Lecture Quizzes

Question 1

Why do we say that a computer is a general-purpose

machine?

Answer 1

It can solve different problems by loading and executing different programs.

Question 2

What are the two main phases of instruction execution?

Answer 2

• Fetch: fetch instruction, increment PC, decode.

- Execute: perform the specified operation.

Question 3

How many memory accesses are usually needed for

executing one instruction?

Answer 3

• At least once for fetching the instruction.

- It may need more for accessing the operands.

Question 4

IBM TS1160 Tape Drive (20TB/tape)

• What type of memory is this?
• What is it mainly used for?
• Can it be used as a main memory? Why?

Answer 4

• Secondary memory, external memory, sequential access
• Backup, archive (secondary memory of archive type)
• No! It is not a RAM.

Question 5

What is the main advantage of the associative

mapping for cache?

Answer 5

A memory block can be mapped to several different

cache lines, which increases cache hit ratio.

Question 6

Is fully associative mapping implemented in

practice?

Answer 6

No! Comparison of all tags is impossible to be done in

reasonable time.

Question 7

Is it true:

• The higher associativity level, the better hit ratio?
• The higher associativity level, the better performance?

Answer 7

• yes

- no

Question 8

What information is specified in an instruction?

Answer 8

• Operation code
• Source operands and destination operand
• Next instruction address, if needed

Question 9

What does it mean by relative addressing?

Answer 9

The effective memory address equals a base address, given in a register, plus an offset.

Question 10

Why is it useful to use relative addressing?

Answer 10

• Program/data can be stored in different MM locations.

- Reduce the number of bits needed for MM addresses.

Question 11

Why do we often
have holes in an
instruction pipeline?

Answer 11

We have different
execution patterns for
different instructions!

Question 12

Which pipeline hazard has the largest impact? Why?

Answer 12

Control hazard, since branches can dramatically reduce pipeline performance, and we have many branch instructions.

Question 13

Programmed I/O, Interrupt-driven I/O, and/or Direct memory
access (DMA)?

• Low-speed I/O
  device, e.g., a
  keyboard

-High-speed I/O
device, e.g., a hard
disk

Answer 13

• No, Yes, No

- No, Maybe, Yes

Question 14

Why is data dependency an important issue for

superscalar architecture?

Answer 14

• A superscalar computer will be efficient only if there are sufficient numbers of instructions to be executed
  concurrently.
• Instructions can be executed concurrently only if there is no data dependency among them.

Question 15

Which data dependencies can be removed?

How?

Answer 15

• Output dependencies and anti-dependencies can be eliminated.
• It is done by using additional registers, which is called