Comp215: Performance 2: Java Internals

Transcription

1

2 Comp215: Performance 2: Java Internals Mack Joyner, Dan S. Wallach (Rice University) Copyright 2016, Mack Joyner, Dan S. Wallach. All rights reserved.

3 Recall: Collision Resolution Chaining: 0 Store all elements that hash to the same slot in a linked list Store a pointer to the head of the linked list in the hash table slot k 5 k 2 k 6 m 1 k 1 k 4 k 7 k 3 k 8 Open Addressing: All elements stored in hash table itself When collisions occur, use a systematic (consistent) procedure to store elements in free slots of the table

4 Recall: Collision Resolution Chaining: Store all elements that hash to the same slot in a linked list 0 k 1 k 4 k 5 k 2 k 6 k 7 k 3 Store a pointer to the head of the linked list in the hash table slot k 8 m 1 Open Addressing: All elements stored in hash table itself When collisions occur, use a systematic (consistent) Linear probing: procedure insert to store in next elements in closest following free location free slots of the table

5 Recall: Collision Resolution Chaining: Store all elements that hash to the same slot in a linked list 0 k 1 k 4 k 5 k 2 k 6 k 7 k 3 Store a pointer to the head of the linked list in the hash table slot Open Addressing: Hash function: h(x) = x % 10 k 8 m 1 All elements stored in hash table itself When collisions occur, use a systematic (consistent) Linear probing: procedure insert to store in next elements in closest following free location free slots of the table

6 Recall: Collision Resolution Chaining: Store all elements that hash to the same slot in a linked list 0 k 1 k 4 k 5 k 2 k 6 k 7 k 3 Store a pointer to the head of the linked list in the hash table slot Open Addressing: Hash function: h(x) = x % 10 All elements stored in hash table itself 12 k 8 m 1 When collisions occur, use a systematic (consistent) Linear probing: procedure insert to store in next elements in closest following free location free slots of the table

7 Recall: Collision Resolution Chaining: Store all elements that hash to the same slot in a linked list 0 k 1 k 4 k 5 k 2 k 6 k 7 k 3 Store a pointer to the head of the linked list in the hash table slot Open Addressing: Hash function: h(x) = x % 10 All elements stored in hash table itself k 8 m 1 When collisions occur, use a systematic (consistent) Linear probing: procedure insert to store in next elements in closest following free location free slots of the table

8 Recall: Collision Resolution Chaining: Store all elements that hash to the same slot in a linked list 0 k 1 k 4 k 5 k 2 k 6 k 7 k 3 Store a pointer to the head of the linked list in the hash table slot Open Addressing: Hash function: h(x) = x % 10 All elements stored in hash table itself k 8 m 1 When collisions occur, use a systematic (consistent) Linear probing: procedure insert to store in next elements in closest following free location free slots of the table

9 Recall: Collision Resolution Chaining: Store all elements that hash to the same slot in a linked list 0 k 1 k 4 k 5 k 2 k 6 k 7 k 3 Store a pointer to the head of the linked list in the hash table slot Open Addressing: Hash function: h(x) = x % 10 All elements stored in hash table itself When collisions occur, use a systematic (consistent) Linear probing: procedure insert to store in next elements in closest following free location free slots of the table k 8 m

10 Recall: Collision Resolution Chaining: Store all elements that hash to the same slot in a linked list 0 k 1 k 4 k 5 k 2 k 6 k 7 k 3 Store a pointer to the head of the linked list in the hash table slot Open Addressing: Hash function: h(x) = x % 10 All elements stored in hash table itself When collisions occur, use a systematic (consistent) Linear probing: procedure insert to store in next elements in closest following free location free slots of the table k 8 m

11 Recall: Collision Resolution Chaining: Store all elements that hash to the same slot in a linked list Store a pointer to the head of the linked list in the hash table slot Open Addressing: Hash function: h(x) = x % 10 All elements stored in hash table itself When collisions occur, use a systematic (consistent) Linear probing: procedure insert to store in next elements in closest following free location free slots of the table k 5 k 2 k 6 m 1 k 1 k 4 k 7 k 3 k 8 Linear probing: delete and move collision chain

12 Recall: Collision Resolution Chaining: Store all elements that hash to the same slot in a linked list Store a pointer to the head of the linked list in the hash table slot Open Addressing: Hash function: h(x) = x % 10 All elements stored in hash table itself When collisions occur, use a systematic (consistent) Linear probing: procedure insert to store in next elements in closest following free location free slots of the table k 5 k 2 k 6 m 1 k 1 k 4 k 7 k 3 k 8 Linear probing: delete and move collision chain

13 Recall: Collision Resolution Chaining: Store all elements that hash to the same slot in a linked list Store a pointer to the head of the linked list in the hash table slot Open Addressing: Hash function: h(x) = x % 10 All elements stored in hash table itself When collisions occur, use a systematic (consistent) Linear probing: procedure insert to store in next elements in closest following free location free slots of the table k 5 k 2 k 6 m 1 k 1 k 4 k 7 k 3 k 8 Linear probing: delete and move collision chain initial search for 22

14 Recall: Collision Resolution Chaining: Store all elements that hash to the same slot in a linked list Store a pointer to the head of the linked list in the hash table slot Open Addressing: Hash function: h(x) = x % 10 All elements stored in hash table itself When collisions occur, use a systematic (consistent) Linear probing: procedure insert to store in next elements in closest following free location free slots of the table k 5 k 2 k 6 m 1 k 1 k 4 k 7 k 3 k 8 Linear probing: delete and move collision chain initial search for 22

15 Recall: Collision Resolution Chaining: Store all elements that hash to the same slot in a linked list Store a pointer to the head of the linked list in the hash table slot Open Addressing: Hash function: h(x) = x % 10 All elements stored in hash table itself 0 22 k 5 k 2 k 6 m 1 32 k 1 k 4 k 7 k 3 k 8 Linear probing: delete and move collision chain initial search for 22 When collisions occur, use a systematic (consistent) Linear probing: procedure insert to store in next elements in closest following free location free slots of the table

16 Live Coding Traditional vs memoized fibonacci

17 Memory Hierarchy Computer Architecture, Fifth Edition: A Quantitative approach by John Hennessy and David Patterson

18 Java Technology The Java programming language The library (JDK) Required to build java files The Java virtual machine (JVM) An instruction set and the meaning of those instructions the bytecodes Needed to run java applications

19 Garbage Collection

20 Where do objects live? You ve ignored the issue of where objects, variables, and methods live The implementation (Python or Java) takes care of these details Fred Kate Julia System.out. println() Java World

21 Where do objects live? The Java System maps Java World onto Processor Resources The implementation (Python or Java) takes care of these details Java compiler and runtime handle mapping Fred Julia Kate System.out. println() Knowing how mapping works can helps understand behavior of program Find ways how to improve it Java World k RAM Processor Core Processor Core Processor Core

22 Fundamentals In this example, what needs storage? The two classes (Point & C) Point s local members (x, y, & draw) C s local members (s, t, & m) m s local variables (a, b, & p) class Point { public int x, y; public void draw(){}; } class C { int s, t; public void m() { int a, b; Point p = new Point(); a = ; b = ; p.draw(); } } A classic example

23 Fundamentals In this example, what needs storage? The two classes (Point & C) Point s local members (x, y, & draw) C s local members (s, t, & m) m s local variables (a, b, & p) Memory in the Java runtime is divided, broadly speaking, into a Heap and a collection of Stacks STACK 0 STACK 1 p: a: b: STACK 2 STACK n Point Hello world! C String Pool new point Heap class Point { public int x, y; public void draw(){}; } class C { int s, t; public void m() { int a, b; Point p = new Point(); a = ; b = ; p.draw(); } } A classic example

24 JVM Memory Layout Stacks Growth space for stacks Heap Globals Code

25 JVM Memory Layout Stacks Growth space for stacks Heap Globals Code created variables go on the thread s stack

26 JVM Memory Layout Stacks Growth space for stacks Heap Globals Code created variables go on the thread s stack created objects with new go on the heap

27 JVM Memory Layout Stacks Growth space for stacks Heap Globals Code created variables go on the thread s stack created objects with new go on the heap Used tail call optimization in past to reduce stack use

28 JVM Memory Layout Stacks Growth space for stacks Heap Globals Code created variables go on the thread s stack created objects with new go on the heap Use garbage collection at runtime to reclaim heap space Used tail call optimization in past to reduce stack use

29 Why Garbage Collection? Today s programs consume storage freely 1GB laptops, 1-4GB deskops, 8-512GB servers 64-bit address spaces (SPARC, Itanium, Opteron) and mismanage it Memory leaks, dangling references, double free, misaligned addresses, null pointer dereference, heap fragmentation Poor use of reference locality, resulting in high cache miss rates and/or excessive demand paging Explicit memory management breaks high-level programming abstraction Luckily, you don t have to do that in Java

30 Sustainable Memory Management (AKA Garbage Collection ) When the heap runs out of space, the system copes Scours the heap looking for objects that are no longer of interest Technical term is live An object is considered live iff it can be reached from the running code Modern garbage collectors are copying collectors

31 Garbage Collection via Copying A copying collector divides the heap into two or more pools New objects are allocated in the current pool When the current pool is full, execution pauses and the collector: copies all live objects from the current pool to the empty pool swaps the designations current and empty Unreachable objects are not copied, so the new pool has free space Current Pool HEAP Empty Pool

32 Garbage Collection via Copying A copying collector divides the heap into two or more pools New objects are allocated in the current pool When the current pool is full, execution pauses and the collector: copies all live objects from the current pool to the empty pool swaps the designations current and empty Unreachable objects are not copied, so the new pool has free space Current Pool HEAP Empty Pool BEFORE Objects left in the empty pool are discarded en masse Empty Pool HEAP Current Pool AFTER

33 Copying Collector Tradeoffs Good: very low cell allocation overhead Out-of-space check requires just an addr comparison Can efficiently allocate variable-sized cells Good: compacting Eliminates fragmentation, good locality of reference Bad: twice the memory footprint When copying, pages of both spaces need to be swapped in Could lead to lots of page faults

34 Generational Garbage Collector Java uses a slightly more complex copying collector All new objects are allocated into Eden Eden is copied, when full, into one of Stable 0 or Stable 1 When Stable is too full, it is added to the Long Term Pool minor swap This is an example of a Generational Garbage Collector HEAP Eden Stable 0 Stable 1 Long Term Pool Key insight: most objects that die, die young Divide heap into generations, GC the younger object more frequently major Periodically reap the older generations Minor collection Major collection Promote young objects to older pool if they survive several garbage collections

35 Implications for Programming If you want performance, pay attention to garbage collection When you are done with an object, make sure you re not referencing it anymore Leaving a reference to the heap object will keep it live Storage can leak, or become un-recyclable Leaving a reference to a large data structure on the stack, or in a global Leads to extra collections and, eventually, an out of memory error

36 Inline Cashing IList<String> emptylist = List.Empty.create(); IList<String> list = emptylist.add( Alice").add("Bob").add("Charlie");... assertequals(list.head(), Charlie );

37 Inline Cashing IList<String> emptylist = List.Empty.create(); IList<String> list = emptylist.add( Alice").add("Bob").add("Charlie");... assertequals(list.head(), Charlie ); IList.head()

38 Inline Cashing IList<String> emptylist = List.Empty.create(); IList<String> list = emptylist.add( Alice").add("Bob").add("Charlie");... assertequals(list.head(), Charlie ); IList.head() List.head()

39 Inline Cashing IList<String> emptylist = List.Empty.create(); IList<String> list = emptylist.add( Alice").add("Bob").add("Charlie");... assertequals(list.head(), Charlie ); + check List.head() Inline cashing of this method in HotSpot JVM

40 Java Profiling Find where in the code execution is spending the most time Improving program hot path could have big impact on overall performance Don t waste time trying to improve performance in rarely executed code Use optimization techniques to improve performance of hot path JIT compilation: compile at runtime to native code and run directly on hardware Static optimizations during compile time (e.g. tail call optimization)

41 Java Profiling

42 Engineering does matter Use efficient data structures Example: Use treap instead of a tree Avoid unnecessary indirection Makes the garbage collector s job more difficult Optimize the common path 80% of work is done by 20% of code