How to use this in vs 2017 and win10 64bit

[superskhm]

No description provided.

[monchin]

Recently I'm doing this thing, but VS2019 instead of 2017.

Basically you need to follow this tutorial. After successfully installation, you should be able to run samples in C:\ProgramData\NVIDIA Corporation\CUDA Samples\v11.0(Notice that C:\PrgramData is defaultly a hidden folder). Note each sample has different VS versions including 2017 and 2019.

If you can run samples successfully, you can compile this library. You need to create a new project with CUDA 11.0 Runtime template. Then you should copy all the source codes here to your new project. You need to add some codes, for example,

#ifdef _MSC_VER 
//not #if defined(_WIN32) || defined(_WIN64) because we have strncasecmp in mingw
#define strncasecmp _strnicmp
#define strcasecmp _stricmp
#define uint unsigned int
#define __intptr_t intptr_t
#endif

You need to add these codes in the right head files to ensure all the codes could be compiled correctly.

Then, as MSVC can't compile C array with variable length, you must change some C arrays to std::vector(std::vector<int>a(length+1); auto ptr_a = a.data()).

Also, you need to open property of this project, and in linker->input->additional dependencies(may be other words because I translated from Chinese), you must add cublas.lib;cuda.lib;cudadevrt.lib;cudart.lib;OpenCL.lib;

Then, I guess you can compile it successfully. Enjoy!

[superskhm]

can you share it with us? Following your suggestion, I can compile it. But I can't get the right result, i get a blank image.

[monchin]

can you share it with us? Following your suggestion, I can compile it. But I can't get the right result, i get a blank image.

That is strange. I used this package's lena.dat in test dir, and ran the main function, and I successfully got res.dat as the result. You can try lena first.