Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 17 Jun 2012 20:02:43 +0400
From: Solar Designer <>
Subject: Re: [patch] optional new raw sha1 implemetation

On Sun, Jun 17, 2012 at 07:27:53PM +0400, Solar Designer wrote:
> There's currently no speedup on XOP:
> user@...l:~/john/magnum-jumbo-cpu3/src$ ../run/john -te -fo=rawsha1_sse4
> Benchmarking: Raw SHA-1 [taviso sse4]... DONE
> Raw:    21591K c/s real, 21591K c/s virtual
> user@...l:~/john/magnum-jumbo-cpu3/src$ ../run/john -te -fo=raw-sha1
> Benchmarking: Raw SHA-1 [SSE2i 8x]... DONE
> Raw:    23513K c/s real, 23751K c/s virtual
> (This is a linux-x86-64-xop build.)  I guess some speedup can be
> achieved by adding use of XOP intrinsics into your new code - that is,
> it should become faster than 23.5M c/s then.

With XOP bit rotates added, I am getting:

Benchmarking: Raw SHA-1 [taviso sse4]... DONE
Raw:    27478K c/s real, 27478K c/s virtual

Benchmarking: Raw SHA-1 [taviso sse4]... DONE
Raw:    28838K c/s real, 28838K c/s virtual

(different invocations of the same code).  We should be able to improve
this further to 30M+ c/s.  I am not making use of _mm_cmov_si128() yet.


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.