Moving average with arrayfire

moving_avg.cpp

/*******************************************************
 * Copyright (c) 2014, ArrayFire
 * All rights reserved.
 *
 * This file is distributed under 3-clause BSD license.
 * The complete license agreement can be obtained at:
 * http://arrayfire.com/licenses/BSD-3-Clause
 ********************************************************/

#include <iostream>
#include <limits>
#include <arrayfire.h>
#include <vector>

using namespace std;
using namespace af;

array moving_average(const array &input, const vector<int> w, const int max_size)
{
    // Create weight matrix of all zeros
    array weights = constant(0, max_size, w.size());

    // Can not use gfor here
    for (int ii = 0; ii < w.size(); ii++) {
        int w_len = w[ii];

        // Calculate the norm factor for current window length
        float norm = 1.0 / (float)(w_len);

        // Only make the first "len" elements equal to the norm

        weights(seq(1, w_len), ii) = constant(norm, w_len, 1);
    }

    // Calculate the moving averages in parallel
    // This is the most compute intensive part
    array res = convolve1(input, weights);
    return res;
}

int main()
{
    af::deviceset(0);
    af::info();

    array a = range(10, 1);
    // 3 window sizes
    std::vector<int> w(3);

    // Set window sizes
    w[0] = 2;
    w[1] = 3;
    w[2] = 4;

    int max_size = 5;

    array b = moving_average(a, w, max_size);

    af_print(a);
    af_print(b);
}

results

$ ./examples/helloworld/helloworld_cuda
ArrayFire v3.0.beta (CUDA, 64-bit Linux, build 102bf9b)
Platform: CUDA Toolkit 7, Driver: 346.47
[0] GeForce GTX 690, 2048 MB, CUDA Compute 3.0
-1- GeForce GTX 690, 2048 MB, CUDA Compute 3.0
a [10 1 1 1]
    0.0000
    1.0000
    2.0000
    3.0000
    4.0000
    5.0000
    6.0000
    7.0000
    8.0000
    9.0000

b [10 3 1 1]
    0.5000     0.3333     0.2500
    1.5000     1.0000     0.7500
    2.5000     2.0000     1.5000
    3.5000     3.0000     2.5000
    4.5000     4.0000     3.5000
    5.5000     5.0000     4.5000
    6.5000     6.0000     5.5000
    7.5000     7.0000     6.5000
    8.5000     8.0000     7.5000
    4.5000     5.6667     6.0000