What is Assembly Language?
Introduction to the GNU/Linux
assembler and linker
for Intel Pentium processors
High-Level Language
• Most programming nowdays is done using
so-called “high-level” languages (such as
FORTRAN, BASIC, COBOL, PASCAL, C,
C++, JAVA, SCHEME, Lisp, ADA, etc.)
• These languages deliberately “hide” from
a programmer many details concerning
HOW his problem actually will be solved
by the underlying computing machinery
The BASIC language
• Some languages allow programmers to
forget about the computer completely
• The language can express a computing
problem with a few words of English, plus
formulas familiar from high-school algebra
• EXAMPLE PROBLEM: Compute 4 plus 5
The example in BASIC
1
2
3
4
5
LET X = 4
LET Y = 5
LET Z = X + Y
PRINT X, “+”, Y, “=“, Z
END
Output:
4+5=9
The C language
• Other high-level languages do require a small
amount of awareness by the program author of
how a computation is going to be processed
• For example, that:
- the main program will get “linked” with a
“library” of other special-purpose subroutines
- instructions and data will get placed into
separate sections of the machine’s memory
- variables and constants get treated differently
- data items have specific space requirements
Same example: written in C
#include <stdio.h>
// needed for printf()
int
int
// initialized variables
// unitialized variable
x = 4, y = 5;
z;
int main()
{
z = x + y;
printf( “%d + %d = %d \n”, x, y, z );
}
“ends” versus “means”
• Key point: high-level languages let programmers
focus attention on the problem to be solved, and
not spend effort thinking about details of “how” a
particular piece of electrical machiney is going to
carry out some desired computation
• Key benefit: their problem gets solved sooner
(because their program can be written faster)
• Programmers don’t have to know very much
about how a digital computer actually works
computer scientist vs. programmer
• But computer scientists DO want to know
how computers actually work:
-- so we can fix computers if they break
-- so we can use the optimum algorithm
-- so we can predict computer behavior
-- so we can devise faster computers
-- so we can build cheaper computers
-- so we can pick one suited to a problem
A machine’s own language
• For understanding how computers work,
we need familiarity with the computer’s
own language (called “machine language”)
• It’s LOW-LEVEL language (very detailed)
• It is specific to a machine’s “architecture”
• It is a language “spoken” using voltages
• Humans represent it with zeros and ones
Example of machine-language
Here’s what a program-fragment looks like:
10100001 10111100 10010011 00000100
00001000 00000011 00000101 11000000
10010011 00000100 00001000 10100011
11000000 10010100 00000100 00001000
It means:
z = x + y;
Incomprehensible?
• Though possible, it is extremely difficult,
tedious (and error-prone) for humans to
read and write “raw” machine-language
• When unavoidable, a special notation can
help (called hexadecimal representation):
A1 BC 93 04 08
03 05 C0 93 04 08
A3 C0 94 04 08
• But still this looks rather meaningless!
Hence: assembly language
• There are two key ideas:
-- mnemonic opcodes: we employ abbreviations
of English language words to denote operations
-- symbolic addresses: we invent “meaningful”
names for memory storage locations we need
• These make machine-language understandable
to humans – if they know their machine’s design
• Let’s see our example-program, rewritten using
actual “assembly language” for Intel’s Pentium
Simplified Block Diagram
Central
Processing
Unit
Main
Memory
system bus
I/O
device
I/O
device
I/O
device
I/O
device
Pentium’s internal “registers”
• Four general-purpose registers:
eax, ebx, ecx, edx
• Four memory-addressing registers:
esp, ebp, esi, edi
• Six memory-segment registers:
cs, ds, es, fs, gs, ss
• The instruction-pointer and flags registers:
eip, eflags
The “Fetch-Execute” Cycle
main memory
central processor
Temporary
Storage
(STACK)
ESP
Program
Variables
(DATA)
Program
Instructions
(TEXT)
EAX
EAX
EAX
EAX
EIP
the system bus
our program’s ‘data’ section
.section
x: .int
y: .int
.comm
fmt: .string
.data
4
5
z, 4
“%d + %d = %d \n”
our program’s ‘text’ section
.section
.text
main:
# comment: assign z = x + y
movl
x, %eax
addl
y, %eax
movl
%eax, z
‘text’ section (continued)
# comment: print the program results
pushl
z
pushl
y
pushl
x
pushl
$fmt
call
printf
addl
$16, %esp
‘text’ section (concluded)
# comment: return control to the caller
ret
# comment: make label visible to linker
.global
main
program translation steps
demo.s
demo.o
demo
program
source
module
assembly
program
object
module
object module library
object module library
object module library
linking
the
executable
program
The GNU Assembler
• With Linux you get free software tools for
compiling your own computer programs
• An assembler (named ‘as’): it translates
assembly language (called the ‘source code’)
into machine language (called the ‘object code’)
$ as demo.s -o demo.o
• A linker (named ‘ld’): it combines ‘object’ files
with function libraries (if you know which ones)
• A C compiler (named ‘gcc’) which invokes both:
$ gcc demo.s -o demo
What must programmer know?
•
•
•
•
•
•
Needed to use CPU register-names (eax)
Needed to know space requirements (int)
Needed to know how stack works (pushl)
Needed to make symbol global (for linker)
Needed to understand how to quit (ret)
And of course how to use system tools:
(e.g., text-editor, assembler, and linker)
Summary
• High-level programming (offers easy and
speedy real-world problem-solving)
• Low-level programming (offers knowledge
and power in utilizing machine capabilities)
• High-level language hides lots of details
• Low-level language reveals the workings
• High-level programs: easily ‘portable’
• Low-level programs: tied to specific CPU