Hello!

A few days ago YT suggested a 6502-emulation video to me, so now it looks like I'm doing an emulator. For fun, as you do .

I've read and seen a couple of explanations of how the SBC works, and that it's an ADC with ~op2, which is easy enough to implement as I already had the ADC implemented.

I've used the emulator at https://www.masswerk.at/6502/ to generate testdata for my instructions with this simple program:

SEC/CLC
LDA #xx
SBC #yy


However, one particular case seems weird to me:
CLC ;edited
LDA #$00
SBC #$FF
=> A: 00 N:0 V:0 Z:0 C:0

Why is Z 0?
I can solve it be only setting Z:1 if C:1, but I haven't seen any documentation mentioning this, just that it reflects the result of A.

Can somebody please enlighten a green 6502 nerd?

edit: CLC, not SEC

_________________
http://WilsonMinesCo.com/ lots of 6502 resources
The "second front page" is http://wilsonminesco.com/links.html .
What's an additional VIA among friends, anyhow?

In this example, A should be 01 not 00,.

You are right, it should be CLC, not SEC. I edited my first post.
I wrote the program in hex and translated to the wrong mnemonic for the forum post, sorry.

So, is ~C considered part of the result and thusly why Z:0, not just what is in register A?

There is a bug in the emulator you are using for that particular case, because Z should always match the 8-bit result. So in this case it should be Z:1.

Here's the emulator's SBC code:

Can you spot what the bug is?

I would not recommend using this emulator as a golden reference!

Ok, great. That certainly clears that.
I thought it be best to assume existing emulators probably had it right and I'm wrong. But in cases like this it seems a second opinion was needed.

I'm guessing the bug is only setting fZER when r==0, when it should check rb?

Thanks for the help!

Yes, except in your example r is -256, hence the r==0 test failing.

Dave

Interesting; I've used that emulator a fair bit when I was away from home and wanted to test some code. That seems like a fairly bad bug!

_________________
"The key is not to let the hardware sense any fear." - Radical Brad

I pinged Norbert and he's fixed the bug!

Nice one Ed.

As this emulator is quite popular, I wonder if it's worth trying to run the Dormann tests?

Hi, everyone!

I just fixed this.

What's somewhat interesting is that nobody noticed or reported this in all those years. Even more so, as the same emulation (with a few additions for cycle accurate timing) is powering the PET 2001 emulator at https://www.masswerk.at/pet and not a single program seems to have been affected by this. (So, for all practical means, this seems to be quite a rare edge case.)

Regarding running any test suites, this may be a bit problematic as the emulator runs at reduced speed in order to convey a glimpse on what's happening. If there's any test suite running on a PET, this would be more realistic and welcome.

As far as I know, the PET emulator runs every program there is, including the 2022 demos. (The only exceptions are those requiring a business keyboard. But I haven't seen a program fail or received any reports of it failing.) Especially "A Bright Shining Star" by Genesis Project is highly critical and VICE used to fail on this with timing issues. (I haven't verified this for the latest version of VICE, so, please, don't take my word for granted. I guess, they should have fixed this in the meantime.)
Still, this somewhat proves that real-life performance is not a perfect test and further tests may be required.

--Norbert

PS: On a positive note, all the applications of the "virtual 6502" suite have recently received an update for improved import of hex-dumps, disassemblies, and Commodore PRG-files. (Meaning, hex-dumps, etc, should be cleaned up properly prior to import and any code start addresses found should be applied automatically. You will be prompted for confirmation, since every automation may fail over something, I haven thought of.)
Moreover, there's now also a small hex-editor for quick for quick edits and/or patch-up jobs. (Its special feature is that it comes with a few text tools.)

Cool!

So, good news: the emulator passed the basic (functional) Dorman test – that is, w/o any BCD mode ADC/SBC tests.
I'll have a look into BCD later.

-- Norbert

That is good news!

(I have to say, Klaus Dormann's test suite has been, and continues to be, an excellent tool for showing up little bugs in emulators. Indeed, it's also been useful in showing up design or implementation bugs in hardware systems, as a consequence of being a thorough exerciser of opcodes. Note that it's not an exhaustive test - one might try Wolfgang Lorenz' suite for that. (See https://web.archive.org/web/20201112023 ... stPrograms))