I'm toying with the creation of a boost.spirit based SX assembler, written in C++. update: I'm reconsidering writing the grammar of the assembler in ANTLR instead of spirit, so that i could use the same grammar in Java (eclipse) and C++ (tool chain).
The sx assembler consists of 2 parts:
- frontend; create a syntax tree out of the program text. There is a working spirit grammar without actions (the actual tree building). Open items are:
recognizing opcodes that start in the label column (the first column). Currently, a word that starts in the first column will be parsed as a label, never as an opcode.
- backend; create a list- and hex-file out of the generated parse tree.
The 'complex instructions' Problem
One challenge in creating an assembler for SX is that the standard assembler knows a few 'complex' instructions that translate into two or more 'primitive' instructions. Extra challenging is that these complex instructions cannot be determined by their opcode alone. A 'mov' instruction, for instance can be primitive (
mov w, #10) or complex (
mov SomeAddress, #10), which translates into
mov w, #10 followed by
mov SomeAddress, w
This project is on halt right now.