Date: Thu, 18 Dec 2014 17:17:48 +0300 From: Solar Designer <solar@...nwall.com> To: john-dev@...ts.openwall.com Subject: scrypt Jim, I went to fix the known (to me) bug in the revision of crypto_scrypt-nosse.c that got into bleeding-jumbo. (I've long since fixed that bug in (y)escrypt, and it is not in yescrypt as actually submitted to PHC.) And I found that you had fixed it in here already - thanks! I am referring to commit 9ac48a24ebe5ecae14bfb248ff2cb057406627f5 and this fix: - XY_size = (size_t)256 * r; + XY_size = (size_t)256 * r + 64; However, I notice that you've included the same change in -sse as well, in the same commit. I think it was not needed. I think the bug was in -nosse only, because in -sse there's no "Z" in smix(). Did you make the change to -sse just in case, knowing that it's harmless even if not needed, or am I missing something? I am also puzzled by: -#ifdef __SSE2__ +#if defined (__SSE2__) || defined (_MSC_VER) (part of the same commit). Doesn't MSVC define __SSE2__ when appropriate? And is it appropriate for us to assume SSE2 whenever we build with MSVC? No one would ever want a new Windows build running on anything older than a Pentium 4? Perhaps this is so, but then wouldn't it be more appropriate to enable the proper compiler flags? Alexander
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.