Hey everyone, a little update from me concerning the work. As I was running some tests with X86 code I first noticed that _mm_set_epi64x() isn't supported by MSVC for X86, so I tried to fix this with a macro redirection to _mm_set_epi32() BUT it didn't pass test vector checks. So I do think that _mm_set_epi32() is the root of all evil as it's used by BLAKE2s, by scrypt and by X86-BLAKE2b. I'll run some tests with an executeble soon (as opposed to unit testing). As reaction to the ongoing difficulties with SSE code I disabled it locally for BLAKE2s, scrypt and X86-BLAKE2b and enforced the use of reference C code. As soon as I get positive results, I'll switch back to SSE.
Now some good news: Fortuna (the CSPRNG) is finished! It doesn't gather entropy by itself yet (-> there's no AutoSeeded version yet) but at least it *should* run. Entropy collector is scheduled after the fix for SSE-errors. Code's not yet on GitHub. BR JPM Am Sonntag, 18. Januar 2015 17:46:28 UTC+1 schrieb Jean-Pierre Münch: > > Hey everyone, > > I have some normal work to do so future releases will come less often as > before. > > But I've found some free hours and finally fully included BLAKE2 family. > But the issue that BLAKE2s' SSE version produces incorrect results is > still live. > > PEM-Pack is now also online although it was kind of broken. > > @Jeffrey: > It might interest what I've changed: > 1. I did include "pch.h" as very first header in all CPP-files as this is > a requirement in VS > 2. I did remove the call to std::transform you make once and replaced it > by an equal loop as the call to transform was causing compile-time errors > for VS with SDL enabled. > > Next thing I'll do is to include the Bouncy-Castle-Patch for ECIES. > Afterwards I'll finally fix scrypt. > > At this point (If the BLAKE2 guys confirm the bug and fix it) I'll finally > contact Wei Dai as I think I've got enough new stuff to trigger a new > release. > > Code's live on GitHub <https://github.com/DevJPM/CryptoJPM> as usual. > > BR > > JPM > > Am Sonntag, 11. Januar 2015 10:37:36 UTC+1 schrieb Jean-Pierre Münch: >> >> Hey everyone, >> >> I've tested BLAKE2b code and it passes the test vectors. >> However, there is an issue with BLAKE2s. >> If i'm using reference code for compression function it'll pass the >> vectors, but if I use the SSE2 and SSSE3 optimized code it will fail. >> I copied the code directly from the reference library >> blake2_code_20140114.zip <https://blake2.net/blake2_code_20140114.zip> >> and manually selected SSE2 and SSSE3 optimizations. >> >> @zooko: >> Could you please check if SSE2 and SSSE3 work correct in BLAKE2s by >> testing yourself? (-> DON'T HAVE XOP, DON'T HAVE SSE41, DON'T HAVE AVX, >> HAVE SSE2, HAVE SSSE3) >> >> BR >> >> JPM >> >> >> Am Donnerstag, 8. Januar 2015 19:35:27 UTC+1 schrieb zooko: >>> >>> >>> >>> On Wed, Jan 7, 2015 at 2:48 PM, Jean-Pierre Münch < >>> [email protected]> wrote: >>> >>>> >>>> @zooko: >>>> Can you please post valid test vectors either here or on the BLAKE2 >>>> website? (two or three per version) >>>> >>> >>> I forwarded your letter to Jean-Philippe Aumasson, and he posted test >>> vectors: >>> >>> https://blake2.net/#ts >>> >>> Regards, >>> >>> Zooko >>> >> -- -- You received this message because you are subscribed to the "Crypto++ Users" Google Group. To unsubscribe, send an email to [email protected]. More information about Crypto++ and this group is available at http://www.cryptopp.com. --- You received this message because you are subscribed to the Google Groups "Crypto++ Users" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. For more options, visit https://groups.google.com/d/optout.
