Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Fri, 4 Sep 2015 00:02:56 +0300
From: Solar Designer <>
Subject: Re: MD5 I() (was: SHA-1 H())

On Thu, Sep 03, 2015 at 10:42:05PM +0200, Lukas Odzioba wrote:
> 2015-09-03 22:06 GMT+02:00 Solar Designer <>:
> > For md5crypt-opencl, we're getting some slowdown - but that kernel is
> > weird (as discussed elsewhere, it currently uses global memory in its
> > inner loop).  We need to optimize it some more, then re-test this change.
> I did not notice a significant slowdown, after a couple of tests it
> seems that results are within measurement error and max seems to be
> the same 1851k c/s.

Is this --test speed on a lower clocked Tahiti?

The slowdown I saw was during actual cracking at length 8, something
like 3000K+ to 2900K (and further down to 2700K when the card heats up
and drops to non-turbo speed, 1050 MHz to 997.5 MHz in this test).

> When we reduce memory bottleneck it should work as expected.

Yes, I expect so.


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.