Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Date: Wed, 10 May 2017 21:51:12 +0200
From: magnum <john.magnum@...hmail.com>
To: john-users@...ts.openwall.com
Subject: Re: CL_OUT_OF_RESOURCES error with --format=dmg-opencl

On 2017-05-10 21:23, B B wrote:
>> On May 9, 2017, at 11:36 PM, magnum <john.magnum@...hmail.com> wrote:
>> On 2017-05-10 00:29, B B wrote:
>>> Output:
>>> Device 0: GeForce GTX 760
>>> Local worksize (LWS) 64, global worksize (GWS) 131072
>>> Loaded 1 password hash (dmg-opencl, Apple DMG [PBKDF2-SHA1 OpenCL 3DES/AES])
>>> Will run 8 OpenMP threads
>>> Warning: only 14 characters available
>>> Press q…….
>>> OpenCL error (CL_OUT_OF_RESOURCES) in file (OpenCL_dmg_fmt_plug.c) at line 764 - (Copy result back)
>>> Now I have tried adjusting LWS and GWS but nothing changes. Looking at the “line 764” I can not discern anything either as it seems it is just a line to declare the error.
>>
>> To me that line number indicates you are using an outdated verson of Jumbo. Try the latest version from Github!
> 
> Thanks for that. I went ahead and compiled the bleeding-jumbo from GitHub. Unfortunately I am getting the same error, line 782 now.
> 
> The first attempt output:
> 
> Build log:
> ptxas info		: 0 bytes gem
> ptxas info		: Compiling entry function ‘derive_key for ‘sm_30’
> ptxas info		: Function properties for derive_key
> ptxas 		. 	176 bytes stack frame, 12 bytes spill stores, 12 bytes spill loads
> ptxas info		: Used 63 registers, 344 bytes cmem[0], 8 bytes cmem[2]
> 
> This output happened only once. I toggled the sm_30 line in john.conf with no change

The compiled kernel is cached by the nvidia runtime so you wont see the 
build statistics again unless or our code changes. And that sm_30 line 
is irrelevant for this format.

> The NVIDIA_CUDA env variable was not set so I looked and  found 4 different versions on my system and set the path for the cuda version from the cuda toolkit, then tried the driver version. I did not try the mesa or Intel version. I set it for /opt/cuda/lib64/libOpenCL.so, thinking the include would be found but I’m unsure about that. They were not in the same directory.

There should be no need for any environment variables. Those were for 
CUDA many years ago, this is OpenCL. And anyway your problem happens at 
run-time of kernel, not when building it.

Does the CL_OUT_OF_RESOURCES happen almost immediately, or after a while?

Please post your output from this:

./john -test —format=dmg-opencl -v=5

and in case the test fails, post output from this as well:

./john -dev=0 -list=opencl-devices


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.