Date: Mon, 31 Dec 2012 09:18:44 +0530 From: Dhiru Kholia <dhiru.kholia@...il.com> To: john-dev@...ts.openwall.com Subject: Re: New self-test for maximum length On Monday 31 December 2012 02:14 AM, magnum wrote: > On 30 Dec, 2012, at 18:11 , Dhiru Kholia <dhiru.kholia@...il.com> wrote: >> On Sunday 30 December 2012 10:26 PM, magnum wrote: >>> On 30 Dec, 2012, at 3:49 , magnum <john.magnum@...hmail.com> wrote: >>>> Unfortunately things get worse if you build with OMP. The test is more effective with more than one keys per crypt(). Not that I have seen any more of your formats yet, it segfaults on IPB so the rest are not tested with a --test=0 run. >>>> >>>> We may have a lot to do. I'll keep focusing on OpenCL formats for now. >>> Benchmarking: Clipperz SRP SHA256 [32/64 oSSL-exp]... (8xOMP) FAILED (get_hash(6)) >>> >> I will look into this one tomorrow. > I had a look at it but saw nothing. It's intermittent, works once out of about four with OMP build, but no problems seen with non-OMP. Very trivial code, I really can't see what could be wrong with it. > ilf > Apart from that one, the dynamics are the only ones left that I know of (and I just heard from Jim, he's on it!). We've nailed quite a few bugs the last couple of days 8^) > 1. Are we using non-thread safe OpenSSL functions in this format? OpenSSL could be breaking OMP builds (just guessing). 2. This format is a copy of wow_srp format. There isn't much difference between them. So why doesn't wow_srp format exhibit similar problems? Dhiru
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.