thecomputekid / premake5-cuda Goto Github PK

Howdy,

Thanks very much for taking the time to do this project!

Your example project works fine here. But my usual MO is to make a single premake5.lua to hold the solution setting and add separate project folders, each with it's own settings stored in a premake5.lua file and "include" them in the solution premake5 file.

Here's a simple example using your own Example project. If you run "generateVS2019.bat", you'll see that the *.cu files are not added to the Example project. The solution and projects live in the "build" folder

Is that just a limitation or is it something that can be fixed. Thanks!

https://github.com/Hurleyworks/cudaPremake

Sorry to bother you again! I use NVIDIA OptiX in my projects and it needs to create .ptx files from .cu files. Below is the relevant vs2019 project settings. I've gotten just about everything working by adding the necessary additions to the "cuda-exported-variables" file. But I can't figure out how create "CompileOut" setting so that a different .ptx file will be saved for each .cu file. Could you tell me how I might do this? Thanks!

 <ItemGroup>
    <CudaCompile Include="deform.cu" />
    <CudaCompile Include="optix_kernels.cu">
      <GPUDebugInfo Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">false</GPUDebugInfo>
      <CodeGeneration Condition="'$(Configuration)|$(Platform)'=='Debug|x64'">compute_75,sm_75</CodeGeneration>
      <CodeGeneration Condition="'$(Configuration)|$(Platform)'=='Release|x64'">compute_75,sm_75</CodeGeneration>
    </CudaCompile>
  </ItemGroup>

 <CudaCompile>
      <GenerateRelocatableDeviceCode>true</GenerateRelocatableDeviceCode>
    </CudaCompile>
    <CudaCompile>
      <NvccCompilation>ptx</NvccCompilation>
    </CudaCompile>
    <CudaCompile>
      <TargetMachinePlatform>64</TargetMachinePlatform>
      <CudaRuntime>Shared</CudaRuntime>
      <CompileOut>$(SolutionDir)$(Platform)\$(Configuration)\$(TargetName)\ptxes\%(Filename).ptx</CompileOut>
      <AdditionalOptions>-std=c++17 --use_fast_math -Xcompiler "/wd 4819" -D_DEBUG %(AdditionalOptions)</AdditionalOptions>
      <FastMath>true</FastMath>
      <GenerateLineInfo>true</GenerateLineInfo>
    </CudaCompile>
    <CudaLink>
      <PerformDeviceLink>false</PerformDeviceLink>
    </CudaLink>

When you have cudaFile with relative path sometimes the module failed to load them.

Hi,
How can I defined in premake the compiler output :

I try to add to -o=path/to/my/output to the cudaCompilerOptions variable, but it didn't work.
Is it possible to defined this parameters with your module ?