[vtk-developers] How to write CMakeLists.txt for a library using CUDA?

[Moreland, Kenneth]

I don't know if I understand your first question, but you should always specify all the libraries that any target depends on with target_link_libraries. I think you are getting away with not specifying the shared object because the linker is picking up the .so in your LD_LIBRARY_PATH, but in general you should not assume that will work.

For the second question, I don't have experience with Visual Studio, but a quick Google search came up with this page that seems to answer the same question: http://stackoverflow.com/questions/5856929/configuring-cmake-to-setup-cuda-on-windows

-Ken

From: Mingcheng Chen <linyufly at gmail.com<mailto:linyufly at gmail.com>>
Date: Monday, March 17, 2014 3:21 PM
To: Kenneth Moreland <kmorel at sandia.gov<mailto:kmorel at sandia.gov>>
Cc: "vtk-developers at vtk.org<mailto:vtk-developers at vtk.org>" <vtk-developers at vtk.org<mailto:vtk-developers at vtk.org>>, vtkusers <vtkusers-bounces at vtk.org<mailto:vtkusers-bounces at vtk.org>>
Subject: [EXTERNAL] Re: [vtk-developers] How to write CMakeLists.txt for a library using CUDA?

Hi Kenneth,

Thanks a lot for your advice!

I followed your advice and successfully built *.so and *.a for the same library, and my following code can use *.so library successfully!

I have two follower questions:
(1) I have link error if I want to compile something using my *.a library, but the CMakeLists.txt is almost the same as the one using *.so library. I use target_link_libraries(myApp ${VTK_SOURCE_DIR}/myLib.a) and target_link_libraries(myApp ${VTK_SOURCE_DIR}/myLib.so) for both. Is there anything I should consider for *.a library?

(2) The CMakeLists.txt that works in Ubuntu does not work in Visual Studio. The *.sln it creates is not a CUDA project. How to solve that problem?

Thanks again!

Best regards,
Mingcheng Chen
March 17th, 2014


On Mon, Mar 17, 2014 at 12:59 PM, Moreland, Kenneth <kmorel at sandia.gov<mailto:kmorel at sandia.gov>> wrote:
This question is really more applicable to the VTK users list, so I'm responding to there as well.

The short answer is that if you add the command "find_package(CUDA REQUIRED)" to your CMakeLists.txt, that will load the CMake code to find the CUDA install. Then you replace add_library with cuda_add_library. You also have to use a .cu extension for your C or C++ files that use CUDA instead of .c or .cxx or .cpp. The cuda_add_library will only use the requisite nvcc compiler on files with a .cu extension.

For more information, try this command prompt command: "cmake --help-module FindCUDA | less". You can also find several examples on the web.

-Ken

From: Mingcheng Chen <linyufly at gmail.com<mailto:linyufly at gmail.com>>
Date: Sunday, March 16, 2014 11:36 AM
To: "vtk-developers at vtk.org<mailto:vtk-developers at vtk.org>" <vtk-developers at vtk.org<mailto:vtk-developers at vtk.org>>
Subject: [EXTERNAL] [vtk-developers] How to write CMakeLists.txt for a library using CUDA?

Hi everyone,

Thanks for your time!

I write a class which uses VTK and CUDA, and I want to write a CMakeLists.txt for others to build my library. I only know how to write CMakeLists.txt for executables. Could anybody please show me how to write for a library?

Thanks!

Best regards,
Mingcheng Chen
March 16th, 2014

--
Research Assistant in Graphics Group
University of Illinois at Urbana-Champaign
http://mingchengchen.org



--
Research Assistant in Graphics Group
University of Illinois at Urbana-Champaign
http://mingchengchen.org
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://public.kitware.com/pipermail/vtk-developers/attachments/20140317/2a1eb14c/attachment-0002.html>