2011.12.08 12:41:30 (144743400472723456) from Daniel J. Bernstein, replying to "zooko❤ⓩ🛡🦓🦓🦓 (@zooko)" (143749111915610114):
@zooko @sevenps @aumasson Tegra update: Skein down to 37.62 cpb. But most of the ARM implementations still aren't Cortex-optimized asm.
2011.12.05 07:51:57 (143583370096541696) from "zooko❤ⓩ🛡🦓🦓🦓 (@zooko)", replying to "Samuel Neves (@sevenps)" (143349879874387969):
@sevenps @aumasson So Skein is *that* much more amenable to SIMD than Blake is?
2011.12.05 08:25:10 (143591729268924416) from "JP Aumasson (@veorq)", replying to "zooko❤ⓩ🛡🦓🦓🦓 (@zooko)" (143583370096541696):
@zooko @sevenps in theory blake256 seems more SIMD-friendly than Skein (32b vs 64b)
2011.12.05 12:26:43 (143652516264034304) from "Samuel Neves (@sevenps)", replying to "JP Aumasson (@veorq)" (143591729268924416):
@aumasson @zooko I wouldn't say Skein is more SIMD-friendly. Variable rotations in rounds are incompatible with most vector instruction sets
2011.12.05 18:50:33 (143749111915610114) from "zooko❤ⓩ🛡🦓🦓🦓 (@zooko)", replying to "Samuel Neves (@sevenps)" (143652516264034304):
@sevenps @aumasson Then why skein512256-h5tegra: 62cpb, -h1mx515: 15, blake256-h5tegra: 30, -h1mx515: 32? NEON makes Skein 4x as efficient‽