What can assemble AssemPro sources?

[Bruce Abbott]

Originally Posted by phx

But that's compatibility again.

CNOP

could also use

NOP

instructions instead of zeros for alignment (hopefully in the code section only). But that depends on the assembler.

Hopefully. In the KS3.0 ROM there is a $4E71 (NOP) that is not word-aligned. How the hell did that happen?

ProAsm has CNOP, EVEN, ALIGN and ALIGN.L, none of which generate NOPs.

ProAsm is very fast, but it does have a few bugs. It may crash if you have a single character label not on the first column. Some 68020 instructions are coded wrong, others can't take labels when they should be able to.

Devpac has its own issues. The editor screws up if a source file has more than ~32,000 lines. Filenames can't have spaces in them.

Barfly assembler is even faster than ProAsm, but it has one annoying 'feature' - it 'optimizes' code with immediate operands to another 'equivalent' form (but different opcode) even when all optimization is off.

But this is all minor stuff. I have some x86 asm source that I can't find an assembler for. The ones I tried threw up an error on every line!

[phx]

Originally Posted by Bruce Abbott

In the KS3.0 ROM there is a $4E71 (NOP) that is not word-aligned. How the hell did that happen?

Was this the only unaligned instruction in this region? Some assemblers can write instructions to unaligned addresses. So a bad CNOP is not the only source for it.

Besides the restriction to code sections, a NOP used for alignment purposes should never be placed at an odd address. Example:

frank@altair cat tst.asm 
        section code,code
        dc.b    -1
        cnop    0,4
        rts
frank@altair vasmm68k_mot -quiet -Fbin tst.asm
frank@altair hexdump -C a.out 
00000000  ff 00 4e 71 4e 75                                 |..NqNu|

Data:

frank@altair cat tst.asm 
        section data,data
        dc.b    -1
        cnop    0,4
        dc.w    -1
frank@altair vasmm68k_mot -quiet -Fbin tst.asm
frank@altair hexdump -C a.out 
00000000  ff 00 00 00 ff ff                                 |......|

ProAsm has CNOP, EVEN, ALIGN and ALIGN.L, none of which generate NOPs.

Good to know! Then ProAsm is already the second assembler, besides AssemPro, which supports this form of ALIGN directives. Maybe I will implement it in vasm for compatibility.

Barfly assembler is even faster than ProAsm, but it has one annoying 'feature' - it 'optimizes' code with immediate operands to another 'equivalent' form (but different opcode) even when all optimization is off.

You mean CMP to CMPI, ANDI to AND, etc.? Yes, that's really annoying and I made sure in PhxAss (and so in vasm as well) that the assembler always generates the code which matches the mnemonic given in the source.

But this is all minor stuff. I have some x86 asm source that I can't find an assembler for. The ones I tried threw up an error on every line!

All I know is that there are at least two different syntax variants for x86. The Intel-syntax with destination operand first and the MIT-syntax (GNU-assembler) with source operand first. Unlike m68k the MIT-syntax looks cleaner for x86, IMHO (which is what vasmx86 implements).

[meynaf]

Originally Posted by Bruce Abbott

But this is all minor stuff. I have some x86 asm source that I can't find an assembler for. The ones I tried threw up an error on every line!

There are variants for x86. Perhaps you just need to specify which one is the target - 16, 32 or 64-bit (but don't ask me how to do that !). There is also Dos vs Windows, things like that. In short, 'x86' by itself does not mean much (apart: you can expect a big mess).
Best asm for x86 seems to be NASM but don't expect it to be anywhere near what we have for 68k.
It would be great if we had x86 asm/disasm that takes/outputs Mot' syntax.

[phx]

Originally Posted by meynaf

It would be great if we had x86 asm/disasm that takes/outputs Mot' syntax.

Not sure if it makes much sense. Otherwise, that's easy:

frank@nerthus cd vasm
frank@nerthus gmake CPU=x86 SYNTAX=mot
...
frank@nerthus ./vasmx86_mot -v
vasm 1.9b (c) in 2002-2023 Volker Barthelmann
vasm x86 cpu backend 0.7a (c) 2005-2006,2011,2015-2019 Frank Wille
vasm motorola syntax module 3.16a (c) 2002-2022 Frank Wille
vasm test output module 1.0 (c) 2002 Volker Barthelmann

[meynaf]

And for the disassembler ?

[phx]

Maybe Bruce writes one?

[meynaf]

That would be great. Could allow me to start porting PC games.

[Hypex]

Originally Posted by phx

Don't know. It's the name of the assembler used in the official Atari SDK for 6502 and 68000. It became open source in 2008 and created offsprings like smac and rmac for Atari, which also supported the JRISC from the Atari Jaguar. As vasm also supports JRISC, 68000 and 6502, the syntax became important and different enough for a separate module.

See: [URL]http://rmac.is-slick.com/history/history/[/URL

Thanks. Interesting history. Also interesting is the confusing names of smac and rmac as they look like Mac assemblers. :-D

The oldstyle syntax module is intended for old 8-bit targets, with directives used in the 70s and early 80s. Not many of these old assemblers supported macros, so named arguments hardly caused any conflict.

Named arguments would also take up more space. Been a while since I've seen 8-bit ASM which might have had other limits on labels. But I only used the built in assembler on my C16.