Transcript Set 2: Architecture - CS Course Webpages

CSCE 121:509-512
Simple Computer Model
Spring 2015
Based on slides created by Bjarne Stroustrup and Jennifer Welch
1
HOW DATA IS REPRESENTED
CSCE 121:509-512 -- Set 2: Architecture
Fundamental unit of computer storage
is a bit, a data entity that can take on
one of two values, 0 or 1
CSCE 121:509-512 -- Set 2: Architecture
Given a sequence of bits (01010001), what
does it represent?
Depends on the code
– Base 10 integer: 1,010,001
– Base 2 integer: 1*20 + 1*24 + 1*26 = 81 (in base 10)
– ASCII: The letter ‘Q’
Must know the code to decode a sequence
There are many other codes…
CSCE 121:509-512 -- Set 2: Architecture
2
HOW DATA IS STORED
CSCE 121:509-512 -- Set 2: Architecture
Bytes Make Memory
•
•
•
•
•
Byte: a group of 8 bits
How many different values?
00000000, 00000001, 00000010, …, 11111110, 11111111
28 = 256 possibilities
Memory: long sequence of bytes, numbered 0, 1, 2, 3,…
11 00 1 10 0 11 00 1 10 000 11 0 011 0 10 0 110 0
byte 0
byte 1
byte 2
CSCE 121:509-512 -- Set 2: Architecture
byte 3
…..
Addresses and Words
Address: the number of a byte in memory
Contents of a byte can change
Use consecutive locations to store longer sequences of
information
4 bytes = 1 word
a word
11 00 1 10 0 11 00 1 10 000 11 0 011 0 10 0 110 0
byte 0
byte 1
byte 2
CSCE 121:509-512 -- Set 2: Architecture
byte 3
…..
Limitations of Finite Data Encodings
Overflow: number is too large
– suppose 1 byte stores integers in base 2, from 0
(00000000) to 255 (11111111)
– if the byte holds 255, then adding 1 to it results in
256, which is too large to be stored in the byte
CSCE
Set --2:Set
Architecture
CSCE 121-200:
121:509-512
2: Architecture
Limitations of Finite Data Encodings
Roundoff error:
– insufficient precision (size of word): try to store
1/8, which is .001 in base 2, with only two bits
– nonterminating expansions in current base: try to
store 1/3 in base 10, which is .333…
– nonterminating expansions in every base:
irrational numbers such as π
CSCE 121:509-512 -- Set 2: Architecture
Kinds of Storage
Cache:
super-fast
Main memory:
random access,
equally fast to
access any
address
Disk: random
access,
significantly
slower than
main memory
Tape: sequential
access,
significantly
slower than disk
CSCE 121:509-512 -- Set 2: Architecture
3
HOW DATA IS OPERATED ON
CSCE 121:509-512 -- Set 2: Architecture
Main Players
• Data in main memory cannot be operated on: must be
copied (loaded) to special locations called registers
• Once data is in registers, it can be operated on by
circuitry called arithmetic logic unit (ALU)
– e.g., add, multiply
• Result of operation can then be copied (stored) back
to main memory
• Procedure is organized by control unit circuitry
– figures out what the ALU should do next
– transfers data between main memory and registers
• Registers, ALU and control unit are in the CPU
CSCE 121:509-512 -- Set 2: Architecture
ALU
CPU
registers
main
memory
control
unit
CSCE 121:509-512 -- Set 2: Architecture
Machine Instructions
Goal: add the number stored in address 3 and the number stored in address
6; put the result in address 10.
Control unit does the following:
1. copies data from main memory address 3 into some register, say 1:
LOAD 3,1
2. copies data in main memory address 6 into some register, say 4:
LOAD 6,4
3. tells ALU to add the contents of registers 1 and 4, and put result in some
register, say 3:
ADD 1,4,3
4. copies data in register 3 into main memory address 10:
STORE 3,10
LOAD, ADD and STORE are machine instructions.
How does the control unit know which instruction is next? The program!
CSCE 121:509-512 -- Set 2: Architecture
4
HOW A PROGRAM IS STORED
CSCE 121:509-512 -- Set 2: Architecture
Program: list of machine instructions using some
agreed upon coding convention.
Example:
ADD
1
0 0 1 0
0 0 0 1
opcode
1st operand
4
0 1 0 0
2nd operand
3
0 0 1 1
3rd operand
Program is stored in memory the same way data
is stored!
CSCE 121:509-512 -- Set 2: Architecture
5
HOW A PROGRAM IS EXECUTED ON
THE DATA
CSCE 121:509-512 -- Set 2: Architecture
How a Program is Executed
• The control unit has
– instruction register: holds current instruction to be executed
– program counter: holds address of next instruction in the
program to be fetched from memory
• Program counter tells where the computer is in the
program. Usually the next instruction to execute is the
next instruction in memory
• Sometimes we want to JUMP to another instruction
(e.g., if or while)
– unconditional JUMP: always jump to address given
– conditional JUMP: only jump if a certain condition is true
(e.g., some register contains 0)
CSCE 121:509-512 -- Set 2: Architecture
ALU
CPU
registers
main
memory
control
unit
CSCE 121:509-512 -- Set 2: Architecture
instruction
register
program
counter
Machine Cycle
• fetch next instruction into instruction register
(IR), as indicated by program counter (PC), and
increment PC
• decode bit pattern in IR to figure out which
circuitry needs to be activated to perform
instruction
• execute instruction by copying data into
registers and activating ALU to do the right
thing
– a JUMP may cause the PC to be altered
CSCE 121:509-512 -- Set 2: Architecture
Diagram of Architecture
CPU:
PC:
R1:
IR:
R2:
ALU
R3:
control unit
R4:
bus
0
main
memory:
1
2
3
4
5
data
…
first instr.
program
second instr.
…
third instr.
95
96
97
CSCE 121:509-512 -- Set 2: Architecture
98
99
100
…
6
HOW A PROGRAM IS TRANSLATED
INTO MACHINE LANGUAGE
CSCE 121:509-512 -- Set 2: Architecture
Evolution of Programming Languages
Machine languages: all in binary
machine dependent
painful for people
no translation needed
Assembly languages: allow symbols for operators and addresses
still machine dependent
slightly less painful
assembler translates assembly
language programs into machine
language
High-level languages: such as Fortran, C, Java, C++
machine independent
easier for people
compiler translates high-level
language programs into machine
language
CSCE 121:509-512 -- Set 2: Architecture
Why Compiling is Hard
one high-level instruction
can correspond to several
machine instructions
• x = (y+3)/z;
fancier control structures
than JUMP
•
•
•
•
while
repeat
if-then-else
case/switch
fancier data structures
• arrays: must calculate addresses for
references to array elements
• structs: similar to arrays, but less
regular
functions/subroutines/ methods:
generate machine language to:
• copy input parameter values
• save return address
• start executing at beginning of function
code
• copy output parameter values at the end
• set PC to return address
CSCE 121:509-512 -- Set 2: Architecture
Compilation Process
Lexical
Analysis
• break up strings of characters into logical
components, called tokens, and discard
comments and spaces
• Ex: total = sum + 55.32 has 5 tokens
Parsing
• decide how the tokens are related
• Ex: sum + 55.32 is an arithmetic expression
• Ex: total = sum + 55.32 is an assignment
statement
Code
Generation
• generate machine instructions for each high-level
instruction
• resulting machine language program, called object
code, is written to disk
CSCE 121:509-512 -- Set 2: Architecture
Linking
Linker combines results of compiling different pieces of
the program separately.
If pieces refer to each other, these references cannot be resolved during
independent compilation
Combined code is written to disk.
function p
…
refers to x
…
main
…
declares x
…
invokes p
…
CSCE 121:509-512 -- Set 2: Architecture
Loading
To run the program, the loader copies the object code
from disk into main memory (location in main memory
is determined by operating system, not programmer)
Loader initializes PC to starting location of program
and adjusts JUMP addresses
Result is an executable
CSCE 121:509-512 -- Set 2: Architecture
Source code for
your program
Compiler
Object code for
your program
Object code for
libraries
Linker/Loader
Executable for your
program
CSCE 121:509-512 -- Set 2: Architecture
Credits
• Slide 1: “NEC APC” by Niv Singer, licensed under CC BY-SA 2.0
• Slide 3: “La tecnología...” by Infocux Technologies, licensed under CC BYNC 2.0
• Slide 4: http://commons.wikimedia.org/wiki/File:ASCII_Code_Chart.svg
• Slide 8: “Overflow” by Braveheart, licensed under CC BY-NC-ND 2.0
• Slide 9: “pumpkin pi” by Craig Damlo, licensed under CC BY-NC-ND 2.0
CSCE 121:509-512 -- Set 2: Architecture