Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <CAF5391JUVGmwnTyJ13hBO4jTYqT-uo-PEe7omkUH++psDoND_g@mail.gmail.com>
Date: Sun, 30 Mar 2014 03:25:19 +0100
From: "Joe ." <dy3afg9@...il.com>
To: john-users@...ts.openwall.com
Subject: Re: pbkdf2-hmac-sha512 on GPU

I've just running it on the CPU, and it worked. I don't think it worked
before on the CPU, so that's great progress!


Non-opencl CPU: (I can then use --fork=8)

./john --format=pbkdf2-hmac-sha512 -t

Raw: 58.8 c/s real, 58.8 c/s virtual


Device 0, (my cpu via opencl)

./john --format=grub-opencl -t --devices=0

Raw: 419 c/s real, 53.3 c/s virtual


--devices set to 1 or 2 gives the same error as before. (I've tried both as
one of them could of been occupied with the display)

Why such a huge difference between them? 419c/s real vs 58.8c/s real. With
a real ciphertext, non-opencl + --fork seams much quicker.

In theory, would I be able to use all 3 opencl devices?


I'll try it again with the GPUs if any software updates appear in the app
store. It's a pretty new machine so apple may update firmware and such down
the road.

> you did all the right things and the ball is ours.

Okay. No rush, but please keep me updated. :)


Thanks,

Joe.


On Sun, Mar 30, 2014 at 2:54 AM, magnum <john.magnum@...hmail.com> wrote:

> OK, that might be a driver bug or a format bug. The Build log warnings
> about signs are benign. The self-test failures are obviously not. Try
> adding "--device=CPU" just for checking if it works then. If it does, you
> did all the right things and the ball is ours.
>
> magnum
>
>
>
> On 2014-03-30 03:35, Joe . wrote:
>
>> Thankyou both,
>>
>> However I am still receiving errors.
>>
>> When running:
>> ./john --format=grub-opencl -t
>>
>> I get:
>> Build log: <program source>:219:16: warning: comparison of integers of
>> different signs: 'int' and 'unsigned int'
>>
>>          for (i = 0; i < keylen; i++)
>>
>>               ~ ^ ~~~~~~
>>
>> <program source>:258:11: warning: comparison of integers of different
>> signs: 'unsigned int' and 'int'
>>
>>          for(i=0;i<saltlen;i++)
>>
>>           ~^~~~~~~~
>>
>> <program source>:351:24: warning: comparison of integers of different
>> signs: 'int' and 'unsigned int'
>>
>>          for (round = 1; round < rounds; round++) {
>>
>>                   ~~~~~ ^ ~~~~~~
>>
>>
>> Benchmarking: grub-opencl, grub-opencl [PBKDF2-SHA512]... FAILED
>> (get_hash[0](0))
>>
>>
>> When I try it out with the test I provided here earlier, I get
>>
>> Self test failed (get_hash[0](0))
>>
>>
>> It could well be my fault, so here's my exact process:
>>
>> Download latest bleeding-jumbo (without modifying anything), Launch
>> terminal
>>
>> cd into /src
>>
>> build macosx-x86-64-opencl
>>
>> wait for completion, no errors,
>>
>> then cd into /run.
>>
>>
>> Thanks,
>>
>> Joe.
>>
>>
>>
>>
>>
>> On Sun, Mar 30, 2014 at 2:03 AM, Lukas Odzioba <lukas.odzioba@...il.com>
>> wrote:
>>
>>  2014-03-29 23:09 GMT+01:00 magnum <john.magnum@...hmail.com>:
>>>
>>>> I committed a fix for your problem, pull bleeding-jumbo now and it will
>>>>
>>> work
>>>
>>>> fine.
>>>>
>>>
>>> magnum, thank you very much for taking care of that!
>>>
>>>
>>
>
>

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.