Zstd Compressed Linux Kernel Images Proposed Once More

[phoronix]

Phoronix: Zstd Compressed Linux Kernel Images Proposed Once More

Going back to at least late 2017 have been proposals for Zstd-compressing the Linux kernel images for the Facebook-developed Zstandard compression algorithm. In 2020 perhaps we will finally see the support mainlined...

http://www.phoronix.com/scan.php?pag...ompressed-2020

[Paradigm Shifter]

While I'll admit that it's probably the sort of data I'm compressing... I've not noticed Zst being "10x" faster to decompress than xz or similar. Maybe 2-3x...

[thelongdivider]

Zstd is awesome for transparent compression on btrfs. It's far more compressed than lzo, but has equal or greater decompression rates.

[Chugworth]

But isn't LZ4 still slightly faster than ZSTD at decompression?

[stiiixy]

And the compression times during updates take less time as well?

[ms178]

As more distros change their compression format to ZSTD, it is about time these patches get merged as the algorithm has proven itself and its usefulness in the field. They can kick out less efficient algorithms if they don't want to support too many. Albeit I don't know if that would be too much of a maintanance burden anyway.

[milkylainen]

Yes please!

[archsway]

But isn't LZ4 still slightly faster than ZSTD at decompression?

An lz4 compressed kernel is about twice as large as xz or zstd, so the extra time to load it from disk would cancel out any improvement in decompression time.

[log0]

But isn't LZ4 still slightly faster than ZSTD at decompression?

ZSTD is LZ4 with some extra stuff to achieve higher compression ratios.

[AlB80]

An lz4 compressed kernel is about twice as large as xz or zstd, so the extra time to load it from disk would cancel out any improvement in decompression time.

lz4hc has good compression ratio (~zstd) with best decompression speed (memcpy/2).