oneMKL Summary Statistics Usage Model#


A typical algorithm for summary statistics is as follows:

  1. Create and initialize an object for dataset.

  2. Call the summary statistics routine to calculate the appropriate estimate.

The following example demonstrates how to calculate mean values for a 3-dimensional dataset filled with random numbers. For dataset creation, the make_dataset helper function is used.

USM-based example#

#include "oneapi/mkl/stats.hpp"

int main() {
    sycl::queue queue;

    constexpr std::size_t n_observations = 1000;
    constexpr std::size_t n_dims = 3;

    // allocate Unified Shared Memory for the dataset of the size n_observations * n_dims and fill it with any data
    // allocate Unified Shared Memory for the mean output of the size n_dims

    // create oneapi::mkl::stats::dataset
    auto dataset = oneapi::mkl::stats::make_dataset<oneapi::mkl::stats::layout::row_major>(n_dims, n_observations, dataset_ptr);

    // call statistics computation routine
    auto event = oneapi::mkl::stats::mean(queue, dataset, mean_ptr);

    // wait until computations are completed

    // ...

Parent topic: Summary Statistics