Hi, <a class="user-mention notranslate" data-hovercard-type="user" data-hovercard-url=

The Compute Capability of NVIDIA GPUs can be found here. <a href="https://develope

Error: no kernel image is available for execution on the device about fusibile HOT 2 CLOSED

philleer commented on July 28, 2024

Error: no kernel image is available for execution on the device

from fusibile.

Comments (2)

philleer commented on July 28, 2024 5

Got it, finally! And here is my solution:
i google my problem and find these issues
cuda runtime error (48): no kernel image is available for execution on the device and
GPU 0: no kernel image is available for execution on the device
according to the reply, i find line 10 in CMakeLists.txt, here is obviously commented for different gpu architecture, so i uncomment line 11 and comment line 10.

#set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};-O3 --use_fast_math --ptxas-options=-v -std=c++11 --compiler-options -Wall -gencode arch=compute_30,code=sm_30 -gencode arch=compute_52,code=sm_52)
set(CUDA_NVCC_FLAGS ${CUDA_NVCC_FLAGS};-O3 --use_fast_math --ptxas-options=-v -std=c++11 --compiler-options -Wall -gencode arch=compute_52,code=sm_52 -gencode arch=compute_60,code=sm_60 -gencode arch=compute_61,code=sm_61)

arch=60 and arch=61 is added just for compatibility according to nvcc fatal
Then i recompile the project and now it works.
BTW, i think a practical pipeline is needed for those who want to run the project quickly for experiments.

git clone https://github.com/kysucix/fusibile.git
cd fusibile
mkdir build
cd build
cmake ..
make

remember to config the CMakeLists.txt to make it works well on your machine, especially with graphical card. And after compilation you can run the executable as

./fusibile -input_folder /path/to/point_folder/ \
-p_folder /path/to/cam_folder/ \
-images_folder=/path/to/image_folder/ \
--depth_min=depth_min_value \
--depth_max=depth_max_value \
--normal_thresh=normal_thresh_value \
--disp_thresh=disp_thresh_value \
--num_consistent=num_consistent_value

from fusibile.