john-users - Re: New format: krb5ng-opencl

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Sat, 8 Dec 2012 18:05:03 +0100
From: magnum <john.magnum@...hmail.com>
To: john-users@...ts.openwall.com
Subject: Re: New format: krb5ng-opencl - cracking speed?

On 8 Dec, 2012, at 14:50 , Dhiru Kholia <dhiru.kholia@...il.com> wrote:
> On Sat, Dec 8, 2012 at 7:12 PM, buawig <buawig@...il.com> wrote:
>> thanks for adding krb5ng-opencl! What speeds do you reach with it?
>> Anyone tried yet?
> 
> ATI 6970 (non-OMP build)
> 
> ✗ ../run/john -fo:krb5ng-opencl -t
> OpenCL platform 0: AMD Accelerated Parallel Processing, 2 device(s).
> Using device 0: Cayman
> Local worksize (LWS) 64, Global worksize (GWS) 32768
> Benchmarking: Kerberos 5 AS-REQ Pre-Auth aes256-cts-hmac-sha1-96
> [OpenCL 4x]... DONE
> Raw:    54274 c/s real, 364088 c/s virtual
> 
> ✗ ../run/john -fo:krb5ng-opencl -t --request-scalar
> OpenCL platform 0: AMD Accelerated Parallel Processing, 2 device(s).
> Using device 0: Cayman
> Local worksize (LWS) 64, Global worksize (GWS) 131072
> Benchmarking: Kerberos 5 AS-REQ Pre-Auth aes256-cts-hmac-sha1-96
> [OpenCL]... DONE
> Raw:    55894 c/s real, 369216 c/s virtual

So maybe I should make Cayman default to scalar for this format. Here's a couple more:

A Macbook Pro laptop (not OMP here but it's got AES-NI):
OpenCL platform 0: Apple, 2 device(s).
Using device 1: GeForce GT 650M
Local worksize (LWS) 768, Global worksize (GWS) 1536
Benchmarking: Kerberos 5 AS-REQ Pre-Auth aes256-cts-hmac-sha1-96 [OpenCL]... DONE
Raw:	6193 c/s real, 128000 c/s virtual

Openwall's test rig:
OpenCL platform 0: NVIDIA CUDA, 1 device(s).
Using device 0: GeForce GTX 570
Local worksize (LWS) 128, Global worksize (GWS) 65536
Benchmarking: Kerberos 5 AS-REQ Pre-Auth aes256-cts-hmac-sha1-96 [OpenCL]... DONE
Raw:    35617 c/s real, 35520 c/s virtual

OpenCL platform 1: AMD Accelerated Parallel Processing, 2 device(s).
Using device 0: Tahiti
Local worksize (LWS) 64, Global worksize (GWS) 262144
Benchmarking: Kerberos 5 AS-REQ Pre-Auth aes256-cts-hmac-sha1-96 [OpenCL]... DONE
Raw:    102001 c/s real, 429744 c/s virtual

OMP and/or AES-NI capabilities currently make a difference but I'll probably move the post processing to GPU at some point in time:

OpenCL platform 0: NVIDIA CUDA, 1 device(s).
Using device 0: GeForce GTX 570
Local worksize (LWS) 128, Global worksize (GWS) 65536
Benchmarking: Kerberos 5 AS-REQ Pre-Auth aes256-cts-hmac-sha1-96 [OpenCL]... (8xOMP) DONE
Raw:    37991 c/s real, 32932 c/s virtual

OpenCL platform 1: AMD Accelerated Parallel Processing, 2 device(s).
Using device 0: Tahiti
Local worksize (LWS) 256, Global worksize (GWS) 262144
Benchmarking: Kerberos 5 AS-REQ Pre-Auth aes256-cts-hmac-sha1-96 [OpenCL]... (8xOMP) DONE
Raw:    125427 c/s real, 262144 c/s virtual

magnum