• Source Language: Python

• Target Language: MIPS Assembly

• Implementation Language: Python

• Authors: Abhilash Kumar, Arnab Ghosh and Saurav Kumar

• Tool Used : PLY (Python Lex and Yacc)

1. Representation: opcode rd rs rt

2. Operators:

   • ADD -- Add (with overflow)
   • ADDI -- Add immediate (with overflow)
   • AND -- Bitwise and
   • B-- Unconditional Branch
   • BEQ -- Branch on equal
   • BGEZ -- Branch on greater than or equal to zero
   • BGTZ -- Branch on greater than zero
   • BLEZ -- Branch on less than or equal to zero
   • BNE -- Branch on not equal
   • DIV -- Divide
   • JAL -- Jump and link
   • JR -- Jump register
   • LW -- Load word
   • LI -- Load Immediate
   • LA -- Load Address
   • MULT -- Multiply
   • NOOP -- no operation
   • OR -- Bitwise or
   • SLL -- Shift left logical
   • SUB -- Subtract
   • SW -- Store word
   • SYSCALL -- System call
   • SRL -- Shift right logical
   • LI -- Load Immediate
   • SNE -- Signed Not Equal
   • SLL -- Signed Left Shift
   • SLT -- Signed Less Than
   • SGT -- Signed Greater Than
   • SLE -- Signed Less Than or Equal to
   • SGE -- Signed Greater Than or Equal to
   • SEQ -- Signed Equal To
   • SUB -- Subtraction
   • MFLO -- Move from LO
   • MFHI -- Move from HI

1. Run the makefile

make

2. To run the Assembly Code Generator, pass the path of filename as argument.

bin/codegen test/<filename>.py

The output will be saved in build directory with name .s. This can be run using SPIM simulator or on a MIPS machine.

3. To clean the executables and other helper files, run make clean.

make clean

• bin:
  • codegen [Python dependent bytecode for assembly code generation]
  • converter.py [Python source file to convert the dump of parser into dot file: may be needed for debugging]
  • lex.py [Python source file from PLY for lexing]
  • lexer.py [Python source file to specify language lexemes]
  • parser.py [Python source file to specify grammar]
  • yacc.py [Python source file from PLY for parsing ]
  • runTimeCode.py [Python source which is used for efficient allocation & maintenance of registers as a helper function for codegen.py]
  • symbolTable.py [Python source file with necessary functions related to symbol table]
  • tac.py [Python scource file with necessary functions related to Three Address Code Representation]
• lib:
  • code.s [Assembly source code to generate useful labels]
  • data.s [Assembly source code to allocate space for useful data spaces]
  • library.py [Python source code to be compiled using our compiler and to be used as library for the language]
• src:
  • runTimeCode.py [ Python source which is used for efficient allocation & maintenance of registers as a helper function for codegen.py ]
  • codegen.py [ Python source file which runs the parser and uses the TAC codes to generate the final assembly code ]
  • converter.py [Python source file to convert the dump of parser into dot file: may be needed for debugging]
  • lex.py [Python source file from PLY for lexing]
  • lexer.py [Python source file to specify language lexemes]
  • irgen [Python dependent bytecode for parsing and semantic analysis]
  • parser.py [Python source file to specify grammar]
  • yacc.py [Python source file from PLY for parsing ]
  • symbolTable.py [Python source file with necessary functions related to symbol table]
  • tac.py [Python scource file with necessary functions related to Three Address Code Representation]
• test:
  • 'filename'.py [Test files]
• build: This directory is used to keep assembly code
• .gitignore
• makefile [To move the source files to bin directory and compile bytecode for lexer and making it executable]
• readme.md

The compiler supports the basic structures of the Programming languages such as assignment , operators , loops ans function calls.

*Note : Indentation is through tabs or spaces

Assignment

a = 10
b = 20

Operators

a = 10
b = 20
print a
print b
c = a + b
print c

Relational & Logical Operators

a = 4 > 3
if a == 1 :
  print 8
c = 3 > 4
if c == 0 :
  print 9
if a or c :
  print 10
if a and c :
  print 11
if a and c or a or c :
  print 12  

If Statement

a = 10
if a == 10:
  print 1
else :
  print 2

While Statement

a = 100
summ = 0
while a > 0:
  summ = summ + a
  a = a-1
print summ

Nested Looping

a = 100
summ = 0
while a > 0:
  b = 40
  while b >= 0:
    summ = summ +b
    b = b -1
    a = a - 2
  print summ

Some Library function calls

• max

print max(4,5)

• min

print min(4,5)

Function Without Parameters

def myfun():
  a = 10
  summ = 0
  while a > 0 :
    summ = summ + a
    a = a - 1
  print summ

myfun()

Function With Parameters

def myfun(c, d, e, f):
  a = 35
  b = a + c
  return c+d+e+f
a = myfun(1000, 100, 10, 1)
print a

Function call from another function

def f():
	return 1
def f1():
	a = f()
	return 2*a
x = f1()
print x

Recursion

def sum(n):
	if n == 0 :
		return 0
	a = sum(n-1)
	return a + n
x = sum(100)
print x

Break Statement

i=1
while i<=3:
  if i == 2:
    break
  else :
    i = 3

Continue Statement

i=1
while i<=3:
  if i == 2:
    continue
  else :
    i = 3

def f():
	return 9

def fib(n):
	def f4():
		print 7
		def funinception():
			print 42
			return 44
		a = funinception()
		print a
		return a
	n = f4()
	return n	
	print 3

x = fib(3)
print x

• Library Function calls -some library calls such as max and min have support from our compiler

• Recursive Functions

  • Python supports recursive functions and so does our compiler

• Function call

  • Function call can be made from another function

• Nested Functions -Python language supports a function to be nested within a function -Our compiler supports one function to be only accessible within the scope of another function

• Register Allocation:

  • Remove redundancy as both caller and callee are flushing registers.
  • Make use of dirty bit
  • Update flushtemporary and flsuh registers to make used of function name and dirty bit
  • Use nextReg heuristic from the Dragon Book for better optimization

• Arguments of functions

  • Number of parameters can be varied
  • At function call time, we know the number of parameters passed

• Code Motion
• Common Sub expression removal
• The nextReg and getReg functions work well after doing block analysis of the code block
• All the operations are computed with the help of Registers which are asymptotically
• The stack is used generously used which is faster in an amortized manner

• Multidimensional Array
• Class
• Lambda Calculus
• More libraries