Picongpu: cupla_cuda_async error while building picongpu on Nvidia

[saipavankalyan]

I'm trying to build picongpu on Nvidia. The CUDA version I am using is 10.2.89. Alpaka and cupla present in this picongpu repository doesn't support CUDA 10.2. So, I replaced alpaka and cupla with the corresponding standalone reposritory which supports CUDA 10.2

I'm unable to build the application and here is the error log.
log1.txt

I'm getting similar error repeated in many files.
/home/src/picongpu/include/pmacc/../pmacc/eventSystem/events/kernelEvents.hpp(209): error #1767: function "cupla::cupla_cuda_async::KernelCudaLike<T_KernelType>::operator() [with T_KernelType=pmacc::idDetail::KernelSetNextId]" cannot be called with the given argument list argument types are: (pmacc::DataSpace<1U>, pmacc::DataSpace<1U>, const size_t, cuplaStream_t) object type is: cupla::cupla_cuda_async::KernelCudaLike<pmacc::idDetail::KernelSetNextId> detected during: instantiation of "void pmacc::exec::KernelStarter<T_Kernel, T_VectorGrid, T_VectorBlock>::operator()(const T_Args &...) const [with T_Kernel=pmacc::exec::Kernel<pmacc::idDetail::KernelSetNextId>, T_VectorGrid=int, T_VectorBlock=int, T_Args=<uint64_t>]" (241): here instantiation of "void pmacc::exec::KernelStarter<T_Kernel, T_VectorGrid, T_VectorBlock>::operator()(const T_Args &...) [with T_Kernel=pmacc::exec::Kernel<pmacc::idDetail::KernelSetNextId>, T_VectorGrid=int, T_VectorBlock=int, T_Args=<uint64_t>]" /home//src/picongpu/include/pmacc/../pmacc/particles/IdProvider.hpp(173): here instantiation of "void pmacc::IdProvider<T_dim>::setNextId(uint64_t) [with T_dim=3U]" /home/src/picongpu/include/pmacc/../pmacc/particles/IdProvider.hpp(82): here instantiation of "void pmacc::IdProvider<T_dim>::init() [with T_dim=3U]" /home/src/picongpu/include/pmacc/../picongpu/simulationControl/MySimulation.hpp(423): here

Please inform if I'm missing something here.
Thanks in advance :)

[psychocoderHPC]

There is a hard to catch warning in your logs:

CMake Warning at /home/tcs/src/picongpu/include/pmacc/PMaccConfig.cmake:393 (message):
  Untested CUDA release! Maybe use a newer PIConGPU?
Call Stack (most recent call first):
  CMakeLists.txt:95 (find_package)

You did the right step and tested the last stable release, our master branch. This branch is older than CUDA 10.2 and looks like it shows issues with CUDA 10.2. More information can be found in #3135
Please test the development branch dev, the issue is fixed there. We will do a new stable release in two or three weeks.

[psychocoderHPC]

So, I replaced alpaka and cupla with the corresponding standalone reposritory which supports CUDA 10.2

PIConGPU 0.4.X depends on a older alpaka/cupla version. We did a few breaking API changes in alpaka therefore the usage of the external newer alpaka/cupla version do not work.

[saipavankalyan]

Yes @psychocoderHPC , the issue got resolved when I used dev branch. Thanks :)