The cr.yp.to microblog: 2012.02.11 04:14:45

2012.02.11 04:14:45 (168171085462896641) from Daniel J. Bernstein, replying to "zooko❤ⓩ🛡🦓🦓🦓 (@zooko)" (168167842326380546):

@zooko You need better eyeballs. :-) Those graphs compare implementations of one primitive; they don't compare primitives. Read the tables.

Context

2012.02.10 23:07:29 (168093759505043457) from "JP Aumasson (@veorq)":

BLAKE-256 tree mode in SUPERCOP http://bench.cr.yp.to/primitives-sha3.html

2012.02.10 23:44:10 (168102989515587584) from "zooko❤ⓩ🛡🦓🦓🦓 (@zooko)", replying to "JP Aumasson (@veorq)" (168093759505043457):

@aumasson I can't tell... does the tree mode get compared to the linear modes in benchmarks?

2012.02.11 03:46:54 (168164074490839040) from Daniel J. Bernstein, replying to "zooko❤ⓩ🛡🦓🦓🦓 (@zooko)" (168102989515587584):

@zooko @aumasson But people who say that tree modes can't help for single cores are drawing bogus conclusions from naive performance models.

2012.02.11 04:01:52 (168167842326380546) from "zooko❤ⓩ🛡🦓🦓🦓 (@zooko)":

@hashbreaker @aumasson Interesting! Eyeballing http://bench.cr.yp.to/impl-hash/bblake256.html vs. http://bench.cr.yp.to/impl-hash/blake256.html, the tree mode is slower so far.