ReduceSubarraysSum.h

Go to the documentation of this file.

// ----------------------------------------------------------------------------
// -                        CloudViewer: www.cloudViewer.org                  -
// ----------------------------------------------------------------------------
// Copyright (c) 2018-2024 www.cloudViewer.org
// SPDX-License-Identifier: MIT
// ----------------------------------------------------------------------------
 
#pragma once
 
#include <tbb/parallel_for.h>
 
namespace cloudViewer {
namespace ml {
namespace impl {
 
template <class T>
void ReduceSubarraysSumCPU(const T* const values,
                           const size_t values_size,
                           const int64_t* const row_splits,
                           const size_t num_arrays,
                           T* out_sums) {
    tbb::parallel_for(tbb::blocked_range<size_t>(0, num_arrays),
                      [&](const tbb::blocked_range<size_t>& r) {
                          for (size_t i = r.begin(); i != r.end(); ++i) {
                              size_t begin_idx = row_splits[i];
                              size_t end_idx = row_splits[i + 1];
 
                              T sum = T(0);
                              for (size_t j = begin_idx; j < end_idx; ++j) {
                                  sum += values[j];
                              }
                              out_sums[i] = sum;
                          }
                      });
}
 
}  // namespace impl
}  // namespace ml
}  // namespace cloudViewer