Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Tue, 26 Jun 2012 06:14:52 +0200
From: Frank Dittrich <frank_dittrich@...mail.com>
To: john-dev@...ts.openwall.com
Subject: Re: precompiled sse-intrinsics vs. -march=native

On 06/26/2012 03:41 AM, Solar Designer wrote:
> The attached patch removes the use of precompiled sse-intrinsics from
> GPU targets.  The alternative would have been to remove -march=native
> from them.  I don't know which is the better choice; neither is perfect.
> Since modern systems tend to have at least AVX (and maybe also XOP),
> -march=native may be preferable over the precompiled SSE2 code.  Yet
> another alternative would be to have more GPU targets for the different
> combinations, but that would be confusing.

I noticed that on my 64bit system some formats were slower with -native
than they were with -avx, so you might use relbench -v to decide...
(BTW: on my 32bit system I noticed that clang produced faster code than
gcc (4.6.3) for several formats.)

Frank

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.