[CMake] Future direction for FindCUDA

[Robert Maynard]

> Can I link to CUDA libraries with GCC while not using FindCUDA

Yes you can if you have enabled the CUDA language. You will need to
use find_package and can use the CMAKE_CUDA_IMPLICIT_LINK_DIRECTORIES
as the location(s) to search.

> Instead of modifying what FindCUDA does, which would break backwards compatibility, FindCUDA could be split into FindCUDA and FindCUDALibraries or something like that.

That is something I have wanted to do, but haven't had the time to
formalize and finish. You can find my initial implementation at
https://gitlab.kitware.com/robertmaynard/find_cuda_wrappers

On Wed, Aug 30, 2017 at 5:06 AM, Mueller-Roemer, Johannes Sebastian
<Johannes.Sebastian.Mueller-Roemer at igd.fraunhofer.de> wrote:
> Hi,
>
> we have encountered this issue as well, in our case when using thrust on the host side. We currently solve this by still using FindCUDA, even when using CMake 3.8/3.9, but only using the CUDA_*_LIBRARY and CUDA_TOOLKIT_INCLUDE variables and nothing else.
>
> Instead of modifying what FindCUDA does, which would break backwards compatibility, FindCUDA could be split into FindCUDA and FindCUDALibraries or something like that.
>
> Regards
> Johannes
>
> Fraunhofer-Institut für Graphische Datenverarbeitung IGD
> Fraunhoferstr. 5  |  64283 Darmstadt  |  Germany
> Tel +49 6151 155-606  |  Fax +49 6151 155-139
> johannes.mueller-roemer at igd.fraunhofer.de | www.igd.fraunhofer.de
>
>
> -----Original Message-----
> From: CMake [mailto:cmake-bounces at cmake.org] On Behalf Of Siyuan Liu
> Sent: Wednesday, August 30, 2017 07:05
> To: cmake at cmake.org
> Subject: [CMake] Future direction for FindCUDA
>
> Dear all,
>
> I've noticed the change to CUDA as a language in CMake 3.8. I've migrated to this new syntax, but this change leads to some difficulties for me with CMake. I think we should treat CUDA as not just a language but rather a language with some libraries.
>
> Now that the FindCUDA module is deprecated, I cannot compile a file with GCC while linking to a CUDA library (since the CUDA_xxx_LIBRARY variables are gone). The reason for not using NVCC is to save compilation time. Personally, I think FindCUDA should become a module for purely finding CUDA libraries for linking without anything related to the NVCC compiler (e.g. the cuda_add_executable macro).
>
> My questions in short:
> 1. Can I link to CUDA libraries with GCC while not using FindCUDA?
> 2. What's your opinion on my suggestion about FindCUDA?
>
> Regards,
> Siyuan
> --
>
> Powered by www.kitware.com
>
> Please keep messages on-topic and check the CMake FAQ at: http://www.cmake.org/Wiki/CMake_FAQ
>
> Kitware offers various services to support the CMake community. For more information on each offering, please visit:
>
> CMake Support: http://cmake.org/cmake/help/support.html
> CMake Consulting: http://cmake.org/cmake/help/consulting.html
> CMake Training Courses: http://cmake.org/cmake/help/training.html
>
> Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/mailman/listinfo/cmake
> --
>
> Powered by www.kitware.com
>
> Please keep messages on-topic and check the CMake FAQ at: http://www.cmake.org/Wiki/CMake_FAQ
>
> Kitware offers various services to support the CMake community. For more information on each offering, please visit:
>
> CMake Support: http://cmake.org/cmake/help/support.html
> CMake Consulting: http://cmake.org/cmake/help/consulting.html
> CMake Training Courses: http://cmake.org/cmake/help/training.html
>
> Visit other Kitware open-source projects at http://www.kitware.com/opensource/opensource.html
>
> Follow this link to subscribe/unsubscribe:
> http://public.kitware.com/mailman/listinfo/cmake