Hi, I encountered this error when I executed cv::cuda::orb:detectAndComputeAsync to do feature point detection. $ cuda-memcheck ./test_orb ========= CUDA-MEMCHECK ========= Program hit cudaErrorDevicesUnavailable (error 46) due to "all CUDA-capable devices are busy or unavailable" on CUDA API cal…

Changing the permissions for file /dev/nvhost-dbg-gpu appears to resolve the issue. This may help un-block developers who come across this thread, but NVIDIA, @Robert_Crovella could this be addressed in a future JetPack version? sudo chmod a+rw /dev/nvhost-dbg-gpu

All CUDA-capable devices busy or unavailable

Robotics & Edge Computing Jetson & Embedded Systems Jetson TX2

Robert_Crovella December 13, 2021, 3:05pm 5

Perhaps there is something specific to using cuda-memcheck with recent JetPack versions. You might get better help by posting on one of the Jetson forums for your device. You could also file a bug.

Topic		Replies	Views
All CUDA-capable devices busy or unavailable Jetson TX2 cuda , jetson-inference	3	936	January 5, 2022
why "all CUDA-capable devices are busy or unavailable" ? CUDA Programming and Performance	34	64633	April 20, 2011
CUDA 3.2 on GTX 480 is "busy or unavailable" CUDA Programming and Performance	19	73514	March 21, 2011
all CUDA-capable devices are busy or unavailable problem in a multi-process Linux application CUDA Programming and Performance	24	3426	January 7, 2011
all cuda-capable devices are busy or unavailable CUDA Programming and Performance	0	2658	February 24, 2011
deviceQuery passes but other demos fail CUDA Programming and Performance	7	2558	January 22, 2009
Cannot run any CUDA kernels CUDA runtime doesn't recognize NVIDIA GPU CUDA Programming and Performance	26	12525	August 24, 2010
Cuda libraries have memory errors CUDA Programming and Performance	0	3598	August 20, 2011
failed to enqueue convolution on stream: CUDNN_STATUS_EXECUTION_FAILED Jetson TX2	10	1276	March 1, 2018
CUDA Toolkit 3.2 release candidate available to registered developers CUDA Programming and Performance	68	63259	December 3, 2010

All CUDA-capable devices busy or unavailable

Related topics