无法从Windows服务(会话0隔离)访问GPU [英] Cannot access GPUs from windows services (session0-isolation)
问题描述
我们无法将软件作为服务运行在
We were unable to run our software as a service on an NV6 azure GPU-server using process-managers like pm2 or by registering it as windows-service. This is because of the limitations on processes running in session0, which prevented the launched instances of our mesh-simplification software from having full access to the GPU-driver (specifically, we were unable to compile shaders using opengl shader level 4.3). We tried putting the azure server's NVIDIA Tesla-GPUs into TCC mode (using nvidia-smi -g 0 -dm 1). This is supposed to give windows services access to the GPU, which it didn't. With TCC mode on, GLViewer reveals that not even desktop applications have access to OpenGL higher than 3.3. Is there a simple way to handle starting, stopping, logging and crash-handling our software automatically without running into issues with session0-limitations? Or alternatively is there a way to access the GPU from a windows service?
推荐答案
要确认，设置虚拟机后是否正确设置了GPU驱动程序? ;
To confirm, did you properly setup the GPU drivers after setting up the VM?
https://docs.microsoft.com/zh-CN -us/azure/virtual-machines/windows/n-series-driver-setup
参考:
这篇关于无法从Windows服务(会话0隔离)访问GPU的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
