Advanced Computer Architecture - Lecture 3: Quantitative principles (Cont’d)

CS 704
Advanced Computer Architecture

Lecture 3

Quantitative Principles … Cont’d

Design for Performance
Prof. Dr. M. Ashraf Chughtai
MAC/VU-Advanced
Computer Architecture

Lecture 3 - Performance... Cont'd

1


Today’s Topics
Recap
I/O performance
Laws and Principles
Performance enhancement
Concluding: quantitative principles
Home work
Summary
MAC/VU-Advanced
Computer Architecture

Lecture 3 - Performance... Cont'd

2

Recap: Lecture 1-2
Computer architecture verses
organization
Technological Developments
Computer design cycle
Performance metrics: time verses
throughput
Price-Performance design
Benchmarks: Performance evaluation
MAC/VU-Advanced
Computer Architecture

Lecture 3 - Performance... Cont'd

3


Computer I/O System
Producer-Server model
– Producer: the device that generates request to be serviced
– Queue: the area where the tasks accumulate waiting to be serviced
– Server: the device performing the requested service
– Response Time: the time a task takes from the moment it is
placed in the buffer to the time server finishes the task

Producer
Arrivals

MAC/VU-Advanced
Computer Architecture

Queue

Server
departures

Lecture 3 - Performance... Cont'd

I/O device/
controller

4


I/O Performance Parameters
Diversity: Which I/O device can
connect to the CPU

Capacity: How many I/O devices can
connect to the CPU

Latency: Overall response time to
complete a task

Bandwidth: Number of task completed
in specified time - throughput
MAC/VU-Advanced
Computer Architecture

Lecture 3 - Performance... Cont'd

5


I/O Transaction Time
The interaction time or transaction time of
a computer is sum of three times:

– Entry Time:

the time for user to enter a

command – average 0. 25 sec; from keyboard 4.0 sec.

– System Response Time:

time between
when user enters the command and system
responds

– Think Time: the time from reception of the
command until the user enters the next
command

MAC/VU-Advanced
Computer Architecture

Lecture 3 - Performance... Cont'd

6


Response time – latency ms

Throughput verses Response time:

200

_

150

_

100

_

50

_








20
0

Performance Measures .. Cont’d

|
0%

|
20%


|
40%

|
60%

|
80%

|
100%

% of maximum throughput - bandwidth
MAC/VU-Advanced
Computer Architecture

Lecture 3 - Performance... Cont'd

7


Response time and throughput calculation
Arrivals

Departures

If the system is in steady state, then the
number of tasks entering the system must be
equal to the number of tasks leaving the system

Little’s Law:
Mean number of tasks in system =
Mean response time x Arrival rate
MAC/VU-Advanced
Computer Architecture

Lecture 3 - Performance... Cont'd

8


Little’s Law – A Little queuing theory
Mean number of tasks in the system =
(Time accumulated) / (Time observe)
Mean response time =
(Time accumulated) / (Number tasks)
Arrival rate λ =

(Number tasks) / (Time observe)
The expression for mean number of task may be written as:
Time accumulated
Timeaccumulated x Number tasks
=
Time observe

Number tasks

Time observe

Mean number of tasks
response
time x Arrival rate
9
Lecture=
3 - mean
Performance...
Cont'd

MAC/VU-Advanced
Computer Architecture


Amdahl's Law
Suppose that enhancement E accelerates a
fraction F of the task by a factor S, and the
remainder of the task is unaffected
Time for Fraction F to be
Enhanced by factor S

Original
Execution
Time of Task
MAC/VU-Advanced
Computer Architecture

Execution time of the
Fraction Enhanced

Execution
Time
after fraction F
Enhanced by factor S
Lecture 3 - Performance... Cont'd

10


Amdahl's Law
Speedup due to enhancement E:
Ex Time without E
Speedup (E) =
Ex Time with E
Performance with E
=
MAC/VU-Advanced
Computer Architecture

Performance without E

Lecture 3 - Performance... Cont'd

11


Amdahl’s Law
Ex Time new = Ex Time old x (1 – Fraction enhanced) + Fraction enhanced
Speedup enhanced

MAC/VU-Advanced
Computer Architecture

Lecture 3 - Performance... Cont'd

12


Amdahl’s Law
ExTimenew = ExTimeold x (1 - Fractionenhanced) + Fractionenhanced
Speedupenhanced

Speedupoverall =

MAC/VU-Advanced
Computer Architecture

ExTimeold
ExTimenew

1

=

(1 - Fractionenhanced) + Fractionenhanced

Lecture 3 - Performance... Cont'd

Speedupenhanced

13


Amdahl’s Law
Floating point instructions improved to
run 2X; but only 10% of actual
instructions are FP

ExTimenew =

Speedupoverall =

MAC/VU-Advanced
Computer Architecture

Lecture 3 - Performance... Cont'd

14


Amdahl’s Law
Floating point instructions improved to

run 2X; but only 10% of actual
instructions are FP

ExTimenew = ExTimeold x (0.9 + .1/2) = 0.95 x ExTimeold
Speedupoverall =

MAC/VU-Advanced
Computer Architecture

1
0.95

Lecture 3 - Performance... Cont'd

=

1.053

15


Amdahl’s Law
ExTimenew = ExTimeold x (1 - Fractionenhanced) + Fractionenhanced
Speedupenhanced

Speedupoverall =

MAC/VU-Advanced
Computer Architecture

ExTimeold
ExTimenew

1
=

(1 - Fractionenhanced) + Fractionenhanced

Lecture 3 - Performance... Cont'd

Speedupenhanced

16


Amdahl’s Law
Solution

ExTimenew = ExTimeold x (0.9 + .1/2) = 0.95 x ExTimeold
Speedupoverall =

MAC/VU-Advanced
Computer Architecture

1
0.95

Lecture 3 - Performance... Cont'd

=

1.053

17