how to implement multithreaded Breadth-first search in java? - java

I've done the breadth-first search in a normal way.
now I'm trying to do it in a multithreaded way.
i have one queue which is shared between the threads.
i use synchronize(LockObject) when i remove a node from the queue ( FIFI queue )
so what I'm trying to do is that.
when i thread finds a solution all the other threads will stop immediately.

i assume you are traversing a tree for your BFS.
create a thread pool.
for each unexplored children in the node, retrieve a thread from the thread pool (perhaps using a Semaphore). mark the child node as 'explored' and explore the node's children in a BFS manner. when you have found a solution or done exploring all the nodes, release the semaphore.
^ i've never done this before so i might have missed out something.

Assuming you want to do this iteratively (see note at the bottom why there may be better closed solutions), this is not a great problem for exercising multi threading. The problem is that multithreading is great if you don't depend on previous results, but here you want the minimum amount of coins.
As you point out, a breadth first solution guarantees that once you reach the desired amount, you won't have any further solutions with less coins in a single threaded environment. However, in a multithreaded environment, once you start calculating a solution, you cannot guarantee that it will finish before some other solution. Let's imagine for the value 21: it can be a 20c coin and a 1c or four 5c coins and a 1c; if both are calculating simultaneously, you cannot guarantee that the first (and correct) solution will finish first. In practice, it is unlikely the situation will happen, but when you work with multithreading you want the solution to work in theory, because multithreads always fail in the demonstration, no matter if they should not have failed until the death heat of the universe.
Now you have 2 possible solutions: one is to introduce choke points at the beginning of each level; you don't start that level until the previous level is finished. The other is once you reach a solution continue doing all the calculations with a lower level than the current result (which means you cannot purge the others). Probably with all the synchronization needed you get better performance by going single threaded, but let's go on.
For the first solution, the natural form is to iterate increasing the level. You can use the solution provided by happymeal, with a Semaphore. An alternative is to use the new classes provided by java.
CoinSet getCoinSet(int desiredAmount) throws InterruptedException {
// Use whatever number of threads you prefer or another member of Executors.
final ExecutorService executor = Executors.newFixedThreadPool(10);
ResultContainer container = new ResultContainer();
container.getNext().add(new Producer(desiredAmount, new CoinSet(), container));
while (container.getResult() == null) {
executor.invokeAll(container.setNext(new Vector<Producer>()));
}
return container.getResult();
}
public class Producer implements Callable<CoinSet> {
private final int desiredAmount;
private final CoinSet data;
private final ResultContainer container;
public Producer(int desiredAmount, CoinSet data, ResultContainer container) {
this.desiredAmount = desiredAmount;
this.data = data;
this.container = container;
}
public CoinSet call() {
if (data.getSum() == desiredAmount) {
container.setResult(data);
return data;
} else {
Collection<CoinSet> nextSets = data.addCoins();
for (CoinSet nextSet : nextSets) {
container.getNext().add(new Producer(desiredAmount, nextSet, container));
}
return null;
}
}
}
// Probably it is better to split this class, but you then need to pass too many parameters
// The only really needed part is to create a wrapper around getNext, since invokeAll is
// undefined if you modify the list of tasks.
public class ResultContainer {
// I use Vector because it is synchronized.
private Vector<Producer> next = new Vector<Producer>();
private CoinSet result = null;
// Note I return the existing value.
public Vector<Producer> setNext(Vector<Producer> newValue) {
Vector<Producer> current = next;
next = newValue;
return current;
}
public Vector<Producer> getNext() {
return next;
}
public synchronized void setResult(CoinSet newValue) {
result = newValue;
}
public synchronized CoinSet getResult() {
return result;
}
}
This still has the problem that existing tasks are executed; however, it is simple to fix that; pass the thread executor into each Producer (or the container). Then, when you find a result, call executor.shutdownNow. The threads that are executing won't be interrupted, but the operation in each thread is trivial so it will finish fast; the runnables that have not started won't start.
The second option means you have to let all the current tasks finish, unless you keep track of how many tasks you have run at each level. You no longer need to keep track of the levels, though, and you don't need the while cycle. Instead, you just call
executor.submit(new Producer(new CoinSet(), desiredAmount, container)).get();
And then, the call method is pretty similar (assume you have executor in the Producer):
public CoinSet call() {
if (container.getResult() != null && data.getCount() < container.getResult().getCount()) {
if (data.getSum() == desiredAmount)) {
container.setResult(data);
return data;
} else {
Collection<CoinSet> nextSets = data.addCoins();
for (CoinSet nextSet : nextSets) {
executor.submit(new Producer(desiredAmount, nextSet, container));
}
return null;
}
}
}
and you also have to modify container.setResult, since you cannot depend that between the if and setting the value it has not been set by some other threads (threads are really annoying, aren't they?)
public synchronized void setResult(CoinSet newValue) {
if (newValue.getCount() < result.getCount()) {
result = newValue;
}
}
In all previous answers, CoinSet.getSum() returns the sum of the coins in the set, CoinSet.getCount() returns the number of coins, and CoinSet.addCoins() returns a Collection of CoinSet in which each element is the current CoinSet plus one coin of each possible different value
Note: For the problem of the coins with the values 1, 5, 10 and 20, the simplest solution is take the amount and divide it by the largest coin. Then take the modulus of that and use the next largest value and so on. That is the minimum amount of coins you are going to need. This rule applies (AFAICT) when the following property if true: if for all consecutive pairs of coin values (i.e. in this case, 1-5, 5-10, 10-20) you can reach any int multiple of the lower element in the pair with with a smaller number of coins using the larger element and whatever coins are necessary. You only need to prove it to the min common multiple of both elements in the pair (after that it repeats itself)

I gather from your comment on happymeal's anwer that you are trying to find how to reach a specific amount of money by adding coins of 1c, 5c, 10c and 20c.
Since each coin denomination divides the denomination of the next bigger coin, this can be solved in constant time as follows:
int[] coinCount(int amount) {
int[] coinValue = {20, 10, 5, 1};
int[] coinCount = new int[coinValue.length];
for (int i = 0; i < coinValue.length; i++) {
coinCount[i] = amount / coinValue[i];
amount -= coinCount[i] * coinValue[i];
}
return coinCount;
}
Take home message: Try to optimize your algorithm before resorting to multithreading, because algorithmic improvements can yield much greater improvements.

I've successfully implemented it.
what i did is that i took all the nodes in the first level, let's say 4 nodes.
then i had 2 threads. each one takes 2 nodes and generate their children. whenever a node finds a solution it has to report the level that it found the solution in and limit the searching level so other threads don't exceed the level.
only the reporting method should be synchronized.
i did the code for the coins change problem. this is my code for others to use
Main Class (CoinsProblemBFS.java)
package coinsproblembfs;
import java.util.ArrayList;
import java.util.LinkedList;
import java.util.List;
import java.util.Queue;
import java.util.Scanner;
/**
*
* #author Kassem M. Bagher
*/
public class CoinsProblemBFS
{
private static List<Item> MoneyList = new ArrayList<Item>();
private static Queue<Item> q = new LinkedList<Item>();
private static LinkedList<Item> tmpQ;
public static Object lockLevelLimitation = new Object();
public static int searchLevelLimit = 1000;
public static Item lastFoundNode = null;
private static int numberOfThreads = 2;
private static void InitializeQueu(Item Root)
{
for (int x = 0; x < MoneyList.size(); x++)
{
Item t = new Item();
t.value = MoneyList.get(x).value;
t.Totalvalue = MoneyList.get(x).Totalvalue;
t.Title = MoneyList.get(x).Title;
t.parent = Root;
t.level = 1;
q.add(t);
}
}
private static int[] calculateQueueLimit(int numberOfItems, int numberOfThreads)
{
int total = 0;
int[] queueLimit = new int[numberOfThreads];
for (int x = 0; x < numberOfItems; x++)
{
if (total < numberOfItems)
{
queueLimit[x % numberOfThreads] += 1;
total++;
}
else
{
break;
}
}
return queueLimit;
}
private static void initializeMoneyList(int numberOfItems, Item Root)
{
for (int x = 0; x < numberOfItems; x++)
{
Scanner input = new Scanner(System.in);
Item t = new Item();
System.out.print("Enter the Title and Value for item " + (x + 1) + ": ");
String tmp = input.nextLine();
t.Title = tmp.split(" ")[0];
t.value = Double.parseDouble(tmp.split(" ")[1]);
t.Totalvalue = t.value;
t.parent = Root;
MoneyList.add(t);
}
}
private static void printPath(Item item)
{
System.out.println("\nSolution Found in Thread:" + item.winnerThreadName + "\nExecution Time: " + item.searchTime + " ms, " + (item.searchTime / 1000) + " s");
while (item != null)
{
for (Item listItem : MoneyList)
{
if (listItem.Title.equals(item.Title))
{
listItem.counter++;
}
}
item = item.parent;
}
for (Item listItem : MoneyList)
{
System.out.println(listItem.Title + " x " + listItem.counter);
}
}
public static void main(String[] args) throws InterruptedException
{
Item Root = new Item();
Root.Title = "Root Node";
Scanner input = new Scanner(System.in);
System.out.print("Number of Items: ");
int numberOfItems = input.nextInt();
input.nextLine();
initializeMoneyList(numberOfItems, Root);
System.out.print("Enter the Amount of Money: ");
double searchValue = input.nextDouble();
int searchLimit = (int) Math.ceil((searchValue / MoneyList.get(MoneyList.size() - 1).value));
System.out.print("Number of Threads (Muste be less than the number of items): ");
numberOfThreads = input.nextInt();
if (numberOfThreads > numberOfItems)
{
System.exit(1);
}
InitializeQueu(Root);
int[] queueLimit = calculateQueueLimit(numberOfItems, numberOfThreads);
List<Thread> threadList = new ArrayList<Thread>();
for (int x = 0; x < numberOfThreads; x++)
{
tmpQ = new LinkedList<Item>();
for (int y = 0; y < queueLimit[x]; y++)
{
tmpQ.add(q.remove());
}
BFS tmpThreadObject = new BFS(MoneyList, searchValue, tmpQ);
Thread t = new Thread(tmpThreadObject);
t.setName((x + 1) + "");
threadList.add(t);
}
for (Thread t : threadList)
{
t.start();
}
boolean finish = false;
while (!finish)
{
Thread.sleep(250);
for (Thread t : threadList)
{
if (t.isAlive())
{
finish = false;
break;
}
else
{
finish = true;
}
}
}
printPath(lastFoundNode);
}
}
Item Class (Item.java)
package coinsproblembfs;
/**
*
* #author Kassem
*/
public class Item
{
String Title = "";
double value = 0;
int level = 0;
double Totalvalue = 0;
int counter = 0;
Item parent = null;
long searchTime = 0;
String winnerThreadName="";
}
Threads Class (BFS.java)
package coinsproblembfs;
import java.util.ArrayList;
import java.util.LinkedList;
import java.util.List;
/**
*
* #author Kassem M. Bagher
*/
public class BFS implements Runnable
{
private LinkedList<Item> q;
private List<Item> MoneyList;
private double searchValue = 0;
private long start = 0, end = 0;
public BFS(List<Item> monyList, double searchValue, LinkedList<Item> queue)
{
q = new LinkedList<Item>();
MoneyList = new ArrayList<Item>();
this.searchValue = searchValue;
for (int x = 0; x < queue.size(); x++)
{
q.addLast(queue.get(x));
}
for (int x = 0; x < monyList.size(); x++)
{
MoneyList.add(monyList.get(x));
}
}
private synchronized void printPath(Item item)
{
while (item != null)
{
for (Item listItem : MoneyList)
{
if (listItem.Title.equals(item.Title))
{
listItem.counter++;
}
}
item = item.parent;
}
for (Item listItem : MoneyList)
{
System.out.println(listItem.Title + " x " + listItem.counter);
}
}
private void addChildren(Item node, LinkedList<Item> q, boolean initialized)
{
for (int x = 0; x < MoneyList.size(); x++)
{
Item t = new Item();
t.value = MoneyList.get(x).value;
if (initialized)
{
t.Totalvalue = 0;
t.level = 0;
}
else
{
t.parent = node;
t.Totalvalue = MoneyList.get(x).Totalvalue;
if (t.parent == null)
{
t.level = 0;
}
else
{
t.level = t.parent.level + 1;
}
}
t.Title = MoneyList.get(x).Title;
q.addLast(t);
}
}
#Override
public void run()
{
start = System.currentTimeMillis();
try
{
while (!q.isEmpty())
{
Item node = null;
node = (Item) q.removeFirst();
node.Totalvalue = node.value + node.parent.Totalvalue;
if (node.level < CoinsProblemBFS.searchLevelLimit)
{
if (node.Totalvalue == searchValue)
{
synchronized (CoinsProblemBFS.lockLevelLimitation)
{
CoinsProblemBFS.searchLevelLimit = node.level;
CoinsProblemBFS.lastFoundNode = node;
end = System.currentTimeMillis();
CoinsProblemBFS.lastFoundNode.searchTime = (end - start);
CoinsProblemBFS.lastFoundNode.winnerThreadName=Thread.currentThread().getName();
}
}
else
{
if (node.level + 1 < CoinsProblemBFS.searchLevelLimit)
{
addChildren(node, q, false);
}
}
}
}
} catch (Exception e)
{
e.printStackTrace();
}
}
}
Sample Input:
Number of Items: 4
Enter the Title and Value for item 1: one 1
Enter the Title and Value for item 2: five 5
Enter the Title and Value for item 3: ten 10
Enter the Title and Value for item 4: twenty 20
Enter the Amount of Money: 150
Number of Threads (Muste be less than the number of items): 2

Related

How can I ensure that the processes of the threads are completely finished?

I am working on Thread for the first time and I tried to code an example I saw on the internet. An ArrayList of numbers must be divided into 4 parts, and 4 separate threads need to find the odd and even numbers in those parts and add them to the "evens" or "odds" list. Although I do not have any problems with the algorithm, I have problems with Threads.
Since the codes are not very long, I am adding them completely.
My Runnable Class:
package ThreadRace;
public class OddEvenFinder implements Runnable {
private final int id;
private final int size;
public OddEvenFinder(int id, int size) {
this.id = id;
this.size = size;
}
#Override
public void run() {
int start = id * this.size;
int end = start + this.size;
while (start < end) {
if (Starter.numbers.get(start) % 2 == 0) {
Starter.evens.add(start);
}
else {
Starter.odds.add(start);
}
start++;
}
}
}
My testing class:
package ThreadRace;
import java.util.ArrayList;
import java.util.List;
public class Starter {
public static List<Integer> numbers = new ArrayList<>();
public static List<Integer> evens = new ArrayList<>();
public static List<Integer> odds = new ArrayList<>();
public static void main(String[] args) throws InterruptedException {
for (int i = 1; i <= 10000; i++) {
numbers.add(i);
}
OddEvenFinder f1 = new OddEvenFinder(0, numbers.size() / 4);
OddEvenFinder f2 = new OddEvenFinder(1, numbers.size() / 4);
OddEvenFinder f3 = new OddEvenFinder(2, numbers.size() / 4);
OddEvenFinder f4 = new OddEvenFinder(3, numbers.size() / 4);
Thread thread1 = new Thread(f1);
Thread thread2 = new Thread(f2);
Thread thread3 = new Thread(f3);
Thread thread4 = new Thread(f4);
thread1.start();
thread2.start();
thread3.start();
thread4.start();
thread1.join();
thread2.join();
thread3.join();
thread4.join();
System.out.println(evens.size());
System.out.println(odds.size());
}
}
When I run the application this way, the length of the evens and odds lists should be 5000-5000, but I get a result between 3000-4000.
Shouldn't the .join() function wait for threads to finish? How can there be numbers that are not included in the lists?
The interesting part is that the problem is almost resolved when I add a few words to debug.
When I edit the code like this:
#Override
public void run() {
int start = id * this.size;
int end = start + this.size;
while (start < end) {
System.out.println("Thread number " + (this.id + 1) + " is working");
if (Starter.numbers.get(start) % 2 == 0) {
System.out.println(start + " added to evens");
Starter.evens.add(start);
}
else {
System.out.println(start + " added to odds");
Starter.odds.add(start);
}
start++;
}
}
The output I get gives almost accurate results like 4999-5000. When I set the size of the numbers array to a smaller value such as 4000-5000, it gives the correct result.
I have 2 questions:
1- Why .join() is not working or what am I wrong about .join()?
2- How is it that printing a few texts makes the program run more accurately?
In the JavaDocs of ArrayList, it says in bold "Note that this implementation is not synchronised". So if several threads want to add an element at the same time, only the last call of the method will set the real value. The values of the other threads are simply overwritten. Therefore, you will get fewer numbers than expected.
In order for the list to be filled in a synchronised way, you should use the keyword "synchronized" as shown below.
synchronized (Starter.evens) {
Starter.evens.add(start);
}
and
synchronized (Starter.odds) {
Starter.odds.add(start);
}

Round Robin Scheduler

I need to implement a "round-robin" scheduler with a job class that I cannot modify. Round-robin scheduler should process the job that has been waiting the longest first, then reset timer to zero. If two jobs have same wait time, lower id is processed first. The job class only gives three values (job id, remaining duration, and priority(which is not needed for this). each job has a start time, so only a couple of jobs may be available during first cycle, few more next cycle, etc. Since the "job array" I am calling is different every time I call it, I'm not sure how to store the wait times.
This is the job class:
public class Jobs{
private int[] stas = new int[0];
private int[] durs = new int[0];
private int[] lefs = new int[0];
private int[] pris = new int[0];
private int[] fins = new int[0];
private int clock;
public Jobs()
{
this("joblist.csv");
}
public Jobs(String filename)
{
BufferedReader fp = null;
String line = "";
String[] b = null;
int[] tmp;
try
{
fp = new BufferedReader(new FileReader(filename));
while((line = fp.readLine()) != null)
{
b = line.split(",");
if(b.length == 3)
{
try
{
int sta = Integer.parseInt(b[0]);
//System.out.println("sta: " + b[0]);
int dur = Integer.parseInt(b[1]);
//System.out.println("dur: " + b[1]);
int pri = Integer.parseInt(b[2]);
//System.out.println("pri: " + b[2]);
stas = app(stas, sta);
//System.out.println("stas: " + Arrays.toString(stas));
durs = app(durs, dur);
//System.out.println("durs: " + Arrays.toString(durs));
lefs = app(lefs, dur);
//System.out.println("lefs: " + Arrays.toString(lefs));
pris = app(pris, pri);
//System.out.println("pris: " + Arrays.toString(pris));
fins = app(fins, -1);
//System.out.println("fins: " + Arrays.toString(fins));
}
catch(NumberFormatException e) {}
}
}
fp.close();
}
catch(FileNotFoundException e) { e.printStackTrace(); }
catch(IOException e) { e.printStackTrace(); }
clock = 0;
}
public boolean done()
{
boolean done = true;
for(int i=0; done && i<lefs.length; i++)
if(lefs[i]>0) done=false;
return done;
}
public int getClock() { return clock; }
public int[][] getJobs()
{
int count = 0;
for(int i=0; i<stas.length; i++)
if(stas[i]<=clock && lefs[i]>0)
count++;
int[][] jobs = new int[count][3];
count = 0;
for(int i=0; i<stas.length; i++)
if(stas[i]<=clock && lefs[i]>0)
{
jobs[count] = new int[]{i, lefs[i], pris[i]};
count++;
}
return jobs;
}
public int cycle() { return cycle(-1); }
public int cycle(int j)
{
if(j>=0 && j<lefs.length && clock>=stas[j] && lefs[j]>0)
{
lefs[j]--;
if(lefs[j] == 0) fins[j] = clock+1;
}
clock++;
return clock;
}
private int[] app(int[] a, int b)
{
int[] tmp = new int[a.length+1];
for(int i=0; i<a.length; i++) tmp[i] = a[i];
tmp[a.length] = b;
return tmp;
}
public String report()
{
String r = "JOB,PRIORITY,START,DURATION,FINISH,DELAY,PRI*DELAY\n";
float dn=0;
float pdn=0;
for(int i=0; i<stas.length; i++)
{
if(fins[i]>=0)
{
int delay = ((fins[i]-stas[i])-durs[i]);
r+= ""+i+","+pris[i]+","+stas[i]+","+durs[i]+","+fins[i]+","+delay+","+(pris[i]*delay)+"\n";
dn+= delay;
pdn+= pris[i]*delay;
}
else
{
int delay = ((clock*10-stas[i])-durs[i]);
r+= ""+i+","+pris[i]+","+stas[i]+","+durs[i]+","+fins[i]+","+delay+","+(pris[i]*delay)+"\n";
dn+= delay;
pdn+= pris[i]*delay;
}
}
if(stas.length>0)
{
r+= "Avg,,,,,"+(dn/stas.length)+","+pdn/stas.length+"\n";
}
return r;
}
public String toString()
{
String r = "There are "+stas.length+" jobs:\n";
for(int i=0; i<stas.length; i++)
{
r+= " JOB "+i+": START="+stas[i]+" DURATION="+durs[i]+" DURATION_LEFT="+lefs[i]+" PRIORITY="+pris[i]+"\n";
}
return r;
}
I don't need full code, just an idea of how to store wait times and cycle the correct job.
While a array based solution 'may' work, I would advocate a more object oriented approach. Create 'Job' class with the desire attributes (id, start_time, wait etc). Using the csv file, create Job objects and hold them in a list. Write a comparator to sort this jobs-list (in this case based on job wait/age would be the factor).
The job executor then has to do the following:
while(jobs exist) {
iterate on the list {
if job is executable // start_time > current sys_time
consume cycles/job for executable jobs
mark completed jobs (optional)
}
remove the completed jobs
}
//\ This loop will add +1 to each job
for(int i = 0; i < jobs.length; i++)
{
waitTime[jobs[i][0]] += 1;
}
int longestWait = 0;//\ This holds value for greatest wait time
int nextJob = 0; //\ This holds value for index of job with greatest wait time
//\ this loop will check for the greatest wait time and and set variables accordingly
for(int i = 0; i < waitTime.length; i++)
{
if(waitTime[i] > longestWait)
{
longestWait = waitTime[i];
nextJob = i;
}
}
//\ this cycles the job with the highest wait time
jobsource.cycle(nextJob);
//\ this resets the wait time for processed job
waitTime[nextJob] = 0;

All possible ways to reach an ending position

http://www.cstutoringcenter.com/problems/problems.php?id=103
For those who doesn't want to click it, it basically says there's a stepping stone, "-" and soldier "#", soldiers can only move right. If the soldier is behind another soldier, he must wait for the soldier to move first. The ending condition is when all soldiers reaches the end.
The number of ways 2 soldier can move across 5 stepping stones.
1) ##--- #-#-- -##-- -#-#- --##- --#-# ---##
2) ##--- #-#-- -##-- -#-#- -#--# --#-# ---##
3) ##--- #-#-- #--#- -#-#- --##- --#-# ---##
4) ##--- #-#-- #--#- -#-#- -#--# --#-# ---##
5) ##--- #-#-- #--#- #---# -#--# --#-# ---##
I'm using a breadth first search, with 5 stones, it's running within seconds, but with 10 stones, it's taking hours, the time is increasing exponentially with the depth. How can I deal with this?
My Codes:
States.java
import java.util.ArrayList;
public class State {
public int stones;
public Soldiers[] soldiers;
public String currentState ="";
public boolean visited = false;
public State(int stones,int Numsoldiers){
System.out.println(Numsoldiers);
this.stones = stones;
soldiers = new Soldiers[Numsoldiers];
System.out.println("length" + soldiers.length);
initState();
}
public State(int stones,Soldiers[] soldiers){
this.stones = stones;
this.soldiers = soldiers;
paintState();
}
public void initState(){
for(int i=0;i<soldiers.length;i++)
{
soldiers[i] = new Soldiers();
soldiers[i].position =i;
currentState+="#";
}
for(int j=soldiers.length;j<stones;j++)
{
currentState+="-";
}
}
private void paintState(){
for(int j=0;j<stones;j++)
{
currentState+="-";
}
char[] stateChar = currentState.toCharArray();
currentState = "";
for(int i=0;i<soldiers.length;i++){
stateChar[soldiers[i].position] = '#';
}
for(int k=0; k<stateChar.length;k++){
currentState += stateChar[k];
}
}
public void printState(){
System.out.println(currentState);
}
public ArrayList<State> getNextStates(){
ArrayList<State> States = new ArrayList<State>();
for(int i=0;i<soldiers.length;i++){
Soldiers[] newSoldiers = new Soldiers[soldiers.length];
for(int j=0;j<soldiers.length;j++){
newSoldiers[j] = new Soldiers(soldiers[j].position);
}
if(!((newSoldiers[i].position+1)==stones))
{
if((currentState.charAt((newSoldiers[i].position+1))=='-'))
{
newSoldiers[i].move();
States.add(new State(stones,newSoldiers));
}
}
}
if(States.size()==0)
{
TestSoldiers.count++;
}
return States;
}
}
Soldiers.java
public class Soldiers {
int position = 0;
public Soldiers(){
position =0;
}
public Soldiers(int pos){
position = pos;
}
public void move(){
position ++;
}
}
TestSoldiers.java
import java.util.LinkedList;
import java.util.Queue;
public class TestSoldiers {
public static int count=0;
public static void main(String[] args){
TestSoldiers t = new TestSoldiers();
}
public TestSoldiers()
{
State s = new State(10,3);
breadthFirstTraversal(s);
System.out.println(count);
}
public void breadthFirstTraversal(State rootNode){
Queue<State> q = new LinkedList<State>();
q.add(rootNode);
while(!q.isEmpty()){
State n = (State)q.poll();
n.printState();
for(State adj : n.getNextStates()){
q.add(adj);
}
}
}
}
How can I make it so that I will only consider each State once while maintaining the integrity of the total number of ways to end (counts in TestSoldiers.java)?
For those of you who want to modify the parameters, it's the new State(n,k) where n is the number of stones and k is the number of soldiers.
Memoization might come in handy.
The idea would be to run depth-first search to count the number of ways to get from the current state to the end, and store this result, then look up the already-calculated value if ever that state is repeated.
For instance, there are 2 ways to reach the end from -#-#-, so, storing this result when we get there via -##--, we could simply look up 2 when we get there via #--#-.
The simplest (but far from most efficient) way to store these would simply be to have a:
Map<Pair<Integer (Position1), Integer (Position2)>, Integer (Count)>
More generically, you could perhaps make that Pair a List.
A more efficient approach would be to have a bitmap where each bit corresponds to whether or not there's a soldier at some given position. So -#-#- would correspond to 01010, which could simply be stored in an int as 10 in decimal - if there are more than 64 stones (i.e. what would fit into a long), you could use a BitSet.
You might be better using combinatorics to compute the number of paths.
For example, suppose there are 2 soldiers and 5 steps.
Represent the distance the first soldier has moved by y, and the distance the second soldier has moved by x.
You are trying to count the number of monotonic paths from 0,0 to 3,3 such that y is never greater than x.
This is a well known problem and the answer is given by the Catalan numbers. In this case, the answer is given by the Catalan number for n=3, which is 5.
When you have more than 2 soldiers you will need to use multidimensional Catalan numbers. A useful guide and formula can be found on OEIS:
T(m, n) = 0! * 1! * .. * (n-1)! * (m * n)! / ( m! * (m+1)! * .. * (m+n-1)! )
My solution runs 10 positions in less than 1 second. The solution is quick and dirty, but the algorithm is what you should be interested in right?
The idea of my algorithm is:
manage a set of paths to compute. start with the path where both soldiers are at the left most positions.
if the set of paths to compute is not empty pick any path and remove it from the set.
if the path is terminated (both soldiers are at the most right positions) print the path. continue with 2.
extend the path by moving the head soldier if possible and put it into the set.
extend the path by moving the tail soldier if possible and put it into the set.
That's it.
public static void main(String[] args) {
List<Node> nodes = Node.newRootNode(10);
while (!nodes.isEmpty()) {
Node node = nodes.remove(0);
if (node.isLeaf()) node.printPath();
else {
if (node.headSoldierCanMove()) nodes.add(node.moveHeadSoldier());
if (node.tailSoldierCanMove()) nodes.add(node.moveTailSoldier());
}
}
}
static final class Node {
static List<Node> newRootNode(final int maxPos) {
return new ArrayList<Node>() {{
add(new Node(1, 2, maxPos, ""));
}};
}
private final int maxPos;
private final String path;
private int tailPos = 1;
private int headPos = tailPos + 1;
private Node(int tailPos, int headPos, int maxPos, String path) {
this.maxPos = maxPos;
this.tailPos = tailPos;
this.headPos = headPos;
this.path = addPath(path);
}
boolean tailSoldierCanMove() {
return tailPos < headPos - 1;
}
Node moveTailSoldier() {
return new Node(tailPos + 1, headPos, maxPos, path);
}
boolean headSoldierCanMove() {
return headPos < maxPos;
}
Node moveHeadSoldier() {
return new Node(tailPos, headPos + 1, maxPos, path);
}
void printPath() {
System.out.println(path);
}
boolean isLeaf() {
return headPos == maxPos && tailPos == headPos - 1;
}
private String addPath(String prefix) {
StringBuilder builder = new StringBuilder(prefix);
for (int pos = 1; pos <= maxPos; pos++) {
builder.append(tailPos == pos || headPos == pos ? "#" : "-");
}
return builder.append(" ").toString();
}
}

Thread-safe circular buffer in Java

Consider a few web server instances running in parallel. Each server holds a reference to a single shared "Status keeper", whose role is keeping the last N requests from all servers.
For example (N=3):
Server a: "Request id = ABCD" Status keeper=["ABCD"]
Server b: "Request id = XYZZ" Status keeper=["ABCD", "XYZZ"]
Server c: "Request id = 1234" Status keeper=["ABCD", "XYZZ", "1234"]
Server b: "Request id = FOO" Status keeper=["XYZZ", "1234", "FOO"]
Server a: "Request id = BAR" Status keeper=["1234", "FOO", "BAR"]
At any point in time, the "Status keeper" might be called from a monitoring application that reads these last N requests for an SLA report.
What's the best way to implement this producer-consumer scenario in Java, giving the web servers higher priority than the SLA report?
CircularFifoBuffer seems to be the appropriate data structure to hold the requests, but I'm not sure what's the optimal way to implement efficient concurrency.
Buffer fifo = BufferUtils.synchronizedBuffer(new CircularFifoBuffer());
Here's a lock-free ring buffer implementation. It implements a fixed-size buffer - there is no FIFO functionality. I would suggest you store a Collection of requests for each server instead. That way your report can do the filtering rather than getting your data structure to filter.
/**
* Container
* ---------
*
* A lock-free container that offers a close-to O(1) add/remove performance.
*
*/
public class Container<T> implements Iterable<T> {
// The capacity of the container.
final int capacity;
// The list.
AtomicReference<Node<T>> head = new AtomicReference<Node<T>>();
// TESTING {
AtomicLong totalAdded = new AtomicLong(0);
AtomicLong totalFreed = new AtomicLong(0);
AtomicLong totalSkipped = new AtomicLong(0);
private void resetStats() {
totalAdded.set(0);
totalFreed.set(0);
totalSkipped.set(0);
}
// TESTING }
// Constructor
public Container(int capacity) {
this.capacity = capacity;
// Construct the list.
Node<T> h = new Node<T>();
Node<T> it = h;
// One created, now add (capacity - 1) more
for (int i = 0; i < capacity - 1; i++) {
// Add it.
it.next = new Node<T>();
// Step on to it.
it = it.next;
}
// Make it a ring.
it.next = h;
// Install it.
head.set(h);
}
// Empty ... NOT thread safe.
public void clear() {
Node<T> it = head.get();
for (int i = 0; i < capacity; i++) {
// Trash the element
it.element = null;
// Mark it free.
it.free.set(true);
it = it.next;
}
// Clear stats.
resetStats();
}
// Add a new one.
public Node<T> add(T element) {
// Get a free node and attach the element.
totalAdded.incrementAndGet();
return getFree().attach(element);
}
// Find the next free element and mark it not free.
private Node<T> getFree() {
Node<T> freeNode = head.get();
int skipped = 0;
// Stop when we hit the end of the list
// ... or we successfully transit a node from free to not-free.
while (skipped < capacity && !freeNode.free.compareAndSet(true, false)) {
skipped += 1;
freeNode = freeNode.next;
}
// Keep count of skipped.
totalSkipped.addAndGet(skipped);
if (skipped < capacity) {
// Put the head as next.
// Doesn't matter if it fails. That would just mean someone else was doing the same.
head.set(freeNode.next);
} else {
// We hit the end! No more free nodes.
throw new IllegalStateException("Capacity exhausted.");
}
return freeNode;
}
// Mark it free.
public void remove(Node<T> it, T element) {
totalFreed.incrementAndGet();
// Remove the element first.
it.detach(element);
// Mark it as free.
if (!it.free.compareAndSet(false, true)) {
throw new IllegalStateException("Freeing a freed node.");
}
}
// The Node class. It is static so needs the <T> repeated.
public static class Node<T> {
// The element in the node.
private T element;
// Are we free?
private AtomicBoolean free = new AtomicBoolean(true);
// The next reference in whatever list I am in.
private Node<T> next;
// Construct a node of the list
private Node() {
// Start empty.
element = null;
}
// Attach the element.
public Node<T> attach(T element) {
// Sanity check.
if (this.element == null) {
this.element = element;
} else {
throw new IllegalArgumentException("There is already an element attached.");
}
// Useful for chaining.
return this;
}
// Detach the element.
public Node<T> detach(T element) {
// Sanity check.
if (this.element == element) {
this.element = null;
} else {
throw new IllegalArgumentException("Removal of wrong element.");
}
// Useful for chaining.
return this;
}
public T get () {
return element;
}
#Override
public String toString() {
return element != null ? element.toString() : "null";
}
}
// Provides an iterator across all items in the container.
public Iterator<T> iterator() {
return new UsedNodesIterator<T>(this);
}
// Iterates across used nodes.
private static class UsedNodesIterator<T> implements Iterator<T> {
// Where next to look for the next used node.
Node<T> it;
int limit = 0;
T next = null;
public UsedNodesIterator(Container<T> c) {
// Snapshot the head node at this time.
it = c.head.get();
limit = c.capacity;
}
public boolean hasNext() {
// Made into a `while` loop to fix issue reported by #Nim in code review
while (next == null && limit > 0) {
// Scan to the next non-free node.
while (limit > 0 && it.free.get() == true) {
it = it.next;
// Step down 1.
limit -= 1;
}
if (limit != 0) {
next = it.element;
}
}
return next != null;
}
public T next() {
T n = null;
if ( hasNext () ) {
// Give it to them.
n = next;
next = null;
// Step forward.
it = it.next;
limit -= 1;
} else {
// Not there!!
throw new NoSuchElementException ();
}
return n;
}
public void remove() {
throw new UnsupportedOperationException("Not supported.");
}
}
#Override
public String toString() {
StringBuilder s = new StringBuilder();
Separator comma = new Separator(",");
// Keep counts too.
int usedCount = 0;
int freeCount = 0;
// I will iterate the list myself as I want to count free nodes too.
Node<T> it = head.get();
int count = 0;
s.append("[");
// Scan to the end.
while (count < capacity) {
// Is it in-use?
if (it.free.get() == false) {
// Grab its element.
T e = it.element;
// Is it null?
if (e != null) {
// Good element.
s.append(comma.sep()).append(e.toString());
// Count them.
usedCount += 1;
} else {
// Probably became free while I was traversing.
// Because the element is detached before the entry is marked free.
freeCount += 1;
}
} else {
// Free one.
freeCount += 1;
}
// Next
it = it.next;
count += 1;
}
// Decorate with counts "]used+free".
s.append("]").append(usedCount).append("+").append(freeCount);
if (usedCount + freeCount != capacity) {
// Perhaps something was added/freed while we were iterating.
s.append("?");
}
return s.toString();
}
}
Note that this is close to O1 put and get. A Separator just emits "" first time around and then its parameter from then on.
Edit: Added test methods.
// ***** Following only needed for testing. *****
private static boolean Debug = false;
private final static String logName = "Container.log";
private final static NamedFileOutput log = new NamedFileOutput("C:\\Junk\\");
private static synchronized void log(boolean toStdoutToo, String s) {
if (Debug) {
if (toStdoutToo) {
System.out.println(s);
}
log(s);
}
}
private static synchronized void log(String s) {
if (Debug) {
try {
log.writeLn(logName, s);
} catch (IOException ex) {
ex.printStackTrace();
}
}
}
static volatile boolean testing = true;
// Tester object to exercise the container.
static class Tester<T> implements Runnable {
// My name.
T me;
// The container I am testing.
Container<T> c;
public Tester(Container<T> container, T name) {
c = container;
me = name;
}
private void pause() {
try {
Thread.sleep(0);
} catch (InterruptedException ex) {
testing = false;
}
}
public void run() {
// Spin on add/remove until stopped.
while (testing) {
// Add it.
Node<T> n = c.add(me);
log("Added " + me + ": " + c.toString());
pause();
// Remove it.
c.remove(n, me);
log("Removed " + me + ": " + c.toString());
pause();
}
}
}
static final String[] strings = {
"One", "Two", "Three", "Four", "Five",
"Six", "Seven", "Eight", "Nine", "Ten"
};
static final int TEST_THREADS = Math.min(10, strings.length);
public static void main(String[] args) throws InterruptedException {
Debug = true;
log.delete(logName);
Container<String> c = new Container<String>(10);
// Simple add/remove
log(true, "Simple test");
Node<String> it = c.add(strings[0]);
log("Added " + c.toString());
c.remove(it, strings[0]);
log("Removed " + c.toString());
// Capacity test.
log(true, "Capacity test");
ArrayList<Node<String>> nodes = new ArrayList<Node<String>>(strings.length);
// Fill it.
for (int i = 0; i < strings.length; i++) {
nodes.add(i, c.add(strings[i]));
log("Added " + strings[i] + " " + c.toString());
}
// Add one more.
try {
c.add("Wafer thin mint!");
} catch (IllegalStateException ise) {
log("Full!");
}
c.clear();
log("Empty: " + c.toString());
// Iterate test.
log(true, "Iterator test");
for (int i = 0; i < strings.length; i++) {
nodes.add(i, c.add(strings[i]));
}
StringBuilder all = new StringBuilder ();
Separator sep = new Separator(",");
for (String s : c) {
all.append(sep.sep()).append(s);
}
log("All: "+all);
for (int i = 0; i < strings.length; i++) {
c.remove(nodes.get(i), strings[i]);
}
sep.reset();
all.setLength(0);
for (String s : c) {
all.append(sep.sep()).append(s);
}
log("None: " + all.toString());
// Multiple add/remove
log(true, "Multi test");
for (int i = 0; i < strings.length; i++) {
nodes.add(i, c.add(strings[i]));
log("Added " + strings[i] + " " + c.toString());
}
log("Filled " + c.toString());
for (int i = 0; i < strings.length - 1; i++) {
c.remove(nodes.get(i), strings[i]);
log("Removed " + strings[i] + " " + c.toString());
}
c.remove(nodes.get(strings.length - 1), strings[strings.length - 1]);
log("Empty " + c.toString());
// Multi-threaded add/remove
log(true, "Threads test");
c.clear();
for (int i = 0; i < TEST_THREADS; i++) {
Thread t = new Thread(new Tester<String>(c, strings[i]));
t.setName("Tester " + strings[i]);
log("Starting " + t.getName());
t.start();
}
// Wait for 10 seconds.
long stop = System.currentTimeMillis() + 10 * 1000;
while (System.currentTimeMillis() < stop) {
Thread.sleep(100);
}
// Stop the testers.
testing = false;
// Wait some more.
Thread.sleep(1 * 100);
// Get stats.
double added = c.totalAdded.doubleValue();
double skipped = c.totalSkipped.doubleValue();
//double freed = c.freed.doubleValue();
log(true, "Stats: added=" + c.totalAdded + ",freed=" + c.totalFreed + ",skipped=" + c.totalSkipped + ",O(" + ((added + skipped) / added) + ")");
}
Maybe you want to look at Disruptor - Concurrent Programming Framework.
Find a paper describing the alternatives, design and also a performance comparement to java.util.concurrent.ArrayBlockingQueue here: pdf
Consider to read the first three articles from BlogsAndArticles
If the library is too much, stick to java.util.concurrent.ArrayBlockingQueue
I would have a look at ArrayDeque, or for a more concurrent implementation have a look at the Disruptor library which is one of the most sophisticated/complex ring buffer in Java.
An alternative is to use an unbounded queue which is more concurrent as the producer never needs to wait for the consumer. Java Chronicle
Unless your needs justify the complexity, an ArrayDeque may be all you need.
Also have a look at java.util.concurrent.
Blocking queues will block until there is something to consume or (optionally) space to produce:
http://docs.oracle.com/javase/1.5.0/docs/api/java/util/concurrent/BlockingQueue.html
Concurrent linked queue is non-blocking and uses a slick algorithm that allows a producer and consumer to be active concurrently:
http://docs.oracle.com/javase/1.5.0/docs/api/java/util/concurrent/ConcurrentLinkedQueue.html
Hazelcast's Queue offers almost everything you ask for, but doesn't support circularity. But from your description I am not sure if you actually need it.
If it were me, I would use the CircularFIFOBuffer as you indicated, and synchronize around the buffer when writing (add). When the monitoring application wants to read the buffer, synchronize on the buffer, and then copy or clone it to use for reporting.
This suggestion is predicated on the assumption that latency is minimal to copy/clone the buffer to a new object. If there are large number of elements, and copying time is slow, then this is not a good idea.
Pseudo-Code example:
public void writeRequest(String requestID) {
synchronized(buffer) {
buffer.add(requestID);
}
}
public Collection<String> getRequests() {
synchronized(buffer) {
return buffer.clone();
}
}
Since you specifically ask to give writers (that is web servers) higher priority than the reader (that is monitoring), I would suggest the following design.
Web servers add request information to a concurrent queue which is read by a dedicated thread, which adds requests to a thread-local (therefore non-synchronized) queue that overwrites the oldest element, like EvictingQueue or CircularFifoQueue.
This same thread checks a flag which indicates if a report has been requested after every request processed, and if positive, produces a report from all elements present in the thread-local queue.

While Parallelizing QuickSort in Java, Threads never returns back at join(). Why?

I am having simple code of paralellizing QuickSort algorithm in Java, in run method I everytime create two seperate new threads for parallelizing processing of Array elements. But as it encounters join() statements for both created threads, threads never backs and halts on joins(), seems join() never releases them.
Below is the code.
class StartJoinQuickSort implements Runnable
{
private int m_Low, m_High;
private int[] m_Array = null;
private final static int NR_OF_VALUES = 10; // TOTAL_NO_VALUES
private int PivotElement;
private static Random m_random = new Random( );
public StartJoinQuickSort(int[] a_Array,int a_Low,int a_High)
{
this.m_Array = a_Array;
this.m_Low = a_Low;
this.m_High = a_High;
}
private void SwapArrayElements(int a_i,int a_j)
{
int temp = this.m_Array[a_i];
this.m_Array[a_i] = this.m_Array[a_j];
this.m_Array[a_j] = temp;
}// end of SwapArrayElements
private static int nextRandomFunctionValue(int aStart, int aEnd)
{
if ( aStart > aEnd )
{
throw new IllegalArgumentException("Start cannot exceed End.");
}
//get the range, casting to long to avoid overflow problems
long range = (long)aEnd - (long)aStart + 1;
// compute a fraction of the range, 0 <= frac < range
long fraction = (long)(range * m_random.nextDouble());
int randomNumber = (int)(fraction + aStart);
return randomNumber;
}// end of nextRandomFunctionValue
private static int[] GetArrayWithRandomValues()
{
int[] ArrayToBePopulatedWithRandomValues = new int[NR_OF_VALUES];
for(int index =0; index<NR_OF_VALUES;index++)
{
int RandomValue = StartJoinQuickSort.nextRandomFunctionValue(0,NR_OF_VALUES);
ArrayToBePopulatedWithRandomValues[index] = RandomValue;
}//end of for
return ArrayToBePopulatedWithRandomValues;
}//end of GetArrayWithRandomValues
private int middleIndex(int left, int right)
{
return left + (right - left) / 2;
}
public int Partition(int a_Start,int a_end)
{
// System.out.println("Partition ..thId : " + Thread.currentThread().getId());
int pivotIndex = 0;
int i = a_Start;
int j = a_end;
try
{
pivotIndex = middleIndex(a_Start , a_end);
this.PivotElement = this.m_Array[pivotIndex];
do
{
while(this.m_Array[i] < PivotElement )
i++;
if(j>0)
{
try
{
while( this.m_Array[j] > PivotElement )
j--;
}
catch(Exception ex){System.out.println(" j : " + j);}
}//end of if
if(i<=j)
{
SwapArrayElements(i,j);
// System.out.println("Swap .." + + Thread.currentThread().getId());
i++;
j--;
}//end of if
}while(i<=j);
}
catch(Exception except)
{
System.out.println("exception in Partition " + except);
}
return j;
}
public void run()
{
//System.out.println("in run..");
//System.out.println("after PARTITION");
StartJoinQuickSort oStartQuickSort_1 = null;
StartJoinQuickSort oStartQuickSort_2 = null;
if(this.m_Low < this.m_High )
{
int Index = Partition(this.m_Low,this.m_High);
Thread thPart_1 = new Thread ( new StartJoinQuickSort( this.m_Array,this.m_Low,Index ) );
Thread thPart_2 = new Thread ( new StartJoinQuickSort( this.m_Array,Index + 1,this.m_High ) );
thPart_1.start(); thPart_2.start();
//}//end of if
//if( Index + 1 < this.m_High)
//{
try
{
thPart_1.join(); thPart_2.join();
}catch (InterruptedException e) { e.printStackTrace();}
}
}//end of run
Regards
Usman
Hmmm, it is never a good idea to implement a recursive algorithm in parallel like this. You will end up creating a huge number of threads (exponential at every level) and will eventually oversubscribe the system.
The best idea is to have a cutoff point, which let's say is equal to the number of available cores. Then when the current level of recursion has a number of branches equal to the cutoff point switch to a sequential quicksort. Some very rough pseudocode of the flow:
parallel_quicksort(level, interval) {
// compute subintervals interval1, interval2
if(level < cutoff) {
spawn1: parallel_quicksort(level + 1, interval1);
spawn2: parallel_quicksort(level + 1, interval2);
join1();
join2();
} else {
quicksort(interval1);
quicksort(interval2);
}
}
Also have a look over this implementation to see if you've missed something: http://www.inf.fh-flensburg.de/lang/algorithmen/sortieren/quick/quicken.htm
What happens if you have e.g. low=0, high=1? If your Partition returns 1, you'll have an infinite loop of threads, right?
join() blocks if your thread doesn't finish. You need to determine why your threads are not finishing. Can you try debugging your program with a debugger?
Thanks all for your kind suggestions and advices.
I myself detected the problem. it was with Partition function which was not working fine, it was having some problems, I choose another one and it worked fine for me..
New Partition procedure is :
public int Partition(int[] a_Array, int a_Left, int a_Right)
{
// chose middle value of range for our pivot
int pivotValue = a_Array[middleIndex(a_Left, a_Right)];
--a_Left;
++a_Right;
while (true)
{
do
++a_Left;
while (a_Array[a_Left] < pivotValue);
do
--a_Right;
while (a_Array[a_Right] > pivotValue);
if (a_Left < a_Right)
SwapArrayElements(a_Left,a_Right);
else
{
return a_Right;
}
}
}

Categories