Tiger Compiler from "Modern Compiler Implementation in ML" by Andrew W. Appel
This is a Tiger compiler with all modules completed until Chapter 12 in the book. Basically I followed all the instructions from the book without much alternation. I also tried to adhere to the functional style as much as possible, that means, minimize the usage of mutable data structures.
So far the compiler works with either mac or linux. Besides that, you need to have Standard ML and spim installed. Both of them can be fetched via homebrew.
Standard ML is obviously need to compile the programs. Since the compiler generates spim code, you'll need the SPIM simulator to actually run the executables.
To build the compiler, simply run
make main
This will build the image using sources.cm. Note that it currently
doesn't support windows.
Given a tiger source program foo.tig, to compile it, you'll need to first copy
the file to the project directory, and run
make foo.spim
This will compile the file, append runtime.s to it, and generate a
output SPIM program called foo.spim. Then to execute the program,
run
spim -file foo.spim
If foo.spim exists, make will not overwrite it. You must remove the file before running make again.
The second half of the book covers a lot of interesting stuff, and I'm thinking about extending the compile with the following features:
- Garbage Collection
- OO Features
- Functional Features
- Polymorphism and Type Inference
- Optimizations