n) not given hence in this scenario we consider ideal case where we assume that numbers of operations is very very large hence
$T_{n}$ = n / $T_{p}$(n+2) so n and n+2 are cancelled out as they are equal(approximately)
hence
$T_{n}$= 1 / $T_{p}$
$T_{n}$= 1 / 90 ps
$T_{n}$= 1 / (90 * 10$^{-12})$ s
$T_{n}$= 10$^{9}$ / (90 * 10$^{-3})$s
$T_{n}$= 1 / (90 * 10$^{-3})$ Giga Operations Per Second(GOPS)
$T_{n}$= 1000 / 90 GOPS
$T_{n}$= 11.11 GOPS
