User Profile

[zxy_thf]

In order to exploit those hardware bugs, the attacker must be run able to run custom code that does crazy things.
If the attacker could run arbitrary...

[zxy_thf]

I'm talking about servers, not desktops....

[zxy_thf]

This is a very welcome change! Those mitigations make little sense if only trusted code will be executed.

[zxy_thf]

TF heavily exploits NVIDIA's cuDNN, and I agree with you there isn't any equivalent library for AMD's GPU....

[zxy_thf]

That's weird, I'm using R7 250, Fedora, Gnome and haven't noticed any serious issues with GLAMOR.
This might be a hardware issue imo....

[zxy_thf]

The good news might be deep learning guys are not using CUDA directly.
As long as AMD made their cards perform well under various Torch/TF workloads...

[zxy_thf]

and won't choose nvidia card from the beginning because of nvidia driver will never be in the kernel source tree.
using non-mainlined driver ==...