You are on page 1of 2

Assignment2

CS 282: Computer Architecture & Organization

Tanvir Alam
KAUST ID: 113348
tanvir.alam@kaust.edu.sa

Answer to question no 1(a)

Computer A (MFLOPS) Computer B(MFLOPS)


Program 1 200/20 = 10 200/200 = 1
Program 2 200/100 = 2 200/20 = 10
Program 3 200/400 = .5 200/100= 2

Answer to question no 1(b)

Mean Computer A Computer B


Arithmetic (10+2+.5)/3 = 4.167 (1+10+2)/3 =4.334
Harmonic 3/ ( 1/10 + 1/2 + 1/.5) = 1.153 3/ ( 1/1 + 1/10 + 1/2) =1.875
1/3
Geometric (based on ( 1 * 1*1) =1 ( .1 * 5 * 4) 1/3 = 1.26
A)
Geometric (based on ( 10 * .2 * .25) 1/3 = .793 ( 1 * 1*1) 1/3 = 1
B)

Based on Computer A Computer A Computer B


Program 1 1 .1
Program 2 1 5
Program 3 1 4

Based on Computer B Computer B Computer A


Program 1 1 10
Program 2 1 .2
Program 3 1 .25
Answer to question no 2(a)
Let ,x = the speed of FPD be improved to make the program run 4 times faster

speed up S = (Execution Time without enhancement)/(Execution Time with enhancement)

=> S = (100)/(10 + 90/x)


=> 4 = (100)/(10 + 90/x)
=> (10 + 90/x)= 25
=> 90/x = 15
=> x = 90/15
=> x = 6

So, speed of FPD be improved SIX TIMES to make the program run 4 times faster

Answer to question no 2(b)


Theoritically, the maximum speedup in FPD is x, that tends to INFINITY.

Let x = INFINITY.

So, using equation from answer 2(a)

=> S = (100)/(10 + 90/INFINITY)


=> S = 10

So, whatever the amount of improvement applied in FPD , we can achieve maximum 10 fold
increase of performance.

You might also like