[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sun, 3 Feb 2013 22:32:29 +0100
From: magnum <john.magnum@...hmail.com>
To: john-dev@...ts.openwall.com
Subject: Re: NetNTLMv1
On 3 Feb, 2013, at 8:10 , Solar Designer <solar@...nwall.com> wrote:
> On Sun, Feb 03, 2013 at 03:01:54AM +0100, magnum wrote:
>> Lol, I hit a luxury problem:
>
> I fully expected you would. :-)
>
>> Benchmarking: NTLMv1 C/R MD4 DES (ESS MD5) [128/128 SSE2 intrinsics 12x]... DONE
>> Many salts: 4294M c/s real, 4294M c/s virtual
>> Only one salt: 38731K c/s real, 38348K c/s virtual
>>
>>
>> I can't tune BLOCK_LOOPS,
>
> Besides tuning BLOCK_LOOPS, you may also try re-enabling the memset()
> approach to cleaning the bitmap in crypt_all() when max_keys_per_crypt
> is known to be large enough at compile time.
>
>> because I hit the 32 bit limit and always see 4294M. I am pretty sure I fixed this for MPI, maybe we should always use the MPI version of that code in bench.c.
>
> I haven't looked into this yet. I think you'll take care of it. ;-)
Done, hope I did not b0rk it somehow. That MPI code was not there anymore, I must have dropped it at some point.
I get speeds over 8.8G c/s for many salts with one core SSE2 on my laptop i7. On Bull it's about 6.5G c/s with current settings. I have no time to try to fine-tune this, I left BLOCK_LOOPS aiming at 256 keys per crypt (it's 252 for 12x), and enable the memset for >200. I'm not sure the latter is good, the figures fluctuates. Maybe it should be a higher threshold. If I get some time I'll try it in Callgrind.
I can't get OMP+SIMD to do good at all. I think when I achieved 80M c/s for NT2, it was with 2 cores of an i7-3820 and a pretty high BLOCK_LOOPS.
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.
