Lecture Assembly language and assemblers

1
Lecture Assembly language and assemblers

• In this lecture I will cover
• Assembly language and its relation to machine
  languages
• Examination of structure and operation od an
  assembler as an example of a language translator

2
An assembler

• We are going to look at Assemblers as an example
  of a translator
• Remember a translator converts the whole of a
  source code file contents into an output file of
  executable code for the given target machine
• as our example we are using an assembler for an
  assembly language version of Simple Machine
  Language
• Assemblers translate assembly language source
  code into machine code

3
Assembly language

• so what is an assembly language?
• when writing at low level you could write
  programs using numerical values for the
  instructions and the address locations, etc. e.g.
  in SML we might write,
• 2007
• 3008
• 2107

4

• for very small programs and using SML which only
  has a small number of instructions this might be
  doable, but for anything else it becomes very
  time consuming and prone to error
• a programmer would have to remember
• 1. all the numerical values for all of the op
  codes
• 2. the addresses of the locations being used for
  data and in particular which locations are being
  used to hold which data
• 3. the addresses of locations within the code to
  branch to

5

• to solve this problem assembly languages were
  developed
• the op code numeric values are replaced with
  mnemonic names which represent symbolically the
  operation the op code performs e.g. ADD instead
  of 30, BRANCHNEG instead of 41,etc
• the numeric address values are replaced with
  label names which identify a given location,
  which can be
• 1. location of instruction - where an instruction
  is held (locations to branch to) or
• 2. location of data - where data is held

6

• So assuming that address location 0A is the
  address of the next instruction after the end of
  a loop, then we can give that location a label
  name like ENDOFLOOP1 and thus write
• BRANCHNEG ENDOFLOOP1
• to specify the instruction, instead of writing
• 410A

7

• Equally assuming that address location 08 holds
  some data we can give it a label like NUM2,
  then in our assembly language we can write
• ADD NUM2
• instead of 3008
• it is much easier to read and to write assembly
  language code rather than machine code.

8
Op code mnemonics
9
Labels

• As we have seen labels can be used to represent
  address values - to locate instructions or data
• So the question is - how are we to assign address
  values to the labels.
• There are 2 methods for doing this, one for each
  of the 2 different types of location that a label
  might represent.

10

• Assigning address to labels when label
  represents
• 1. Location of instructions. The label name is
  placed at the start of a line or immediately
  before a line which contains the instruction
  which you want to be labelled (e.g. so we can
  branch to it at some time) - label is given
  address value of location in code of the
  instruction it is labelling.
• 2. Location of data. The label name is placed at
  the start of a line and a special directive to
  the assembler is used to indicate that the
  location in the code where the label is placed
  is to be used to hold data and to permit the user
  to initialise that location with a data value
• We will see some examples later

11
Assembler directives

• An assembler directive is an instruction to the
  assembler program about something the assembler
  must do, it is not a source code instruction
  that will be translated into machine code for
  later execution.
• INITIALISE is an assembler directive to allow us
  to initialise data at a specific location, which
  we can access using a label
• if INITIALISE follows a label name then the
  number following INITIALISE is placed in the
  location identified by the label

12
Characteristics of assembly languages

• Characteristics of an assembly language include
• 1.one-to-one correspondence between assembly
  language mnemonics and machine code instructions
• 2.one-to-one correspondence between label names
  and memory addresses
• in an assembly language for a real processor
  there are many more instructions and thus
  mnemonics, and many more sophisticated directives
  to the assembler

13
Format of assembly language code

• Typically source code is arranged into 4 fields
• 1. An optional label field - must start in the
  first column of a line
• 2. An operation field - for mnemonics or
  assembler directives - must NOT start in first
  column of a line
• 3. An operand field - for label names that
  represent locations of data or locations to
  branch to - some operations e.g. HALT do not have
  an operand
• 4. An optional comment field
• Fields are separated by at least one space or tab

14
Example assembly language program for SML

• READ FIRST
• READ SECOND
• LOAD FIRST
• SUBTRACT SECOND first-second
• BRANCHNEG OUT2ND if neg then
• WRITE FIRST 2nd number
• HALT is larger
• OUT2ND WRITE SECOND otherwise
• HALT first larger or
• FIRST INITIALISE 0 same size
• SECOND INITIALISE 0

15
Object code for example

• 1009
• 100A
• 2009
• 310A
• 4107
• 1109
• 4300
• 110A
• 4300
• 0000
• 0000
• Previous slide easier to read, understand and
  write

16
Major functions of assembler

• 1. Replace symbolic op-codes with numeric
  op-codes
• 2. Replace symbolic location names (e.g. labels)
  with actual numeric addresses
• 3. Allocate addresses into which numeric code is
  to be placed i.e what to use as the start address
  of code
• 4. Detect and report errors encountered

17
Implementation of an assembler

• We are now going to look at how an assembler is
  written (implemented in code). We will look at
  how
• 1. To hold the information about the mapping
  between instruction mnemonics and opcode values
  and between label names and address values. This
  is done using a symbol table.
• 2. To break up (analyse) the program text into
  meaningful components so we can identify the
  instruction mnemonics and label names, etc. This
  is done using a tokenizer.

18
1. Use a symbol table

• Symbol table,
• 1. Associates opcode mnemonic with numeric opcode
  value and
• 2. Associates label name with numeric address e.g.

19
2. Use a tokenizer

• During the assembly process it is necessary to
  analyse(i.e. break down) into its components each
  line of assembly source code
• a Tokenizer takes a string as a parameter and
  separates out from the string all the significant
  components that make up the string
• it analyses the string into a series of tokens
  that are separated by delimiters

20

• the default delimiters are white space characters
  like ltspacegt and lttabgt so given a string from
  earlier assembly language program,
• OUT2ND WRITE SECOND otherwise
• it would break it up into 5 tokens
• OUT2ND
• WRITE
• SECOND
• otherwise

21
Operation of an assembler

• Structure of assembler determined by forward
  referencing problem - a symbol may be used
  (referenced) in an operand field before it has
  been defined. This is a problem because the
  assembler cannot generate machine code for the
  instruction until the reference has been resolved
  i.e. it knows what address value symbolic forward
  reference represents.
• Normally handled by reading source file in 2
  passes.

22
First pass - to sort out the references

• First pass reads through assembly language
  program one line at a time, noting the occurrence
  of labels (i.e. symbolic references) that occur
  at the beginning of a line and storing them into
  the symbol table, together with the location
  address value they identify
• to enter correct address value for a label the
  assembler must know the address of the location
  of the instruction being labelled or address of
  location of data word being labelled

23

• To do this, as the assembler works through one
  line after another it calculates how much space
  is required for each instruction and operand it
  comes across and keeps the address of current
  location in a location counter which is updated
  as the assembler moves from one instruction to
  the next
• So when the assembler finds a label it knows what
  the address of the current location is.
• In SML this is easy to do because all
  instructions take up the same space i.e. one
  word, but in other assembly languages this is not
  the case and the calculation is more complex.

24
Second pass - to output the translation

• Second pass reads through source code one line at
  a time, constructing the numeric instruction
  values which are to be output. It does this by,
• 1. looking up the mnemonic op-codes in the symbol
  table and reading out the numeric value that has
  been associated with the given mnemonic.
• 2. looking up the operand labels in the symbol
  table and reading out the address values which it
  appends onto numeric opcode values to give
  complete instruction word
• 3.Instruction word then output to object file

25
Relocatable code

• Currently Simpletron loads a machine code program
  into consecutive locations starting from address
  0. It would be nice to be able to load a program
  anywhere in memory and be able to run the program
  from that location.
• We can easily tell the virtual machine to load a
  program starting from some address other than 0.
• BUT - our machine language programs use address
  values for the location of data and the location
  of places to branch to. Loading the program
  starting from a memory location other than 0 will
  then mean that all the addresses in the machine
  code will now be wrong.

26

• To solve this problem the assembler must output
  relocatable code, so that a loader (called a
  relocating loader) can load the program anywhere
  in memory. To do this the assembler
• 1. produces code as if it was going to be loaded
  starting from address 0 as before but also,
• 2. adds information to each instruction record in
  the object file indicating whether or not
  instruction operand values need modification
• The relocating loader then will add to operand
  address values the value of the new start address
  (often called an offset). This will give the
  correct address.