[john-users] Re: "No OpenCL devices was found" and "dri2 connection failed!" errors, when using NVidia gpus

[Claudio André]

Any idea what may be causing `No OpenCL devices was found` and `X server
found. dri2 connection failed!` errors when trying to use NVidia cards?

I bet you have more than one OpenCL implementation installed (could be
mesa, pocl, ...)

You can try to get more information doing something like this: (try 0,
then 1,2, ...)
`john --verbosity=5 --list=opencl-devices --device=0`

And,
`_SKIP_OCL_INITIALIZATION=1 john --verbosity=5 --list=opencl-devices`

---

If nothing shines, you could also post the result of:
lspci -k | grep -iA 3 vga
dnf list installed | grep opencl | sort

dnf list installed | grep beignet | sort

Claudio

[Piotr Dobrogost]

Post by Claudio André
I bet you have more than one OpenCL implementation installed (could be

mesa, pocl, ...)

Thank you for taking time to reply. It looks like you are right.

Post by Claudio André
You can try to get more information doing something like this: (try 0,

then 1,2, ...)

Post by Claudio André
`john --verbosity=5 --list=opencl-devices --device=0`

At https://paste.fedoraproject.org/paste/Tuab~zsBF6DDSfpR-ViENA there's an
output from the following:

for device in {0..10}; do echo -e "----- DEVICE $device -----\n";
./john --verbosity=5 --list=opencl-devices --device=$device; done

Post by Claudio André
And,
`_SKIP_OCL_INITIALIZATION=1 john --verbosity=5 --list=opencl-devices`

Please see https://paste.fedoraproject.org/paste/VhE-Pn1ardN2UYXpffc84w

Post by Claudio André
lspci -k | grep -iA 3 vga

Please see https://paste.fedoraproject.org/paste/sTTM~6mpPBYhFa10DG1bDA

Post by Claudio André
dnf list installed | grep opencl | sort

Please see https://paste.fedoraproject.org/paste/U3b4BqKYftVPSCDIesMr2w

Post by Claudio André
dnf list installed | grep beignet | sort

Please see https://paste.fedoraproject.org/paste/BdQyYitXGpy2-pdugQFz9Q


It looks like device 0 is CPU, device 1 is glitch in the Matrix :), devices
2-9 are 8 Nvidia GPUs and device 10 is built-in/on-board GPU.
I guess running john with --devices=2,3,4,5,6,7,8,9 option should result in
work being distributed among all 8 GPUs.
How can I check if it's really the case?
I have all GPUs mining and when I run john with --devices=2,3,4,5,6,7,8,9
option I see only the first GPU has lower hashrate which I take to mean john
is running only on it and not on any other GPU.

Regards,
Piotr

[magnum]

Post by Piotr Dobrogost
It looks like device 0 is CPU, device 1 is glitch in the Matrix :), devices
2-9 are 8 Nvidia GPUs and device 10 is built-in/on-board GPU.
I guess running john with --devices=2,3,4,5,6,7,8,9 option should result in
work being distributed among all 8 GPUs.
How can I check if it's really the case?
I have all GPUs mining and when I run john with --devices=2,3,4,5,6,7,8,9
option I see only the first GPU has lower hashrate which I take to mean john
is running only on it and not on any other GPU.

Please read doc/README-OPENCL. You need to use the -fork option so you
get one process per GPU. Something like this (but pause mining before this):

./john -fork=8 -dev=2,3,4,5,6,7,8,9 (...rest of command line)

magnum