[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 08 Jul 2012 16:24:52 +0200
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: RAR's SHA-1 speed on GPU
On 2012-07-08 12:04, Solar Designer wrote:
> On Sat, Jul 07, 2012 at 07:47:37AM +0200, magnum wrote:
>> An 8-character password results in 7,078,320 bytes of data fed to SHA-1,
>> that is 110,599 blocks of 64 bytes.
A more correct figure is 7,077,888 bytes, 110592 blocks. Then 16 extra
blocks are ran for IV, so 110608 should be the correct figure. I just
fixed a slight bug in rar_fmt for that calculation so it's now like this:
(strlen(plaintext) * 2 + 8 + 3) * 0x40000 / 64 + 16
The *2 is for Unicode, 8 bytes salt, 3 bytes "serial". 16 extra
sha1final's are made for calculating IV.
>> gws 33152 4404 c/s 414914052 sha1/s 7.527 sec per crypt_all()
>
> Hmm, but 4404*110599 = 487M, why does the above show only 415M?
> Is it counting SHA-1s of sometimes more than one block maybe?
Because this is a six-char password: It was 4404*94213, and after my
bugfix it's 4404*94224. Current figures for Tahiti:
gws 65536 7175 c/s 676057200 sha1/s 9.133 sec per crypt_all()+
If we can get this figure closer to the 2 billion mentioned for mscash2,
it would be a nice boost :)
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.
