[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 18 Dec 2011 13:41:27 -0600
From: "JimF" <jfoug@....net>
To: <john-dev@...ts.openwall.com>
Subject: Re: Test results for 179 jumbo-4
From: "Solar Designer" <solar@...nwall.com>
>> >Thus, I am puzzled as to why you're getting them defined on SPARC, which
>> >is also big-endian.
>>
>>...
>> which within the new test suite, was causing it to think dynamic_27 was a
>> valid format on the BE systems, and also NOT making dynamic_29 a valid
>> format on the same build.
>>
>> I have put a patch on the wiki. It is based on jumbo-4, but I am 99% sure it
>> will be just fine, applied to jumbo-5.
>
> Thanks. This is post-jumbo-5 material now, and I think this issue is
> minor enough that it does not warrant a -jumbo-6 yet.
That patch may have been minor enough to skip. But the patch I did last night is not. The fix to hmacmd5. It is failing on SSE2i. It will benchmark test just fine, but when you run it with 'real' data it will fail badly. It will only crack hashes if the password index is within PARA=1. If PARA is more than that, then only the first MMX_COEF limbs will crack, the others will fail, due to the cmp function not properly working past the first block of PARAs.
There is a patch on the wiki, 0002. Without that patch, PARA=2 is only 50% likely to find a password, and PARA=3 is only 1/3 likely. Thus, without patch 0002, this format is pretty poor, on any SSE2i build (which includes most 64 bit builds)
Jim.
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.
