john-dev - Re: PHC: Argon2 on GPU

Follow @Openwall on Twitter for new release announcements and other news

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date: Fri, 14 Aug 2015 17:22:05 +0200
From: Agnieszka Bielec <bielecagnieszka8@...il.com>
To: john-dev@...ts.openwall.com
Subject: Re: PHC: Argon2 on GPU

2015-08-14 16:44 GMT+02:00 Agnieszka Bielec <bielecagnieszka8@...il.com>:
> this can have something common with MEM_SIZE/4 (now I have removed /4)
> http://www.openwall.com/lists/john-dev/2015/08/06/22
> sorry, couldn't find my original e-mail

removed /4 because I had problems on super on Tahiti with this

(this is my previous version from commit
ada77dbb6a8a967271590d7879c39214df92e434 and with nother costs)

[a@...er run]$ ./john --test --format=argon2i-opencl --v=4
Benchmarking: argon2i-opencl [Blake2 OpenCL]...
memory per hash : 1000.00 kB
Device 0: Tahiti [AMD Radeon HD 7900 Series]
Calculating best global worksize (GWS); max. 1s single kernel invocation.
gws:       256         580 c/s         580 rounds/s 441.045ms per crypt_all()!
gws:       512        1077 c/s        1077 rounds/s 475.273ms per crypt_all()+
gws:      1024        1953 c/s        1953 rounds/s 524.191ms per crypt_all()+
OpenCL error (CL_MEM_OBJECT_ALLOCATION_FAILURE) in file
(opencl_argon2i_fmt_plug.c) at line (524) - (failed in
clEnqueueNDRangeKernel)

[a@...er run]$ GWS=1024 ./john --test --format=argon2i-opencl --v=4
Benchmarking: argon2i-opencl [Blake2 OpenCL]...
memory per hash : 1000.00 kB
Device 0: Tahiti [AMD Radeon HD 7900 Series]
Local worksize (LWS) 64, global worksize (GWS) 1024
using different password for benchmarking
DONE
Speed for cost 1 (t) of 3, cost 2 (m) of 1000, cost 3 (l) of 1
Many salts:     1969 c/s real, 204800 c/s virtual
Only one salt:  1950 c/s real, 204800 c/s virtual

[a@...er run]$ GWS=2048 ./john --test --format=argon2i-opencl --v=4
Benchmarking: argon2i-opencl [Blake2 OpenCL]...
memory per hash : 1000.00 kB
Device 0: Tahiti [AMD Radeon HD 7900 Series]
Local worksize (LWS) 64, global worksize (GWS) 2048
using different password for benchmarking
DONE
Speed for cost 1 (t) of 3, cost 2 (m) of 1000, cost 3 (l) of 1
Many salts:     3471 c/s real, 409600 c/s virtual
Only one salt:  3471 c/s real, 409600 c/s virtual

I checked if I forgot about deallocating some memory but I don't think so

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.