Hacker News Reader

Hacker News

Comment by blueflow | Hacker News Reader

Parent Full thread

blueflow

Infuriatingly differently from the 8086 i used to program for.

I tried to enter a "db CDh, 20h" which is the second-shortest valid DOS program.

- The "DB" instruction does not support multiple operands

- Programs needs to have start labels

- Start label can't point to bytes defined by "DB"

- NOP instruction not recognized

- DB cant be used without a label??

So i changed my mind and tried to write a simple 'ret' one-byte program. Which is a valid way to exit a program in DOS and CP/M.

  Unrecognized token `` found at 0:3 Expected one of Error : ret is encountered without corresponding call

This is not an error in assembly!

Also the terms. "compiling" is the process that generates assembly from a higher language. "assembling" is the process of turning assembly into machine code.

Whatever this is, this is not for me.

View on HN

Ooops, yeah it probably needs some work. I _think_ it is the work of a student and I found it surprisingly beautiful and well done. The actual grammar (which is how I stumbled upon this project) is here:

https://github.com/YJDoc2/8086-Emulator/blob/master/src/lib/...

Looks like it's not emulating x86 at all, just interpreting assembly mnemonics as if they were BASIC.

edit

Here's how to define a 'procedure', which is different from a 'label' and can be called:

    def count_up {
        INC AX
        CALL count_up
    }

    start:
        XOR AX,AX
        CALL count_up

Now the 'C'ultists are polluting assembly language with their curly braces! Notice that SP doesn't change while it counts up, the call stack exists only inside the interpreter.

I actually had a calculator as a kid that could be programmed in both BASIC and a similar pretend-assembler.

Hey, yes I do like the braces ;}

The decision to differentiate procedures from labels was intentional, and taken due to some restrictions of our IR (which is not the best, issue is open to change that).

Also, the way to define procedure is this way (separate form labels), as I felt that might be a bit more familiar to newcomers to assembly, who are coming from other languages.

Call stack exists inside assembler due to the same reason, as in case someone accidentally bashes their stack , it would still not corrupt the whole code. Maybe we should do a better job at reporting this so users can know they have corrupted the stack, instead of silently ignoring.

The SP should be changing, it is probably a bug. Thanks for reporting, I'll try to take a look at it.

>The SP should be changing, it is probably a bug. Thanks for reporting, I'll try to take a look at it.

Since the code doesn't have an address inside the emulated memory space, it kind of makes sense for there to be a separate return stack. Problem is that you're trying to emulate 8086 and not a Forth machine :)

One solution might be to have 'CALL' decrement SP (perhaps writing some dummy value like 0000 or FFFF to memory), and also remember where SS:SP points now. The interpreter's return stack could store that address, along with the reference to the line of code it should return to.

When executing 'RET', check if the stack pointer is at the same location as it was after the last 'CALL', and halt with an error message if it is different. I suggest you also allow 'JMP' to procedures, or just get rid of the separation entirely and make everything a label. The procedure syntax could still be allowed as an alternative but just declare a normal label.

Another small but impactful bug: REP seems to always perform one more iteration when CX is zero!

> Looks like it's not emulating x86 at all, just interpreting assembly mnemonics as if they were BASIC.

Wow. Reading the authors other comments, it sounds like they couldn't find an assembler they liked for their operating system, and instead decided to spend their course building their own instead, as they imagined it would work. Stay away from this tool if your goal is to learn about 8086. This project is an 8086-inspired programming language.

I was a bit disappointed too, but wouldn't be quite as negative:

It is clearly a student project, intended to run very simple code and show how the registers and memory are affected. They did read the Intel manual, and most of the instructions seem to be implemented correctly including flag behaviour; for example the SBB AL,0x69 + DAS trick works.

Generating and running actual machine code might well be "out of scope" for something like this.

Some typos:

    (segment-register) [bs/bp/si/di , signed word number]; The offset is taken from the registers, and the number is added to it
    (segment-register) [bs/bp , si/di , (signed word number) ];

Hey, I'm sorry but can you be more clear in what are the typos? I'd be glad to fix them, but just seem to miss them in this comment :sweat_smile:

If your site hadn't been so "advanced", I could just have given you a URL ;)

It's in the "Memory" part of the instructions page.

Hey, I can totally understand that this did not work out for you. I am glad nonetheless that you gave it a try anyways, and mentioned the issues you faced, so we can improve on them. Thanks for that!

This was originally made primarily for our own use, so I can definitely understand that the choices we made for this is not everyone's cup of tea.

However, I would just like to reply to the points, in case anyone else is also facing the same things.

- DB indeed doesn't support setting multiple different values, and that is definitely a thing we should change when we update his. Alternatively you can set multiple bytes to a same value using `DB [value,count]`

- Yes, this is intentional. As this emulator is not a "true" 8086 emulator that can just start from any memory address, we need a start label, much similar to how C needs a main function.

- Again, intentional. DB is strictly for data storage, and the whole code is (internally) split into data section (at start) and code section (after data section). Thus you cannot define DB once you start the code section. And as said, start indicates entry point of code, so any DB instruction cannot have start label.

- NOP is not implemented, and is an open issue at the repo.

- DB can be used without a label. for example ``` db [7,5] DW 0x7857 start: mov byte[0], 5

``` is a valid and compiling code. In case you had any issues using DB without label, feel free to comment here, or open an issue on github.

- Yes, ret cannot be used outside procedures. This is again intentional. As this emulator was aimed at students and newcomers to assembly, we wanted to kind of safeguard from accidentally using ret outside of a procedure context. In real assembly, it will let you use ret anywhere, but this does not.

> Also the terms. "compiling" is the process that generates assembly from a higher language. "assembling" is the process of turning assembly into machine code.

You are very much correct on this one. However, in this particular case I would suggest use of 'compiling' is also correct. As noted, this is an emulator of 8086, and we use an string based IR generated from the input assembly code. (This is not the best choice in hind-sight and there is an open issue for changing this). In that sense, we are indeed "compiling" the assembly code you have given to an executable/runnable representation, which then the emulator (which is basically an interpreter in this case) can run.

That said, I do agree that in general, (not in particular context of this emulator) "compiling" is not suited.

---

All said, This might not be everyone's choice for emulator, and I am glad that you gave it a try. Thanks!

>Yes, ret cannot be used outside procedures. This is again intentional. As this emulator was aimed at students and newcomers to assembly, we wanted to kind of safeguard from accidentally using ret outside of a procedure context.

This decision seems more likely to teach newcomers the wrong things, like expecting the following code to work:

    def plus {
        ; add two numbers on stack, return result on stack
        pop ax
        pop bx
        add ax,bx
        push ax
        ret
    }

To catch errors like this, you might want to implement a "shadow stack" in the interpreter that keeps track of the type of each stack element (number or return address).

edit, some more thoughts on how to implement this:

https://news.ycombinator.com/item?id=35517012

/edit

But the most essential feature of assembly / machine language is that there is in reality no distinction between characters or numbers and memory addresses, procedures and labels, or between code and data. Even compiler-generated code often has jumps to procedures (tail-call optimization).

Also by not exposing any code addresses, your virtual machine becomes less powerful than C - how would you call a function pointer?

It's an 8086 emulator, not a DOS emulator.

You don't need DOS for any of the things i tried to do above.

Even on a non-DOS 8086 platform, you would get the instructions for it to assemble.

It's more like a PC 8086 emulator (or at least it tries to be). I mean the example is a call to the BIOS...

http://www.ctyme.com/intr/rb-0210.htm

Long time no see Ralf Brown's Interrupt List...

Reply on news.ycombinator.com