cuda - setting cudaDeviceProperty: cudaKernelExecTimeoutEnabled -


Is it possible to set device properties to cudaKernelExecTimeoutEnabled at runtime 0? This is currently 1 because I am also running X-server on it The problem is that my thread has timed out.

On Linux, you can deactivate timeouts (approximately 6 me 7 seconds) for your X11 configuration By adding an option to the modern Linux system, there is no X11 configuration, everything is automatically detected, but you can create it manually:

  mkdir -p / etc / X11 / xorg.conf.d Touch / etc /X11/xorg.conf.d/20-nvidia.conf   

Then add the following to this file:

  section "device" identifier Interpretation "My GPU" driver "Nvidia" alternative "interactive", "0" Andsait   

The key "interactive" option. It works for me even without X11 restart. It was clear that as soon as my OpenCL application starts, this configuration is read and applied. "Identifier" can be an arbitrary value "Driver" is "Nvidia" in my case, because I am using a proprietary Nvidia driver for Linux (part of official Debian package repository).

Comments