Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Thu, 3 Sep 2015 21:25:58 +0300
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: 23% performance regression for brypt (Intel i5-4570 CPU)

On Thu, Sep 03, 2015 at 07:59:40PM +0200, magnum wrote:
> On 2015-09-02 02:00, magnum wrote:
> >https://github.com/magnumripper/JohnTheRipper/issues/1199
> 
> Some more tests on various gear:
> 
> i7-4790 (core i7 Haswell, AVX2) gcc 4.8.2:
> Huge boost for X3 except when running HT (slight loss)
> 
> E5-2699 v3 (Xeon Haswell, AVX2) gcc 4.8.2:
> Huge boost for X3 except when running HT (slight loss)
> 
> E5-2670 (Xeon Sandy Bridge, AVX) gcc 4.8.2:
> Huge boost for X3, except with HT: Still a boost but very small.
> 
> "Huge boost" again means 20-30% while "slight loss" means < 5%.

On CPUs with HT, we should primarily care about behavior with HT
enabled.  5% loss with HT enabled is more important than 30% boost with
HT not used or disabled.  Another 5% loss with X3:

http://www.openwall.com/lists/john-dev/2014/12/25/3

The real issue are CPUs that lack HT.  Since such CPUs that nevertheless
have AVX appear to be fairly common (e.g., Frank has one), maybe we
should bite the bullet and enable X3 for all x86-64 builds, after all.
Preferably with a configure option to disable that.

I was hoping to arrive at asm code that would make this problem
obsolete, but it isn't ready yet and I have no time for it now.

Alexander

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.