Testing my NES compressor on Shakespeare's King Lear. It's actually doing pretty well, not too far off from DEFLATE.

The diagnostic text is also showing me the most common bits of words in it, which is kinda fun.

My compressor turned up ['st thou] as a common word fragment in a work of Shakespeare so I have to believe I'm doing something right.

So I've finally got this NES compressor and decompressor working soup to nuts.
- RAM requirement is _very_ low
- Compression is decent (not as good as DEFLATE but not too far off either)
- Speed is slow but usable

I'd say it's a moderate success. Will prepare it for public soon.

Follow

The NES decoding speed is ~600 cycles per byte, so this is about 3 seconds to decode the full 8k CHR space. A far cry from "instant" but fast enough to be useful for many purposes.

This is just a rough first implementation, too. There's still room to optimize.

For the last part of this test, I implemented a canonical huffman variation of this, which creates data that's ~5% smaller but takes about 7x as long to decode. That's probably too much for too little. :S

I'm back on my "compress Shakespeare for 6502" bullshit...

Feeding random bits to a decompressor that's trained on the text of a short story.

An only slightly buggy decompressor produces some interesting text...

Sign in to participate in the conversation
Mastodon

Server run by the main developers of the project 🐘 It is not focused on any particular niche interest - everyone is welcome as long as you follow our code of conduct!