iiinova — where your digital self keeps evolving

Zooko🛡️ @zooko.bsky.social · 07/07/2023

On bench.cr.yp.to/results-hash.html#aa…, BLAKE3 is one of the most efficient secure hash functions, but in reality BLAKE3 is much *more* efficient than shown there, because it uses ARM NEON (which is not shown in that benchmark) and because of the new BLAKE3 v1.4.1 speedups.

131

Zooko🛡️ @zooko.bsky.social · 01/07/2023

@elonmusk I demand the return of the Fail Whale!

141

Zooko🛡️ @zooko.bsky.social · 01/07/2023

New OpenZFS release candidate just dropped with BLAKE3 as a checksum. phoronix.com/news/OpenZFS-2.2-rc1 This is another example of software engineers upgrading from weak/insecure algorithms to BLAKE3 (which is secure) due to BLAKE3's performance. ⤵️

040

Zooko🛡️ @zooko.bsky.social · 01/07/2023

Hey im back cuz twitter is ratelimiting me and i wanna tell everyone how awesome BLAKE3 is

051

Zooko🛡️ @zooko.bsky.social · 27/05/2023

Thanks, tqbf. I'm guessing you missed the second half because of this UI on this app, where the second half of the thread is hidden behind an innocuous "Continue thread…": /cc @pfrazee.com

110

Zooko🛡️ @zooko.bsky.social · 07/05/2023

Mashing reload on bench.cr.yp.to/web-impl/amd64-beeli… to see when the new hash function benchmarks are going to drop. (Jack @oconnor663.bsky.social O'Connor has submitted a patch for a problem in the bench.cr.yp.to benchmarks of BLAKE3.)

010

Zooko🛡️ @zooko.bsky.social · 05/05/2023

Sorry deep links to specific posts on my blog are broken, but if you search long and hard you might be able to find this post: tahoe-lafs.org/~warner/blog.html#%5…

110

Zooko🛡️ @zooko.bsky.social · 02/05/2023

According to bench.cr.yp.to/web-impl/amd64-icela… the SSE impl of BLAKE3 should be only about 25% slower than the AVX512 impl on IceLake. So something really doesn't add up.

100

Zooko🛡️ @zooko.bsky.social · 02/05/2023

Hmmm… am i reading this paper right? This seems to say that the SSE impl of BLAKE3 on IceLake took 57 cpb on 64 b, but the AVX512 impl of BLAKE3 on IceLake on bench.cr.yp.to/results-hash.html#am… took <3 cpb on 64 b.

100

Zooko🛡️ @zooko.bsky.social · 02/05/2023

You can see SHA3 in this graph. In the future there will probably be modest improvements in SHA3 performance, due to SHA3 circuits in chips, but there will probably be bigger improvements in BLAKE3 performance, due to parallelization: bsky.app/profile/zooko.bsky.social/…

130

Zooko🛡️ @zooko.bsky.social · 01/05/2023

Contrast with the legacy web site, where when i post two links the first one disappears and will probably be overlooked by 99% of readers: twitter.com/zooko/status/1652869524…

050

Zooko🛡️ @zooko.bsky.social · 30/04/2023

BLAKE3 is *much* more efficient (in time and energy) than SHA256, like 14 *times* as efficient in typical use cases on typical platforms.

170

Zooko🛡️ @zooko.bsky.social · 30/04/2023

BLAKE (the original) was very well-studied during the SHA3 competition. In NIST's final report on the SHA3 process, they stated that the depth of scientific analysis applied to BLAKE exceeded even that applied to Keccak (the final SHA3 winner)

1110