Date: Wed, 10 Apr 2013 15:02:21 -0500 From: "jfoug" <jfoug@....net> To: <john-dev@...ts.openwall.com> Subject: RE: default_cmp_all From: magnum [mailto:john.magnum@...hmail.com] > You mis-read it. Here's the unstable code: > ... clip > It tests salt->hash_size, not the same as "salted or not" :-) So it is not using cmp_all, if there are very few items. If that is the case, then the original issue reported within raw-sha256-ng's cmp_all requires correction. cmp_all should return a true value if any value is found, and false otherwise. I think a simple change to this will 'fix' it: return _mm_movemask_epi8 (_mm_cmpeq_epi32 (tmp, zero)) != 0xffff; Also, I think we should look at making sure there are checks done during self test, where there are NO matches expected, making sure that cmp_all returns false. It's cmp_all is not right. It will always return true. I was running against the 1500 items in a test suite file, and cmp_all was not being called.
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.