Getting OpenCV to work with CUDA, "no kernel image is available..."

[theosib]

I have a Jetson Orin Nano, and I’d like to get OpenCV to use CUDA. I understand that this isn’t built in, so I followed instructions that I found at https://qengineering.eu/install-opencv-4.5-on-jetson-nano.html to install OpenCV 4.6. Since those instructions are for an older Jetson Nano, I modified the script to select the correct GPU:

CUDA_ARCH_BIN=8.0

This is specifically what I have found in forums for dealing with the error I’m getting, but it isn’t helping. Despite getting the version right, as far as I can tell, I still get this error when I try to do anything CUDA-related:

error: (-217:Gpu API call) no kernel image is available for execution on the device in function 'call' Aborted

Is 8.0 the right architecture/GPU? Are there tools I can install to diagnose this? What else can I do to figure out what’s going wrong here?

Thanks.

[dusty_nv]

Hi @theosib, try CUDA_ARCH_BIN=8.7 for Orin instead. If you check the CUDA deviceQuery output, it is listed there:

/usr/local/cuda/samples/1_Utilities/deviceQuery$ ./deviceQuery
./deviceQuery Starting...

 CUDA Device Query (Runtime API) version (CUDART static linking)

Detected 1 CUDA Capable device(s)

Device 0: "Orin"
  CUDA Driver Version / Runtime Version          11.4 / 11.4
  CUDA Capability Major/Minor version number:    8.7
  Total amount of global memory:                 6434 MBytes (6746046464 bytes)
  (008) Multiprocessors, (128) CUDA Cores/MP:    1024 CUDA Cores
  GPU Max Clock rate:                            624 MHz (0.62 GHz)
  Memory Clock rate:                             624 Mhz
  Memory Bus Width:                              64-bit
  L2 Cache Size:                                 2097152 bytes
  Maximum Texture Dimension Size (x,y,z)         1D=(131072), 2D=(131072, 65536), 3D=(16384, 16384, 16384)
  Maximum Layered 1D Texture Size, (num) layers  1D=(32768), 2048 layers
  Maximum Layered 2D Texture Size, (num) layers  2D=(32768, 32768), 2048 layers
  Total amount of constant memory:               65536 bytes
  Total amount of shared memory per block:       49152 bytes
  Total shared memory per multiprocessor:        167936 bytes
  Total number of registers available per block: 65536
  Warp size:                                     32
  Maximum number of threads per multiprocessor:  1536
  Maximum number of threads per block:           1024
  Max dimension size of a thread block (x,y,z): (1024, 1024, 64)
  Max dimension size of a grid size    (x,y,z): (2147483647, 65535, 65535)
  Maximum memory pitch:                          2147483647 bytes
  Texture alignment:                             512 bytes
  Concurrent copy and kernel execution:          Yes with 2 copy engine(s)
  Run time limit on kernels:                     No
  Integrated GPU sharing Host Memory:            Yes
  Support host page-locked memory mapping:       Yes
  Alignment requirement for Surfaces:            Yes
  Device has ECC support:                        Disabled
  Device supports Unified Addressing (UVA):      Yes
  Device supports Managed Memory:                Yes
  Device supports Compute Preemption:            Yes
  Supports Cooperative Kernel Launch:            Yes
  Supports MultiDevice Co-op Kernel Launch:      Yes
  Device PCI Domain ID / Bus ID / location ID:   0 / 0 / 0
  Compute Mode:
     < Default (multiple host threads can use ::cudaSetDevice() with device simultaneously) >

deviceQuery, CUDA Driver = CUDART, CUDA Driver Version = 11.4, CUDA Runtime Version = 11.4, NumDevs = 1
Result = PASS

[theosib]

Thanks! I didn’t know about that. I guess I should spend some time looking through those samples. I’ve changed the version to 8.7 and will report back how it goes. Thanks!