Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 02 Sep 2015 20:19:02 +0200
From: magnum <>
Subject: Re: SHA-1 H()

On 2015-09-02 17:20, Solar Designer wrote:
> SHA-1's H() aka F3() is the same as SHA-2's Maj(), yet somehow we're
> using less optimal expressions for it on systems with bitselect().
> In opencl_sha1.h we have:
> #define F3(x, y, z)     (bitselect(x, y, z) ^ bitselect(x, 0U, y))
> I've just tried changing this to:
> #define F3(x, y, z)     bitselect(x, y, (z) ^ (x))
> and got some speedup for pbkdf2-hmac-sha1-opencl on GCN (1200K to
> 1228K c/s).
> (...)
> magnum, will you take care of this, please?  And test on GPUs and on XOP.

Good find, I totally missed this one. I'm glad I moved dozens of code 
copies into opencl_sha1.h and similar headers: Much less code to 
maintain. I'll look into it (including the stuff I snipped from my quote).


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.