What's wrong with my multi-threaded Matrix Mult. code? 1 thread is fastest

Question

For some reason, 1 thread performs the best. What's wrong with my implementation?
import java.util.Random;
import java.util.Date;
import java.util.concurrent.ExecutorService;
import java.util.concurrent.Executors;
import java.util.concurrent.TimeUnit;
public class Matrix {     
     private int values[][];
     private int rows;
     private int columns;
     public Matrix(int r, int c) {
          this.rows = r;
          this.columns = c;
          this.values = new int[r][c];
     private void randomize() {
          Random generator = new Random();
          for (int r = 0; r < this.rows; r++) {
               for (int c = 0; c < this.columns; c++) {
                    this.values[r][c] = generator.nextInt(10);
     public String toString() {
          String out = "";
          for (int r = 0; r < this.rows; r++) {
               for (int c = 0; c < this.columns; c++) {
                    if (c == 0) out += "[";
                    else out += "\t";
                    out += this.values[r][c];
               out += "]\n";
          return out;
     public boolean equals(Object obj) {
          Matrix other = (Matrix) obj;
          if (this.columns != other.columns || this.rows != other.rows)  {
               return false;
          for (int r = 0; r < this.rows; r++) {
               for (int c = 0; c < this.columns; c++) {
                    if (this.values[r][c] != other.values[r][c]) {
                         return false;
          return true;
     // matrix multiplication using single thread
     public Matrix times(Matrix other) {
          assert(this.columns == other.rows);
          Matrix out = new Matrix(this.rows, other.columns);
          for (int r = 0; r < this.rows; r++) {
               for (int c = 0; c < other.columns; c++) {
                    int dotProduct = 0;
                    for (int z = 0; z < this.columns; z++) {
                         dotProduct += this.values[r][z] * other.values[z][c];
                    out.values[r][c] = dotProduct;
          return out;
     // matrix multiplication with many threads
     public Matrix ptimes(Matrix other, int numberOfThreads) throws InterruptedException { // parallel
          assert(this.columns == other.rows);
          Matrix out = new Matrix(this.rows, other.columns);
          ExecutorService threadExecutor = Executors.newFixedThreadPool(numberOfThreads);
          for (int r = 0; r < this.rows; r++) {
               for (int c = 0; c < other.columns; c++) {
                    threadExecutor.execute(new HelperThread(r, c, this, other, out));
          threadExecutor.shutdown();
          threadExecutor.awaitTermination(2, TimeUnit.DAYS);
          return out;
     private class HelperThread implements Runnable {
          private int row;
          private int col;
          private Matrix a;
          private Matrix b;
          private Matrix out;
          HelperThread(int r, int c, Matrix a, Matrix b, Matrix o) {
               this.row = r;
               this.col = c;
               this.a = a;
               this.b = b;
               this.out = o;
          public void run() {
               int dotProduct = 0;
               for (int z = 0; z < a.columns; z++) {
                    dotProduct += this.a.values[row][z] * this.b.values[z][col];
               this.out.values[row][col] = dotProduct;
     public static void main(String[] args) throws InterruptedException {
          int size = 100;
          Matrix a = new Matrix(size, size);
          a.randomize();     
          Matrix b = new Matrix(size, size);
          b.randomize();
          for (int t = 1; t < 15; t++) {
               long start = new Date().getTime();
               System.out.print(t + " threads: ");
               Matrix c = a.ptimes(b, t);
               //System.out.println(c);
               long finish = new Date().getTime();
               System.out.println((finish - start) + " milliseconds");
               Matrix d = a.times(b);
               assert(c.equals(d));
}

Answer 1

This one is even faster. On my dual core I get:
Warmup
Single Threaded
5.20616 milliseconds
5.52872 milliseconds
5.12708 milliseconds
5.59048 milliseconds
5.16104 milliseconds
5.1838 milliseconds
5.37104 milliseconds
5.1788 milliseconds
5.18636 milliseconds
5.15736 milliseconds
Multi Threaded with 2 threads
3.22184 milliseconds
2.86552 milliseconds
2.86284 milliseconds
3.67032 milliseconds
3.08032 milliseconds
2.97388 milliseconds
2.93084 milliseconds
3.44012 milliseconds
2.89744 milliseconds
2.88136 milliseconds
As you can see the Multi-Threaded versions are now faster.
    // matrix multiplication with many threads
    ExecutorService threadExecutor = Executors.newFixedThreadPool(Runtime.getRuntime().availableProcessors());
    public Matrix ptimes(Matrix other) throws InterruptedException, ExecutionException {
        assert (this.columns == other.rows);
        Matrix out = new Matrix(this.rows, other.columns);
        Future futures[] = new Future[rows];
        for (int r = 0; r < this.rows; r++) {
            futures[r] = threadExecutor.submit(new HelperThread(r, this, other, out));
        for(Future f : futures) {
            f.get();
        return out;
    private class HelperThread implements Callable<Object> {
        private int row;
        private Matrix a;
        private Matrix b;
        private Matrix out;
        HelperThread(int r, Matrix a, Matrix b, Matrix o) {
            this.row = r;
            this.a = a;
            this.b = b;
            this.out = o;
        public String call() throws Exception {
            int dotProduct = 0;
            for (int c = 0; c < b.columns; c++) {
                for (int z = 0; z < a.columns; z++) {
                    dotProduct += this.a.values[row][z] * this.b.values[z][c];
                this.out.values[row][c] = dotProduct;
            return null;
    public static void main(String[] args) throws InterruptedException, ExecutionException {
        int size = 100;
        Matrix a = new Matrix(size, size);
        a.randomize();
        Matrix b = new Matrix(size, size);
        b.randomize();
        System.out.println("Warmup");
        for (int t = 0; t < 1000; t++) {
            Matrix c = a.ptimes(b);
            Matrix d = a.times(b);
            assert (c.equals(d));
        System.out.println("Single Threaded");
        for (int t = 0; t < 10; t++) {
            long start = System.nanoTime();
            Matrix d = a.times(b);
            long finish = System.nanoTime();
            System.out.println((finish - start)/1000000.0 + " milliseconds");
        System.out.println("Multi Threaded with " + Runtime.getRuntime().availableProcessors() + " threads");
        for (int t = 0; t < 10; t++) {
            long start = System.nanoTime();
            Matrix c = a.ptimes(b);
            long finish = System.nanoTime();
            System.out.println((finish - start)/1000000.0 + " milliseconds");
            Matrix d = a.times(b);
            assert (c.equals(d));
        System.exit(0);
    }