Date: Thu, 17 Feb 2011 06:31:57 +0300 From: Solar Designer <solar@...nwall.com> To: john-users@...ts.openwall.com Subject: 1.7.6-jumbo-12 fixes md5-gen self-test bug, improves MSCash*/OpenMP Hi, I've just released John the Ripper 1.7.6-jumbo-12. It is available both as a tarball ready to be compiled: http://www.openwall.com/john/ and as a patch against 1.7.6: http://www.openwall.com/john/#contrib The changes since -jumbo-11 are: The bug causing self-tests of md5-gen "formats" to fail and then crash, which was introduced in -jumbo-10, has now been fixed. I did this by relaxing the assumptions/requirements of the self-tests. The property of md5-gen that I did not expect was actually of no consequence to uses by JtR itself, or at least this is the current understanding. The MSCash and MSCash2 "formats" have been enhanced to adjust the number of key slots according to the number of threads with OpenMP builds. This appears to have resulted in some speedup. Here are the new numbers on a Core i7 (8 threads): Benchmarking: M$ Cache Hash [Generic 1x]... DONE Many salts: 71221K c/s real, 8891K c/s virtual Only one salt: 19517K c/s real, 2436K c/s virtual Benchmarking: M$ Cache Hash 2 [Generic 1x]... DONE Raw: 392 c/s real, 49.2 c/s virtual Finally, I spotted and fixed a few unaligned accesses on big-endian systems in mscash2_fmt.c, but I did not bother to actually test it on such systems now (so there may be more issues of this nature anyway). Since the changes since -jumbo-11 have been rather conservative (fixing known issues only), -jumbo-12 is sort of a "stable jumbo" (an oxymoron?) JimF has recently contributed improved md5-gen code (thanks!), which I am likely to merge for the next update, thereby potentially/temporarily making it "unstable" again. Thus, -jumbo-12 is a good candidate for "stable" builds now. Thanks, 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.