~~Stolen~~ Cross-posted from here: https://fosstodon.org/@foo/113731569632505985
this post was submitted on 02 Jan 2025
809 points (99.1% liked)
Programmer Humor
32710 readers
152 users here now
Post funny things about programming here! (Or just rant about your favourite programming language.)
Rules:
- Posts must be relevant to programming, programmers, or computer science.
- No NSFW content.
- Jokes must be in good taste. No hate speech, bigotry, etc.
founded 5 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
Has anyone ever come across 8 or 16 bit floats? What were they used for?
Neural net evaluation mainly, but FP16 is used in graphics too.
I found this: https://en.wikipedia.org/wiki/Minifloat
Actually, you can consider RGB values to be (triplets of) floats, too.
Typically, one pixel takes up up to 32 bits of space, encoding Red, Green, Blue, and sometimes Alpha (opacity) values. That makes approximately 8 bits per color channel.
Since each color can be a value between 0.0 (color is off) and 1.0 (color is on), that means every color channel is effectively a 8-bit float.
Pretty sure what you're describing isn't floating-point numbers, but fixed-point numbers... Which would also work just as well or better in most cases where floats are used.
Aren't they fractions rather than floating point decimals?