9.1 Introduction [1/3]

1
9.1 Introduction 1/3

• Compiler Model
• Front-End - language dependant part
• Back-End - machine dependant part

2
Introduction 2/3

• IL? ???
• Modular Construction
• Automatic Construction
• Easy Translation
• Portability
• Optimization
• Bootstrapping
• IL? ??
• Polish Notation --- Postfix, IR
• Three Address Code --- Quadruple, Triple,
  Indirect triple
• Tree Structured Code --- PT, AST, TCOL
• Abstract Machine Code --- P-code, EM-code,
  U-code, Byte-code

3
Introduction 3/3

• Two level Code Generation
• ILS
• ????? ???? ?? ?? ? ?? ??
• ?? ??? ????? high level??.
• ILT
• ???? ???? ?? ?????? ??? ?? ?? ??
• ????? ????? low level??.
• ILS to ILT
• ILS?? ILT?? ??? ?? ???.

4
9.2 Polish Notation 1/2

• Polish mathematician Lucasiewiez invented the
  parenthesis-free notation.
• Postfix(Suffix) Polish Notation
• earliest IL
• popular for interpreted language - SNOBOL, BASIC
• general form
• e1 e2 ... ek OP (k 1)
• where, OP k_ary operator
• ei any postfix expression (1 i k)

5
Polish Notation 2/2

• example
• if a then if c-d then ac else ac else ab
• ?gta L1 BZ c d - L2 BZ a c L3 BR
• L2 a c L3 BR L1 a b L3
• note
• 1) high level source to IL - fast easy
  translation
• IL to target - difficulty
• 2) easy evaluation - operand stack
• 3) optimization ??? - ?? IL?? translation ??
• 4) parentheses free notation - arithmetic
  expression
• interpretive language? ??

6
9.3 Three Address Code 1/3

• most popular IL, optimizing compiler
• General form
• where, A result address
• B, C operand addresses
• op operator
• (1) Quadruple - 4-tuple notation
• ltoperatorgt,ltoperand1gt,ltoperand2gt,ltresultgt
• (2) Triple - 3-tuple notation
• ltoperatorgt,ltoperand1gt,ltoperand2gt
• (3) Indirect triple - execution order table
  triples

A B op C
7
Three Address Code 2/3

• example
• a b c d / e
• f c d

8
Three Address Code 3/3

• Note
• Quadruple vs. Triple
• quadruple - optimization ??
• triple - removal of temporary addresses
• ? Indirect Triple
• extensive code optimization ??
• IL rearrange ?? (triple ??)
• easy translation - source to IL
• difficult to generate good code
• quadruple to two-address machine
• triple to three-address machine

9
9.4 Tree Structured Code 1/4

• Abstract Syntax Tree
• parse tree?? redundant? information ??.
• Leaf node -- variable name, constant
• Internal node -- operator
• ?? 9.8
• x 0
• y z 2 y
• while ((xltn) (vx ! z)) x x1
• return x

10
Tree Structured Code 2/4

• Tree Structured Common Language(TCOL)
• Variants of AST - containing the result of
  semantic analysis.
• TCOL operator - type context specific operator
• Context
• value --- rhs of assignment statement
• location --- lhs of assignment statement
• boolean --- conditional control statement
• statement --- statement
• ex) . operand - location
• result - value
• while operand - boolean, statement
• result - statement

11
Tree Structured Code 3/4

• Example) int a float b
• ...
• b a 1
• Representation ---- graph orientation
• internal notation ----- efficient
• external notation ---- debug, interface
• linear graph notation

12
Pascal-P Code 1/2

• Pascal P Compiler --- portable compiler
  producing P_CODE
• for an abstract machine(P_Machine).
• P_Machine ----- hypothetical stack machine
  designed for Pascal language.
  
• (1) Instruction --- closely related to the PASCAL
  language.
• (2) Registers PC --- program counter
• NP --- new pointer
• SP --- stack pointer
• MP --- mark pointer
• (3) Memory CODE --- instruction part
• STORE --- data part(constant area,
  stack, heap)

13
Pascal-P Code 2/2
14
Ucode 1/8

• Ucode
• the intermediate form used by the Stanford
  Portable Pascal compiler.
• stack-based and is defined in terms of a
  hypothetical stack machine.
• Ucode Interpreter Appendix B.
• Addressing
• stack addressing gt a tuple (B, O)
• B the block number containing the address
• O the offset in words from the beginning of the
  block,
• offsets start at 1.
• label
• to label any Ucode instruction with a label
  field.
• All targets of jumps and procedures must be
  labeled.
• All labels must be unique for the entire program.

15
Ucode 2/8

• Example
• Consider the following skeleton
• int x
• void main()
• int i
• int j
• // .....
• block number
• - ???? 1
• - ?? ? ???? 2
• variable addressing
• - x (1,1)
• - i (2,1)
• - j (2,2)

16
Ucode 3/8

• Ucode Operations(39?)
• Unary --- notop, neg, inc, dec, dup
• Binary --- add, sub, mult, div, mod, swp
• and, or, gt, lt, ge, le, eq, ne
• Stack Operations --- lod, str, ldc, lda
• Control Flow --- ujp, tjp, fjp
• Range Checking --- chkh, chkl
• Indirect Addressing --- ldi (load indirect),
  sti (store indirect)
• Procedure --- cal, ret, retv, ldp, proc, end
• Etc. --- nop, bgn, sym

17
Ucode 4/8

• Example
• x a b c
• lod 1 1 / a /
• lod 1 2 / b /
• lod 1 3 / c /
• mult
• add
• str 1 4 / x /
• if (agtb) a a b
• lod 1 1 / a /
• lod 1 2 / b /
• gt
• fjp next
• lod 1 1 / a /
• lod 1 2 / b /
• add
• str 1 1 / a /
• next ...

18
Ucode 5/8

• Indirect Addressing
• is used to access the array elements.
• ldi --- indirect load
• replace stacktop by the value of the item at
  location stacktop.
• to retrieve Ai
• lod i // actually (Bi, Oi))
• lda A // also (block number,
  offset)
• add // effective address
• ldi // indirect load gets contents
  of Ai

19
Ucode 6/8

• sti --- indirect store
• sti stores stacktop into the address at
  stackstacktop-1,
• both items are popped.
• Ai j
• lod i
• lda A
• add
• lod j
• Sti

20
Ucode 7/8

• Procedure Calling Sequence
• function definition
• void func(int x, int array)
• function call
• func(a, list)
• calling sequence
• ldp // load parameter
• lod a // load the value of actual
  parameter
• lda list // load the address of actual
  parameter
• call func // call func