[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <CALW8-7JHGBp-JjVqs2JLQ-GeSn6eVe9u3ViTsmpbxvsxd1sreQ@mail.gmail.com>
Date: Fri, 4 Sep 2015 13:34:08 +0200
From: Dmitry Khovratovich <khovratovich@...il.com>
To: "discussions@...sword-hashing.net" <discussions@...sword-hashing.net>
Subject: Re: [PHC] Low Argon2 performance in L3 cache
We are still exploring various optimizations. A pure C implementation may
also improve in performance, as for these small memory sizes some C++
artifacts may come into play. For example, I just realized that the memory
was value-initialized by 0, which is now disabled with ~10% performance
gain.
On Thu, Sep 3, 2015 at 11:53 PM, Bill Cox <waywardgeek@...il.com> wrote:
>
> Here's a speed comparison of single-thread hashing of 4MiB between Argon2,
> Yescrypt-2p, and TwoCats on my Xeon E5-1650 CPU running at 3.50GHz:
>
> Argon2d: 2.6 ms
> Yescrypt-2p: 1.8 ms
> TwoCats: 0.72 ms
>
> Assuming the attacker is L3 memory bandwidth bound, he will have a 13X
> higher area*time cost if attacking TwoCats instead of Argon2.
>
I think you're mistaken in squaring the difference factor: area is the same
but the time differs, so cost is only 3.5x (or how much?).
> A natural solution is to use more threads. However, those other cores are
> likely running threads of their own:
>
> Argon2 v1.1.1 4 threads: 0.68 ms
> TwoCats 4 threads, no multiplies: .35 ms
>
> Wasn't Alexander getting something like 4,000 Yescrypt 4 MiB hashes per
> second? If true, this is very impressive.
>
> What can be done to Argon2 to improve L3 hashing performance?
>
>
Maybe another multithreading approach?
On my 1.8 GHz machine I get sharp speed increase when switching from 1 to 2
threads: 4-5x for 1,2,4, even 8 MB.
--
Best regards,
Dmitry Khovratovich
Content of type "text/html" skipped
Powered by blists - more mailing lists