MKP-logo-white-transparent Title 4th-edition
Chapter 1
Computer Abstractions and Technology

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 2
The Computer Revolution
nProgress in computer technology
nUnderpinned by Moore’s Law
nMakes novel applications feasible
nComputers in automobiles
nCell phones
nHuman genome project
nWorld Wide Web
nSearch Engines
nComputers are pervasive

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 3
Classes of Computers
nDesktop computers
nGeneral purpose, variety of software
nSubject to cost/performance tradeoff
nServer computers
nNetwork based
nHigh capacity, performance, reliability
nRange from small servers to building sized
nEmbedded computers
nHidden as components of systems
nStringent power/performance/cost constraints

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 4
The Processor Market
f01-01-P374493

Y-axis is in millions

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 5
What You Will Learn
nHow programs are translated into the machine language
nAnd how the hardware executes them
nThe hardware/software interface
nWhat determines program performance
nAnd how it can be improved
nHow hardware designers improve performance
nWhat is parallel processing

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 6
Understanding Performance
nAlgorithm
nDetermines number of operations executed
nProgramming language, compiler, architecture
nDetermine number of machine instructions executed per operation
nProcessor and memory system
nDetermine how fast instructions are executed
nI/O system (including OS)
nDetermines how fast I/O operations are executed

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 7
Below Your Program
nApplication software
nWritten in high-level language
nSystem software
nCompiler: translates HLL code to machine code
nOperating System: service code
nHandling input/output
nManaging memory and storage
nScheduling tasks & sharing resources
nHardware
nProcessor, memory, I/O controllers
f01-02-P374493

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 8
Levels of Program Code
nHigh-level language
nLevel of abstraction closer to problem domain
nProvides for productivity and portability
nAssembly language
nTextual representation of instructions
nHardware representation
nBinary digits (bits)
nEncoded instructions and data
f01-03-P374493

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 9
f01-04-P374493
Components of a Computer
nSame components for
all kinds of computer
nDesktop, server,
embedded
nInput/output includes
nUser-interface devices
nDisplay, keyboard, mouse
nStorage devices
nHard disk, CD/DVD, flash
nNetwork adapters
nFor communicating with other computers
The BIG Picture

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 10
f01-05-P374493
Anatomy of a Computer
Output device
Input device
Input device
Network cable

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 11
Anatomy of a Mouse
nOptical mouse
nLED illuminates desktop
nSmall low-res camera
nBasic image processor
nLooks for x, y movement
nButtons & wheel
nSupersedes roller-ball mechanical mouse
n
optical-mouse-exploded optical-mouse

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 12
Through the Looking Glass
nLCD screen: picture elements (pixels)
nMirrors content of frame buffer memory
f01-06-P374493

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 13
Opening the Box
f01-07-P374493 f01-08-P374493

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 14
Inside the Processor (CPU)
nDatapath: performs operations on data
nControl: sequences datapath, memory, ...
nCache memory
nSmall fast SRAM memory for immediate access to data

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 15
Inside the Processor
nAMD Barcelona: 4 processor cores
f01-09-P374493

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 16
Abstractions
nAbstraction helps us deal with complexity
nHide lower-level detail
nInstruction set architecture (ISA)
nThe hardware/software interface
nApplication binary interface
nThe ISA plus system software interface
nImplementation
nThe details underlying and interface
The BIG Picture

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 17
flash-cards
A Safe Place for Data
nVolatile main memory
nLoses instructions and data when power off
nNon-volatile secondary memory
nMagnetic disk
nFlash memory
nOptical disk (CDROM, DVD)
hard-disk-drive flash-memory-exploded dvd-drive

Floppy disks – capacity 100-200 KB
Macintosh Diskettes – 1.44MB

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 18
Networks
nCommunication and resource sharing
nLocal area network (LAN): Ethernet
nWithin a building
nWide area network (WAN: the Internet
nWireless network: WiFi, Bluetooth
ethernet-cables wireless-router

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 19
Technology Trends
nElectronics technology continues to evolve
nIncreased capacity and performance
nReduced cost
Year
Technology
Relative performance/cost
1951
Vacuum tube
1
1965
Transistor
35
1975
Integrated circuit (IC)
900
1995
Very large scale IC (VLSI)
2,400,000
2005
Ultra large scale IC
6,200,000,000
DRAM capacity
f01-12-P374493

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 20
Defining Performance
nWhich airplane has the best performance?

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 21
Response Time and Throughput
nResponse time
nHow long it takes to do a task
nThroughput
nTotal work done per unit time
ne.g., tasks/transactions/… per hour
nHow are response time and throughput affected by
nReplacing the processor with a faster version?
nAdding more processors?
nWe’ll focus on response time for now…

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 22
Relative Performance
nDefine Performance = 1/Execution Time
n“X is n time faster than Y”
nExample: time taken to run a program
n10s on A, 15s on B
nExecution TimeB / Execution TimeA
= 15s / 10s = 1.5
nSo A is 1.5 times faster than B

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 23
Measuring Execution Time
nElapsed time
nTotal response time, including all aspects
nProcessing, I/O, OS overhead, idle time
nDetermines system performance
nCPU time
nTime spent processing a given job
nDiscounts I/O time, other jobs’ shares
nComprises user CPU time and system CPU time
nDifferent programs are affected differently by CPU and system performance

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 24
CPU Clocking
nOperation of digital hardware governed by a constant-rate clock
Clock (cycles)
Data transfer
and computation
Update state
Clock period
nClock period: duration of a clock cycle
ne.g., 250ps = 0.25ns = 250×10–12s
nClock frequency (rate): cycles per second
ne.g., 4.0GHz = 4000MHz = 4.0×109Hz

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 25
CPU Time
nPerformance improved by
nReducing number of clock cycles
nIncreasing clock rate
nHardware designer must often trade off clock rate against cycle count

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 26
CPU Time Example
nComputer A: 2GHz clock, 10s CPU time
nDesigning Computer B
nAim for 6s CPU time
nCan do faster clock, but causes 1.2 × clock cycles
nHow fast must Computer B clock be?

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 27
Instruction Count and CPI
nInstruction Count for a program
nDetermined by program, ISA and compiler
nAverage cycles per instruction
nDetermined by CPU hardware
nIf different instructions have different CPI
nAverage CPI affected by instruction mix

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 28
CPI Example
nComputer A: Cycle Time = 250ps, CPI = 2.0
nComputer B: Cycle Time = 500ps, CPI = 1.2
nSame ISA
nWhich is faster, and by how much?
A is faster…
…by this much

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 29
CPI in More Detail
nIf different instruction classes take different numbers of cycles
nWeighted average CPI
Relative frequency

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 30
CPI Example
nAlternative compiled code sequences using instructions in classes A, B, C
Class
A
B
C
CPI for class
1
2
3
IC in sequence 1
2
1
2
IC in sequence 2
4
1
1
nSequence 1: IC = 5
nClock Cycles
= 2×1 + 1×2 + 2×3
= 10
nAvg. CPI = 10/5 = 2.0
nSequence 2: IC = 6
nClock Cycles
= 4×1 + 1×2 + 1×3
= 9
nAvg. CPI = 9/6 = 1.5

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 31
Performance Summary
nPerformance depends on
nAlgorithm: affects IC, possibly CPI
nProgramming language: affects IC, CPI
nCompiler: affects IC, CPI
nInstruction set architecture: affects IC, CPI, Tc
The BIG Picture

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 32
Power Trends
nIn CMOS IC technology
×1000
×30
5V → 1V
f01-15-P374493

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 33
Reducing Power
nSuppose a new CPU has
n85% of capacitive load of old CPU
n15% voltage and 15% frequency reduction
nThe power wall
nWe can’t reduce voltage further
nWe can’t remove more heat
nHow else can we improve performance?

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 34
Uniprocessor Performance
f01-16-P374493
Constrained by power, instruction-level parallelism, memory latency

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 35
Multiprocessors
nMulticore microprocessors
nMore than one processor per chip
nRequires explicitly parallel programming
nCompare with instruction level parallelism
nHardware executes multiple instructions at once
nHidden from the programmer
nHard to do
nProgramming for performance
nLoad balancing
nOptimizing communication and synchronization

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 36
Manufacturing ICs
nYield: proportion of working dies per wafer
f01-18-P374493

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 37
AMD Opteron X2 Wafer
nX2: 300mm wafer, 117 chips, 90nm technology
nX4: 45nm technology
f01-19-P374493

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 38
Integrated Circuit Cost
nNonlinear relation to area and defect rate
nWafer cost and area are fixed
nDefect rate determined by manufacturing process
nDie area determined by architecture and circuit design

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 39
SPEC CPU Benchmark
nPrograms used to measure performance
nSupposedly typical of actual workload
nStandard Performance Evaluation Corp (SPEC)
nDevelops benchmarks for CPU, I/O, Web, …
nSPEC CPU2006
nElapsed time to execute a selection of programs
nNegligible I/O, so focuses on CPU performance
nNormalize relative to reference machine
nSummarize as geometric mean of performance ratios
nCINT2006 (integer) and CFP2006 (floating-point)

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 40
CINT2006 for Opteron X4 2356
Name
Description
IC×109
CPI
Tc (ns)
Exec time
Ref time
SPECratio
perl
Interpreted string processing
2,118
0.75
0.40
637
9,777
15.3
bzip2
Block-sorting compression
2,389
0.85
0.40
817
9,650
11.8
gcc
GNU C Compiler
1,050
1.72
0.47
24
8,050
11.1
mcf
Combinatorial optimization
336
10.00
0.40
1,345
9,120
6.8
go
Go game (AI)
1,658
1.09
0.40
721
10,490
14.6
hmmer
Search gene sequence
2,783
0.80
0.40
890
9,330
10.5
sjeng
Chess game (AI)
2,176
0.96
0.48
37
12,100
14.5
libquantum
Quantum computer simulation
1,623
1.61
0.40
1,047
20,720
19.8
h264avc
Video compression
3,102
0.80
0.40
993
22,130
22.3
omnetpp
Discrete event simulation
587
2.94
0.40
690
6,250
9.1
astar
Games/path finding
1,082
1.79
0.40
773
7,020
9.1
xalancbmk
XML parsing
1,058
2.70
0.40
1,143
6,900
6.0
Geometric mean
11.7
High cache miss rates

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 41
SPEC Power Benchmark
nPower consumption of server at different workload levels
nPerformance: ssj_ops/sec
nPower: Watts (Joules/sec)

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 42
SPECpower_ssj2008 for X4
Target Load %
Performance (ssj_ops/sec)
Average Power (Watts)
100%
231,867
295
90%
211,282
286
80%
185,803
275
70%
163,427
265
60%
140,160
256
50%
118,324
246
40%
920,35
233
30%
70,500
222
20%
47,126
206
10%
23,066
180
0%
0
141
Overall sum
1,283,590
2,605
∑ssj_ops/ ∑power
493

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 43
Pitfall: Amdahl’s Law
nImproving an aspect of a computer and expecting a proportional improvement in overall performance
nCan’t be done!
nExample: multiply accounts for 80s/100s
nHow much improvement in multiply performance to get 5× overall?
nCorollary: make the common case fast

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 44
Fallacy: Low Power at Idle
nLook back at X4 power benchmark
nAt 100% load: 295W
nAt 50% load: 246W (83%)
nAt 10% load: 180W (61%)
nGoogle data center
nMostly operates at 10% – 50% load
nAt 100% load less than 1% of the time
nConsider designing processors to make power proportional to load

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 45
Pitfall: MIPS as a Performance Metric
nMIPS: Millions of Instructions Per Second
nDoesn’t account for
nDifferences in ISAs between computers
nDifferences in complexity between instructions
nCPI varies between programs on a given CPU

MKP-logo
Chapter 1 — Computer Abstractions and Technology — 46
Concluding Remarks
nCost/performance is improving
nDue to underlying technology development
nHierarchical layers of abstraction
nIn both hardware and software
nInstruction set architecture
nThe hardware/software interface
nExecution time: the best performance measure
nPower is a limiting factor
nUse parallelism to improve performance