Skip to content

Add lifthrasiir codec at 5,707,152 bytes.#24

Merged
agavra merged 4 commits intoagavra:mainfrom
lifthrasiir:main
Feb 13, 2026
Merged

Add lifthrasiir codec at 5,707,152 bytes.#24
agavra merged 4 commits intoagavra:mainfrom
lifthrasiir:main

Conversation

@lifthrasiir
Copy link
Contributor

At the moment, this is the best codec that doesn't use any non-zstd entropy coder (which was my own, intentional constraint). It also somehow beats the current winner when zstd is replaced with lpaq1, but that's no fun ;-)

Disclaimer: Almost entirely written with GLM-4.7 and Claude Code. Compression codecs are famously hard to get right, but modern coding agents make them much more doable!

This is not strictly necessary because the submission expects the input
to be sorted, but small tweaks can make it work with unsorted inputs as
well. This commit improves encoding for those cases (only).
@agavra
Copy link
Owner

agavra commented Feb 13, 2026

Confirmed in CI:

β”Œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”¬β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”
β”‚ Codec                  β”‚           Size β”‚ vs Naive   β”‚
β”œβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”Όβ”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€
β”‚ Naive                  β”‚    210,727,389 β”‚   baseline β”‚
β”‚ lifthrasiir            β”‚      5,707,266 β”‚     -97.3% β”‚
β””β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”΄β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”€β”˜

Very nicely done, impressed that it works without zstd! I'll update the leaderboard to have the CI-confirmed score and merge.

@agavra agavra merged commit 5d97ade into agavra:main Feb 13, 2026
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants

Comments