Post ALljyCyDngEYR8aa80 by jordonr@mastodon.technology
(DIR) More posts by jordonr@mastodon.technology
(DIR) Post #ALlLKh7trbKNTsk2j2 by atoponce@fosstodon.org
2022-07-22T21:47:20Z
0 likes, 0 repeats
/dev/zero vs GNU coreutils yes(1) performance:% dd if=/dev/zero bs=32M | pv -s 100G -S > /dev/null 100GiB 0:00:55 [1.80GiB/s] [================================>] 100%% yes | pv -s 100G -S > /dev/null 100GiB 0:00:20 [4.87GiB/s] [================================>] 100%
(DIR) Post #ALlLVbbiiuqfY7tP3h by theavidhorizon@fosstodon.org
2022-07-22T21:49:19Z
0 likes, 0 repeats
@atoponce hot damn that's quite a bit faster. Reminds me of standard gzip vs pigz.
(DIR) Post #ALlLZu9bi40WEFnDo8 by atoponce@fosstodon.org
2022-07-22T21:50:06Z
0 likes, 0 repeats
@theavidhorizon Yeah. It certainly surprised me, to the point of "is it actually faster than /dev/zero? Holy shit."
(DIR) Post #ALlLuZAcvAyFEN7ckC by theavidhorizon@fosstodon.org
2022-07-22T21:53:50Z
0 likes, 0 repeats
@atoponce I think a lot of these programs will be reimplemented with more efficient multicore/thread capable alternatives. I would't even have given "yes" a single glance if not for your post.
(DIR) Post #ALlMZdVvg0ncZifaQC by wedaly@fosstodon.org
2022-07-22T22:01:15Z
0 likes, 0 repeats
@atoponce the first one has double the number of syscalls (read + write instead of just write)
(DIR) Post #ALlNxm5hXULe52TYYa by emacsen@emacsen.net
2022-07-22T22:16:49Z
0 likes, 0 repeats
@atoponce You've proven that userspace is faster than the kernel ;)
(DIR) Post #ALlO3k5Ttw1HbVDbaC by atoponce@fosstodon.org
2022-07-22T22:17:54Z
0 likes, 0 repeats
@wedaly Still less than half the performance however.
(DIR) Post #ALljyCyDngEYR8aa80 by jordonr@mastodon.technology
2022-07-23T02:23:24Z
0 likes, 0 repeats
Wondering how fast you can “yes” a drive (if you even can, verses zeroing with /dev/zero# yes | dd of=/dev/sdX