Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 13 Sep 2015 19:35:25 +0300
From: Solar Designer <>
Subject: Re: experiment with functions to reject computed hashes

On Sun, Sep 13, 2015 at 07:13:56PM +0300, Solar Designer wrote:
> Oh, OK.  I see we're not even skipping the last 3 steps for SHA-2, even
> though we were already doing that on GPU a couple of years ago, but seem
> to have lost that since.

We do have SHA256_SHORT and SHA512_SHORT macros, which skip the last 3
rounds, but we're not using them anywhere.  myrice's code actually had
that working.

> BTW, I think we might in fact want to revert
> the non-iterated SHA-2 GPU formats to prior to Claudio's implementation
> of hash comparisons on GPU, which mostly made things worse (way too slow
> when cracking more than a handful of hashes - turned out to be unusably
> slow in the recent contest).  myrice's implementations worked better
> (and IIRC had no GPL restrictions on them too).

I think we should at the very least re-introduce myrice's latest code as
an alternative.  Until Claudio corrects the performance regression when
cracking many hashes, this older code will run faster.  And when Claudio
does correct the performance regression, this older code should still be
kept for its relaxed license (and should be the basis for further
enhancements for that reason).


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.