Here are the Swift 2.3 and Swift 3.0 versions of the ColorCubeHelper class.

ColorCubeHelper-swift2.swift

//
//  ColorCubeHelper.swift
//
//  Created by Joshua Sullivan on 10/01/16.
//  Copyright © 2016 Joshua Sullivan. All rights reserved.
//

import UIKit
import Accelerate

public final class ColorCubeHelper {
    
    public enum ColorCubeError: ErrorType {
        case IncorrectImageSize
        case MissingImageData
        case UnableToCreateDataProvider
    }
    
    public static func createColorCubeData(inputImage image: UIImage, cubeDimension: Int) throws -> NSData {
        
        // Set up some variables for calculating memory size.
        let imageSize = image.size
        let dim = Int(imageSize.width)
        let pixels = dim * dim
        let channels = 4

        // If the number of pixels doesn't match what's needed for the supplied cube dimension, abort.
        guard pixels == cubeDimension * cubeDimension * cubeDimension else {
            throw ColorCubeError.IncorrectImageSize
        }
        
        // We don't need a sizeof() because uint_8t is explicitly 1 byte.
        let memSize = pixels * channels
        
        // Get the UIImage's backing CGImageRef
        guard let img = image.CGImage else {
            throw ColorCubeError.MissingImageData
        }
        
        // Get a reference to the CGImage's data provider.
        guard let inProvider = CGImageGetDataProvider(img) else {
            throw ColorCubeError.UnableToCreateDataProvider
        }
        
        let inBitmapData = CGDataProviderCopyData(inProvider)
        let inBuffer = CFDataGetBytePtr(inBitmapData)
        
        // Calculate the size of the float buffer and allocate it.
        let floatSize = memSize * sizeof(Float.self)
        let finalBuffer = unsafeBitCast(malloc(floatSize), UnsafeMutablePointer<Float>.self)
        
        // Convert the uint_8t to float. Note: a uint of 255 will convert to 255.0f.
        vDSP_vfltu8(inBuffer, 1, finalBuffer, 1, UInt(memSize))

        // Divide each float by 255.0 to get the 0-1 range we are looking for.
        var divisor = Float(255.0)
        vDSP_vsdiv(finalBuffer, 1, &divisor, finalBuffer, 1, UInt(memSize))
        
        // Don't copy the bytes, just have the NSData take ownership of the buffer.
        let cubeData = NSData(bytesNoCopy: finalBuffer, length: floatSize, freeWhenDone: true)
        
        return cubeData
        
    }
}

ColorCubeHelper-swift3.swift

//
//  ColorCubeHelper.swift
//
//  Created by Joshua Sullivan on 10/10/16.
//  Copyright © 2016 Joshua Sullivan. All rights reserved.
//

import UIKit
import Accelerate

public class ColorCubeHelper {
    
    public enum ColorCubeError: Error {
        case incorrectImageSize
        case missingImageData
        case unableToCreateDataProvider
        case unableToGetBitmpaDataBuffer
    }
    
    public static func createColorCubeData(inputImage image: UIImage, cubeDimension: Int) throws -> Data {
        
        // Set up some variables for calculating memory size.
        let imageSize = image.size
        let dim = Int(imageSize.width)
        let pixels = dim * dim
        let channels = 4
        
        // If the number of pixels doesn't match what's needed for the supplied cube dimension, abort.
        guard pixels == cubeDimension * cubeDimension * cubeDimension else {
            throw ColorCubeError.incorrectImageSize
        }
        
        // We don't need a sizeof() because uint_8t is explicitly 1 byte.
        let memSize = pixels * channels
        
        // Get the UIImage's backing CGImageRef
        guard let img = image.cgImage else {
            throw ColorCubeError.missingImageData
        }
        
        // Get a reference to the CGImage's data provider.
        guard let inProvider = img.dataProvider else {
            throw ColorCubeError.unableToCreateDataProvider
        }
        
        let inBitmapData = inProvider.data
        guard let inBuffer = CFDataGetBytePtr(inBitmapData) else {
            throw ColorCubeError.unableToGetBitmpaDataBuffer
        }
        
        // Calculate the size of the float buffer and allocate it.
        let floatSize = memSize * MemoryLayout<Float>.size
        let finalBuffer = unsafeBitCast(malloc(floatSize), to:UnsafeMutablePointer<Float>.self)
        
        // Convert the uint_8t to float. Note: a uint of 255 will convert to 255.0f.
        vDSP_vfltu8(inBuffer, 1, finalBuffer, 1, UInt(memSize))
        
        // Divide each float by 255.0 to get the 0-1 range we are looking for.
        var divisor = Float(255.0)
        vDSP_vsdiv(finalBuffer, 1, &divisor, finalBuffer, 1, UInt(memSize))
        
        // Don't copy the bytes, just have the NSData take ownership of the buffer.
        let cubeData = NSData(bytesNoCopy: finalBuffer, length: floatSize, freeWhenDone: true)
        
        return cubeData as Data
        
    }
}

in_use.swift

guard let colorCubeImage = UIImage(named: "MyColorEffect-64.png") else {
  preconditionFailure("Couldn't load the cube data image.")
}
do {
    let data = try ColorCubeHelper.createColorCubeData(inputImage: colorCubeImage, cubeDimension: 64)
    let params: [String : Any] = [ "inputCubeData" : data ]
    let filter = CIFilter(name: "CIColorCube", withInputParameters: params)
    // Now you can use the filter.
} catch {
    debugPrint("Unable to create cube data: \(error)")
}

@JoshuaSullivan
I have another question.
Your code uses CIColorCube, but we have also CIColorCubeWithColorSpace.
Do you know the differences between them?

I'm working on this muukii/Brightroom#78

@JoshuaSullivan
I have another question.
Your code uses CIColorCube, but we have also CIColorCubeWithColorSpace.
Do you know the differences between them?

I don't know the technical differences between them, but when I use CIColorCubeWithColorSpace the results are the expected and when I use CIColorCube they aren't.