Guide to making a CHIP-8 emulator

[utterances-bot]

Guide to making a CHIP-8 emulator - Tobias V. Langhoff

A high-level guide to making a CHIP-8 emulator.

https://tobiasvl.github.io/blog/write-a-chip-8-emulator/

[cbreezier]

This was a great write-up - I found that it provided a good level of detail while leaving enough of the implementation details for me to figure out.

Two things that were a little unclear to me, both in the DXYN instruction:

1. While drawing the sprite, I wasn't sure whether I should be actually modifying I or just reading from successive memory locations. I ended up not modifying it, which seemed to work fine.
2. The written explanation didn't say anything about setting the VF flag - I had to find that out from the pseudocode section which I had initially tried to avoid.

My attempt (seems to be working, sans sound) is in Rust: https://github.com/cbreezier/rchip8

[tobiasvl]

@cbreezier Hey, thanks for the feedback!

1. Thanks, I've attempted to clarify this. You're correct, I should not be modified (and neither should VX and VY, for that matter).
2. Hmm, the written explanation does have a bullet point about setting VF, though? It says "If the current pixel in the sprite row is on and the pixel at coordinates X,Y on the screen is also on, turn off the pixel and set VF to 1". Let me know if anything is unclear about this, or if you were talking about some other part of the tutorial.

Again, thanks for the feedback, and I'm glad you found it useful!

[DasStone]

Thanks for writing this great article! I’m sure it helped many people get into the topic of emulators, as it did with me.

But there is one thing I’m still not sure about (it has to do with FX0A Instruction). It says that the instruction “blocks” execution. Does this mean that the timers should stop decrementing as well or do they continue at their normal rate of 60Hz (I’ve scanned the Internet for any mention of this, but I haven’t found a clear, unambiguous answer. But it seems like that the timers should continue)?

Maybe this is something you find worth clarifying in your article.

Thanks for your time!

Here is my attempt in Rust for anyone interested: https://github.com/DasStone/chip8emu_rs

[tobiasvl]

@DasStone Thanks for the kind words and the feedback! You're right, this is definitely worth clarifying in the blog post. In the meantime, I can confirm that yes, you're correct, the timers should continue while the FX0A instruction is blocking for input.

In the CHIP-8 interpreter on the COSMAC VIP, FX0A was simply an infinite loop that ran until a key was pressed. The timers were decremented in the COSMAC VIP's built-in display interrupt routine, which is why they're decremented 60 times per second – the display interrupt occurs each frame drawn to the display, which refreshes at 60 Hz. You can read more about it here: https://laurencescotford.com/chip-8-on-the-cosmac-vip-interrupts/

[jakubito]

Hi @tobiasvl, thanks for posting this excellent guide. I referenced it many times while writing my own interpreter.

I have a question regarding one of your statements:

However, the actual drawing of the sprite should not wrap. If a sprite is drawn near the edge of the screen, it should be clipped, and not wrap. The sprite should be partly drawn near the edge, and the other part should not reappear on the opposite side of the screen.

But this reference says:

If the sprite is positioned so part of it is outside the coordinates of the display, it wraps around to the opposite side of the screen.

For example, paddles in Pong seem to wrap in other interpreters I tried (e.g. Octo IDE), but based on your instructions, they shouldn't. Am I missing something here?

Edit: Just found out Octo IDE has an option for wrapping/clipping sprites. I think I'll make it configurable too.

[tobiasvl]

@jakubito Thanks, interesting! I can confirm that my statement is correct – sprites shouldn't partially wrap in either the original CHIP-8 interpreter (source) or in CHIP-48/S-CHIP (source).

However! By "Pong", I assume you're referring to David Winter's Pong game. It was written for his own DOS interpreter, which was based on the CHIP-48 specification. This specification seems to be what Cowgod's reference is based on as well. It's possible David Winter introduced sprite wrapping himself, and this is the reason for the disparity.

[JamesWWalker]

Thanks for this helpful post on creating a CHIP-8 emulator! I just developed an interest in emulator creation and, after a little research, discovered that writing a CHIP-8 emulator (/interpreter) is apparently a rite of passage. With the help of this guide, I put together a working CHIP-8 emulator in Java that passes the BonCoder test, and it only took around 8 hours. I'm planning to try to make an NES emulator next. I expect that to be a big step up in complexity!

[jasonnelson618]

Hello! Love the article and I've been working on my own interpreter, however I noted you said for the IBM logo program only the opcodes you mentioned were needed, but there are several opcodes starting with "F" in the versions of the IBM logo I can find, so I'm quite confused if more opcodes are needed or if I'm messing up something. Any clarity would be appreciated! Thanks in advance.

[tobiasvl]

Hello! Love the article and I've been working on my own interpreter, however I noted you said for the IBM logo program only the opcodes you mentioned were needed, but there are several opcodes starting with "F" in the versions of the IBM logo I can find, so I'm quite confused if more opcodes are needed or if I'm messing up something. Any clarity would be appreciated! Thanks in advance.

@jasonnelson618 Thanks for the kind words! There should be no FXXX opcodes in the IBM program. How are you reaching those opcodes? My guess is that you're reading the graphics data (for the IBM logo) as code somehow, ie. that you have a bug that has caused your Program Counter to point into the graphics, attempting to execute it.

If you push your code to GitHub, I could have a look at it and see if I spot the bug. Otherwise, good luck bug hunting! Might be a good idea to write some debug stuff to the terminal, such as the current PC and opcode at each step, so you see what's happening.

[minchopaskal]

A small spelling error

For this instruction, this is not the case. If V0 contains FF and you execute 6001, the CHIP-8’s flag register VF is not affected.

I guess you meant 7001

[tobiasvl]

@minchopaskal Thanks, fixed!

[fubuki4649]

I cannot seem to find this IBM logo file that you are talking about. The ones I can find all have different instructions that require implementing a much larger portion of the CHIP-8 instruction set to run.

It would be nice if you could link the ROM that you are referring to. Thanks!

[tobiasvl]

I cannot seem to find this IBM logo file that you are talking about. The ones I can find all have different instructions that require implementing a much larger portion of the CHIP-8 instruction set to run.

It would be nice if you could link the ROM that you are referring to. Thanks!

@bloodandcoffee Sure, I can put a link in the guide. In the meantime, it's this one: https://loktar00.github.io/chip8/roms/IBM%20Logo.ch8

(I'm not aware of any other IBM logo programs with more instructions - are you sure you weren't looking at the correct one, but interpreted the graphics data as instructions?)

[mengstr]

This is a nice guide, it really helps me a lot while implementing a Chip-8 interpreter in (microcontroller-less) hardware. Thanks.

I found a small omission in the pseudocode for DXYN - The X variable should be reset in the outer loop after incrementing Y, or else the sprite would smeared out towards the right.

[dedraks]

Thanks for the guide.
I just made my first emulador in Rust.

https://github.com/dedraks/rchip8