Transcript PPT

Computer Architecture
ECE 4801
Berk Sunar
Erkay Savas
Outline
 Brief Overview
 How is a computer program executed?
 Computer organization
 Roadmap for this class
Things You Learn in this Course
 How computers work; the basic foundation
 How to analyze their performance (and how not to)
 Key technologies determining the performance of modern
processors






Datapath Design
Pipelining
Cache Systems
Memory Hierarchy
I/O
Multiprocessors
Instruction Set Architecture
 Important abstraction
 Interface between hardware and low-level software
 Or features available to programmers
 instructions set architecture (ISA)
 e.g. does the processor have an multiply instruction?
 instruction encoding
 Data representation
 I/O mechanism.
 addressing mechanism
 Modern instruction set architectures:
 80x86/Pentium/K6, PowerPC, DEC Alpha, MIPS, SPARC, HP,
ARM.
Computer Organization
 Computer Organization is how features are implemented in
hardware
 Transparent to programmers
 Different implementations are possible for the same
architecture (affects performance/price)
 Determines how memory, CPU, peripherals, busses are
interconnected and how control signals routed.
 Has HUGE impact on performance.
 Performance of the organization is usually application
dependent. (e.g. I/O intensive, computation intensive, memory
bound etc.)
How to Program a Computer?
 A simple but universal interface
 Machine Code (binary images)
 Assembly language
 Uses mnemonics that map directly to ISA e.g. addw, lb, jmp etc.
 More readable than machine languages
 Error prone but excellent for low-level optimization
 High-level languages
 E.g. C/C++, Pascal, Fortran, Java, C#
 Much easier to use and program
 Promotes code portability
 Not as efficient as custom assembly
Processing a C Program
High-level language
program (in C)
swap (int v[], int k){
int temp;
temp = v[k];
v[k] = v[k+1];
v[k+1] = temp;
}
Binary machine
language program
for MIPS
Assembly language
program for MIPS
C compiler
0000000010100001000000000001
10000000000010001110000110000
01000011000110001100010000000
00000000001000110011110010000
00000000001001010110011110010
00000000000000001010110001100
01000000000000001000000001111
1000000000000000001000
swap:
muli
add
lw
lw
sw
sw
jr
$2,
$2,
$15,
$16,
$16,
$15,
$31
$5, 4
$4, $2
0($2)
4($2)
0($2)
4($2)
Assembler
Functions of a Computer
 Data processing , e.g. sort entries of a spreadsheet
 Data storage, e.g. personal files, applications,
movies, music etc.
 Data movement, e.g. play a music file, display a
picture
 Control, (applies to all examples above)
Five Classic Components
Computer
Processor
Datapath
Control
Memory
Input
Output
System Interconnection
Bridges
Inside the Processor Chip
Instruction
Cache
Data
Cache
Control
Bus
branch
prediction
integer floating-point
datapath
datapath
An Actual View
22nm Intel Core CPU
Source: Intel Corp.
6 core CPU with
L3 caches
https://computing.llnl.gov/tutorials/parallel_comp/
Memory
 Nonvolatile:
 ROM
 Hard disk, floppy disk, magnetic tape, CDROM, USB Memory
 Flash memory
 Volatile
 DRAM used usually for main memory
 SRAM used mainly for on-chip memory such as register and cache
 DRAM is much cheaper than SRAM
 SRAM is much faster than DRAM
 How about solid state drives?
DRAM and Processor Characteristics
Solutions to Memory Problems
 Increase number of bits retrieved at one time
 Make DRAM “wider” rather than “deeper”
 Change DRAM interface
 Cache
 Reduce frequency of memory access
 More complex cache and cache on chip
 Increase interconnection bandwidth
 High speed buses
 Hierarchy of buses
Computer Networks
 Very essential aspect of computer systems
 Communication
 Resource sharing
 Remote access
 Ethernet is the most popular LAN
 Range is limited to 1 kilometer
 10/100 Mbit/s
 Wide Area Networks (WAN)
 Cross continents and backbone of the Internet
Roadmap
 Performance issues
 Instruction set of MIPS
 Arithmetic and ALU
 Constructing a processor to execute our instructions
(datapath design)
 Pipelining
 Memory hierarchy: caches and virtual memory
 I/O