flat assembler - fasm68k - a Motorola 68000 instruction set for fasmg

Tomasz Grysztar wrote:

Thank you for sharing! I'm adding it to the list.

Awesome.

The 68000 was both easier and harder to implement than I thought. Most of the stuff was very straight-forward because the relatively simple and relatively small and relatively consistent uniform instructions.

Had a few problems, first some addressing modes in 68000 assembly contains the , (comma) character for example:

Code:

ori.b #0, $0F(a5,d2.l)

fasmg would interpret that as 3 arguments, so I had to use something like this (could probably be refactored):

Code:

define m68k.split_operands_result
namespace m68k.split_operands_result
    define num_operands
    define op1
    define op2
end namespace

calminstruction m68k.split_operands? line&
        local op1, op2
        match op1 =, op2, line, ()
        jyes two_operands
    ; one_operand:
        call m68k.parse_operand@op1, line
        compute m68k.split_operands_result.num_operands, 1
        arrange m68k.split_operands_result.op1, line
        exit
    two_operands:
        compute m68k.split_operands_result.num_operands, 2
        arrange m68k.split_operands_result.op1, op1
        arrange m68k.split_operands_result.op2, op2
end calminstruction

calminstruction chk.w? operands&
    call m68k.split_operands, operands
    call m68k.parse_operand@op1, m68k.split_operands_result.op1
    call m68k.parse_operand@op2, m68k.split_operands_result.op2
    emit 1, 01000001b + @op2.ea_reg_bits shl 1
    emit 1, 10000000b + @op1.ea_mode_bits shl 3 + @op1.ea_reg_bits
    call emit_ext_words, @op1.ext_bytes, @op1.ext_bytes_bits
end calminstruction

Since famg outputs in little-endian I needed this:

Code:

macro calminstruction?.emit? num_bytes*, value*
    emit (num_bytes), (value) bswap (num_bytes)
end macro

I suppose I also will need to override other commands such as `store` and `load` which I have not done yet.

I want it to be as compatible with other 68k assemblers as possible but I don't want to spend unreasonable amount of time on that. For example in 68k assemblers they seems to use * (asterisk) as the value of current address (like $ in fasmg), so I think you can have something like this in other 68k assemblers:

Code:

***

which would mean current address times current address. That's probably too complex to support I think.

Another case would be things being comments automatically in the "last column" without any preceeding semicolon. I don't know maybe that's possible to support we'll see.

Colon-less labels seems to be a thing too but the solution for that is in the fasmg manual if I recall correctly.