We Keep Coding

The performance of the HashMap in JAVA8

I had a question when I learned the HashMap source code in java8。
Source code is so complicated, how much efficiency?
So I wrote a code about the hash conflict。
public class Test {
final int i;
public Test(int i) {
this.i = i;
}
public static void main(String[] args) {
java.util.HashMap<Test, Test> set = new java.util.HashMap<Test, Test>();
long time;
Test last;
Random random = new Random(0);
int i = 0;
for (int max = 1; max < 200000; max <<= 1) {
long c1 = 0, c2 = 0;
int t = 0;
for (; i < max; i++, t++) {
last = new Test(random.nextInt());
time = System.nanoTime();
set.put(last, last);
c1 += (System.nanoTime() - time);
last = new Test(random.nextInt());
time = System.nanoTime();
set.get(last);
c2 += (System.nanoTime() - time);
}
System.out.format("%d\t%d\t%d\n", max, (c1 / t), (c2 / t));
}
}
public int hashCode() {
return 0;
}
public boolean equals(Object obj) {
if (obj == null)
return false;
if (!(obj instanceof Test))
return false;
Test t = (Test) obj;
return t.i == this.i;
}
}
I show the results in Excel。
enter image description here
I am using java6u45 java7u80 java8u131。
I do not understand why the performance of java8 will be so bad
I'm trying to write my own HashMap.
I would like to learn HashMap in java8 which is better, but I did not find it.

Your test scenario is non-optimal for Java 8 HashMap. HashMap in Java 8 optimizes collisions by using binary trees for any hash chains longer than a given threshold. However, this only works if the key type is comparable. If it isn't then the overhead of testing to see if the optimization is possible actually makes Java 8 HashMap slower. (The slow-down is more than I expected ... but that's another topic.)
Change your Test class to implement Comparable<Test> ... and you should see that Java 8 performs better than the others when the proportion of hash collisions is large enough.
Note that the tree optimization should be considered as a defensive measure for the case where the hash function doesn't perform. The optimization turns O(N) worst-case performance to O(logN) worst-case.
If you want your HashMap instances to have O(1) lookup, you should make sure that you use a good hash function for the key type. If the probability of collision is minimized, the optimization is moot.
Source code is so complicated, how much efficiency?
It is explained in the comments in the source code. And probably other places that Google can find for you :-)

Using Java 7 HashMap in Java 8

I have updated a Java application to Java 8. The application heavily relies on HashMaps.
When I run the benchmarks, I see unpredictable behavoir. For some inputs, the application runs faster than before, but for larger inputs, it's constantly slower.
I've checked the profiler and the most time consuming operation is HashMap.get. I suspect the changes
are due to the HashMap modification in Java 8, but it may not be true, as I have changed some other parts.
Is there an easy way that I hook in the original Java 7 HashMap into my Java 8 application so that I only change the hashmap implementation to see if I still observe the change in performance.
The following is a minimal program that tries to simulate what my application is doing.
The basic idea is that i need to share nodes in the application. At some runtime point, a node
should be retrieved or created if it already does not exist based on some integer properties. The following only uses two integer, but in the real application I have one, two and three integer keys.
import java.util.HashMap;
import java.util.Map;
import java.util.Random;
public class Test1 {
static int max_k1 = 500;
static int max_k2 = 500;
static Map<Node, Node> map;
static Random random = new Random();
public static void main(String[] args) {
for (int i = 0; i < 15; i++) {
long start = System.nanoTime();
run();
long end = System.nanoTime();
System.out.println((end - start) / 1000_000);
}
}
private static void run() {
map = new HashMap<>();
for (int i = 0; i < 10_000_000; i++) {
Node key = new Node(random.nextInt(max_k1), random.nextInt(max_k2));
Node val = getOrElseUpdate(key);
}
}
private static Node getOrElseUpdate(Node key) {
Node val;
if ((val = map.get(key)) == null) {
val = key;
map.put(key, val);
}
return val;
}
private static class Node {
private int k1;
private int k2;
public Node(int k1, int k2) {
this.k1 = k1;
this.k2 = k2;
}
#Override
public int hashCode() {
int result = 17;
result = 31 * result + k1;
result = 31 * result + k2;
return result;
}
#Override
public boolean equals(Object obj) {
if (this == obj)
return true;
if (!(obj instanceof Node))
return false;
Node other = (Node) obj;
return k1 == other.k1 && k2 == other.k2;
}
}
}
The benchmarking is primitive, but still, this is the result of 15 runs on Java 8:
8143
7919
7984
7973
7948
7984
7931
7992
8038
7975
7924
7995
6903
7758
7627
and this is for Java 7:
7247
6955
6510
6514
6577
6489
6510
6570
6497
6482
6540
6462
6514
4603
6270
The benchmarking is primitive, so I appreciate if someone who is familiar with JMH or other benchmarking tools run it, but from what I observe the results are better for Java 7. Any ideas?

Your hashCode() is very poor. In example you posted you have 250000 unique values but only 15969 unique hash codes. Because of lot of collisions, Java 8 swaps lists with trees. In your case it only adds overhead, because many elements not only have the same position in hash table but also the same hash code. The tree ends up as a linked list anyway.
There are couple of ways to fix this:
Improve your hashCode. return k1 * 500 + k2; resolves the issue.
Use THashMap. Open addressing should work better in case of collisions.
Make Node implement Comparable. This will be used by HashMap to construct balanced tree in case of conflicts.

What's the fastest way to initialize a large list of integers?

I need to pre-populate a List with a large number of integer values.
Is there are faster way to do this other than iteration?
Current Code:
class VlanManager {
Queue<Integer> queue = Lists.newLinkedList();
public VlanManager(){
for (int i = 1; i < 4094; i++) {
queue.add(i);
}
}
This code is in the constructor of a class that is created pretty frequently so I'd like this to be as efficient (read:performance not lines of code) as possible

4094 isnt to many items to loop but if it is getting called very frequently you might look at doing something with a static variable.
private static Integer[] theList;
static {
theList = new Integer[4094];
for (int i = 1; i < 4094; i++) {
theList[i-1] = i;
}
}
then make that list a List
Queue<Integer> intQue = new LinkedList(Arrays.asList(theList));
There is a danger of using this method if you have a list of mutable objects. Heres an example of what can happen. Integers are immutable so this doesnt actually apply to your question as it stands
class MyMutableObject {
public int theValue;
}
class Test {
private static MyMutableObject[] theList;
static {
theList = new MyMutableObject[4094];
for (int i = 1; i <= 4094; i++) {
theList[i-1] = new MyMutableObject();
theList[i-1].theValue = i;
}
}
public static void main(String [] args) {
Queue<MyMutableObject> que = new LinkedList(Arrays.asList(theList));
System.out.println(que.peek().theValue); // 1
// your actually modifing the same object as the one in your static list
que.peek().theValue = -100;
Queue<MyMutableObject> que2 = new LinkedList(Arrays.asList(theList));
System.out.println(que2.peek().theValue); // -100
}
}
#Bohemian Has some good points on using a static List instead of an array, while the performance gains are very small they are none the less performance gains. Also because the 'array' is actually only ever being used as a List not an array it should be declared as such.
private static List<Integer> theList;
static {
theList = new ArrayList(4094);
for (Integer i = 0; i < 4094; i++) {
theList.add(i+1);
}
}

The fastest way would be to create a reference list (initialized using an instance block - neatly wrapping it all up in one statement):
private static final List<Integer> LIST = new ArrayList<Integer>(4094) {{
for (int i = 1; i < 4094; i++)
LIST.add(i);
}};
Then in your constructor, initialize the queue using the copy constructor:
Queue<Integer> queue;
public VlanManager(){
queue = new LinkedList<Integer>(LIST);
}
You will not write a faster implementation than what's in the JDK.

I realize this question has already been answered. But I think one important answer is missing: The fastest way to initialize a LinkedList with the values 0..4093 is .. DON'T DO IT AT ALL. Especially if speed is an issue.
What you basically are doing is creating a structure consisting of 4093 Node elements each consiting of two pointers to prev/next element and one pointer to an Integer object. Each of this Nodes must be created (and free). In addition nearly each contained Integer must be created (and freed). 'Nearly' because Java uses a cache for Integer but normally (you can change this with system properties) in the range of -127..127.
This is a lot to do in order to get a simple list of integer and if used intensively gives the GC a lot to do afterwards.
That being said there are numerous possible ways of doing this in a more efficient way. But they depend on what your concrete usage pattern is. Just to name a few:
Use an Array: boolean [] inUse' and set the taken vlan-id totrue` if it's taken
Even better use a BitSet instead of the array
Don't store which vlan is free, but which vlan is taken. I think they tend to be free and so there are much more free as there are taken ones. (this means much less to keep track of).
If you insist on using a LinkedList don't initialize it with your class but have it already initialized. This depends on how much of them you would need. You could keep a pool of them. Or perhaps your codes allows reusage of old lists. (yes, you could sort them after usage.)
Surely there are more...
All of this methods require you to build your own 'Queue' interface. But perhaps this has not to be as rich as Java's. And it really isn't that difficult. If you really use this intensively you could reach perfomance improvement factor 10x-1000x++.
A possible implementation using BitSet with an instantiation cost of nearly nothing could be:
import java.util.BitSet;
import org.testng.annotations.Test;
public class BitSetQueue {
// Represents the values 0..size-1
private final BitSet bitset;
private final int size;
private int current = 0;
private int taken = 0;
public BitSetQueue( int size ){
this.bitset = new BitSet( size );
this.size = size;
this.current = size-1;
}
public int poll(){
// prevent endless loop
if( taken == size ) return -1;
// seek for next free value.
// can be changed according to policy
while( true ){
current = (current+1)%size;
if( ! bitset.get( current ) ){
bitset.set( current );
taken++;
return current;
}
}
}
public boolean free( int num ){
if( bitset.get( num ) ){
bitset.clear( num );
taken--;
return true;
}
return false;
}
#Test
public static void usage(){
BitSetQueue q = new BitSetQueue( 4094 );
for( int i = 0; i < 4094; i++ ){
assertEquals( q.poll(), i );
}
assertEquals( q.poll(), -1 ); // No more available
assertTrue( q.free( 20 ) );
assertTrue( q.free( 51 ) );
assertEquals( q.poll(), 20 );
assertEquals( q.poll(), 51 );
}
}

Java multi-thread scalability issue

more updates
As is explained in the selected answer, the problem is in JVM's garbage collection algorithm.
JVM uses card marking algorithm to keep track of modified references in object fields. For each reference assignment to a field, it marks an associated bit in the card to be true -- this causes a false-sharing hence blocks scaling. The details are well described in this article: https://blogs.oracle.com/dave/entry/false_sharing_induced_by_card
The option -XX:+UseCondCardMark (in Java 1.7u40 and up) mitigates the problem, and makes it scale almost perfectly.
updates
I found out (hinted from Park Eung-ju) that assigning an object into a field variable makes the difference. If I remove the assignment, it scales perfectly.
I think probably it has something to do with Java memory model -- such as, an object reference must point to a valid address before it gets visible, but I am not completely sure. Both double and Object reference (likely) have 8 bytes size on 64 bit machine, so it seems to me that assigning a double value and an Object reference should be the same in terms of synchronization.
Anyone has a reasonable explanation?
Here I have a weird Java multi-threading scalability problem.
My code simply iterates over an array (using the visitor pattern) to compute simple floating-point operations and assign the result to another array. There is no data dependency, nor synchronization, so it should scale linearly (2x faster with 2 threads, 4x faster with 4 threads).
When primitive (double) array is used, it scales very well. When object type (e.g. String) array is used, it doesn't scale at all (even though the value of the String array is not used at all...)
Here's the entire source code:
import java.util.ArrayList;
import java.util.Arrays;
import java.util.concurrent.CyclicBarrier;
class Table1 {
public static final int SIZE1=200000000;
public static final boolean OBJ_PARAM;
static {
String type=System.getProperty("arg.type");
if ("double".equalsIgnoreCase(type)) {
System.out.println("Using primitive (double) type arg");
OBJ_PARAM = false;
} else {
System.out.println("Using object type arg");
OBJ_PARAM = true;
}
}
byte[] filled;
int[] ivals;
String[] strs;
Table1(int size) {
filled = new byte[size];
ivals = new int[size];
strs = new String[size];
Arrays.fill(filled, (byte)1);
Arrays.fill(ivals, 42);
Arrays.fill(strs, "Strs");
}
public boolean iterate_range(int from, int to, MyVisitor v) {
for (int i=from; i<to; i++) {
if (filled[i]==1) {
// XXX: Here we are passing double or String argument
if (OBJ_PARAM) v.visit_obj(i, strs[i]);
else v.visit(i, ivals[i]);
}
}
return true;
}
}
class HeadTable {
byte[] filled;
double[] dvals;
boolean isEmpty;
HeadTable(int size) {
filled = new byte[size];
dvals = new double[size];
Arrays.fill(filled, (byte)0);
isEmpty = true;
}
public boolean contains(int i, double d) {
if (filled[i]==0) return false;
if (dvals[i]==d) return true;
return false;
}
public boolean contains(int i) {
if (filled[i]==0) return false;
return true;
}
public double groupby(int i) {
assert filled[i]==1;
return dvals[i];
}
public boolean insert(int i, double d) {
if (filled[i]==1 && contains(i,d)) return false;
if (isEmpty) isEmpty=false;
filled[i]=1;
dvals[i] = d;
return true;
}
public boolean update(int i, double d) {
assert filled[i]==1;
dvals[i]=d;
return true;
}
}
class MyVisitor {
public static final int NUM=128;
int[] range = new int[2];
Table1 table1;
HeadTable head;
double diff=0;
int i;
int iv;
String sv;
MyVisitor(Table1 _table1, HeadTable _head, int id) {
table1 = _table1;
head = _head;
int elems=Table1.SIZE1/NUM;
range[0] = elems*id;
range[1] = elems*(id+1);
}
public void run() {
table1.iterate_range(range[0], range[1], this);
}
//YYY 1: with double argument, this function is called
public boolean visit(int _i, int _v) {
i = _i;
iv = _v;
insertDiff();
return true;
}
//YYY 2: with String argument, this function is called
public boolean visit_obj(int _i, Object _v) {
i = _i;
iv = 42;
sv = (String)_v;
insertDiff();
return true;
}
public boolean insertDiff() {
if (!head.contains(i)) {
head.insert(i, diff);
return true;
}
double old = head.groupby(i);
double newval=Math.min(old, diff);
head.update(i, newval);
head.insert(i, diff);
return true;
}
}
public class ParTest1 {
public static int THREAD_NUM=4;
public static void main(String[] args) throws Exception {
if (args.length>0) {
THREAD_NUM = Integer.parseInt(args[0]);
System.out.println("Setting THREAD_NUM:"+THREAD_NUM);
}
Table1 table1 = new Table1(Table1.SIZE1);
HeadTable head = new HeadTable(Table1.SIZE1);
MyVisitor[] visitors = new MyVisitor[MyVisitor.NUM];
for (int i=0; i<visitors.length; i++) {
visitors[i] = new MyVisitor(table1, head, i);
}
int taskPerThread = visitors.length / THREAD_NUM;
MyThread[] threads = new MyThread[THREAD_NUM];
CyclicBarrier barrier = new CyclicBarrier(THREAD_NUM+1);
for (int i=0; i<THREAD_NUM; i++) {
threads[i] = new MyThread(barrier);
for (int j=taskPerThread*i; j<taskPerThread*(i+1); j++) {
if (j>=visitors.length) break;
threads[i].addVisitors(visitors[j]);
}
}
Runtime r=Runtime.getRuntime();
System.out.println("Force running gc");
r.gc(); // running GC here (excluding GC effect)
System.out.println("Running gc done");
// not measuring 1st run (excluding JIT compilation effect)
for (int i=0; i<THREAD_NUM; i++) {
threads[i].start();
}
barrier.await();
for (int i=0; i<10; i++) {
MyThread.start = true;
long s=System.currentTimeMillis();
barrier.await();
long e=System.currentTimeMillis();
System.out.println("Iter "+i+" Exec time:"+(e-s)/1000.0+"s");
}
}
}
class MyThread extends Thread {
static volatile boolean start=true;
static int tid=0;
int id=0;
ArrayList<MyVisitor> tasks;
CyclicBarrier barrier;
public MyThread(CyclicBarrier _barrier) {
super("MyThread"+(tid++));
barrier = _barrier;
id=tid;
tasks = new ArrayList(256);
}
void addVisitors(MyVisitor v) {
tasks.add(v);
}
public void run() {
while (true) {
while (!start) { ; }
for (int i=0; i<tasks.size(); i++) {
MyVisitor v=tasks.get(i);
v.run();
}
start = false;
try { barrier.await();}
catch (InterruptedException e) { break; }
catch (Exception e) { throw new RuntimeException(e); }
}
}
}
The Java code can be compiled with no dependency, and you can run it with the following command:
java -Darg.type=double -server ParTest1 2
You pass the number of worker threads as an argument (the above uses 2 threads).
After setting up the arrays (that is excluded from the measured time), it does a same operation for 10 times, printing out the execution time at each iteration.
With the above option, it uses double array, and it scales very well with 1,2,4 threads (i.e. the execution time reduces to 1/2, and 1/4), but
java -Darg.type=Object -server ParTest1 2
With this option, it uses Object (String) array, and it doesn't scale at all!
I measured the GC time, but it was insignificant (and I also forced running GC before measuring times). I have tested with Java 6 (updates 43) and Java 7 (updates 51), but it's the same.
The code has comments with XXX and YYY describing the difference when arg.type=double or arg.type=Object option is used.
Can you figure out what is going on with the String-type argument passing here?

HotSpot VM generates following assemblies for reference type putfield bytecode.
mov ref, OFFSET_OF_THE_FIELD(this) <- this puts the new value for field.
mov this, REGISTER_A
shr 0x9, REGISTER_A
movabs OFFSET_X, REGISTER_B
mov %r12b, (REGISTER_A, REGISTER_B, 1)
putfield operation is completed in 1 instruction.
but there are more instructions following.
They are "Card Marking" instructions. (http://www.ibm.com/developerworks/library/j-jtp11253/)
Writing reference field to every objects in a card (512 bytes), will store a value in a same memory address.
And I guess, store to same memory address from multiple cores mess up with cache and pipelines.
just add
byte[] garbage = new byte[600];
to MyVisitor definition.
then every MyVisitor instances will be spaced enough not to share card marking bit, you will see the program scales.

This is not a complete answer but may provide a hint for you.
I have changed your code
Table1(int size) {
filled = new byte[size];
ivals = new int[size];
strs = new String[size];
Arrays.fill(filled, (byte)1);
Arrays.fill(ivals, 42);
Arrays.fill(strs, "Strs");
}
to
Table1(int size) {
filled = new byte[size];
ivals = new int[size];
strs = new String[size];
Arrays.fill(filled, (byte)1);
Arrays.fill(ivals, 42);
Arrays.fill(strs, new String("Strs"));
}
after this change, the running time with 4 threads with object type array reduced.

According to http://docs.oracle.com/javase/specs/jls/se7/html/jls-17.html#jls-17.7
For the purposes of the Java programming language memory model, a single write to a non-volatile long or double value is treated as two separate writes: one to each 32-bit half. This can result in a situation where a thread sees the first 32 bits of a 64-bit value from one write, and the second 32 bits from another write.
Writes and reads of volatile long and double values are always atomic.
Writes to and reads of references are always atomic, regardless of whether they are implemented as 32-bit or 64-bit values.
Assigning references are always atomic,
and double is not atomic except when it is defined as volatile.
The problem is sv can be seen by other threads and its assignment is atomic.
Therefore, wrapping visitor's member variables (i, iv, sv) using ThreadLocal will solve the problem.

"sv = (String)_v;" makes the difference. I also confirmed that the type casting is not the factor. Just accessing _v can't make the difference. Assigning some value to sv field makes the difference. But I can't explain why.

Having trouble understanding how to maintain state using classes

I'm new to using OOP, I typically just put all my code in a single class and use methods. But I want to maintain state information and think classes are the best fit but I'm having trouble wrapping my head around it.
Say I have a list of items and I want to stop when the total sum of all previous items in the list equals X(in this case 10 so it takes item 1 + 2, then 2+3.etc..until it hits the threshold 10), I can use a method to calculate it but it involves me doing the entire process all over again when all I really need to do is increment by the last item and then see if my data exceeds the threshold. Here's my code so far but I know its not good because although it works its really just using the class as an independent method and recalculating on every loop. My goal is to,using this structure, reduce loops if not necessary to check thresholds.
Any suggestions?
Code:
public class LearningClassesCounter {
public static void main(String[] args) {
int[] list = new int[]{1,2,3,4,5,6,7,8,9,10};
int[] data_list = new int[list.length];
for (int current_location = 0; current_location<list.length;current_location++) {
//can only put commands in here. Nothing above.
Counter checker = new Counter(data_list);
System.out.println(checker.check_data(current_location));
for (int i =0; i<100; i++){
if (checker.check_data(current_location) == false) {
break;
}
data_list[current_location] = (list[current_location]+1); //this is just a random function, it could be any math function I just put it in here to show that some work is being done.
}
}
//its done now lets print the results
for (Integer item : data_list) {
System.out.println(item);
}
}
}
class Counter {
private int[] data_list;
private int total_so_far;
// create a new counter with the given parameters
public Counter(int[] data_list) {
this.data_list = data_list;
this.total_so_far = 0;
}
public boolean check_data(int current_location) {
// TODO Auto-generated method stub
int total_so_far = 0;
//System.out.println(total_so_far);
for (int item : data_list) {
total_so_far = item + total_so_far;
if (total_so_far >= 10) {
break;
}
}
if (total_so_far>=10) {
return false;
} else {
return true;
}
}
}
I don't need anyone to fix my code or anything(I want to do it myself, the code is just to give an idea of what I'm doing). I'm more interested in the flaw in my logic and maybe a way for me to better think about designing classes so I can apply them to my own situations better.

So the solution is that you do not update the data_list directly. Instead have a setter method in the Counter class that takes the index and value to update. It updates the value in the array and also updates a count value.
Something like this:
class Counter{
private final int[] list;
private count = 0;
private final maxCount = 10;
public Counter(int[] list){
this.list = list;
}
public boolean updateValueAndCheckPastMax(int index, int value){
list[index] = value;
count += value;
return count >= maxCount;
}
}

You are way over thinking this, and a counter class is not really necessary in this case.
I'm also interested as to why you'd be doing this line:
data_list[current_location] = (list[current_location]+1);
Do you want your data_list to be the same as list, but each value is incremented by 1?
If you are merely trying to return a sub-array of the values that are < 10, i would suggest just doing this in a for loop, and using an int as a counter.