Zstd Compression For EROFS Published: Better Than LZ4 But Higher CPU Costs

[ddriver]

...

Looks like your box is whack!

[Anux]

Looks like your box is whack!

? I don't quiet understand that. I made some edits with a tmpfs test, it clearly shows zstd is faster.

Edit: Damn I'm stupid, did all my test in my power saving profile but still doesn't change that zstd is faster
lz4 2258 MB/s
zstd 3204 MB/s

[erniv2]

Why does it state "CPU overhead".

Is it zstd is more heavy on the cpu with more compression "wich is expected".

Or is it the zstd implemention is ineffective and just gobbles up the cpu for no know reason ?

[intelfx]

So I am supposed to dislike the guy cuz he was hired by a another lousy guy to be exploited by his lousy company?

I have no idea what you are supposed to dislike, but if you dislike Zstd because of "Zuck symbology" (whatever that means), then you probably should be disliking LZ4 for the same reasons — because those two projects belong to the exact same person employed by the exact same company.

[evergreen]

Well, if you are comparison CLI speeds,
you are comparing difference I/O implementations,
not just the algorithms.

[Anux]

Well, if you are comparison CLI speeds,
you are comparing difference I/O implementations,
not just the algorithms.

So you think using the library in a project would result in faster lz4 speed? Maybe, but since both are from the same dev, why would he use a superior I/O implementation for one and not the other (especially lz4 which is all about speed)?