Flispcc is a C (subset) compiler written in Java for the FLISP processor, which is an 8-bit processor used in introductionary courses to computer engineering on Chalmers University of Technology. This compiler make use of the BNFC tool, which generates a compiler front-end from a labeled BNF grammar.
The FLISP processor is very similar to the Motorola 68HC11 microcontroller since both are accumulator-based (one-address) CPUs with similar instructions and architechture. This led me to believe that it was easier to translate an exisiting compiler for the 68HC11 into FLISP instructions. After some research I realized that it was harder than what I first thought. I would have to invest many hours learning how GCC works. In the course DAT151 "Programming Language Technology" I recently took, we learned about context-free grammars and some tools that produces compiler front-ends, and this is where I learned about BNFC. The laborations for the course recommended BNFC to handle the parsing of files. So instead of translating GCC code I choose to write a compiler myself.
The precidence and associativity is directly derived from "C--", which is a small fragment of C++ used in the course I mentioned above:
| Level | Expression forms | Assoc | Explanation |
|---|---|---|---|
| 6 | literal | - | literal |
| 6 | identifier x | - | variable |
| 6 | x(e,...,e) | none | function call |
| 6 | x++, x-- | none | post in/decrement |
| 6 | ++x, --x | none | pre in/decrement |
| 5 | e*e, e/e | left | mult, div |
| 4 | e+e, e-e | left | add, sub |
| 3 | e<e, e>e, e>=e, e<=e | none | comparison |
| 3 | e==e, e!=e | none | (in)equality |
| 2 | e&&e | left | conjunction |
| 1 | e||e | left | disjunction |
| 0 | x=e | right | assignment |
Note: Only a subset of C is implemented, since all the features of C11 can not be fully utilized with the FLISP.
| Datatype | Description |
|---|---|
| unsigned char | 8 bit unsigned value [0-255] |
| signed char | 8 bit signed value [-127-126] |
| int | same as signed char |
| bool | 1 bit value (true=1, false=0) |
| string | n+1 bit value (n=token) |
Note: only int implemented so far
Functionality from the C language that will not be implemented or is unnecessary to implement will be listed here:
- No instructions for multiplication and division (could be implemented with functions)
- Including other files (.h files)
Functionality that is planned to be included will be listed here:
- Unary
- Function calls
- All bitwise operations (eg. a&b, a|b)
- String variables, literals and a subset of string operations
- all types are not implemented fully yet.
- I can not guarantee that complex expressions work.
- Branching implemented but not tested
- ..(and more)
-
Linux
- Download binary from git or make from source. (Making requires the dependencies that BNFC requires).
- Download digiflisp from here.
- Add QAflisp to PATH:
export PATH=$PATH:/usr/share/digiflisp/(to the folder where you install digiflisp). - Now the compiler chain should call the assembler after compilation.
java Main <file.c>
-
Windows Same as above but add QAflisp to Path environment variable, different procedure but should give same result.
Note: I have only tested installing the compiler and running the compiler on Linux.
To build from source you need to have BNFC installed together with ANTLR4.
While in the source folder run make.