Bad access exception when using vDSP

Question

I am trying to perform a FFT in swift using vDSP but when I call the absolute function I get a bad access exception: Thread 3: EXC_BAD_ACCESS (code=EXC_I386_GPFLT). I suspect that I have made some error in managing the memory when using unsafe pointers, however I cannot figure out what it is.

static func test(signal: [Float]) -> [Float] {
    let length = vDSP_Length(signal.count)
    let log2n = vDSP_Length(ceil(log2(Float(length * 2))))
    let fftSetup = vDSP.FFT(log2n: log2n, radix: .radix2, ofType: DSPSplitComplex.self)!
    
    var forwardInputReal = [Float](signal) // Copy the signal here
    var forwardInputImag = [Float](repeating: 0, count: Int(length))
    var forwardOutputReal = [Float](repeating: 0, count: Int(length))
    var forwardOutputImag = [Float](repeating: 0, count: Int(length))
    
    var magnitudes = [Float](repeating: 0, count: Int(signal.count))
    forwardInputReal.withUnsafeMutableBufferPointer { forwardInputRealPtr in
        forwardInputImag.withUnsafeMutableBufferPointer { forwardInputImagPtr in
            forwardOutputReal.withUnsafeMutableBufferPointer { forwardOutputRealPtr in
                forwardOutputImag.withUnsafeMutableBufferPointer { forwardOutputImagPtr in
                    // Input
                    let forwardInput = DSPSplitComplex(realp: forwardInputRealPtr.baseAddress!, imagp: forwardInputImagPtr.baseAddress!)
                    // Output
                    var forwardOutput = DSPSplitComplex(realp: forwardOutputRealPtr.baseAddress!, imagp: forwardOutputImagPtr.baseAddress!)

                    fftSetup.forward(input: forwardInput, output: &forwardOutput)
                    vDSP.absolute(forwardOutput, result: &magnitudes)
                }
            }
        }
    }
    
    return magnitudes
}

Answer 1

The same exception happened in my case when initializing a DCT transformer.

let setup = vDSP.DCT(count: windowSize, transformType: vDSP.DCTTransformType.II)

It looks like the library uses an FFT algorithm that requires the window size to be a power of two. Once I met this requirement and padded my data accordingly, the problem went away.