Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 1 Apr 2015 18:56:15 +0200
From: Agnieszka Bielec <>
Subject: Re: [GSoC] John the Ripper support for PHC finalists

On 2015-03-30 02:11, magnum wrote:
> What you should do is this right before opencl_init():
> +  sprintf(build_opts, "-DBINARY_SIZE=%d -DSALT_SIZE=%d -DMEM_SIZE=%d",
> And a corresponding change to the kernel:
> -#define BINARY_SIZE     257
> -#define SALT_SIZE       32
> -#define MEM_SIZE        131072
> +// BINARY_SIZE, SALT_SIZE and MEM_SIZE is passed with -D during build

thanks a lot!
Sorry I was using __local or __global memory because I was thinking
that is impossible to allocate array whose size depends on a parameter of
the kernel. I moved it to the __private memory and now the opencl
implementation is a lot faster. I have actually problems with --dev=1
(CL_OUT_OF_RESOURCES)  and with --dev=4 (Raw: 33.3 c/s real, 6400 c/s

there is a way to pass options by comand line somethig alike: -m_cost=x ?
I need it for the auto_tune and compiling the kernel.
also for good auto_tuning I must automatically generate hashes for specified
options but this is incompatible with the actual scheme, I see this ugly.
what are your opinions?

magnum wrote:
>Would that solve your problem? The first thing you need to do is rebase
>on latest bleeding-jumbo, then apply changes to pomelo-opencl similar to
>what happened to mysqlsha1-opencl in commit ef4b3dbb. Then add whatever
>code you like in reset().

I don't see commit ef4b3dbb anywhere, sorry


Content of type "text/html" skipped

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.