thrust::inclusive_scan_by_key

Defined in thrust/scan.h

template<typename DerivedPolicy, typename InputIterator1, typename InputIterator2, typename OutputIterator> OutputIterator thrust::inclusive_scan_by_key(const thrust::detail::execution_policy_base<DerivedPolicy> &exec, InputIterator1 first1, InputIterator1 last1, InputIterator2 first2, OutputIterator result)

inclusive_scan_by_key computes an inclusive key-value or ‘segmented’ prefix sum operation. The term ‘inclusive’ means that each result includes the corresponding input operand in the partial sum. The term ‘segmented’ means that the partial sums are broken into distinct segments. In other words, within each segment a separate inclusive scan operation is computed. Refer to the code sample below for example usage.

This version of inclusive_scan_by_key assumes equal_to as the binary predicate used to compare adjacent keys. Specifically, consecutive iterators i and i+1 in the range [first1, last1) belong to the same segment if *i == *(i+1), and belong to different segments otherwise.

This version of inclusive_scan_by_key assumes plus as the associative operator used to perform the prefix sum. When the input and output sequences are the same, the scan is performed in-place.

Results are not deterministic for pseudo-associative operators (e.g., addition of floating-point types). Results for pseudo-associative operators may vary from run to run.

The algorithm’s execution is parallelized as determined by exec.

The following code snippet demonstrates how to use inclusive_scan_by_key using the thrust::host execution policy for parallelization:

#include <thrust/scan.h>
#include <thrust/execution_policy.h>
...

int data[10] = {1, 1, 1, 1, 1, 1, 1, 1, 1, 1};
int keys[10] = {0, 0, 0, 1, 1, 2, 3, 3, 3, 3};

thrust::inclusive_scan_by_key(thrust::host, keys, keys + 10, data, data); // in-place scan

// data is now {1, 2, 3, 1, 2, 1, 1, 2, 3, 4};