Skip to content

Latest commit

 

History

History
181 lines (118 loc) · 6.26 KB

README.md

File metadata and controls

181 lines (118 loc) · 6.26 KB

PCL

Build Status

Compiler for the PCL Language written in Python.

This is part of the semester assignment for Compilers course taught in ECE NTUA (Spring 2018-2019).

👥 Authors

This compiler would have never been born without the orderly contributions of its authors


🍅 What is PCL?

PCL is a imperative programming language based on a proper subset of ISO PASCAL, among with some changes. The basic characteristics of PCL include:

  1. Syntax similar to PASCAL
  2. Structured functions similar to PASCAL
  3. Basic data types for integers and real numbers, booleans and characters
  4. Arrays of fixed or variable size
  5. Built-in function library

The complete PCL specification is available under docs/pcl2019.pdf (in Greek).

🔩 Setup

The compiler comes with a Makefile for installation. Install it via

make depend
make compiler

Please note that PCL compiler requires Python >=3.6 to work since it has metaprogramming features (due to SLY) to specify lexers and parsers. Older versions won't work.

🔨 Usage

After you have set up PCL you can use the pclc.py executable of the PCL compiler. You can display the usage of pclc.py via

pclc.py --help

The pcl executable allows the use of its constituent parts independently. Such parts include

  1. The lexer
  2. The parser
  3. The semantic analyzer
  4. The codegen module

Compile PCL programs

For instance assume that you have the following PCL program under example.pcl

program dummmy;
	var x: integer;
begin
	x := 0;
	while x < 10 do
	begin
		writeInteger(x);
		writeChar('\n');
		x := x + 1;
	end;
end.

which prints the numbers 0 to 9. Running

pclc.py example.pcl

will produce IR code at example.imm , the object file at example.o and the final (linked) executable at example.out. If you specify the flag -i then the program should be read from stdin and the IR will be emitted to stdout . So the PCL compiler should be called as

pclc.py -i <example.pcl >example.imm

If one wants to create an object file, he could use UNIX pipes

pclc.py -i <example.pcl | llc -filetype=obj >example.o

Compiling and linking the final can be separately done with the use of gcc. If one wants static linking with the pcl/builtins.h library then one should use

gcc -Wall -Werror -fpic -lm example.o /path/to/builtins.c/builtins.c -o example.out 

Note that builtins depend on stdio.h and math.h so the flag -lm is used to do dynamic linking with the math library. If one wants to perform dynamic linking with pcl/libbuiltins.so one should use

gcc -Wall -Werror -fpic -lm -L/path/to/libbuiltins.so -lbuiltins example.o -o example.out

Again, if one wants to use pipes, it is possible via the -x option as

pclc.py -i <example.pcl | llc -filetype=obj | gcc -Wall -Werror -fpic -lm -L/path/to/libbuiltins.so -lbuiltins -o example.out -x -

The pclc.py executable produces verified LLVM code, throwing an exception otherwise. Successful compilation exits with code 0.

Furthermore one can specify optimization flags using the -O argument. More specifically the acceptable values are 0, 1 and 2 according to the LLVM Reference for PassManagerBuilder. For a complete and fathomable list on the LLVM optimizations performed by the optimizer, we redirect the interested reader to this StackOverflow thread.

Test individual parts of PCL

For testing individual parts of the compiler, one has to specify the --pipeline argument as a list containing a subset of the following (in correct order) arguments:

  • lex to invoke the lexer
  • parse to invoke the parser
  • sem to invoke the semantic analyzer
  • codegen to invoke the codegen module
  • pprint to print the (annotated) AST to stdout.

So in case one wants to do lexical analysis only

pclc.py example.pcl --pipeline lex pprint

to get the tokens to stdout. If one wants to do lexing, parsing and semantic analysis (type checking, label checking) and get the types of the compatible AST nodes.

pclc.py example.pcl --pipeline lex parse sem pprint

More information on the design of the current compiler can be found at the wiki.

📺 Technological Stack

This implementation of PCL is developed using the Python language and the following meta-programs

  1. SLY for lexing and parsing
  2. llvmlite for IR generation
  3. LLVM for producing the object files

Contributing to the project

If you want to contribute to the project, please submit a pull request.

Documentation

The PCL documentation is located at docs/ and the wiki. You can generate the API docs from the docstrings via

pydoc pcl.submodule

where submodule is one of the submodules inside pcl/.

Tests

The PCL Language comes with tests built-in for every "independent" part of the compiler. The tests are located in the tests/ directory and the examples used in the examples/ directory. You will need pytest to run them. You can run the tests via

make test

References

If you want to dive deeper into our compiler we advise you study the following references

  1. Ullman, Jeffrey D., and Alfred V. Aho. "Principles of compiler design." Reading: Addison Wesley (1977).
  2. Skordalakis, Manolis, and Papaspyrou, Nikolaos. "Compilers". Symmetria Publications (2003 - in Greek).
  3. PCL Specification under docs/pcl2019.pdf (in Greek).
  4. llvmlite Reference Manual.
  5. SLY Reference Manual.