0% found this document useful (0 votes)
278 views

Introduction To Computer Organization

This document discusses computer performance and metrics. It covers the following topics in 3 chapters: 1. Computer abstractions and technology, including semiconductors and how feature size affects chip area and cost. 2. Performance concerns like throughput, response time, and energy consumption. Smaller chip size and feature reduce cost but increase design complexity. 3. Performance metrics like time, speedup ratio, and measuring execution time using clock cycles. It discusses sequential and combinational logic in processors.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
278 views

Introduction To Computer Organization

This document discusses computer performance and metrics. It covers the following topics in 3 chapters: 1. Computer abstractions and technology, including semiconductors and how feature size affects chip area and cost. 2. Performance concerns like throughput, response time, and energy consumption. Smaller chip size and feature reduce cost but increase design complexity. 3. Performance metrics like time, speedup ratio, and measuring execution time using clock cycles. It discusses sequential and combinational logic in processors.
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

SYSLABUS:

* -

CHAPTER 1: COMPLETE]
CHAPTER 2:
CumPTER 4
CHAPTER 6 % complete
CAROTER 5 ->
partial topics
COMPUTER PERFORMANCE
#RDUCTION,

* CompuTzR ABSTRACTIONS & TechNonocy

* PERFORMANCE
concerned with the with it.
components & the cost associated
->
performance of
->
Performance: is Throughput ii)Response time / Execution time (These 2 are units measurement)
-

of
-> Be aware of the problem first & then solve it (don't
to have high yield andless power usage
We want computers
go
&
haphazardly).
processors all semiconductions
->
as

semiconductors work well in 19-24 & semiconductors disintegrate in hot temperatives

Energy: -
ECt) JPCt)dE
=

(sum of power over a period of time)


->
Energy problems happen due to consumption.
->
you pay for the energy consumed.

Cost: Here its with his concerned with chip area


also indirectly associated
with
associated design complexity,
-> -

More complex - more


expensive to verify # logic gates
in a
processor.
Things easy to verify: -

: Easy bug design error detect


a. less time debugging design errors.

3.

Main concern: Speed

-> Went to History of Architecture


Computer to World War to
related # now

Chip surface (siteinianDesigning chips is complex as it with each gate


contains million gates having
multiple transistors

&on featuresizeof
technology egnumtechno and
feature size

gates

->
You need to design & produce machines who can design such chips with small feature size => more expense

Feature Size Area Feature size Area

A -N2
5nm 10 nm

10nm 2nm 1/25


=

-> As we want to
architects use a design with smaller no of gates
& area.

>
Chips decrease in
size over time.
need some
economy with design, use less circuits
without compromising performance.
->

Centripetal force -> 0


cost is with gates & chip design
associated

Target Frequency for processors

Binning -> in data warehousing


↳ Here associated
with clock speed of chips.

* PERFORMANCEMETRICS
& Rate
Time
We use time occassionally for performance companion
1 lore: Clock Speed:1 Ga
Get with higher clock speed. Opth
1 to increase frequency
1. a core

Get another of
2. core same
frequency.
↳ - 26 or p
time t -t -> Tells time used
by programs himse OS.
in

Python Time Package Module

->
System Time: -
Time used by sys programs for availing OS services.
-> User CPU Time: Time -

required by men
program during
executh.
Reproduceability:
-> - desirable feature
in CA.

-> For s/0 testing, focus on instruct at hand. Write simulators.

Speedup
=>
at a Performance Ratio.

Performance Ratio: k
=

Sometimes it will be as per usage.

1. Improvement (Speedup -1) =


x180%

·
MEASUREEXECUTION TIME
ARM processors can measure earth time using clock cycles.

Processor Logic Cienif +

2
Types of hogic
1. Sequential: Memory - ->
i/p->State ->
o/p:f(ip, State)
2. Combinational Combinatorial: -
Gates
- no memory
↳ olp:-f(i/p)

Machine
e.g.-
Vending
i/p:-Cash,
Buttons for drink.

clock ....... ->


s9
I
rock cycle/pewod
Frequency a ha. Clock Rate
(requency)
"
Clock Period &
ET truck
Count
#Clock cycles Clock Period
ET=
tochascles on =

ET =

&
or
Clock Rate

Clock period Rock Cycles Time

areasomuc
P.:1x109 2x10 cycles
-

P2 25x189 8x109 cycles

Clock cycle Y.
A Icycles 30%
30%
B
Gcycles
2 I
cycle 40%
10 "instructu

CPlans program (cycles read on ang for progenecuth)


CPLarg 0.3x2 0.3x4 84x)
+

+
=

= 0.6 12
+

+ p.4
=
: 2.2(p]
Instructu cycles 12x (PI
=

A I cycles 109 2x109


-"

cocars-ghomom-iocuo
B 4 cycles
hx109
cycle
ee
C I

cr
Worst CPI (Biggest
= instruct took longer cycles)
Smallest I 1 =
(Smallest instructs took I cycle)

ET Instruct =
C5.x CP X Clock cycles

It. X SPI
ET ct
=

Frequency

POWER
* & ENERGY CONSUMER
Types Of PoWER
1. STATIC: - On standby
2. to
Dynamic:-Heeds do activity

Mostly power consumed.

PD CV.g =
- done in BEE. Thanks to Kamla.
C< capicitance
V VoltageIlsed
results P
=
in

f <-
frequency ↑
Unit (PD): Walt on W -

To reduce Power consumpth, reduce voltage consumpth


Energy (Joules
Biggest frequency is with
associated voltage P Vx1
=

z JPCt)
= db

n-> looks at cycles. Slo:- main


manages of . *
staticpoweralwaysemits even without framesee

Use switches to modulate


power usage on
power
islands. Call sloto
manage
the same.

MIPS
Rating:commercialrequirement for selling chips.
MIPS <- Millions of Instructs Per Second

mpsoxte-micro shares sale

ax10, z 4 x 10" miRs 900 MIps


=

upper bound
s
Companies
often advertise Peak MIPS i.e. highest MIPS a
process can get on
any program auth
Program with lowest CPI-high MIPS.
↳ often meaningless
eg:Cif-continue hep)
->
Compass Exover execute
of program for spy performance.
Point -> funct call for
- Floating.
-> converted
into integes so implementati

Amdahl'shave: -

Ratioof improved old winit. Performance on Enemith.


Target diff customers.

self:I
E'
=

caasios + a
=

If Sideal 1
=

SPEC CPU-yea -> benchmarks


SPEC processors will fest
Reference Machine <-
runthe program shown & not in competiti
Different rations for different
programs
asfor
Here use G.M. i.e.
on calculating the reference median.

You might also like