Modern Computer Architecture

Lecture1 Fundamentals of
Quantitative Design and Analysis (I)
Hongbin Sun
国家集成电路人才培养基地
Xi’an Jiaotong University
 Course Administration
• Instructor: Prof. Hongbin Sun (hsun@mail.xjtu.edu.cn)
• Office: West 4th Building, Qujiang Campus
• TA:
• Lectures:
• Text Book: Computer Architecture: A Quantitative Approach
Hennessey and Patterson, 5th Edition (2012)
• Prerequisite: Digital Logic&Computer Organization
• Course Webpage:
http://gr.xjtu.edu.cn/web/hsun/3
 1.1 Introduction

• ENIAC was the first electronic general-purpose computer

announced in 1946. ENIAC was designed to calculate artillery
firing tables (火炮射击图表) for the US Army‘s Ballistic
Research Laboratory (弹道研究实验室, BRL).
• Computer technology has made incredible progress in the
roughly 65 years since ENIAC was created.
Computers then
Computers now
Major Technology Generations
Growth in Processor Performance
Move to multi-processor

Pinnacle of uniprocessor:
RISC Intel Penium4@3.4GHz

Hardware renaissance
 The Moore’s Law

• “Cramming More Components onto Integrated Circuits”

– Gordon Moore, Electronics, 1965
• # of transistors on cost-effective integrated circuit double every 18
months
 CISC vs. RISC
• CISC (Complex Instruction Set Computer), 1970s
• Two significant changes in the computer marketplace
– The emergence of high level language
– The creation of standardized, vendor-independent operating
systems, such as UNIX and Linux
• RISC (Reduced Instruction Set Computer), 1980s
– The exploitation of instruction-level parallelism (pipelining and
multiple instruction issue)
– The use of cache
• The RISC-based computers raised the performance bar,
forcing prior architectures to keep up or disappear
– Digital Equipment Vax
– Intel x86
 RISC Architecture
• RISC: A fixed (32-bit) instruction size with few format;
• CISC: typically had variable length instruction sets with many
format.
• RISC: A load-store arch. where data processing instructions
operate only on registers, separate from MA instruction;
• CISC: typically allowed values in memory to be used as
operands in data processing instructions.
• RISC: A large register bank of thirty two 32-bit registers, all
of which could be used for any purpose, to allow the load-
store architecture to operate efficiently;
• CISC: not as large as RISC, and most had different registers
for different purpose.
 RISC Organization
• RISC: hard-wired instruction decode logic;
• CISC: used large microcode ROMs to decode their instructions.
• RISC: pipelined execution;
• CISC: allowed little, if any, overlap between consecutive
instruction (though they do now)
• RISC: single-cycle execution;
• CISC: typically took many clock cycles to complete a single
instruction.

• RISC: MIPS, ARM

• CISC: x86
The Growth Effect in 20th Century
1. It has significantly enhanced the capability available to
computer users.
2. This dramatic improvement in cost-performance leads to
new classes of computers.
– Personal Computer (PC)
– Mobile Client Devices
– Warehouse-scale computer
3. Continuing improvement of semiconductor manufacturing
has led to the dominance of microprocessor-based
computers across the entire range of computer design.
4. Software development, allowed programmers today to
trade performance for productivity.
 SaaS & Cloud Computing

• Software as a Service (SaaS) used over the Internet is

replacing shrink-wrapped software that must be installed
and run on a local computer.
The Diversified Applications

Google’s
Goggles

The nature of applications also changed!

 The Growth after 2003
• Since 2003, uniprocessor performance improvement has
dropped to less than 22% per year.
– Maximum power dissipation of air-cooled chip
– The lack of more instruction-level parallelism to exploit efficiently
• The milestone signal in 2004: Intel canceled its uniprocessor
project.
– Instruction-Level Parallelism (ILP)
– Data-Level Parallelism (DLP)
– Thread-Level Parallelism (TLP)
– Request-Level Parallelism (RLP)
Pinnacle of Single-Core MP
The Future: Processor becomes a transistor?

Intel 48 cores single chip cloud computing

 1.2 Classes of Computers
• Personal Mobile Device (PMD)
• Desktop Computing
• Servers
• Clusters/Warehouse-Scale Computers
• Embedded Computers
 Classes of Parallelism and Arch.
• There are basically two kinds of parallelsim in applications:
– Data-Level Parallelism (DLP)
– Task-Level Parallelism (TLP)

• Computer hardware in turn can exploit these two kinds of

application parallelism in four major ways:
– Instruction-Level Parallelism: pipelining & speculative
execution
– Vector Architecture and Graphic Processor Units (GPU)
– Thread-Level Parallelism
– Request-Level Parallelism
 Flynn’s Taxonomy of CA
• Single instruction stream, single data stream (SISD)
– Uniprocessor
– Instruction Level Parallelism, ILP
• Single instruction stream, multiple data stream (SIMD)
– Vector architecture, multimedia extensions and GPUs
– Data-Level Parallelism, DLP
• Multiple instruction streams, single data stream (MISD)
– No commercial multiprocessor of this type right now
• Multiple instruction streams, multiple data stream (MIMD)
– Multiprocessor, Thread-Level Parallelism, TLP
– Cluster and ware-house scale computers, RLP
1.3 What is Computer Architecture?
Abstraction in modern computer system
Abstraction in modern computer system

The myopic view of

computer architecture:
instruction set architecture,
the interface between
software and hardware.
Abstraction in modern computer system

Computer Architecture
Computer Architecture is Constantly
Changing
Computer Architecture is Constantly
Changing
 Computer Architecture’s Changing Definition

• 1950s to 1960s: Computer Architecture Course:

Computer Arithmetic
• 1970s to mid 1980s: Computer Architecture Course:
Instruction Set Design, especially ISA appropriate
for compilers
• 1990s: Computer Architecture Course:
Design of CPU, memory system, I/O system,
Multiprocessors, Networks
• 2000s: Computer Architecture Course: Non Von-
Neumann architectures, Reconfiguration, Focused
MIPs
 Computer architecture topics
Input/Output and Storage
Disks, WORM, Tape RAID

Emerging Technologies
DRAM Interleaving
Bus protocols

Coherence,
Memory L2 Cache Bandwidth,

Other Processors
Hierarchy Latency
Network
Communication
L1 Cache Addressing,
VLSI Protection,
Instruction Set Architecture
Exception Handling

Pipelining, Hazard Resolution, Pipelining and Instruction

Superscalar, Reordering, Level Parallelism
Prediction, Speculation,
Vector, Dynamic Compilation
 Computer architecture topics

Shared Memory,
P M P M P M P M
Message Passing,
° ° ° Data Parallelism

S Interconnection Network Network Interfaces

Processor-Memory-Switch Topologies,
Routing,
Multiprocessors Bandwidth,
Networks and Interconnections Latency,
Reliability
 Instruction Set

software

instruction set

hardware
 ISA vs. Microarchitecture
• Architecture covers all three aspects of computer design –
instruction set architecture (ISA), microarchitecture or
organization and hardware.
• Instruction Set Architecture:
– Programmer visible state (register and memory)
– Operations (Instructions and how they work)
– Execution semantics (Interrupts)
– Input/Output
– Data types/sizes
• Microarchitecture:
– Trade-offs on how to implement ISA for some metrics (speed, energy
and cost).
– Examples: pipeline depths, cache size, execution order, bus widths
and ALU widths.
Same ISA, Different Microarchitecture
Diff. ISA, Diff. Microarchitecture
 A Review of ISA (1)
• Class of ISA: General-purpose register architectures
– Register-memory ISAs: 80x86
– Load-store ISAs: ARM and MIPS
• Memory address: byte addressing, aligned

Little endian Big endian

A Review of ISA (2)
 A Review of ISA (3)
• Type and size of operands
– 8-bit (ASCII character)
– 16-bit (Unicode character or half word)
– 32-bit (Integer or word)
– 64-bit (double word or long integer)
– IEEE 754 floating point in 32-bit (single precision) and 64-bit (double
precision)
• Operations
– Data transfer
– Arithmetic/logical
– Control
– Floating point
 A Review of ISA (4)
• Control flow instructions: PC relative addressing
– Conditional branches
– Unconditional jumps
– Procedure calls and returns
• Encoding on ISA
– Fixed length: easy to decode, RISC arch, eg. ARM, MIPS, PowerPC
– Variable length: less space in memory and caches, CISC arch
• eg. 80x86 (1 byte up to 17 bytes)
– Mostly fixed or compressed:
• eg. MIPS16, Thumb
• eg. PowerPC and some VLIW (store instructions compressed and,
decompress into instruction cache)
– Very Long Instruction Word (VLIW): multiple instructions in a fixed
length bundle
• eg. TI C6000
A Review of ISA (5)
X86 (IA-32) Instruction Encoding
A Review of ISA (6)
MIPS Instruction Encoding
A Review of ISA (7)
 A Review of ISA (8)
• Technology influenced ISA
– Storage is expensive, tight encoding important
– Reduced Instruction Set Computer
– Multicore/Manycore: Transistors not turning into sequential
performance
• Application influenced ISA
– Instructions for applications: embedded, DSP
– Compiler technology has improved
• SPARC register windows no longer needed
• Compiler can do register allocation efficiently
• The other challenges beyond ISA design are particularly
acute at the present, when differences among ISAs are small
and when there are distinct application areas.