slidingWindowSum(_:usingWindowLength:)

Returns the single-precision sliding window sum of a vector.

Declaration

static func slidingWindowSum<U>(_ vector: U, usingWindowLength windowLength: Int) -> [Float] where U : AccelerateBuffer, U.Element == Float

Parameters

vector:
The input vector.
windowLength:
The length of the window, which must be greater than zero.

Discussion

Use this function to populate an output vector with the consecutive sums of values in a sliding window. The number of elements in the input vector must be one subtracted from the sum of the window length and the number of window positions.

The number of elements in the output vector must be the same as the number of elements in the input vector. This is because although the function only writes the number of window positions to the output vector, it may require additional elements for intermediate computation.

The following code shows how to compute the sums of values in a three-element sliding window moved through a ten-element array. The code processes every element in the input array, and the output result contains eight values.

let input: [Float] = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10]

var output = [Float](repeating: .nan,
                     count: input.count)

let windowLength = vDSP_Length(3)
let outputCount = vDSP_Length(input.count) - windowLength + 1
let stride = vDSP_Stride(1)

vDSP_vswsum(input, stride,
            &output, stride,
            outputCount,
            windowLength)

// Prints "[6.0, 9.0, 12.0, 15.0, 18.0, 21.0, 24.0, 27.0]".
print(output[0 ..< Int(outputCount)])

The figure below illustrates the process the function uses to calculate the result:

[Image]

Alternatively, the following code shows how to process input elements to generate a specific number of output elements. Use this technique, for example, to process streaming input data. The code below takes count elements from the input collection and returns the sums at eight window positions:

let input: [Float] = [1, 2, 3, 4, 5, 6, 7, 8, 9, 10,
                      11, 12, 13, 14, 15, 16, 17, 18, 19, 20,
                      21, 22, 23, 24, 25, 26, 27, 28, 29, 30]
let inputPageNumber = 1

let windowLength = vDSP_Length(3)
let outputCount = vDSP_Length(8)

let count = Int(outputCount + windowLength - 1)

var output = [Float](repeating: .nan,
                     count: count)

let stride = vDSP_Stride(1)

input.withUnsafeBufferPointer { srcPtr in
    vDSP_vswsum((srcPtr.baseAddress!.advanced(by: count * inputPageNumber)),
                stride,
                &output, stride,
                outputCount,
                windowLength)
}

// Prints "[36.0, 39.0, 42.0, 45.0, 48.0, 51.0, 54.0, 57.0]"
print(output[0 ..< Int(outputCount)])