Pipelining

Question

Approach :

Comments

@MiNiPanda

 

Conceptually if we see then i think, in case of sequential execution there is no need of buffers then are we taking it just for sake of answer matching ?

Plz correct if my concept going wrong somewhere.

---

@MiNiPanda

I tried it by taking random values ( i consider same clock time) , may be i should have tried using different clock rate but will it be a correct approach , solving by values ?

Plz have a look.

---

Yes in sequential execution we don't need buffers. I just showed for pipelining not for non-pipelining system.

I got D) for next question

Let no. of instructions in the entire code be n. And time taken be t secs.

Also i consider each instruction takes same time so here 1 inst will take n/t secs.

n/2 insts. will take t/2 secs.

The speed up is done on these n/2 instructions i.e. in the current scenario the time taken for n/2 instructions be x secs.

Then speedup=(time taken before on n/2 insts)/(time taken now on n/2 insts)

=> 50= (t/2)/x => x= t/100

For rest of the n/2 insts. in current model will take t/2 secs as no improvement is done on that.

So total time taken is  t/100+t/2=51t/100

Over all speed up= (total time before)/(total time now)

=t/(51t/100) =1.96

So atmost 2