// ----------------------------------------------------------------------------
// -                        Open3D: www.open3d.org                            -
// ----------------------------------------------------------------------------
// Copyright (c) 2018-2023 www.open3d.org
// SPDX-License-Identifier: MIT
// ----------------------------------------------------------------------------
//

#define EIGEN_USE_GPU
#include "FixedRadiusSearchOpKernel.h"
#include "open3d/core/CUDAUtils.h"
#include "open3d/core/nns/FixedRadiusSearchImpl.cuh"

using namespace open3d;
using namespace fixed_radius_search_opkernel;
using namespace tensorflow;

template <class T, class TIndex>
class FixedRadiusSearchOpKernelCUDA : public FixedRadiusSearchOpKernel {
public:
    explicit FixedRadiusSearchOpKernelCUDA(OpKernelConstruction* construction)
        : FixedRadiusSearchOpKernel(construction) {
        texture_alignment =
                open3d::core::GetCUDACurrentDeviceTextureAlignment();
    }

    void Kernel(tensorflow::OpKernelContext* context,
                const tensorflow::Tensor& points,
                const tensorflow::Tensor& queries,
                const tensorflow::Tensor& radius,
                const tensorflow::Tensor& points_row_splits,
                const tensorflow::Tensor& queries_row_splits,
                const tensorflow::Tensor& hash_table_splits,
                const tensorflow::Tensor& hash_table_index,
                const tensorflow::Tensor& hash_table_cell_splits,
                tensorflow::Tensor& query_neighbors_row_splits) {
        auto device = context->eigen_gpu_device();

        OutputAllocator<T, TIndex> output_allocator(context);

        void* temp_ptr = nullptr;
        size_t temp_size = 0;

        // determine temp_size
        open3d::core::nns::impl::FixedRadiusSearchCUDA<T, TIndex>(
                device.stream(), temp_ptr, temp_size, texture_alignment,
                (int64_t*)query_neighbors_row_splits.flat<int64>().data(),
                points.shape().dim_size(0), points.flat<T>().data(),
                queries.shape().dim_size(0), queries.flat<T>().data(),
                radius.scalar<T>()(), points_row_splits.shape().dim_size(0),
                (int64_t*)points_row_splits.flat<int64>().data(),
                queries_row_splits.shape().dim_size(0),
                (int64_t*)queries_row_splits.flat<int64>().data(),
                hash_table_splits.flat<uint32_t>().data(),
                hash_table_cell_splits.shape().dim_size(0),
                hash_table_cell_splits.flat<uint32_t>().data(),
                hash_table_index.flat<uint32_t>().data(), metric,
                ignore_query_point, return_distances, output_allocator);

        Tensor temp_tensor;
        TensorShape temp_shape({ssize_t(temp_size)});
        OP_REQUIRES_OK(context,
                       context->allocate_temp(DataTypeToEnum<uint8_t>::v(),
                                              temp_shape, &temp_tensor));
        temp_ptr = temp_tensor.flat<uint8_t>().data();

        // actually run the search
        open3d::core::nns::impl::FixedRadiusSearchCUDA<T, TIndex>(
                device.stream(), temp_ptr, temp_size, texture_alignment,
                (int64_t*)query_neighbors_row_splits.flat<int64>().data(),
                points.shape().dim_size(0), points.flat<T>().data(),
                queries.shape().dim_size(0), queries.flat<T>().data(),
                radius.scalar<T>()(), points_row_splits.shape().dim_size(0),
                (int64_t*)points_row_splits.flat<int64>().data(),
                queries_row_splits.shape().dim_size(0),
                (int64_t*)queries_row_splits.flat<int64>().data(),
                hash_table_splits.flat<uint32_t>().data(),
                hash_table_cell_splits.shape().dim_size(0),
                hash_table_cell_splits.flat<uint32_t>().data(),
                hash_table_index.flat<uint32_t>().data(), metric,
                ignore_query_point, return_distances, output_allocator);
    }

private:
    int texture_alignment;
};

#define REG_KB(type, itype)                                               \
    REGISTER_KERNEL_BUILDER(Name("Open3DFixedRadiusSearch")               \
                                    .Device(DEVICE_GPU)                   \
                                    .TypeConstraint<type>("T")            \
                                    .TypeConstraint<itype>("index_dtype") \
                                    .HostMemory("radius")                 \
                                    .HostMemory("points_row_splits")      \
                                    .HostMemory("queries_row_splits")     \
                                    .HostMemory("hash_table_splits"),     \
                            FixedRadiusSearchOpKernelCUDA<type, itype>);
REG_KB(float, int)
REG_KB(float, long)
#undef REG_KB