Furry Technologists

1371 readers

1 users here now

Science, Technology, and pawbs

founded 2 years ago

MODERATORS

[email protected]

AI language models can exceed PNG and FLAC in lossless compression, says study | Ars Technica (arstechnica.com)

submitted 1 year ago by [email protected] to c/[email protected]

5 comments fedilink hide all child comments

you are viewing a single comment's thread
view the rest of the comments

[–] [email protected] 2 points 1 year ago (1 children)

QOI should actually be worse than JPEG-XL in both speed and size, if you turn JPEG-XL's effort level down to a speed-oriented level like -e 1. I wrote a quick script to compare them, and this is the output from 10 random (fully-optimized) PNGs I had: https://pastebin.com/raw/nSkBSePB

I have a lot of cores on my CPU so you probably want to be looking at the user time metric, but consider that parallelization is a strong and intentional benefit to JPEG-XL.

The results are not exactly scientific since I only picked 10 PNGs, but it shows a general trend.

[–] [email protected] 1 points 1 year ago (1 children)

Ooh, i didn't take parallelization into consideration, that's very interesting, thank you

[–] [email protected] 0 points 1 year ago

Not sure if I made it clear but the real time metric is equivalent to time passing in real life, and user time is the total CPU time spent by the application, e.g. 2 seconds of user time across 4 cores might be 0.5 seconds of real time. Parallelization is really good for end users (browsers loading an image, someone saving a file from GIMP, etc), but single-threaded performance can still be very important for enterprise applications where you're converting a ton of images all the time, and you can just give every conversion its own thread instead. I mainly focused on user time when looking at the difference between JXL and QOI, since that's more indicative of the total amount of work being done in a CPU-agnostic way.