Architecture Qualifying Exam Fall 2000


1. (A)

40 - 14 = 26
26 + 6 = 32 bits (4 bytes)

1. (B)

58 - 14 = 44
each level a page 16 K / 4 = 4 K entries per page, which is 12 bit.
So, devide 44 by 12 , we get 12, 12, 12, 8.
Hence the multi-level paging scheme, looks like the following:
------------------------------------------
8 12 12 12 14
------------------------------------------

1. (C)

64 K / 16 k = 4 (way)

2. A.

In MSI coherence strategy, M means Modify, S means Shared, I means Invalidated.
1. P1 read 100      100 is S in P1.
2. P2 read 104      104 is S in P2.
3. P1 send I 100      100 is M in P1.
4. P3 read 100      100 is S in P3.
    P1 flush 100      100 is S in P1 (also P1 could send flushed, and P3 could have it as M. )
5. P4 read & send I 104      104 is I in P2, 104 is M in P4.
6.                       100 is I in P3 (if P1 is flushed, then P3 has to write back).

2. B.

1. P1 read 100      100 is S in P1
2. P2 read 100      100 is S in P2
3. P1 send I 100      100 is I in P2, M in P1
4. P3 read 100      100 is S in P1, S in P3
5. P4 read & I 100     100 is I in P1, P3, M in P4
6. nothing

3. A.

        Bandwidth     Latency    cost
-----------------------------------------------------------------------------------------
Bus           O(1)     O(1)    O(n)
-----------------------------------------------------------------------------------------
Ring          O(n)     O(n)    O(n)
-----------------------------------------------------------------------------------------
Crossbar     O(n)    O(1)    O(n^2)
-----------------------------------------------------------------------------------------
Binary Hypercube     O(n)    O(logn)    O(nlogn)
-----------------------------------------------------------------------------------------
2D Mesh    O(n)    O(n^1/2)    O(n)

3. B.

Latency become much less important if using wormhole routing
Must use bisection bandwidth for bandwidth

4. A

1.    lw r1, 100(r2)
2.    lw r3, 200(r1)
3.    sw r3, 100(r4)
4.    lw r4, 100(r5)
5.    add r6, r3, r4
6.    sw r6, 300(r5)
7.    addi r2, r2, 4
8.    add r4, r2, r4

r1 from instruction 1 to instruction 2
r3 from instruction 2 to instruction 3
r3 from instruction 2 to instruction 5
r4 from instruction 4 to instruction 5
r4 from instruction 4 to instruction 8
r6 from instruction 5 to instruction 6
r2 from instruction 7 to instruction 8

4. B

1.    lw r1, 100(r2)
2.    lw r3, 200(r1)
3.    sw r3, 100(r4)
4.    lw r4, 100(r5)
5.    add r6, r3, r4
6.    sw r6, 300(r5)
7.    addi r2, r2, 4
8.    add r4, r2, r4

r2 from instruction 1 to instruction 7
r4 from instruction 3 to instruction 4
r4 from instruction 5 to instruction 8

4. B

lw r1, 100(r2)   F D X M W
lw r3, 200(r1)       F D    X M X
sw r3, 100(r4)             F D X M W
lw r4, 100(r5)                  F D X M W
add r6, r3, r4                        F D    X M W
sw r6, 300(r5)                              F D X M W
addi r2, r2, 4                                     F D X M W
add r4, r2, r4                                         F D X M W