[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20130127015906.GA3503@openwall.com>
Date: Sun, 27 Jan 2013 05:59:06 +0400
From: Solar Designer <solar@...nwall.com>
To: john-dev@...ts.openwall.com
Subject: Re: mask-mode speed
On Mon, Jan 07, 2013 at 07:22:54PM -0500, jfoug@....net wrote:
> ---- Solar Designer <solar@...nwall.com> wrote:
> > Actually, we can do a bit faster.
>
> I put your code in, and the speed did increase, some. It went from 12410K/s to 12610K/s.
It was more like 12k to 13k for me (or a ~7% speedup).
> A little faster but still a huge overhead for a fast format. But this is still about 1/2 the speed of natural CRC32 (on this PC), and about 6x the overhead of the current mask mode code.
Of course. I am not arguing that external can be as fast as builtin
mask mode - for fast hashes it can't be.
> Again, the 'OT' for this post is about mask mode, and it certainly IS much faster than even existing incremental, and is very simple code, and hopefully will be very easy to implement within the GPU. A less overhead than even incremental will be very nice, when we get to ultra fast formats, like the simple formats in GPU.
Right.
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.
