JVM Workshop

1. @jkubrynski / kubrynski.blog @jkubrynski / kubrynski.blog JVM DIVE FOR MERE MORTALS JVM DIVE FOR MERE MORTALS WORKSHOPS WORKSHOPS JAKUB KUBRYNSKI JAKUB KUBRYNSKI jakub@devskiller.com / @jkubrynski / https://kubrynski.blog
2. $ WHOAMI $ WHOAMI 14+ YEARS PROFESSIONAL EXPERIENCE 14+ YEARS PROFESSIONAL EXPERIENCE DEVSKILLER CO-FOUNDER DEVSKILLER CO-FOUNDER BOTTEGA TRAINER & AUDITOR BOTTEGA TRAINER & AUDITOR DEVOXX.PL PROGRAM COMMITTEE MEMBER DEVOXX.PL PROGRAM COMMITTEE MEMBER OPENSOURCE CONTRIBUTOR OPENSOURCE CONTRIBUTOR
3. WHO ARE YOU? WHO ARE YOU? Bytecode ? Hashtable vs ConcurrentHashMap ? JIT ? Concurrent Mark-Sweep ? Re ection vs MethodHandle ? Invokedynamic ?
4. PERFORMANCE TUNING PRINCIPLES PERFORMANCE TUNING PRINCIPLES measurement one change at a time deep understanding over stackover ow.com hints
5. JAVA VIRTUAL MACHINE JAVA VIRTUAL MACHINE
6. LIFE CYCLE LIFE CYCLE source -> javac -> bytecode bytecode -> classloader -> interpreter interpreter -> JIT -> optimized native code
7. SOURCE CODE SOURCE CODE package com.random.company.app; public class StringUtilsHelper { public boolean isEmpty(String str) { return str != null && str.length() > 0; } }
8. JAVAC JAVAC converts source code into byte code checks simple optimizations
9. CLASS FILE CLASS FILE ClassFile u4 magic; // CAFEBABE u2 minor_version; u2 major_version; u2 constant_pool_count; cp_info constant_pool[constant_pool_count-1]; u2 access_flags; u2 this_class; u2 super_class; u2 interfaces_count; u2 interfaces[interfaces_count]; u2 fields_count; field_info fields[fields_count]; u2 methods_count; method_info methods[methods_count]; u2 attributes count;
10. BYTECODE BYTECODE list of operation codes $xxd -p Test.class ...1b04a0000504ac2a1b0464b600021b68ac... 1b => 27 => iload_1 04 => 4 => iconst_1 a0 => 160 => if_icmpne 7 04 => 4 => iconst_1 ac => 172 => ireturn 2a => 42 => aload_0 1b => 27 => iload_1 04 => 4 => iconst_1 64 => 100 => isub b6 => 182 => invokevirtual #5 1b => 27 => iload_1 68 => 104 => imul ac => 172 => ireturn
11. BYTECODE - MORE INFO BYTECODE - MORE INFO current usage: 239/255 $JDK/src/hotspot/share/interpreter/bytecodes.hpp class sun.jvm.hotspot.interpreter.Bytecodes http://www.javaworld.com/article/2077233/core-java/bytecode- basics.html
12. JAVAC DEMO JAVAC DEMO HTTPS://GITHUB.COM/JKUBRYNSKI/JVM-WORKSHOP HTTPS://GITHUB.COM/JKUBRYNSKI/JVM-WORKSHOP
13. CLASSLOADER CLASSLOADER dynamically loads classes hierarchies Bootstrap classloader Extension classloader Application classloader Custom classloader
14. CLASSLOADING PHASES CLASSLOADING PHASES loading -> reads class le linking verifying -> veri es bytecode correctness preparing -> allocates memory resolving -> links with classes, interfaces, elds, methods initializing -> static initializers
15. INTERPRETER INTERPRETER detects the critical hot spots in the program template interpreter java -XX:+UnlockDiagnosticVMOptions -XX:+PrintInterpreter -version
16. INTERPRETER - MORE INFO INTERPRETER - MORE INFO $JDK/src/hotspot/share/interpreter/templateTable.cpp#TemplateTable::initialize https://openjdk.java.net/groups/hotspot/docs/RuntimeOverview.html#Interpreter|outline
17. JIT JIT Just-In-Time optimizes code compiles methods into native code -client (C1) / -server (C2) runs up to 20 times faster
18. INLINING INLINING public String getStringFromSupplier(Supplier<String> supplier) { return supplier.get(); } public String businessMethod(String param) { Supplier<String> stringSupplier = new StringSupplier(”my” + param); return getStringFromSupplier(stringSupplier); } // turns to public String businessMethod(String param) { Supplier<String> stringSupplier = new StringSupplier(”my” + param); return stringSupplier.get(); }
19. UNROLLING UNROLLING private static String[] options = {"yes", "no", "true", "false"} public void someMethod() { for (String opt : options) { process(opt); } } //turns into public void someMethod() { process("yes"); process("no"); process("true"); process("false"); }
20. SCALAR REPLACEMENT SCALAR REPLACEMENT public record(int x, int y) { Point point = new Point(x, y); storePoint(point); } // inlining public record(int x, int y) { Point point = new Point(x, y); events.store("Added point", point.x, point.y); } // scalar replacement public record(int x, int y) { events.store("Added point", x, y); }
21. DEAD CODE ELIMINATION DEAD CODE ELIMINATION public void myMethod() { for (int i = 0; i < THRESHOLD; i++) { new String("test"); } } // turns into public void myMethod() { }
22. LOCK ELISION LOCK ELISION public void process(List<User> users) { List<User> result = new ArrayList<>(); synchronized(result) { fillResult(users); } } //turns into public void process(List<User> users) { List<User> result = new ArrayList<>(); fillResult(users); }
23. TYPE SHARPENING TYPE SHARPENING List<User> users = new ArrayList<>(); // turns into ArrayList<User> users = new ArrayList<>();
24. ON STACK REPLACEMENT ON STACK REPLACEMENT happens when the interpreter discovers that a method is looping converts an interpreted stack frame into a native compiled stack frame
25. TIERED COMPILATION TIERED COMPILATION LEVELS LEVELS 0: Interpreted code 1: Simple C1 compiled code 2: Limited C1 compiled code 3: Full C1 compiled code 4: C2 compiled code
26. WHY SHOULD I CARE? WHY SHOULD I CARE? JIT does most of the optimizations we could do manually without "obfuscating" source code Performance/load tests should run only on "hot" application
27. HOW TO TRACK? HOW TO TRACK? When after restarting your app is at the full speed? $ jstat -compiler <PID> 1s // or JDK9+ => -Xlog:jit*=debug JDK8 => -XX:+PrintCompilation
28. -XLOG:JIT -XLOG:JIT $JDK/src/hotspot/share/compiler/compileTask.cpp#CompileTask::print_impl const char compile_type = is_osr_method ? '%' : ' '; const char sync_char = is_synchronized ? 's' : ' '; const char exception_char = has_exception_handler ? '!' : ' '; const char blocking_char = is_blocking ? 'b' : ' '; const char native_char = is_native ? 'n' : ' '; compile_id, compile_type, sync_char, exception_char, blocking_char, native_char, comp_level [0,090s][debug][jit,compilation] 104 3 java.lang.module.ModuleDescriptor$Exports::isQualified (18 bytes) [0,096s][debug][jit,compilation] 122 4 java.lang.module.ModuleDescriptor$Exports::isQualified (18 bytes) [0,097s][debug][jit,compilation] 104 3 java.lang.module.ModuleDescriptor$Exports::isQualified (18 bytes) made not entrant [0,032s][debug][jit,compilation] 1 3 java.util.concurrent.ConcurrentHashMap::tabAt (22 bytes) [0,032s][debug][jit,inlining ] @ 15 jdk.internal.misc.Unsafe::getObjectAcquire (7 bytes) [0,059s][debug][jit,compilation] 35 n 0 java.lang.Object::hashCode (native) [0,065s][debug][jit,compilation] 41 ! 3 java.util.concurrent.ConcurrentHashMap::putVal (432 bytes)
29. JIT DEMO JIT DEMO
30. EXECUTION COMPONENTS EXECUTION COMPONENTS program counter frame stack
31. STACK TRACE STACK TRACE "main@1" prio=5 tid=0x1 nid=NA runnable java.lang.Thread.State: RUNNABLE at io.codearte.BlockBuilder.startBlock(BlockBuilder.groovy:21) at io.codearte.Generator.process(Generator.java:318) at io.codearte.ImportantApp.do(ImportantApp.java:64) at sun.reflect.NativeMethodImpl.invoke(NativeMethodImpl.java:18) at sun.reflect.NativeMethodImpl.invoke(NativeMethodImpl.java:62) at java.lang.reflect.Method.invoke(Method.java:497)
32. DEBUGGING DEBUGGING
33. DEBUGGING DEBUGGING
34. MEMORY LAYOUT MEMORY LAYOUT
35. OBJECT LAYOUT OBJECT LAYOUT com.eshop.model.Product object internals: OFFSET SIZE TYPE DESCRIPTION VALUE 0 12 (object header) N/A 12 4 int Product.id N/A 16 4 String Product.name N/A 20 4 (loss due to the next object alignment) Instance size: 24 bytes (estimated, the sample instance is not available) Space losses: 0 bytes internal + 4 bytes external = 4 bytes total
36. OBJECT LAYOUT OBJECT LAYOUT com.eshop.model.Product object internals: OFFSET SIZE TYPE DESCRIPTION VALUE 0 12 (object header) N/A 12 4 int Product.id N/A 16 4 int Product.price N/A 20 4 String Product.name N/A Instance size: 24 bytes (estimated, the sample instance is not available) Space losses: 0 bytes internal + 0 bytes external = 0 bytes total
37. OBJECT LAYOUT OBJECT LAYOUT com.eshop.model.Product object internals: OFFSET SIZE TYPE DESCRIPTION VALUE 0 12 (object header) N/A 12 4 int Product.id N/A 16 4 int Product.price N/A 20 1 boolean Product.available N/A 21 3 (alignment/padding gap) N/A 24 4 String Product.name N/A 28 4 (loss due to the next object alignment) Instance size: 32 bytes (estimated, the sample instance is not available) Space losses: 3 bytes internal + 4 bytes external = 7 bytes total
38. OBJECT LAYOUT OBJECT LAYOUT com.eshop.model.Product object internals: OFFSET SIZE TYPE DESCRIPTION VALUE 0 16 (object header) N/A 16 4 int Product.id N/A 20 4 int Product.price N/A 24 1 boolean Product.available N/A 25 7 (alignment/padding gap) N/A 32 8 String Product.name N/A Instance size: 40 bytes (estimated, the sample instance is not available) Space losses: 3 bytes internal + 0 bytes external = 3 bytes total
39. JAVA OBJECT LAYOUT JAVA OBJECT LAYOUT https://openjdk.java.net/projects/code-tools/jol/
40. JOL DEMO JOL DEMO
41. GARBAGE COLLECTOR GARBAGE COLLECTOR cleans memory important performance factor vector algorithm stop the world in safepoints
42. GC ALGORITHMS GC ALGORITHMS Serial Parallel Concurrent Mark Sweep G1 Epsilon* (No-Op) - JEP 318 ZGC* (Low Latency) - JEP 333 * Experimental
43. VECTOR ALGORITHM VECTOR ALGORITHM mark_gc_roots() for (each_root_object) { mark_all_referenced_objects() } for (each_object_in_memory) { if (is_marked_as_reacheable) { unmark_the_object() } else { remove_object_and_reclaim_memory() } }
44. GENERATIONAL HYPOTESIS GENERATIONAL HYPOTESIS Infant mortality young objects are much more probable to die Idea is to process both generations separately Tenuring threshold
45. GENERATIONAL HYPOTESIS GENERATIONAL HYPOTESIS Does not work for caches Do not keep caches in the same heap as the application Go off-heap
46. DISPATCH TYPES DISPATCH TYPES invokevirtual invokestatic invokespecial invokedynamic
47. LAMBDAS LAMBDAS generated by javac bootstraped by LambdaMetafactory called with invokedynamic
48. LAMBDA UNDER THE HOOD LAMBDA UNDER THE HOOD BigDecimal sumCreditEntries(Client client) { return sumEntries(client.getAccounts(), account -> account.getCreditEntries()); } private static java.util.List lambda$sumCreditEntries$0(com.sandbox.Account); private Period period; BigDecimal sumCreditEntries(Client client) { return sumEntries(client.getAccounts(), account -> account.getCreditEntries(period)); } private java.util.List lambda$sumCreditEntries$0(com.sandbox.Account); BigDecimal sumCreditEntries(Client client, Period period) { return sumEntries(client.getAccounts(), account -> account.credit(period)); } private static java.util.List lambda$sumCreditEntries$0 (java.time.Period, com.sandbox.Account);
49. METHOD REFERENCE METHOD REFERENCE SIMILAR TO LAMBDAS, BUT NO NEED TO GENERATE A METHOD SIMILAR TO LAMBDAS, BUT NO NEED TO GENERATE A METHOD BECAUSE WE'RE CALLING A METHOD BECAUSE WE'RE CALLING A METHOD
50. BENCHMARKS BENCHMARKS CallTypes.baseline avgt 30 4.163 ± 0.009 ns/op CallTypes.lambda avgt 30 4.174 ± 0.015 ns/op CallTypes.methodRef avgt 30 4.244 ± 0.049 ns/op CallTypesExternal.baseline avgt 30 50.055 ± 0.275 ns/op CallTypesExternal.lambda avgt 30 50.980 ± 0.650 ns/op CallTypesExternal.methodRef avgt 30 50.655 ± 0.376 ns/op
51. METHODHANDLES METHODHANDLES
52. METHODHANDLES METHODHANDLES is not a replacement for re ection (re ection is more for introspection of classes) re ection does access control during invocation while MethodHandle checks all during lookup
53. EXAMPLE EXAMPLE MethodHandle toUpperCase = MethodHandles.lookup() .findVirtual(String.class, "toUpperCase", MethodType.methodType(String.class)) Object result = toUpperCase.invoke("test")); String result = (String) toUpperCase.invokeExact("test"));
54. TIME TO BENCHMARK TIME TO BENCHMARK
55. BENCHMARKS BENCHMARKS 1ns = 0.000 001 ms = 0.000 000 001 s Benchmark Mode Cnt Score Error Units Invocations.baseline avgt 30 35.761 ± 0.113 ns/op Invocations.reflectionWithoutLookup avgt 30 41.413 ± 0.223 ns/op Invocations.handleWithoutLookup avgt 30 42.002 ± 1.002 ns/op Invocations.handleExactWithoutLookup avgt 30 38.134 ± 0.153 ns/op Invocations.reflection avgt 30 71.207 ± 1.558 ns/op Invocations.handle avgt 30 858.148 ± 18.135 ns/op
56. EXCEPTIONS EXCEPTIONS public class ClientAlreadyExistsException extends Throwable { @Override public synchronized Throwable fillInStackTrace() { return this; } }
57. EXCEPTIONS EXCEPTIONS DEEP MEANS THERE ARE 4 MORE FRAMES DEEP MEANS THERE ARE 4 MORE FRAMES Benchmark Mode Cnt Score Error Units Exceptions.standardExcept avgt 30 1029.919 ± 5.026 ns/op Exceptions.standardExceptDeep avgt 30 1121.771 ± 6.615 ns/op Exceptions.stacklessExcept avgt 30 18.827 ± 0.066 ns/op Exceptions.stacklessExceptDeep avgt 30 19.835 ± 0.053 ns/op
58. STREAMS STREAMS TIME TO BENCHMARK TIME TO BENCHMARK
59. FURTHER READING FURTHER READING Optimizing Java - Benjamin J Evans, James Gough The Well-Grounded Java Developer - Benjamin J. Evans, Martijn Verburg Java Performance - Charlie Hunt, Binu John Java Performance: The De nitive Guide - Scott Oaks
60. TOOLS TOOLS jdk Visual VM Mission Control JPro ler Honest Pro ler Java Object Layout
61. I WANT MORE! I WANT MORE! THE JAVA® VIRTUAL MACHINE SPECIFICATION THE JAVA® VIRTUAL MACHINE SPECIFICATION HG CLONE HG CLONE HTTP://HG.OPENJDK.JAVA.NET/JDK11/JDK11/ HTTP://HG.OPENJDK.JAVA.NET/JDK11/JDK11/ HTTP://OPENJDK.JAVA.NET/PROJECTS/CODE-TOOLS/JMH HTTP://OPENJDK.JAVA.NET/PROJECTS/CODE-TOOLS/JMH
62. BENCHMARKS BENCHMARKS HTTPS://GITHUB.COM/JKUBRYNSKI/JVM-WORKSHOP HTTPS://GITHUB.COM/JKUBRYNSKI/JVM-WORKSHOP
63. QUESTIONS? QUESTIONS?
64. THANKS! THANKS!