Date: Wed, 9 Sep 2015 00:05:45 +0200 From: magnum <john.magnum@...hmail.com> To: john-dev@...ts.openwall.com Subject: Re: Large stack alignment On 2015-09-08 23:52, magnum wrote: > On 2015-09-08 22:11, magnum wrote: >> On 2015-09-08 21:45, magnum wrote: >>> All this is exactly what we've seen so far IRL in Jumbo. All problems >>> we've had was aligned declarations local to a parallel for. >>> >>> Also, I do not see *any* problem using gcc-4.9.2 (only OSX tested as of >>> yet). I'll dig into their list of changes (but I already googled this so >>> much it would be strange if I missed it). >>> >>> I'll do some more tests on various systems and compilers. Jim, please >>> test under Cygwin and, if possible, under MinGW64. Several google hits >>> indicate problem is (or was) worse under MinGW64. >> >> gcc 4.4, 4.5 and 4.6 (bull) shows the same problems and same workaround >> (using function). >> gcc 4.4.7 (super) shows the same problems and same workaround (using >> function). >> gcc 4.8.3 (well) shows the same problems and same workarounds (function >> or vtype). >> gcc 4.9.1 (super) shows the same problems and same workarounds. > > Oh, and icc 14.0.0 (super) does not have the problem at all. Also tested clang 3.0-6ubuntu3 (bull), gcc 5.1.0 (OSX), the native OSX' compilers (although like Bull's clang they lack OpenMP). No problem seen. I'm done with this for now. Unless someone debunks my view of this I suggest we revert the alignment changes to md5cryptsse() and simply keep using mem_align() for declarations within OMP for loops in the future. For all we know from real-life, that will be Good Enough[tm]. 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.