Originally posted by **jaegon-kim-77** August 26, 2021
Hello. My name is Jaegon Kim.
Last few weeks I have tried build and execute TIGRE toolkit on Matlab. My environment is like follow.
- Window 10 x64
- Matlab 2016b
- CUDA 9.2
- RTX 3060 (12GB global memory, 8.6 of CUDA compute capability)
- Visual studio 2015
I have followed step by step instruction on "MATLAB_installation.md" and there were no errors when I execute "Compile.m".
After this step I have execute "InitTIGRE.m" also.
But when I try execute demo file like "d03_generateData.m" on Demos folder there were problem on execution CUDA kernel function (I guess. Error message is "no kernel image is available for execution on the device"). Since I could not resolve this problem, I have build CUDA kernel function execution file(exe) and then call that execution file(before execute that I have to write parameters and image data to file.) instead of direct call CUDA kernel function on mex file, and I can see results on MATLAB.
Anyway I can execute some algorithms but there are huge overheads since my method introduce file read/write frequently.
If anybody can give me some information or direction then it will be a big help for me.
Thanks for read this question.
Discussed in https://github.com/CERN/TIGRE/discussions/322