Error in building ITK by enabling RTK

[navami123nair]

Hi,
I am new in using ITK-RTK for image reconstruction. I wanted to run the example code in RTK using CPU and GPU, so initially I configured ITK V5.3.0 for CPU and it was successful. But now I am trying to run the CUDA based example code in RTK for which I have configured ITK_USE_GPU as well as RTK_USE_CUDA. But unfortunately the build fails.

MY ITK version was 5.3.0 and then I tried to configure using Version 5.1.0 and my CUDA version was V12.3, then I tried to configure with V11.8, but still build fails. The error is "Texture is not a Template". Now when I tried to build ITK only with ITK_USE_GPU enabled, the following build error occurred.

 "810>itkGPUDemonsRegistrationFilterTest.obj : error LNK2019: unresolved external symbol "public: static char const * __cdecl itk::GPUDenseFiniteDifferenceImageFilterKernel::GetOpenCLSource(void)" (?GetOpenCLSource@GPUDenseFiniteDifferenceImageFilterKernel@itk@@SAPEBDXZ) referenced in function "public: static char const * __cdecl itk::GPUDenseFiniteDifferenceImageFilter<class itk::GPUImage<class itk::Vector<float,2>,2>,class itk::GPUImage<class itk::Vector<float,2>,2>,class itk::DemonsRegistrationFilter<class itk::GPUImage<float,2>,class itk::GPUImage<float,2>,class itk::GPUImage<class itk::Vector<float,2>,2> > >::GetOpenCLSource(void)" (?GetOpenCLSource@?$GPUDenseFiniteDifferenceImageFilter@V?$GPUImage@V?$Vector@M$01@itk@@$01@itk@@V12@V?$DemonsRegistrationFilter@V?$GPUImage@M$01@itk@@V12@V?$GPUImage@V?$Vector@M$01@itk@@$01@2@@2@@itk@@SAPEBDXZ)
810>itkGPUDemonsRegistrationFilterTest2.obj : error LNK2001: unresolved external symbol "public: static char const * __cdecl itk::GPUDenseFiniteDifferenceImageFilterKernel::GetOpenCLSource(void)" (?GetOpenCLSource@GPUDenseFiniteDifferenceImageFilterKernel@itk@@SAPEBDXZ)" 

This is just two of the error messages shown. Please help me with a solution. Thanks in advance.

[github-actions]

Thank you for contributing an issue! 🙏

Welcome to the ITK community! 🤗👋☀️

We are glad you are here and appreciate your contribution. Please keep in mind our community participation guidelines. 📜
Also, please check existing open issues and consider discussion on the ITK Discourse. 📖

This is an automatic message. Allow for time for the ITK community to be able to read the issue and comment on it.

[SimonRit]

The issue title is surprising since the reported error is not related to RTK. RTK is not compatible with Cuda 12 yet, see RTKConsortium/RTK#549. RTK is only compatible with the latest ITK release and ITK master at the time of the release.

[navami123nair]

Yes I have read this https://discourse.itk.org/t/about-gpu-version-compilation-of-rtk/5905/9 So I have configured using CUDA version 11.8, but still the same error persists.

[SimonRit]

Do you mean the one reported above? As you point out, it's here without RTK and I suggest renaming the issue

[navami123nair]

Actually I did not understand what you said. I need to do a project using RTK for which I am trying to study it by executing the example codes. Since I used ITK v5.3.0 and you suggest to use the latest version, I am going to try with the same.

[navami123nair]

While configuring ITK in cmake, should I enable Module_CudaCommon.

[navami123nair]

With the latest ITK version I am not able to configure RTK_USE_CUDA. The error coming are

CMake Error at 753 (message):  Compiling the CUDA compiler identification source file
  "CMakeCUDACompilerId.cu" failed.  Build flags: 
   Id flags: --keep;--keep-dir;tmp -v
 
 
 
   The output was:
 
   1
 
   MSBuild version 17.7.2+d6990bcfa for .NET Framework
 
   Build started 1/17/2024 2:45:03 PM. Project on
   node 1 (default targets).
 
   PrepareForBuild: Creating directory "Debug\".
     Creating directory "Debug\CompilerIdCUDA.tlog\".
 
   InitializeBuildStatus: Creating "Debug\CompilerIdCUDA.tlog\unsuccessfulbuild" because "AlwaysCreate" was specified.
     Touching "Debug\CompilerIdCUDA.tlog\unsuccessfulbuild". error : The CUDA Toolkit v12.3 directory does not exist.  Please
   verify the CUDA Toolkit is installed properly or define the CudaToolkitDir
   property to resolve this error.  Build FAILED.(ValidateCudaBuild target) -> error : The CUDA Toolkit v12.3 directory does not exist.  Please verify the CUDA Toolkit is installed properly or define the CudaToolkitDir property to resolve this error.  0 Warning(s)
       1 Error(s)
 
  
 
   Time Elapsed 00:00:00.17
 
 Call Stack (most recent call first): 8 (CMAKE_DETERMINE_COMPILER_ID_BUILD)
 53 (__determine_compiler_id_test)
 307 (CMAKE_DETERMINE_COMPILER_ID)
 29 (enable_language)
 374 (include)
   CMakeLists.txt:583 (include)
 
 
 Configuring incomplete, errors occurred!

How to resolve this issue.

[SimonRit]

There seems to be a problem with your Cuda installation?
CudaCommon is automatically enabled by RTK.

[navami123nair]

I once again installed CUDA V11.8, and tried to configure ITK-Master but at the initial stage itself it failed to configure. So next I tried with ITK V5.4. Now Module_RTK and ITK_USE_GPU were configured and build, but moving on to build RTK_USE_CUDA it fails. What should I do now?

[SimonRit]

The error you posted above suggests that CMake cannot find Cuda. This should be reproducible with a simple CMakeLists.txt file such as

cmake_minimum_required(VERSION 3.17 FATAL_ERROR)
project(CudaTest)
find_package(CUDAToolkit REQUIRED)

Can you check if you can configure this project with cmake? Which version of cmake are you using BTW?

[navami123nair]

My Cmake version is 3.27.8.

[navami123nair]

The
cmake_minimum_required(VERSION 3.17 FATAL_ERROR) project(CudaTest) find_package(CUDAToolkit REQUIRED)
is configuring without any issue. In ITK configuration issue is persisting when RTK_USE_CUDA is enabled.

[navami123nair]

Hi,
The issue was resolved. I had CUDA v12.3 as well as v11.8 in my system, due to which the CUDA issue occurred. I have uninstalled CUDA v12.3 and then again configured ITK using CMake and now I am able to configure and build RTK_USE_CUDA and was able to run the GPU example program in RTK.

Thank you for all the support. Since the problem is resolved, I am closing this issue.

Regards,
Navami S Nair.