Date: Tue, 24 Sep 2013 20:26:55 +0200 From: magnum <john.magnum@...hmail.com> To: john-dev@...ts.openwall.com Subject: Re: Daniel's reports #13, #14 On 24 sep 2013, at 19:50, magnum <john.magnum@...hmail.com> wrote: > On 24 sep 2013, at 16:45, Dániel Bali <balijanosdaniel@...il.com> wrote: >> 2013/9/24 Lukas Odzioba <lukas.odzioba@...il.com> >>> 2013/9/23 Dániel Bali <balijanosdaniel@...il.com>: >>>> Now I'm almost done with the OpenCL implementation, except I get "FAILED >>>> (cmp_one(6))". The two hashes not matching are 872d6c8b and 5ff8c745. >>> >>> Could you please move current code to your public >>> JohnTheRipper:bleeding-jumbo repository on github and keep it updated? >> >> Okay, I pushed the current code to the repository. >> Currently I don't have any ideas about why the check fails. > > You announce a PLAINTEXT_LENGTH of 55 although you only support exactly 8. The set_key() you use "packs" keys (no trailing nor padding zeros) which is a very good idea but your kernel does not really unpack them. As long as all test vectors are exactly length 8 it might not be a problem but if I were you I would fix this before debugging further. For example, there are max-length self-tests going on that will be using length 55 as it is now. From a quick glance this should not break things in this case but who knows. I love it when I'm right. The only thing you need to do is change PLAINTEXT_LENGTH to 8 in line 25 of opencl_crc32_fmt.c, then it works like a champ. magnum
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.