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

KAIST 04 Machine Level Programming Basics

The document discusses machine-level programming and assembly basics. It provides an overview of Intel processors and architectures, and covers the basics of assembly including registers, operands, instructions, and the x86-64 architecture. It also discusses compiling C code to assembly and machine code.

Uploaded by

김민재
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)
39 views

KAIST 04 Machine Level Programming Basics

The document discusses machine-level programming and assembly basics. It provides an overview of Intel processors and architectures, and covers the basics of assembly including registers, operands, instructions, and the x86-64 architecture. It also discusses compiling C code to assembly and machine code.

Uploaded by

김민재
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/ 47

KAIST

Machine‐Level Programming I: Basics
CS230: System Programming
4th Lecture

Instructor:
Jaehyuk Huh (허재혁)

Lecture slides based on CS:App from CMU and CS230 KAIST (Prof. Kim Daeyoung)
1
KAIST

Today: Machine Programming I: Basics
• History of Intel processors and architectures
• C, assembly, machine code
• Assembly Basics: Registers, operands, move
• Intro to x86‐64
KAIST

Intel x86 Processors
• Totally dominate laptop/desktop/server market

• Evolutionary design
– Backwards compatible up until 8086, introduced in 1978
– Added more features as time goes on

• Complex instruction set computer (CISC)
– Many different instructions with many different formats
• But, only small subset encountered with Linux programs
– Hard to match performance of Reduced Instruction Set Computers 
(RISC)
– But, Intel has done just that!
• In terms of speed.  Less so for low power.
KAIST

Intel x86 Evolution: Milestones
Name Date Transistors MHz
• 8086 1978 29K 5‐10
– First 16‐bit processor.  Basis for IBM PC & DOS
– 1MB address space
• 386 1985 275K 16‐33
– First 32 bit processor , referred to as IA32
– Added “flat addressing”
– Capable of running Unix
– 32‐bit Linux/gcc uses no instructions introduced in later models
• Pentium 4F 2004 125M 2800‐3800
– First 64‐bit processor, referred to as x86‐64
• Core i7 2008 731M 2667‐3333
– Our lab machines
KAIST

Intel x86 Processors: Overview
Architectures Processors
X86‐16 8086

286
X86‐32/IA32 386
486
Pentium
MMX Pentium MMX

SSE Pentium III

SSE2 Pentium 4

SSE3 Pentium 4E
X86‐64 / EM64t Pentium 4F time

Core 2 Duo
SSE4 Core i7
IA: often redefined as latest Intel architecture
KAIST

Intel x86 Processors, contd.
• Machine Evolution
– 386 1985 0.3M
– Pentium 1993 3.1M
– Pentium/MMX 1997 4.5M
– PentiumPro 1995 6.5M
– Pentium III 1999 8.2M
– Pentium 4 2001 42M
– Core 2 Duo 2006 291M
– Core i7 2008 731M
• Added Features
– Instructions to support multimedia operations
• Parallel operations on 1, 2, and 4‐byte data, both integer & FP
– Instructions to enable more efficient conditional operations
• Linux/GCC Evolution
– Two major steps: 1) support 32‐bit 386.  2) support 64‐bit x86‐64
KAIST

More Information
• Intel processors (Wikipedia)
• Intel microarchitectures
KAIST

New Species: ia64, then IPF, then 
Itanium,… 
Name DateTransistors
• Itanium 2001 10M
– First shot at 64‐bit architecture: first called IA64
– Radically new instruction set designed for high performance
– Can run existing IA32 programs
• On‐board “x86 engine”
– Joint project with Hewlett‐Packard
• Itanium 2 2002 221M
– Big performance boost
• Itanium 2 Dual‐Core 2006 1.7B
• Itanium has not taken off in marketplace
– Lack of backward compatibility, no good compiler support, Pentium 4 got 
too good
KAIST

x86 Clones: Advanced Micro Devices 
(AMD)
• Historically
–AMD has followed just behind Intel
–A little bit slower, a lot cheaper
• Then
–Recruited top circuit designers from Digital Equipment Corp. and other 
downward trending companies
–Built Opteron: tough competitor to Pentium 4
–Developed x86‐64, their own extension to 64 bits
KAIST

Intel’s 64‐Bit
• Intel Attempted Radical Shift from IA32 to IA64
– Totally different architecture (Itanium)
– Executes IA32 code only as legacy
– Performance disappointing
• AMD Stepped in with Evolutionary Solution
– x86‐64 (now called “AMD64”)
• Intel Felt Obligated to Focus on IA64
– Hard to admit mistake or that AMD is better
• 2004: Intel Announces EM64T extension to IA32
– Extended Memory 64‐bit Technology
– Almost identical to x86‐64!
• All but low‐end x86 processors support x86‐64
– But, lots of code still runs in 32‐bit mode
KAIST

Our Coverage
• IA32
– The traditional x86

• x86‐64/EM64T
– The emerging standard

• Presentation
– Book presents IA32 in Sections 3.1—3.12
– Covers x86‐64 in 3.13
– We will cover both simultaneously
– Some labs will be based on x86‐64, others on IA32
KAIST

Today: Machine Programming I: Basics
• History of Intel processors and architectures
• C, assembly, machine code
• Assembly Basics: Registers, operands, move
• Intro to x86‐64
KAIST

Definitions
• Architecture: (also instruction set architecture: ISA) The parts 
of a processor design that one needs to understand to write 
assembly code. 
– Examples:  instruction set specification, registers.
• Microarchitecture: Implementation of the architecture.
– Examples: cache sizes and core frequency.

• Example ISAs (Intel): x86, IA, IPF
KAIST

Assembly Programmer’s View
CPU Memory
Addresses
Registers Object Code
PC
Data Program Data
Condition OS Data
Instructions
Codes

Stack
• Programmer‐Visible State
– PC: Program counter
• Address of next instruction
• Called “EIP” (IA32) or “RIP” (x86‐64)
– Register file
• Heavily used program data
– Memory
• Byte addressable array
– Condition codes
• Code, user data, (some) OS data
• Store status information about most 
recent arithmetic operation • Includes stack used to support 
procedures
• Used for conditional branching
KAIST

Turning C into Object Code
– Code in files p1.c p2.c
– Compile with command: gcc –O1 p1.c p2.c -o p
• Use basic optimizations (-O1)
• Put resulting binary in file p

text C program (p1.c p2.c)

Compiler (gcc -S)

text Asm program (p1.s p2.s)

Assembler (gcc or as)

binary Object program (p1.o p2.o) Static libraries 


(.a)
Linker (gcc or ld)

binary Executable program (p)
KAIST

Compiling Into Assembly
C Code Generated IA32 Assembly
int sum(int x, int y) sum:
{ pushl %ebp
int t = x+y; movl %esp,%ebp
return t; movl 12(%ebp),%eax
} addl 8(%ebp),%eax
popl %ebp
ret

Some compilers use instruction 
“leave”

Obtain with command
/usr/local/bin/gcc –O1 -S code.c
Produces file code.s
KAIST

Assembly Characteristics: Data Types
• “Integer” data of 1, 2, or 4 bytes
– Data values
– Addresses (untyped pointers)

• Floating point data of 4, 8, or 10 bytes

• No aggregate types such as arrays or structures
– Just contiguously allocated bytes in memory
KAIST

Assembly Characteristics: Operations
• Perform arithmetic function on register or memory data

• Transfer data between memory and register
– Load data from memory into register
– Store register data into memory

• Transfer control
– Unconditional jumps to/from procedures
– Conditional branches
KAIST

Object Code
Code for sum
• Assembler
0x401040 <sum>:
0x55
– Translates .s into .o
0x89 – Binary encoding of each instruction
0xe5 – Nearly‐complete image of executable code
0x8b
0x45
– Missing linkages between code in different 
0x0c files
0x03 • Linker
0x45
0x08 – Resolves references between files
• Total of 11 bytes
0x5d – Combines with static run‐time libraries
0xc3 • Each instruction  • E.g., code for malloc, printf
1, 2, or 3 bytes
– Some libraries are dynamically linked
• Starts at address 
0x401040 • Linking occurs when program begins 
execution
KAIST

Machine Instruction Example
• C Code
int t = x+y;
– Add two signed integers
• Assembly
– Add 2 4‐byte integers
addl 8(%ebp),%eax
• “Long” words in GCC parlance
Similar to expression:  • Same instruction whether signed 
x += y or unsigned
More precisely: – Operands:
int eax; x: Register %eax
int *ebp; y: Memory M[%ebp+8]
eax += ebp[2] t: Register %eax
– Return function value in %eax

0x80483ca: 03 45 08 • Object Code
– 3‐byte instruction
– Stored at address 0x80483ca
KAIST

Disassembling Object Code
Disassembled
080483c4 <sum>:
80483c4: 55 push %ebp
80483c5: 89 e5 mov %esp,%ebp
80483c7: 8b 45 0c mov 0xc(%ebp),%eax
80483ca: 03 45 08 add 0x8(%ebp),%eax
80483cd: 5d pop %ebp
80483ce: c3 ret

• Disassembler
objdump -d p
– Useful tool for examining object code
– Analyzes bit pattern of series of instructions
– Produces approximate rendition of assembly code
– Can be run on either a.out (complete executable) or .o file
KAIST

Alternate Disassembly
Disassembled
Object
0x401040:
0x55 Dump of assembler code for function sum:
0x89 0x080483c4 <sum+0>: push %ebp
0xe5 0x080483c5 <sum+1>: mov %esp,%ebp
0x8b 0x080483c7 <sum+3>: mov 0xc(%ebp),%eax
0x45 0x080483ca <sum+6>: add 0x8(%ebp),%eax
0x0c 0x080483cd <sum+9>: pop %ebp
0x03 0x080483ce <sum+10>: ret
0x45
0x08
0x5d • Within gdb Debugger
0xc3 gdb p
disassemble sum
– Disassemble procedure
x/11xb sum
– Examine the 11 bytes starting at sum
KAIST

What Can be Disassembled?
% objdump -d WINWORD.EXE

WINWORD.EXE: file format pei-i386

No symbols in "WINWORD.EXE".
Disassembly of section .text:

30001000 <.text>:
30001000: 55 push %ebp
30001001: 8b ec mov %esp,%ebp
30001003: 6a ff push $0xffffffff
30001005: 68 90 10 00 30 push $0x30001090
3000100a: 68 91 dc 4c 30 push $0x304cdc91

• Anything that can be interpreted as executable code
• Disassembler examines bytes and reconstructs assembly source
KAIST

Today: Machine Programming I: Basics
• History of Intel processors and architectures
• C, assembly, machine code
• Assembly Basics: Registers, operands, move
• Intro to x86‐64
KAIST

Integer Registers (IA32) Origin
(mostly obsolete)
%eax %ax %ah %al accumulate

%ecx %cx %ch %cl counter


general purpose

%edx %dx %dh %dl data

%ebx %bx %bh %bl base

source
%esi %si index

destination
%edi %di index
stack
%esp %sp
pointer
base
%ebp %bp
pointer

16‐bit virtual registers
(backwards compatibility)
KAIST

Moving Data: IA32 %eax

• Moving Data %ecx
movl Source, Dest: %edx

• Operand Types %ebx
– Immediate: Constant integer data %esi
• Example: $0x400, $-533
%edi
• Like C constant, but prefixed with ‘$’
• Encoded with 1, 2, or 4 bytes %esp
– Register: One of 8 integer registers %ebp
• Example: %eax, %edx
• But %esp and %ebp reserved for special use
• Others have special uses for particular instructions
– Memory: 4 consecutive bytes of memory at address given by register
• Simplest example: (%eax)
• Various other “address modes”
KAIST

movl Operand Combinations
Source Dest Src,Dest C Analog

Reg movl $0x4,%eax temp = 0x4;


Imm
Mem movl $-147,(%eax) *p = -147;

movl Reg movl %eax,%edx temp2 = temp1;


Reg
Mem movl %eax,(%edx) *p = temp;

Mem Reg movl (%eax),%edx temp = *p;

Cannot do memory‐memory transfer with a single instruction
KAIST

Simple Memory Addressing 
Modes
• Normal (R) Mem[Reg[R]]
– Register R specifies memory address

movl (%ecx),%eax

• Displacement D(R) Mem[Reg[R]+D]


– Register R specifies start of memory region
– Constant displacement D specifies offset

movl 8(%ebp),%edx
KAIST

Using Simple Addressing Modes
swap:
pushl %ebp
void swap(int *xp, int *yp) Set
movl %esp,%ebp
{ Up
pushl %ebx
int t0 = *xp;
int t1 = *yp;
*xp = t1; movl 8(%ebp), %edx
*yp = t0; movl 12(%ebp), %ecx
} movl (%edx), %ebx Body
movl (%ecx), %eax
movl %eax, (%edx)
movl %ebx, (%ecx)

popl %ebx
popl %ebp Finish
ret
KAIST

Using Simple Addressing Modes
swap:
pushl %ebp
void swap(int *xp, int *yp) Set
movl %esp,%ebp
{ Up
pushl %ebx
int t0 = *xp;
int t1 = *yp;
*xp = t1; movl 8(%ebp), %edx
*yp = t0; movl 12(%ebp), %ecx
} movl (%edx), %ebx Body
movl (%ecx), %eax
movl %eax, (%edx)
movl %ebx, (%ecx)

popl %ebx
popl %ebp Finish
ret
KAIST

Understanding Swap
void swap(int *xp, int *yp) •
{ • Stack
int t0 = *xp; • (in memory)
Offset
int t1 = *yp;
*xp = t1; 12 yp
*yp = t0; 8 xp
}
4 Rtn adr
0 Old %ebp %ebp
-4 Old %ebx %esp
Register Value
%edx xp
%ecx yp
movl 8(%ebp), %edx # edx = xp
%ebx t0
movl 12(%ebp), %ecx # ecx = yp
%eax t1 movl (%edx), %ebx # ebx = *xp (t0)
movl (%ecx), %eax # eax = *yp (t1)
movl %eax, (%edx) # *xp = t1
movl %ebx, (%ecx) # *yp = t0
KAIST
Address
Understanding Swap 123 0x124
456 0x120
0x11c
%eax 0x118
%edx Offset
0x114
%ecx yp 12 0x120 0x110
xp 8 0x124 0x10c
%ebx
4 Rtn adr 0x108
%esi
%ebp 0
0x104
%edi -4
0x100
%esp
movl 8(%ebp), %edx # edx = xp
%ebp 0x104 movl 12(%ebp), %ecx # ecx = yp
movl (%edx), %ebx # ebx = *xp (t0)
movl (%ecx), %eax # eax = *yp (t1)
movl %eax, (%edx) # *xp = t1
movl %ebx, (%ecx) # *yp = t0
KAIST
Address
Understanding Swap 123 0x124
456 0x120
0x11c
%eax 0x118
%edx 0x124 Offset
0x114
%ecx yp 12 0x120 0x110
xp 8 0x124 0x10c
%ebx
4 Rtn adr 0x108
%esi
%ebp 0
0x104
%edi -4
0x100
%esp
movl 8(%ebp), %edx # edx = xp
%ebp 0x104 movl 12(%ebp), %ecx # ecx = yp
movl (%edx), %ebx # ebx = *xp (t0)
movl (%ecx), %eax # eax = *yp (t1)
movl %eax, (%edx) # *xp = t1
movl %ebx, (%ecx) # *yp = t0
KAIST
Address
Understanding Swap 123 0x124
456 0x120
0x11c
%eax 0x118
%edx 0x124 Offset
0x114
%ecx 0x120 yp 12 0x120 0x110
xp 8 0x124 0x10c
%ebx
4 Rtn adr 0x108
%esi
%ebp 0
0x104
%edi -4
0x100
%esp
movl 8(%ebp), %edx # edx = xp
%ebp 0x104 movl 12(%ebp), %ecx # ecx = yp
movl (%edx), %ebx # ebx = *xp (t0)
movl (%ecx), %eax # eax = *yp (t1)
movl %eax, (%edx) # *xp = t1
movl %ebx, (%ecx) # *yp = t0
KAIST
Address
Understanding Swap 123 0x124
456 0x120
0x11c
%eax 0x118
%edx 0x124 Offset
0x114
%ecx 0x120 yp 12 0x120 0x110
xp 8 0x124 0x10c
%ebx 123
4 Rtn adr 0x108
%esi
%ebp 0
0x104
%edi -4
0x100
%esp
movl 8(%ebp), %edx # edx = xp
%ebp 0x104 movl 12(%ebp), %ecx # ecx = yp
movl (%edx), %ebx # ebx = *xp (t0)
movl (%ecx), %eax # eax = *yp (t1)
movl %eax, (%edx) # *xp = t1
movl %ebx, (%ecx) # *yp = t0
KAIST
Address
Understanding Swap 123 0x124
456 0x120
0x11c
%eax 456 0x118
%edx 0x124 Offset
0x114
%ecx 0x120 yp 12 0x120 0x110
xp 8 0x124 0x10c
%ebx 123
4 Rtn adr 0x108
%esi
%ebp 0
0x104
%edi -4
0x100
%esp
movl 8(%ebp), %edx # edx = xp
%ebp 0x104 movl 12(%ebp), %ecx # ecx = yp
movl (%edx), %ebx # ebx = *xp (t0)
movl (%ecx), %eax # eax = *yp (t1)
movl %eax, (%edx) # *xp = t1
movl %ebx, (%ecx) # *yp = t0
KAIST
Address
Understanding Swap 456 0x124
456 0x120
0x11c
%eax 456
456 0x118
%edx 0x124 Offset
0x114
%ecx 0x120 yp 12 0x120 0x110
xp 8 0x124 0x10c
%ebx 123
4 Rtn adr 0x108
%esi
%ebp 0
0x104
%edi -4
0x100
%esp
movl 8(%ebp), %edx # edx = xp
%ebp 0x104 movl 12(%ebp), %ecx # ecx = yp
movl (%edx), %ebx # ebx = *xp (t0)
movl (%ecx), %eax # eax = *yp (t1)
movl %eax, (%edx) # *xp = t1
movl %ebx, (%ecx) # *yp = t0
KAIST
Address
Understanding Swap 456 0x124
123 0x120
0x11c
%eax 456 0x118
%edx 0x124 Offset
0x114
%ecx 0x120 yp 12 0x120 0x110
xp 8 0x124 0x10c
%ebx 123
4 Rtn adr 0x108
%esi
%ebp 0
0x104
%edi -4
0x100
%esp
movl 8(%ebp), %edx # edx = xp
%ebp 0x104 movl 12(%ebp), %ecx # ecx = yp
movl (%edx), %ebx # ebx = *xp (t0)
movl (%ecx), %eax # eax = *yp (t1)
movl %eax, (%edx) # *xp = t1
movl %ebx, (%ecx) # *yp = t0
KAIST

Complete Memory Addressing Modes
• Most General Form
D(Rb,Ri,S) Mem[Reg[Rb]+S*Reg[Ri]+ D]
– D:  Constant “displacement” 1, 2, or 4 bytes
– Rb:  Base register: Any of 8 integer registers
– Ri: Index register: Any, except for %esp
• Unlikely you’d use %ebp, either
– S:  Scale: 1, 2, 4, or 8 (why these numbers?)

• Special Cases
(Rb,Ri) Mem[Reg[Rb]+Reg[Ri]]
D(Rb,Ri) Mem[Reg[Rb]+Reg[Ri]+D]
(Rb,Ri,S) Mem[Reg[Rb]+S*Reg[Ri]]
KAIST

Today: Machine Programming I: Basics
• History of Intel processors and architectures
• C, assembly, machine code
• Assembly Basics: Registers, operands, move
• Intro to x86‐64
KAIST

Data Representations: IA32 + x86‐64
• Sizes of C Objects (in Bytes)
– C Data Type Generic 32‐bit Intel IA32 x86‐64
• unsigned 4 4 4
• int 4 4 4
• long int 4 4 8
• char 1 1 1
• short 2 2 2
• float 4 4 4
• double 8 8 8
• long double 8 10/12 16
• char * 4 4 8
– Or any other pointer
KAIST

x86‐64 Integer Registers
%rax %eax %r8 %r8d

%rbx %ebx %r9 %r9d

%rcx %ecx %r10 %r10d

%rdx %edx %r11 %r11d

%rsi %esi %r12 %r12d

%rdi %edi %r13 %r13d

%rsp %esp %r14 %r14d

%rbp %ebp %r15 %r15d

– Extend existing registers.  Add 8 new ones.
– Make %ebp/%rbp general purpose
KAIST

Instructions
• Long word l (4 Bytes) ↔ Quad word q (8 Bytes)

• New instructions:
– movl ➙ movq
– addl ➙ addq
– sall ➙ salq
– etc.

• 32‐bit instructions that generate 32‐bit results
– Set higher order bits of destination register to 0
– Example: addl
KAIST

32‐bit code for swap
swap:
pushl %ebp
void swap(int *xp, int *yp) Set
{
movl %esp,%ebp
Up
int t0 = *xp; pushl %ebx
int t1 = *yp;
*xp = t1; movl 8(%ebp), %edx
*yp = t0; movl 12(%ebp), %ecx
} movl (%edx), %ebx Body
movl (%ecx), %eax
movl %eax, (%edx)
movl %ebx, (%ecx)

popl %ebx
popl %ebp Finish
ret
KAIST

64‐bit code for swap
swap:
Set
void swap(int *xp, int *yp) Up
{
int t0 = *xp; movl (%rdi), %edx
int t1 = *yp; movl (%rsi), %eax
Body
*xp = t1; movl %eax, (%rdi)
*yp = t0; movl %edx, (%rsi)
}
ret Finish

• Operands passed in registers (why useful?)
– First (xp) in %rdi, second (yp) in %rsi
– 64‐bit pointers
• No stack operations required
• 32‐bit data
– Data held in registers %eax and %edx
– movl operation
KAIST

64‐bit code for long int swap
swap_l:
Set
void swap(long *xp, long *yp) Up
{
long t0 = *xp; movq (%rdi), %rdx
long t1 = *yp; movq (%rsi), %rax
Body
*xp = t1; movq %rax, (%rdi)
*yp = t0; movq %rdx, (%rsi)
}
ret Finish

• 64‐bit data
– Data held in registers %rax and %rdx
– movq operation
• “q” stands for quad‐word
KAIST

Machine Programming I: Summary
• History of Intel processors and architectures
– Evolutionary design leads to many quirks and artifacts
• C, assembly, machine code
– Compiler must transform statements, expressions, procedures into 
low‐level instruction sequences
• Assembly Basics: Registers, operands, move
– The x86 move instructions cover wide range of data movement forms
• Intro to x86‐64
– A major departure from the style of code seen in IA32

You might also like