design.txt

design


steps

- input: loads from file
- newline: convert all newlines to LF, ensure LF at EOF
- pass0: remove comments, tokenize


program arguments

- target architecture
- output format (a.out, coff, elf, ...)


assembler syntax

- comments
  - line comment: //
  - block comment: /* */
  - nesting block comment: /+ /+ +/ +/
- directives
  - all are prefixed with a dot
  - must be lowercase
  - file inclusion:
    - .include "filename"
  - symbols, conditional:
    - .def name value
    - .undef name
    - .if name|value
    - .ifdef name
    - .ifndef name
    - .else
    - .elif name|value
    - .endif
  - data:
    - .byte val
    - .word val
    - .dword val
    - .qword val
    - .dqword val
    - .range start end [-]increment
  - strings:
    - .str string
    - .strz string+0
  - alignment:
    - .align value
  - addressing:
    - .org addr
  - sections:
    - .section sect
- instructions
  - opcodes must be lowercase
  - in situations where size is ambiguous, a character must be appended to the
    opcode: b for byte, w for word, l for dword
  - operands are in source, destination order
- integers
  - integers behave mostly as in C11:
    - dec starts with nonzero
    - oct starts with 0
    - hex starts with 0x
    - bin starts with 0b
    - no suffixes
  - negative integers are treated as two's compliment numbers
  - integer size will be as large as possible by default, and then narrowed down
    as necessary where used
- registers
  - register names must be lowercase
  - no prefix
  - suffix for size in ambiguous situations
- effective addresses
  - indirect:  [eax]
  - indexed:   [eax+4]
  - sized:     [4*eax+4]
  - postinc:   [...]++   (if applicable)
  - predec:  --[...]     (if applicable)
- macros
  - should have C-style macro replacement (functions)
  - should have do-this-n-times (loops)
- output
  - possibly user-chooseable
  - one of the object formats that ld understands
  - possible formats:
    - a.out
    - coff
    - elf


TODO: scour the gas documentation for useful directives, macro ideas, etc.