gangliao / tiger

implement a full compiler based on c++ 11

TIGER - A Tiny Full Compiler

This tiny compiler includes both front end and back end.

Front end: Grammar Rules, LL(1) Parse Table, Syntax and Semantic Check and Intermediate Code.

Back end: IR Optimization (Intra-block CFG optimization), MIPS Register Allocation, Instruction Selection and Code Generation.

You can access https://github.com/gangliao/TIGER to view the documentation!

How to Build

0. development environment

   Currently, this project repository is maintained on github publicly and also been deployed on Travis CI. It supports both Ubuntu and Mac OS X.

1. build:

   	# cd project dir
   	cd Tiger-Compiler
   	# build scanner, parser, generator
   	mkdir build && cd build
   	# cmake building tool
   	cmake ..
   	make -j4

2. run:

   You can parsing test cases named *.tiger under /testCases2 to generate IR code. Default it will utilize CFG optimized technique to generate MIPS asm code.

   	# verbose mode: "-d to implement a verbose mode"
   	./src/parser <filename> -d

   If you want to use the naive mode to generate asm code, simply issue:

   	./src/parser <filename> -d -naive

3. test:

   In testCases2 directory, it includes a test script test.sh to execute all test cases and generate the corresponding asm files *.naive.s and *.cfg.s.

   After souce code is compiled, you can simply issue the commands:

   cd testCases2
   sh ./test.sh

Demo

NOTE: This demo is gif graph format. If you have problem to view it in the markdown file, you can directly open it which located at img/demo.gif.

This Demo shows that

(1) [Compile Source Code] How to compile and generate parser binary ?

(2) [Compiler Front End] How to transfrom raw tiger program into IR code ?

(3) [Compiler Back End] How to generate optimized MIPS asm code via IR code ?

Desgin Internals

Tiger Compiler Front End - Design Internals

Tiger Compiler Back End - Design Internals

Test Cases

We passed all tests cases which provided by TA.

Please check out the details in report Phase2_Testing_and_Output.pdf from current directory, which includes test cases and their quality comparisons for naive and CFG intra-block register allocation.

Accomplishment

• Register allocation code
  • Naive
  • CFG and intra block allocation
  • EBB and intra-EBB allocation
  • Whole function register allocation
  • Live Range Analysis and Graph Coloring
• Instruction selection and generation code
• Passes tests using generated code executing on simulator.
• Report (desgin Internals, how to build, run, code quality comparisions, etc.)

tiger's Issues

Python to excel

optional: import xlrd

Test case pass rate

Passing test cases ... @Rachnas

• test 1
• test 2
• test 3
• test 4
• test 5
• test 6
• test 7
• test 8
• test 9
• test 10
• test 11
• test 12
• test 13
• test 14
• test 15
• test 16
• test 17 @gangliao while
• test 18 while
• test 19
• test 20
• test 21
• test 22
• test 23
• test 24
• test 25
• test 26

Read Skeleton code for tiger compiler

Read Skeleton code for tiger compiler that can help us to divide this work so that we can work simultaneously.

• reading skeleton code @gangliao

• reading skeleton code @Rachnas

1. corrections is grammar (remove ambiguity)

3. parse table

Progress

• naive mode
• cfg
• ebb
• instruction selection
• code generation
• function call
• pass all test cases

I think we do not have to implement ebb. @Rachnas

Read Report

@Rachnas

This issue is to record and track any doubts of project reporting.

• reading the project report @gangliao

some doubts from gang:

....

• read report @Rachnas

2. modify grammer for to support LL parsing

How to develop Tiger-compiler individually and review code by peers

@Rachnas

1. create local branch

git checkout -b rachnas

2. develop new features in rachnas branch
3. pull request from rachnas branch to master branch, which will let peers review your code automatically.

Way to go !!!

Symbol Table Design Doc

When I read source code from Semantic Analyzer directory, I found two files are important to implement symbol table, which is SymbolTablePair.hpp and SymbolTable.hpp.

The report states that we can utilize hash table data structure to accomplish this. Thus, to the best of my knowledge, SymbolTablePair includes Entry and name, Entry could be Variables, Constants, Types, Functions, Temporaries, name is something like terminal id in this language.

We can define

class SymbolTable {
   std::map<SymbolTablePair, Record> symbol_table_;
....
}

red-black tree in C++ STL, SymbolTablePair can be regarded as table entries, Record is actual stuff in this table.

Symbol table implementation is quite easy.
We can fit all things into map data structure, but the really tricky part is how to implement scoping.

I will read textbook to find some ideas...