Issue compiling DualSPHysics v5.4 for CPU — GPU-related files being compiled

[Pbarbosa92]

Hi guys,

I'm trying to compile DualSPHysics v5.4 for CPU only using a Docker image. I’m using cmake with the following commands:

mkdir /DualSPHysics_v5.4/src/source/build
cd /DualSPHysics_v5.4/src/source/build
CC=gcc CXX=g++ cmake .. && make DualSPHysics5.4CPU_linux64

Initially, I encountered an error stating that cuda_runtime_api.h could not be found. This was unexpected, since I'm building the CPU version and assumed CUDA-related headers wouldn’t be needed.

To work around that, I switched to a CUDA-enabled base image:

FROM nvidia/cuda:11.7.1-devel-ubuntu22.04

This allowed the build to find the CUDA headers, but then I ran into a new set of compilation errors, seemingly from GPU-related files being compiled even though I'm targeting CPU:

/DualSPHysics_v5.4/src/source/JSphGpuSingle_VRes.cpp:277:6: error: 'StrGeomVresGpu' was not declared in this scope; did you mean 'StrGeomVresCpu'?
...
error: 'class JDsPartMotionSave' has no member named 'AddDataExtraGpu'; did you mean 'AddDataExtraCpu'?
...

This suggests that make DualSPHysics5.4CPU_linux64 is incorrectly trying to compile GPU-related files (e.g., JSphGpuSingle_VRes.cpp) which shouldn't be included for a CPU-only build.

Questions:

• Is there a specific CMake flag or build configuration required to ensure that only the CPU source files are compiled?
• Should I be using a different make target altogether for a pure CPU build?

Thanks in advance for your help!

[imestevez]

CMake has been updated to fix the CPU issue in the latest commits on GitHub and also in the full package (v5.4.2) on the website: https://dual.sphysics.org/downloads/