A crude attempt at an editor & assembler/disassembler for Atmel microcontrollers (although not for the tiny ones)... very much a work in progress. Assembly/disassembly works pretty good now (verified by dog fooding and comparing generated disassembly with avr-objdump output and disassembling and re-compiling large amounts of data that cover every possible opcode). Not all preprocessor / assembler directives are implemented yet and the editor is still quite crude

The editor currently supports

• background compilation
• disassemble,save,load
• searching back & forth (using CTRL-k,CTRL-k CTRL-b)
• navigating to a specific line number (CTRL-g)
• navigating to your last edit
• CTRL-clicking on identifiers to jump to their definition
• CTRL-space auto-completion of identifiers

I intend to add a rename refactoring and more goodies though (auto-indent and block editing are sorely needed as well as support for multi-line macros...)

• Maven
• JDK >= 1.8

mvn clean package

java -jar target/javr.jar

Note that for reasons unknown to me the AVR assembler duplicates a lot of preprocessor instructions as assembler directives ... I'm currently only implementing the preprocessor instructions

• full ATmega88 instruction set
• able to convert the AVR assembly syntax to avr-as
• output to RAW / Intel HEX / ELF (executable/relocatable) files
• local labels ( watch out, currently not working properly when used in macros)
  • Just write '.myLabel' instead of 'myLabel:' when declaring them ; referencing local labels works just like with global labels so no 'b' or 'f' suffixes are needed. Note that its illegal to declare a local label that has the same identifier as a global label ; otherwise I would've needed to use the ugly 'b'/'f' suffix solution to resolve the ambiguity)
• Expression correctly handles nested expressions and operator precedence for the following operators:
  • arithmetic operators: -(unary+binary),+,/,*
  • bitwise operators: ~,<<,>>,|,&
  • logical operators: !,&&,||,>, <, >= , <= , == , !=
• Expressions support floating-point numbers (careful, those will be silently truncated to integral numbers during code generation)
• Number literals support the underscore ('_') character to improve readability ( %1010_1111, 0xff_ff_ff, 1234_5678 )
• built-in functions: HIGH(x), LOW(x) that yield the upper/lower 8 bits of a 16-bit word
• Assembler understands X / Y / Z as shorthand notation for r27:r26 / r29:r28 / r31:r30
• Assembler currently will NOT accept just specifying the lower register when an instruction expects a combined / 'compound' register
  • use the syntax as given in the official Atmel ISA documentation instead
• Preprocessor instructions
  • #include
  • #if
  • #ifdef / #ifndef / #endif
  • #define
  • #message / #info / #warning
  • #pragma [ parsed but ignored ]
• Assembler directives
  • .byte
  • .cseg
  • .db
  • .def /.undef
  • .dw / .word
  • .dseg
  • .eseg
  • .equ
  • .irq
    • This directive currently only works when selecting ELF relocatable as output format. It generates a relocation that tell the linker to put a JMP instruction into the given IRQ vector that jumps to the next global function label following the .irq directive. The maximum IRQ vector index (zero-based) depends on the project's target architecture.

• .org directive
• local labels in macros

• the LPM instruction treats the Z register as holding a byte address while SPM treats it as holding a WORD address.

This assembler internally treats all addresses as byte addresses (even in .cseg) so when using LPM you have to write

.cseg

ldi ZH,HIGH( label )
ldi ZL,LOW( label )
lpm r16,Z 

label: .db 0x01,0x02,0x03,0x04

but with SPM you need to write

.cseg

ldi ZH,HIGH( label>>1 )
ldi ZL,LOW( label>>1 )
spm Z+

label: .db 0x01,0x02,0x03,0x04

• You can do forward/backward references to local labels without having to add suffixes like 'f' or 'b' to the name , so stuff like the following compiles fine:

  main: cpi r17,0x20 breq skip ldi r16,0x0f .loop dec r16 brne loop .skip

• Auto-indent is crude and buggy (sometimes corrupts files)
• Gutter icon is off by one line for some compilation messages
• Generated relocation information when output format is ELF is not complete yet (see failing unit tests)
• Syntax coloring is off when using MSDOS-style line endings (won't fix this since it is related to the fact that Swing text components internally convert all EOL sequences to '\n' but my parser uses the 'true' text offsets)
• parsing #define is currently broken when trying to #define stuff like (a+b)/c (gets irritated by the leading parens)
• parsing of .def is quite picky when it comes to whitespace ( '.def x = r16' works but '.def x=r16' doesnt ...)
• Parser throws NPE because of NULL ICompilationContext when certain tokens are present at the end of the top-level file. This is because the PreprocessingLexer prematurely pops the last context while the parser is not yet done parsing. Work-around is to just add a newline to the very end of the top-level file.

• Fix failing ELF relocation unit-tests
• Add support for floating-point arithmetic in expressions
• Fix NPE because of NULL ICompilationContext when certain tokens are present at the end of the top-level file.
• Implement object size calculation for ObjectType.FUNCTION symbols (though I didn't notice any avr-ld glitches caused by the missing size info yet)
• Support having '_' characters as separators in binary literals to improve readability
• Make .irq work when generating RAW output, not just ELF relocatables
• Support emitting ELF relocations for targets other than atmega328p
• Compile source files individually (output ELF relocatable) and add linking step to reduce compilation times and allow C programs to be linked against the output
• Add support for .org directive
• Better parse error recovery
• Add support for #else / #elseif
• Find a way for meaningful error reporting when expanding macros...