Working around CUDA issue

[hkaiser]

Fixes #6407

[hkaiser]

@G-071 Please check whether this solves the issue you reported.

[G-071]

Unfortunately, it seems NVCC has trouble with this one as well:

/home/daissgr/workshop/KokkosTiger/build/hpx/include/hpx/properties/property.hpp(26): error: expected a ">"
          detected during:
            instantiation of "hpx::experimental::tag_fallback_invoke" based on template arguments <hpx::parallel::execution::with_processing_units_count_t, hpx::execution::detail::parallel_policy_shim<hpx::execution::parallel_executor, hpx::execution::experimental::default_parameters> &, const unsigned long &> 
/home/daissgr/workshop/KokkosTiger/build/hpx/include/hpx/functional/invoke_result.hpp(38): here
            instantiation of class "hpx::util::invoke_result<F, Ts...> [with F=const hpx::functional::detail::tag_fallback_invoke_t_ns::tag_fallback_invoke_t, Ts=<hpx::experimental::prefer_t, const hpx::parallel::execution::with_processing_units_count_t &, hpx::execution::detail::parallel_policy_shim<hpx::execution::parallel_executor, hpx::execution::experimental::default_parameters> &, const std::size_t &>]" 
/home/daissgr/workshop/KokkosTiger/build/hpx/include/hpx/parallel/util/detail/chunk_size.hpp(173): here
            instantiation of "hpx::util::iterator_range<hpx::parallel::util::detail::chunk_size_iterator<IterOrR>, hpx::parallel::util::detail::chunk_size_iterator<IterOrR>> hpx::parallel::util::detail::get_bulk_iteration_shape(ExPolicy &, IterOrR &, std::size_t &, Stride) [with ExPolicy=hpx::execution::detail::parallel_policy_shim<hpx::execution::parallel_executor, hpx::execution::experimental::default_parameters>, IterOrR=int64_t, Stride=std::size_t]" 
/home/daissgr/workshop/KokkosTiger/build/hpx/include/hpx/parallel/algorithms/for_loop.hpp(1667): here
            instantiation of "hpx::parallel::util::detail::algorithm_result_t<ExPolicy, void> hpx::experimental::tag_fallback_invoke(hpx::experimental::for_loop_strided_t, ExPolicy &&, std::decay_t<I>, I, S, Args &&...) [with ExPolicy=hpx::execution::detail::parallel_policy_shim<hpx::execution::parallel_executor, hpx::execution::experimental::default_parameters>, I=int64_t, S=int64_t, Args=<lambda [](int64_t)->void>, _concept_requires_1648=42, _concept_check_1648=0]" 
/home/daissgr/workshop/KokkosTiger/build/hpx/include/hpx/functional/detail/tag_fallback_invoke.hpp(296): here
            instantiation of "auto hpx::functional::detail::tag_base_ns::tag_fallback<Tag, Enable>::operator()(Args &&...) const->hpx::functional::detail::tag_fallback_invoke_result_t<Tag, Args &&...> [with Tag=hpx::experimental::for_loop_strided_t, Enable=hpx::meta::constant<std::integral_constant<__nv_bool, true>>, Args=<hpx::execution::detail::parallel_policy_shim<hpx::execution::parallel_executor, hpx::execution::experimental::default_parameters>, int64_t, int64_t, const int64_t &, lambda [](int64_t)->void>, <unnamed>=void]" 
/home/daissgr/workshop/KokkosTiger/src/kokkos/core/src/Kokkos_HPX.hpp(1047): here
            [ 3 instantiation contexts not shown ]
            instantiation of "void Kokkos::Impl::ViewValueFunctor<DeviceType, ValueType, true>::parallel_for_implementation() [with DeviceType=Kokkos::Device<Kokkos::HostSpace::execution_space, Kokkos::HostSpace::memory_space>, ValueType=uint32_t]" 
/home/daissgr/workshop/KokkosTiger/src/kokkos/core/src/impl/Kokkos_ViewMapping.hpp(3082): here
            instantiation of "std::enable_if_t<<expression>, void> Kokkos::Impl::ViewValueFunctor<DeviceType, ValueType, true>::construct_shared_allocation() [with DeviceType=Kokkos::Device<Kokkos::HostSpace::execution_space, Kokkos::HostSpace::memory_space>, ValueType=uint32_t, Dummy=uint32_t]" 
/home/daissgr/workshop/KokkosTiger/src/kokkos/core/src/impl/Kokkos_ViewMapping.hpp(3463): here
            instantiation of "Kokkos::Impl::SharedAllocationRecord<void, void> *Kokkos::Impl::ViewMapping<Traits, std::enable_if_t<<expression>, void>>::allocate_shared(const Kokkos::Impl::ViewCtorProp<P...> &, const Traits::array_layout &, __nv_bool) [with Traits=Kokkos::ViewTraits<uint32_t *, Kokkos::HostSpace>, P=<std::string, Kokkos::HostSpace::memory_space, Kokkos::HostSpace::execution_space>]" 
/home/daissgr/workshop/KokkosTiger/src/kokkos/core/src/Kokkos_View.hpp(1434): here
            instantiation of "Kokkos::View<DataType, Properties...>::View(const Kokkos::Impl::ViewCtorProp<P...> &, const std::enable_if_t<<expression>, Kokkos::ViewTraits<DataType, Properties...>::array_layout> &) [with DataType=uint32_t *, Properties=<Kokkos::HostSpace>, P=<std::string>]" 
/home/daissgr/workshop/KokkosTiger/src/kokkos/core/src/Kokkos_View.hpp(1536): here
            instantiation of "Kokkos::View<DataType, Properties...>::View(const Label &, std::enable_if_t<Kokkos::Impl::is_view_label<Label>::value, const size_t>, size_t, size_t, size_t, size_t, size_t, size_t, size_t) [with DataType=uint32_t *, Properties=<Kokkos::HostSpace>, Label=char [27]]" 
/home/daissgr/workshop/KokkosTiger/src/kokkos/core/src/Kokkos_HPX.hpp(557): here

The other suggestion

#if !defined(HPX_CUDA_VERSION) || (HPX_CUDA_VERSION >= 1102)

seems to work once I adapted the HPX_CUDA_VERSION >= 1300 (I am on 1200 right now, so the ifdef would have not removed the noexcept code otherwise). Since we do not know which CUDA version will work in the future: Should we just drop the second if condition and only use

#if !defined(HPX_CUDA_VERSION)

?

[hkaiser]

The other suggestion

#if !defined(HPX_CUDA_VERSION) || (HPX_CUDA_VERSION >= 1102)

seems to work once I adapted the HPX_CUDA_VERSION >= 1300 (I am on 1200 right now, so the ifdef would have not removed the noexcept code otherwise). Since we do not know which CUDA version will work in the future: Should we just drop the second if condition and only use

#if !defined(HPX_CUDA_VERSION)

Yep, let's do that instead, then.

[hkaiser]

@G-071 I have updated this PR to do exactly that.

[codacy-production]

Coverage summary from Codacy

See diff coverage on Codacy

Coverage variation	Diff coverage
✅ -0.01%	✅ ∅

Coverage variation details

	Coverable lines	Covered lines	Coverage
Common ancestor commit (7b35448)	197211	167924	85.15%
Head commit (ce14789)	190473 (-6738)	162166 (-5758)	85.14% (-0.01%)

Coverage variation is the difference between the coverage for the head and common ancestor commits of the pull request branch: <coverage of head commit> - <coverage of common ancestor commit>

Diff coverage details

	Coverable lines	Covered lines	Diff coverage
Pull request (#6409)	0	0	∅ (not applicable)

Diff coverage is the percentage of lines that are covered by tests out of the coverable lines that the pull request added or modified: <covered lines added or modified>/<coverable lines added or modified> * 100%

See your quality gate settings    Change summary preferences

_{You may notice some variations in coverage metrics with the latest Coverage engine update. For more details, visit the documentation}

[G-071]

This did the trick, thanks! I can build [email protected] with it again. Should be good to merge!

That being said, I was unable to run Octo-Tiger with this: Octo-Tiger is segfaulting either on the first time step or when loading the options (same Octo-Tiger version still works with HPX 1.9.1). This seems to be an unrelated to this PR though as this also happens with master. I will investigate some more and open a separate issue for this.