Date: Sat, 14 May 2011 17:39:57 -0500 From: "JimF" <jfoug@....net> To: <john-dev@...ts.openwall.com> Subject: Jumbo-3 RC1 released I have placed (on the wiki patch page), a new jumbo RC. Here is what has changed, SINCE jumbo-2-RC1: NOTE, this version makes numerous changes to Makefile, and the builds. PLEASE test, make sure envirionments build properly, and follow up with problems, or compiler warnings (especially new ones, not in john core, or john-jumbo1). Fixed several setup 'memory leaks' (HDAA KRB5) Incorporated sse-intrinsics md5 / Apache Apache was not working in the 1.77 patch. All hash functions implemented. fully integrated into md5_gen In md5_gen, got this working: all SHA1 (using intrisics) gen(27) and gen(28) OpenBSD md5/Apache Md5 made changes to SSEmd5body() functions telling them to load IV's, in preparation for handling SSE/Intrisic data longer than 55 bytes. NOTE, on 64 bit builds, there are some formats not working properly. For now, these have been commented out in those builds. Added a new function to .S ssa asm. It makes it easier to dual build for Para and nonPara mode. Changes to makefile. new linux-x86-64-icc new linux-x86-64-clang new linux-x86-sse2i (intrisic 32 bit linux) new win32-mingw-x86-sse2i (intrisic 32 bit linux) adding intrinsics to: linux-x86-64 linux-x86-64-avx linux-x86-64-xop linux-x86-64-clang linux-x86-64-icc linux-x86-64-32-sse2 linux-x86-64-32-mmx (but that version is not buildable) linux-x86-sse2i freebsd-x86-64 solaris-x86-64-cc solaris-x86-64-gcc macosx-x86-64 john-macosx-x86-64 win32-mingw-x86-sse2i Added a new arch file x86-sse2i.h This is 32 bit, but with intrisics Made changes to all intrisic arch files, to distinguish PARA count from gcc and icc. raw-md5 is now a thin format. It links to md5_gen(0). raw-sha1 now has intrisic SSE code. Added sse-intrisics.c/.h Changes (enhancements to Markov processing). Added 2 new params, minlevel and minlen Salted-sha format (same as nsldaps) Moved init of md5_gen down to after the 'default' core john init calls, with the exception of it is before fmt_md5 Added some nice debugging functions to memory.c (to help dump out SSE buffers). Added some new function to SHA (32 bit Asm). shammx_nosizeupdate_nofinalbyteswap Enanced unique.c 1. Added -v Some screen outputs 2. -inp=fname Allows using without being a filter. Fixes some Win32 issued (embedded EOF, last line missing EOL, etc) 3. -ex_file=fname A 2nd file that is also used to unique the data against. Only unique data in input file also NOT in this file will be written. 4. -ex_file_only=fn Same as -ex_file BUT assumes current data is already unique, thus does not perform unique on original input data. 5. -cut=num Cut the lines to num bytes long (or shorter). Also, the -ex_file data is cut when read, so it properly uniques data. 6. -cut=LM This will cut words to 7 bytes, AND if longer than 7 bytes, will grab the next 7 bytes. Also, data is upcased. NOTE, this operation is NOT performed on the -ex_file data. To use -ex_file, it should already be in LM format. 7. -mem=num A number from 13 to 25 (default john build is 20). This controls size of hash buffer, and input building buffer. Default of 20 is hash is 1 million items, and buffer of 64MB. Each number up is 2x larger, each num down is 1/2 size. So these patches are now in: Jumbo-2: md5_gen (JimF) Jumob-3: sse_intrisics (Simon), unique (JimF) I am going to turn over the next jumbo work to magnum, and allow him to get several patches built in. I will start working on 'thin' formats for MD5_gen, and getting several other formats in. (magnum may put these formats in, on the jumbo-4). I will also work on figuring out the 64-bit problems in md5_gen (or find someone who is better at debugging under *nix than I am). Jim.
Powered by blists - more mailing lists
Powered by Openwall GNU/*/Linux - Powered by OpenVZ