I always get a CUDA grid launch failure while debugging with the CUDA Nsight Legacy debugger.
Normal (c++ part) debugging or running the program without kernel (NSight) debugging does work however.
I’m using a laptop with the following configuration
Windows 10 Pro 1709
Visual Studio 2017 (NSight edition)
CUDA toolkit: 10.0
NSight 6.0
The laptop has two gpu’s. A Intel UHD 630 and a NVidia GTX 1050
- I did try running Nsight Monitor as administrator and increased the WDDM TDR delay as described in this thread: https://devtalk.nvidia.com/default/topic/1027241/nsight-visual-studio-edition/cuda-grid-launch-failed/
- I did look into setting the NVidia GPU in TCC mode, but as far as I can tell that's not support on pascal GPU's.
- A weird side note, my NVidia Control panel shows practically no settings, only 3 categories under '3D Settings' are available.
Nothing has resolved my issue so far. Could anybody help? If I need to provide additional information that could help you, please ask.
Error message:
CUDA context created : 206ddb35f60
CUDA module loaded: 206ed14f7a0 kernel.cu
CUDA grid launch failed: CUcontext: 2228512579424 CUmodule: 2228770633632 Function: _Z17d_attention_modelPfPjPKfS2_S2_S2_S2_PKtS4_PKbtjtttffff
NVidia-SMI information
Tue Dec 11 10:42:30 2018
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 417.01 Driver Version: 417.01 CUDA Version: 10.0 |
|-------------------------------+----------------------+----------------------+
| GPU Name TCC/WDDM | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GeForce GTX 1050 WDDM | 00000000:01:00.0 Off | N/A |
| N/A 37C P8 N/A / N/A | 75MiB / 4096MiB | 0% Default |
+-------------------------------+----------------------+----------------------+