[FIX] Fix compile error in distance benchmark

[trxcllnt]

This PR fixes the following compile error:

pairwise_linestring_distance.cu (80): error: too many initializer values
          detected during:
            instantiation of "std::tuple<rmm::device_vector<cuspatial::cartesian_2d<T>>, rmm::device_vector<int32_t>> generate_linestring(int32_t, int32_t, T, cuspatial::cartesian_2d<T>) [with T=float]" 

[harrism]

Did this not get caught by CI because we are not building benchmarks by default in CI? I was going to look at that today, coincidentally.

The old version of the code looks correct to me -- where did this fail? It doesn't fail locally for me or @isVoid .

[vyasr]

Given that all benchmarks compiled successfully in #597 this change is suspicious. Is this perhaps a compiler bug on a new system (Hopper)?

[trxcllnt]

Nope, I encountered this just attempting to build benchmarks with GCC 9.4 and CUDA 11.6.2 in rapids-compose.

[trxcllnt]

Here's the full error:

$ gcc --version && g++ --version && nvcc --version
gcc (conda-forge gcc 9.4.0-16) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

g++ (conda-forge gcc 9.4.0-16) 9.4.0
Copyright (C) 2019 Free Software Foundation, Inc.
This is free software; see the source for copying conditions.  There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.

nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Tue_Mar__8_18:18:20_PST_2022
Cuda compilation tools, release 11.6, V11.6.124
Build cuda_11.6.r11.6/compiler.31057947_0

FAILED: benchmarks/CMakeFiles/DISTANCES_BENCH.dir/pairwise_linestring_distance.cu.o 
/usr/local/bin/nvcc -forward-unknown-to-host-compiler -DCUDF_VERSION=22.08.00 -DGTEST_LINKED_AS_SHARED_LIBRARY=1 -DJITIFY_USE_CACHE -DSPDLOG_ACTIVE_LEVEL=SPDLOG_LEVEL_INFO -DTHRUST_DEVICE_SYSTEM=THRUST_DEVICE_SYSTEM_CUDA -DTHRUST_HOST_SYSTEM=THRUST_HOST_SYSTEM_CPP -I/home/ptaylor/dev/rapids/cuspatial/cpp/benchmarks -I/home/ptaylor/dev/rapids/cuspatial/cpp -I/home/ptaylor/dev/rapids/cuspatial/cpp/src -I/home/ptaylor/dev/rapids/cuspatial/cpp/build/cuda-11.6.2/fea-drop-glue-code/release/_deps/benchmark-src/src/../include -I/home/ptaylor/dev/rapids/cudf/cpp/build/cuda-11.6.2/branch-22.08/release/_deps/libcudacxx-src/include -I/home/ptaylor/dev/rapids/cudf/cpp/build/cuda-11.6.2/branch-22.08/release/_deps/thrust-src -I/home/ptaylor/dev/rapids/cudf/cpp/build/cuda-11.6.2/branch-22.08/release/_deps/thrust-src/dependencies/cub -I/home/ptaylor/dev/rapids/cuspatial/cpp/include -I/home/ptaylor/dev/rapids/cuspatial/cpp/build/cuda-11.6.2/fea-drop-glue-code/release/_deps/nvbench-src -I/home/ptaylor/dev/rapids/cuspatial/cpp/build/cuda-11.6.2/fea-drop-glue-code/release/_deps/nvbench-build -isystem=/home/ptaylor/dev/rapids/cudf/cpp -isystem=/home/ptaylor/dev/rapids/cudf/cpp/src -isystem=/home/ptaylor/dev/rapids/compose/etc/conda/cuda_11.6/envs/rapids/include -isystem=/home/ptaylor/dev/rapids/cudf/cpp/build/cuda-11.6.2/branch-22.08/release/_deps/dlpack-src/include -isystem=/home/ptaylor/dev/rapids/cudf/cpp/build/cuda-11.6.2/branch-22.08/release/_deps/jitify-src -isystem=/home/ptaylor/dev/rapids/cudf/cpp/include -isystem=/home/ptaylor/dev/rapids/cudf/cpp/build/cuda-11.6.2/branch-22.08/release/include -isystem=/home/ptaylor/dev/rapids/rmm/include -isystem=/usr/local/cuda/include -Xcompiler=-fdiagnostics-color=always -Xcompiler=-Wno-deprecated-declarations -O3 -DNDEBUG --generate-code=arch=compute_75,code=[sm_75] -Xcompiler=-fPIE -Xcompiler=-Wno-parentheses --expt-extended-lambda --expt-relaxed-constexpr -Werror=all-warnings -Xcompiler=-Wall,-Werror,-Wno-error=deprecated-declarations -Xfatbin=-compress-all -Xcompiler=-Wno-deprecated-declarations -Werror=cross-execution-space-call -Xcompiler -pthread -std=c++17 -MD -MT benchmarks/CMakeFiles/DISTANCES_BENCH.dir/pairwise_linestring_distance.cu.o -MF benchmarks/CMakeFiles/DISTANCES_BENCH.dir/pairwise_linestring_distance.cu.o.d -x cu -c /home/ptaylor/dev/rapids/cuspatial/cpp/benchmarks/pairwise_linestring_distance.cu -o benchmarks/CMakeFiles/DISTANCES_BENCH.dir/pairwise_linestring_distance.cu.o
/home/ptaylor/dev/rapids/cuspatial/cpp/benchmarks/pairwise_linestring_distance.cu(80): error: too many initializer values
          detected during:
            instantiation of "std::tuple<rmm::device_vector<cuspatial::cartesian_2d<T>>, rmm::device_vector<int32_t>> generate_linestring(int32_t, int32_t, T, cuspatial::cartesian_2d<T>) [with T=float]" 
(100): here
            instantiation of "void pairwise_linestring_distance_benchmark(nvbench::state &, nvbench::type_list<T>) [with T=float]" 
(140): here
            instantiation of "void pairwise_linestring_distance_benchmark_line_140::operator()(nvbench::state &, nvbench::type_list<Ts...>) [with Ts=<float>]" 
_deps/nvbench-src/nvbench/detail/type_list_impl.cuh(100): here
            instantiation of "void nvbench::tl::detail::foreach<TypeList,Functor,Is...>(std::index_sequence<Is...>, Functor &&) [with TypeList=nvbench::type_list<nvbench::type_list<float>, nvbench::type_list<double>>, Functor=lambda [](auto)->auto, Is=<0UL, 1UL>]" 
_deps/nvbench-src/nvbench/detail/type_list_impl.cuh(109): here
            instantiation of "void nvbench::tl::detail::foreach<TypeList,Functor>(Functor &&) [with TypeList=nvbench::type_list<nvbench::type_list<float>, nvbench::type_list<double>>, Functor=lambda [](auto)->auto]" 
_deps/nvbench-src/nvbench/type_list.cuh(140): here
            [ 3 instantiation contexts not shown ]
            instantiation of "void nvbench::benchmark<KernelGenerator, TypeAxes>::do_run() [with KernelGenerator=pairwise_linestring_distance_benchmark_line_140, TypeAxes=nvbench::type_list<floating_point_types>]" 
_deps/nvbench-src/nvbench/benchmark.cuh(65): here
            implicit generation of "nvbench::benchmark<KernelGenerator, TypeAxes>::~benchmark() noexcept [with KernelGenerator=pairwise_linestring_distance_benchmark_line_140, TypeAxes=nvbench::type_list<floating_point_types>]" 
_deps/nvbench-src/nvbench/benchmark.cuh(65): here
            instantiation of class "nvbench::benchmark<KernelGenerator, TypeAxes> [with KernelGenerator=pairwise_linestring_distance_benchmark_line_140, TypeAxes=nvbench::type_list<floating_point_types>]" 
_deps/nvbench-src/nvbench/benchmark.cuh(65): here
            instantiation of "nvbench::benchmark<KernelGenerator, TypeAxes>::benchmark() [with KernelGenerator=pairwise_linestring_distance_benchmark_line_140, TypeAxes=nvbench::type_list<floating_point_types>]" 
/home/ptaylor/dev/rapids/compose/etc/conda/cuda_11.6/envs/rapids/x86_64-conda-linux-gnu/include/c++/9.4.0/bits/unique_ptr.h(857): here
            instantiation of "std::_MakeUniq<_Tp>::__single_object std::make_unique<_Tp,_Args...>(_Args &&...) [with _Tp=nvbench::benchmark<pairwise_linestring_distance_benchmark_line_140, nvbench::type_list<floating_point_types>>, _Args=<>]" 
(140): here

1 error detected in the compilation of "/home/ptaylor/dev/rapids/cuspatial/cpp/benchmarks/pairwise_linestring_distance.cu".

[harrism]

Hmmm, I think the original code should work -- it's initializing the two elements using an implicitly generated constructor. Your change forces it to use a different implicitly generated constructor (initializer list constructor?)

[thomcom]

I had the same error, @trxcllnt just now building cuspatial from branch-22.10. Your fix however doesn't work for me, but ({}) does.

[jrhemstad]

The alternative is to add a constructor to vec_2d and import it in cartesian_2d

https://godbolt.org/z/Gh6v1zWKa

[trxcllnt]

@jrhemstad that didn't work for me, but I pushed a new fix that does seem to be working.

[harrism]

What we really want is just return prev + segment_length * rad; what if you change the auto const& parameters to explicit cartesian_2d<T>.

[trxcllnt]

@gpucibot merge

[trxcllnt]

@gpucibot merge

[harrism]

The original seems to work on the latest ToT NVCC so the bug may be already fixed. In any case, this workaround is not as offensive so approving.