Hi! I would like to get some help!
I’m running a multi-threaded program - in which each display thread calls a series of OpenGL API calls to display the content stored in GPU buffers. There are seven display threads. I have observed that, occasionally (and randomly), GPU workload activities were not triggered after OpenGL API calls are initiated from CPU. This behaviour can be observed as a “short pause” in real time display. As observed in the profile trace, there’s about 10 frames (each 16.67ms) without any GPU workload activity. After that, GPU workload activity recovered and display becomes normal again.
Do you have any idea how to debug this issue? Could it be related to the driver? Here’s some some background
Platform: NVIDIA Quadro P2000
NVIDIA Driver Version 415
OS: Linux (Buildroot)
Language: CUDA 9.1 (CUDA with OpenGL graphics interoperability), OpenGL 4.6.0, Qt5.9.
Profile tool: NVIDIA Nsight 2019.3.6
Window System: X11
Thank you very much for your insights!