Slices

credit for this assignment: Finn Voichick and Dennis Cosgrove

Motivation

Coarsening, or n-way split as we tend to call it in this course, comes up a fair amount. In this studio, we will create our Slices class, that allows us to split up our data n-times (depending on what value we pass in) in a consistent organized fashion. This makes sure that each thread is balanced in the work that it does, minimizing the Critical Path Length.

Since we will be utilizing this class throughout the semester in different studios and assignments, it is important that data is split up in a specific way. The tests should help make sure your solution is in working order.

Mistakes To Avoid

Warning: Do NOT Parallelize

This studio is about creating an easier way to split data so that it can be worked on in parallel. The splitting of the data is still done sequentially. That means at no point in this studio should you be using async() or finish().

Warning: Do NOT Copy The Data Into SubArrays

Our "Slices" will actually be a list of objects called "Slice". A Slice contains a couple things, one of them being the full, original data that is passed through at the start. It is important that the entire array is given to the Slice. Not a copy of the array, nor a shortened version of the array, but the entire thing.

Code to Implement

class:	Slices.java
methods:	createNSlicesForArrayObject
package:	slices.studio
source folder:	student/src/main/java

createNSlicesForArrayObject

method: private static <T> List<Slice<T>> createNSlicesForArrayObject(A data, int numSlices) (sequential implementation only)

Given an Array (A) and a number of slices to make, return a List of object Slice.

In order to support primitive arrays (e.g. byte[], int[], et cetera) and non-primitive arrays (e.g. Object[]) we provide all of the public methods, each which simply call the single method requiring implementation:

	public static <T> List<Slice<T[]>> createNSlices(T[] data, int numSlices) {
		return createNSlicesForArrayObject(data, numSlices);
	}
	public static List<Slice<byte[]>> createNSlices(byte[] data, int numSlices) {
		return createNSlicesForArrayObject(data, numSlices);
	}
	public static List<Slice<char[]>> createNSlices(char[] data, int numSlices) {
		return createNSlicesForArrayObject(data, numSlices);
	}
	public static List<Slice<short[]>> createNSlices(short[] data, int numSlices) {
		return createNSlicesForArrayObject(data, numSlices);
	}
	public static List<Slice<int[]>> createNSlices(int[] data, int numSlices) {
		return createNSlicesForArrayObject(data, numSlices);
	}
	public static List<Slice<long[]>> createNSlices(long[] data, int numSlices) {
		return createNSlicesForArrayObject(data, numSlices);
	}
	public static List<Slice<float[]>> createNSlices(float[] data, int numSlices) {
		return createNSlicesForArrayObject(data, numSlices);
	}
	public static List<Slice<double[]>> createNSlices(double[] data, int numSlices) {
		return createNSlicesForArrayObject(data, numSlices);
	}

Tip: To pull this off we will need to use the Array.getLength(Object) method.

Strict Specification

We are overly strict about the specification of how the data must be sliced up. This is to allow to accurately compare results intermediate results throughout the semester.

Example: array.length=7; numSlices=4

Distribute the remainder 1 each to the lower end slices (the first few slices or the lower index slices).

A

A

B

B

C

C

D

Warning:Do NOT slice up the data by giving all of the remainder to one slice

A |B |C |D |D |D |D

class Slice<T>

use: class Slice

This class has everything you need for n-way split problems, specifically: getSliceIndexId(), getMinInclusive(), and getMaxExclusive().

getOriginalUnslicedData() is not necessary until the K-Mer Counting Lab

Testing Your Solution

Correctness

class:	SlicesTestSuite.java
package:	slice.studio
source folder:	testing/src/test/java