Date: Mon, 7 Jan 2013 03:52:56 +0100 From: magnum <john.magnum@...hmail.com> To: john-dev@...ts.openwall.com Subject: Re: BLAKE2 (was: New plugin load order magic) On 5 Jan, 2013, at 6:03 , Solar Designer <solar@...nwall.com> wrote: > On Sat, Jan 05, 2013 at 10:00:10AM +0530, Dhiru Kholia wrote: >> Speed using XOP on FX-8120. There is lot of improvement but I was >> expecting more. > > Yeah. As you might have noticed, I exchanged some tweets with @aumasson > and @sevenps on this a couple of weeks ago. There's not enough > parallelism in one instance of BLAKE2 to fully hide Bulldozer's > instruction latencies. Things are much better on Sandy Bridge. > > So we'll need to do our own interleaved instructions for multiple > instances for cracking these hashes faster, much like we do for > MD4/MD5/SHA-1. Jim, Do you feel like taking a stab at interleaving per above? I think you are best suited among us. Also, I am not sure raw-blake2 will currently build on non-SIMD platforms, although this should be a trivial fix (there is generic code present too). magnum
Powered by blists - more mailing lists
Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.