@gianni@disobey.net avatar gianni , to random

New Halide blog post on post-processing for chroma as it is relevant to compression efficiency: https://halide.cx/blog/chroma-handling

Seems like you can get a roughly 2% efficiency improvement with 4:2:0 decoder output just with smarter post-processing for chroma. Pretty neat!

@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@h4ckernews@mastodon.social avatar h4ckernews Bot , to random

Post-transformer inference: 224× compression of Llama-70B with improved accuracy

https://zenodo.org/records/17873275

@gianni@disobey.net avatar gianni , to random

Introducing oavif: faster target quality image compression

https://giannirosato.com/blog/post/oavif/

oavif can be up to 63% faster than traditional target quality encoders. Learn how in the attached blog post!

@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@aras@mastodon.gamedev.place avatar aras , to random

New thing from the Zstd (and LZ4/xxhash) crowd: OpenZL. A dozen or so building blocks (shufflers, transposers, filters, compressors) that can be chained together. Can automatically build "best" config based on sample data; same decompressor can decompress any config. Nice! https://openzl.org/ (and whitepaper: https://arxiv.org/abs/2510.03203)

@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@h4ckernews@mastodon.social avatar h4ckernews Bot , to random

Borg - Deduplicating Archiver with Compression and Encryption

https://www.borgbackup.org/

@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@gianni@disobey.net avatar gianni , to random

Introducing Iris-WebP, a better WebP encoder: https://halide.cx/iris/index.html

Iris offers up to >20% better compression than libwebp, and trades blows with libaom for AVIF.

My goal is to build an image-first encoder ecosystem for the modern Web, so that new web image codecs aren't generally superfluous.

@h4ckernews@mastodon.social avatar h4ckernews Bot , to random

7-Zip for Windows can now use more than 64 CPU threads for compression

https://www.7-zip.org/history.txt

@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@h4ckernews@mastodon.social avatar h4ckernews Bot , to random

From Tokens to Thoughts: How LLMs and Humans Trade Compression for Meaning

https://arxiv.org/abs/2505.17117

@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@h4ckernews@mastodon.social avatar h4ckernews Bot , to random
@octade@soc.octade.net avatar octade , to linguistics group

Hexlish Alphabet for English, Constructed Languages and Cryptography: Automatic, Structural Compression with a Phonetic Hexadecimal Alphabet

DOI : https://doi.org/10.5281/zenodo.13139469

Hexlish is a legible, sixteen-letter alphabet for writing the English language and for encoding text as legible base 16 or compressed binary. Texts composed using the alphabet are automatically compressed by exactly fifty percent when converted from Hexlish characters into binary characters. Although technically lossy, this syntactic compression enables recovery of the correct English letters via syntactic reconstruction. The implementer can predict the size of the compressed binary file and the size of the text that will result from decompression. Generally it is intuitive to recognize English alphabet analogues to Hexlish words. This makes Hexlish a legible alternative to the standard hexadecimal alphabet.

@conlang languagelovers@a.gup.pe icon languagelovers group linguistics@a.gup.pe icon linguistics group academicchatter@a.gup.pe icon AcademicChatter group

@vulkan@fosstodon.org avatar vulkan , to random

This new sample demonstrates how to use different types of compressed GPU textures in a Vulkan application showing the timing benefits of each using Vulkan-Hpp.

https://github.com/KhronosGroup/Vulkan-Samples/tree/main/samples/performance/hpp_texture_compression_comparison