org.openjdk.jmh.infra.Blackhole Maven / Gradle / Ivy
/*
* Copyright (c) 2005, 2013, Oracle and/or its affiliates. All rights reserved.
* DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
*
* This code is free software; you can redistribute it and/or modify it
* under the terms of the GNU General Public License version 2 only, as
* published by the Free Software Foundation. Oracle designates this
* particular file as subject to the "Classpath" exception as provided
* by Oracle in the LICENSE file that accompanied this code.
*
* This code is distributed in the hope that it will be useful, but WITHOUT
* ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
* FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
* version 2 for more details (a copy is included in the LICENSE file that
* accompanied this code).
*
* You should have received a copy of the GNU General Public License version
* 2 along with this work; if not, write to the Free Software Foundation,
* Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
*
* Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
* or visit www.oracle.com if you need additional information or have any
* questions.
*/
package org.openjdk.jmh.infra;
import org.openjdk.jmh.util.Utils;
import java.lang.ref.WeakReference;
import java.util.Random;
/*
See the rationale for BlackholeL1..BlackholeL4 classes below.
*/
abstract class BlackholeL0 {
private int markerBegin;
}
abstract class BlackholeL1 extends BlackholeL0 {
private boolean p001, p002, p003, p004, p005, p006, p007, p008;
private boolean p011, p012, p013, p014, p015, p016, p017, p018;
private boolean p021, p022, p023, p024, p025, p026, p027, p028;
private boolean p031, p032, p033, p034, p035, p036, p037, p038;
private boolean p041, p042, p043, p044, p045, p046, p047, p048;
private boolean p051, p052, p053, p054, p055, p056, p057, p058;
private boolean p061, p062, p063, p064, p065, p066, p067, p068;
private boolean p071, p072, p073, p074, p075, p076, p077, p078;
private boolean p101, p102, p103, p104, p105, p106, p107, p108;
private boolean p111, p112, p113, p114, p115, p116, p117, p118;
private boolean p121, p122, p123, p124, p125, p126, p127, p128;
private boolean p131, p132, p133, p134, p135, p136, p137, p138;
private boolean p141, p142, p143, p144, p145, p146, p147, p148;
private boolean p151, p152, p153, p154, p155, p156, p157, p158;
private boolean p161, p162, p163, p164, p165, p166, p167, p168;
private boolean p171, p172, p173, p174, p175, p176, p177, p178;
}
abstract class BlackholeL2 extends BlackholeL1 {
public volatile byte b1;
public volatile boolean bool1;
public volatile char c1;
public volatile short s1;
public volatile int i1;
public volatile long l1;
public volatile float f1;
public volatile double d1;
public byte b2;
public boolean bool2;
public char c2;
public short s2;
public int i2;
public long l2;
public float f2;
public double d2;
public volatile Object obj1;
public volatile BlackholeL2 nullBait = null;
public int tlr;
public volatile int tlrMask;
public BlackholeL2() {
Random r = new Random(System.nanoTime());
tlr = r.nextInt();
tlrMask = 1;
obj1 = new Object();
b1 = (byte) r.nextInt(); b2 = (byte) (b1 + 1);
bool1 = r.nextBoolean(); bool2 = !bool1;
c1 = (char) r.nextInt(); c2 = (char) (c1 + 1);
s1 = (short) r.nextInt(); s2 = (short) (s1 + 1);
i1 = r.nextInt(); i2 = i1 + 1;
l1 = r.nextLong(); l2 = l1 + 1;
f1 = r.nextFloat(); f2 = f1 + Math.ulp(f1);
d1 = r.nextDouble(); d2 = d1 + Math.ulp(d1);
if (b1 == b2) {
throw new IllegalStateException("byte tombstones are equal");
}
if (bool1 == bool2) {
throw new IllegalStateException("boolean tombstones are equal");
}
if (c1 == c2) {
throw new IllegalStateException("char tombstones are equal");
}
if (s1 == s2) {
throw new IllegalStateException("short tombstones are equal");
}
if (i1 == i2) {
throw new IllegalStateException("int tombstones are equal");
}
if (l1 == l2) {
throw new IllegalStateException("long tombstones are equal");
}
if (f1 == f2) {
throw new IllegalStateException("float tombstones are equal");
}
if (d1 == d2) {
throw new IllegalStateException("double tombstones are equal");
}
}
}
abstract class BlackholeL3 extends BlackholeL2 {
private boolean q001, q002, q003, q004, q005, q006, q007, q008;
private boolean q011, q012, q013, q014, q015, q016, q017, q018;
private boolean q021, q022, q023, q024, q025, q026, q027, q028;
private boolean q031, q032, q033, q034, q035, q036, q037, q038;
private boolean q041, q042, q043, q044, q045, q046, q047, q048;
private boolean q051, q052, q053, q054, q055, q056, q057, q058;
private boolean q061, q062, q063, q064, q065, q066, q067, q068;
private boolean q071, q072, q073, q074, q075, q076, q077, q078;
private boolean q101, q102, q103, q104, q105, q106, q107, q108;
private boolean q111, q112, q113, q114, q115, q116, q117, q118;
private boolean q121, q122, q123, q124, q125, q126, q127, q128;
private boolean q131, q132, q133, q134, q135, q136, q137, q138;
private boolean q141, q142, q143, q144, q145, q146, q147, q148;
private boolean q151, q152, q153, q154, q155, q156, q157, q158;
private boolean q161, q162, q163, q164, q165, q166, q167, q168;
private boolean q171, q172, q173, q174, q175, q176, q177, q178;
}
abstract class BlackholeL4 extends BlackholeL3 {
private int markerEnd;
}
/**
* Black Hole.
*
* Black hole "consumes" the values, conceiving no information to JIT whether the
* value is actually used afterwards. This can save from the dead-code elimination
* of the computations resulting in the given values.
*/
public final class Blackhole extends BlackholeL4 {
/*
* IMPLEMENTATION NOTES:
*
* The major things to dodge with Blackholes are:
*
* a) Dead-code elimination: the arguments should be used on every call,
* so that compilers are unable to fold them into constants or
* otherwise optimize them away along with the computations resulted
* in them.
*
* b) False sharing: reading/writing the state may disturb the cache
* lines. We need to isolate the critical fields to achieve tolerable
* performance.
*
* c) Write wall: we need to ease off on writes as much as possible,
* since it disturbs the caches, pollutes the write buffers, etc.
* This may very well result in hitting the memory wall prematurely.
* Reading memory is fine as long as it is cacheable.
*
* To achieve these goals, we are piggybacking on several things in the
* compilers:
*
* 1. Superclass fields are not reordered with the subclass' fields.
* No practical VM that we are aware of is doing this. It is unpractical,
* because if the superclass fields are at the different offsets in two
* subclasses, the VMs would then need to do the polymorphic access for
* the superclass fields.
*
* 2. Compilers are unable to predict the value of the volatile read.
* While the compilers can speculatively optimize until the relevant
* volatile write happens, it is unlikely to be practical to be able to stop
* all the threads the instant that write had happened.
*
* 3. Compilers' code motion usually respects data dependencies, and they would
* not normally schedule the consumer block before the code that generated
* a value.
*
* 4. Compilers are not doing aggressive inter-procedural optimizations,
* and/or break them when the target method is forced to be non-inlineable.
*
* Observation (1) allows us to "squash" the protected fields in the inheritance
* hierarchy so that the padding in super- and sub-class are laid out right before
* and right after the protected fields. We also pad with booleans so that dense
* layout in superclass does not have the gap where runtime can fit the subclass field.
*
* Observation (2) allows us to compare the incoming primitive values against
* the relevant volatile-guarded fields. The values in those guarded fields are
* never changing, but due to (2), we should re-read the values again and again.
* Also, observation (3) requires us to to use the incoming value in the computation,
* thus anchoring the Blackhole code after the generating expression.
*
* Primitives are a bit hard, because we can't predict what values we
* will be fed. But we can compare the incoming value with *two* distinct
* known values, and both checks will never be true at the same time.
* Note the bitwise AND in all the predicates: both to spare additional
* branch, and also to provide more uniformity in the performance. Where possible,
* we are using a specific code shape to force generating a single branch, e.g.
* making compiler to evaluate the predicate in full, not speculate on components.
*
* Objects should normally abide the Java's referential semantics, i.e. the
* incoming objects will never be equal to the distinct object we have, and
* volatile read will break the speculation about what we compare with.
* However, smart compilers may deduce that the distinct non-escaped object
* on the other side is not equal to anything we have, and fold the comparison
* to "false". We do inlined thread-local random to get those objects escaped
* with infinitesimal probability. Then again, smart compilers may skip from
* generating the slow path, and apply the previous logic to constant-fold
* the condition to "false". We are warming up the slow-path in the beginning
* to evade that effect. Some caution needs to be exercised not to retain the
* captured objects forever: this is normally achieved by calling evaporate()
* regularly, but we also additionally protect with retaining the object on
* weak reference (contrary to phantom-ref, publishing object still has to
* happen, because reference users might need to discover the object).
*
* Observation (4) provides us with an opportunity to create a safety net in case
* either (1), (2) or (3) fails. This is why Blackhole methods are prohibited from
* being inlined. This is treated specially in JMH runner code (see CompilerHints).
* Conversely, both (1), (2), (3) are covering in case (4) fails. This provides
* a defense in depth for Blackhole implementation, where a point failure is a
* performance nuisance, but not a correctness catastrophe.
*
* In all cases, consumes do the volatile reads to have a consistent memory
* semantics across all consume methods.
*
* There is an experimental compiler support for Blackholes that instructs compilers
* to treat specific methods as blackholes: keeping their arguments alive. At some
* point in the future, we hope to switch to that mode by default, thus greatly
* simplifying the Blackhole code.
*
* An utmost caution should be exercised when changing the Blackhole code. Nominally,
* the JMH Core Benchmarks should be run on multiple platforms (and their generated code
* examined) to check the effects are still in place, and the overheads are not prohibitive.
* Or, in other words:
*
* IMPLEMENTING AN EFFICIENT / CORRECT BLACKHOLE IS NOT A SIMPLE TASK YOU CAN
* DO OVERNIGHT. IT REQUIRES A SIGNIFICANT JVM/COMPILER/PERFORMANCE EXPERTISE,
* AND LOTS OF TIME OVER THAT. ADJUST YOUR PLANS ACCORDINGLY.
*/
private static final boolean COMPILER_BLACKHOLE = Boolean.getBoolean("compilerBlackholesEnabled");
static {
Utils.check(Blackhole.class, "b1", "b2");
Utils.check(Blackhole.class, "bool1", "bool2");
Utils.check(Blackhole.class, "c1", "c2");
Utils.check(Blackhole.class, "s1", "s2");
Utils.check(Blackhole.class, "i1", "i2");
Utils.check(Blackhole.class, "l1", "l2");
Utils.check(Blackhole.class, "f1", "f2");
Utils.check(Blackhole.class, "d1", "d2");
Utils.check(Blackhole.class, "obj1");
}
public Blackhole(String challengeResponse) {
/*
* Prevent instantiation by user code. Without additional countermeasures
* to properly escape Blackhole, its magic is not working. The instances
* of Blackholes which are injected into benchmark methods are treated by JMH,
* and users are supposed to only use the injected instances.
*
* It only *seems* simple to make the constructor non-public, but then
* there is a lot of infrastructure code which assumes @State has a default
* constructor. One might suggest doing the internal factory method to instantiate,
* but that does not help when extending the Blackhole. There is a *messy* way to
* special-case most of these problems within the JMH code, but it does not seem
* to worth the effort.
*
* Therefore, we choose to fail at runtime. It will only affect the users who thought
* "new Blackhole()" is a good idea, and these users are rare. If you are reading this
* comment, you might be one of those users. Stay cool! Don't instantiate Blackholes
* directly though.
*/
if (!challengeResponse.equals("Today's password is swordfish. I understand instantiating Blackholes directly is dangerous.")) {
throw new IllegalStateException("Blackholes should not be instantiated directly.");
}
}
/**
* Make any consumed data begone.
*
* WARNING: This method should only be called by the infrastructure code, in clearly understood cases.
* Even though it is public, it is not supposed to be called by users.
*
* @param challengeResponse arbitrary string
*/
public void evaporate(String challengeResponse) {
if (!challengeResponse.equals("Yes, I am Stephen Hawking, and know a thing or two about black holes.")) {
throw new IllegalStateException("Who are you?");
}
obj1 = null;
}
/**
* Consume object. This call provides a side effect preventing JIT to eliminate dependent computations.
*
* @param obj object to consume.
*/
public final void consume(Object obj) {
if (COMPILER_BLACKHOLE) {
consumeCompiler(obj);
} else {
consumeFull(obj);
}
}
/**
* Consume object. This call provides a side effect preventing JIT to eliminate dependent computations.
*
* @param b object to consume.
*/
public final void consume(byte b) {
if (COMPILER_BLACKHOLE) {
consumeCompiler(b);
} else {
consumeFull(b);
}
}
/**
* Consume object. This call provides a side effect preventing JIT to eliminate dependent computations.
*
* @param bool object to consume.
*/
public final void consume(boolean bool) {
if (COMPILER_BLACKHOLE) {
consumeCompiler(bool);
} else {
consumeFull(bool);
}
}
/**
* Consume object. This call provides a side effect preventing JIT to eliminate dependent computations.
*
* @param c object to consume.
*/
public final void consume(char c) {
if (COMPILER_BLACKHOLE) {
consumeCompiler(c);
} else {
consumeFull(c);
}
}
/**
* Consume object. This call provides a side effect preventing JIT to eliminate dependent computations.
*
* @param s object to consume.
*/
public final void consume(short s) {
if (COMPILER_BLACKHOLE) {
consumeCompiler(s);
} else {
consumeFull(s);
}
}
/**
* Consume object. This call provides a side effect preventing JIT to eliminate dependent computations.
*
* @param i object to consume.
*/
public final void consume(int i) {
if (COMPILER_BLACKHOLE) {
consumeCompiler(i);
} else {
consumeFull(i);
}
}
/**
* Consume object. This call provides a side effect preventing JIT to eliminate dependent computations.
*
* @param l object to consume.
*/
public final void consume(long l) {
if (COMPILER_BLACKHOLE) {
consumeCompiler(l);
} else {
consumeFull(l);
}
}
/**
* Consume object. This call provides a side effect preventing JIT to eliminate dependent computations.
*
* @param f object to consume.
*/
public final void consume(float f) {
if (COMPILER_BLACKHOLE) {
consumeCompiler(f);
} else {
consumeFull(f);
}
}
/**
* Consume object. This call provides a side effect preventing JIT to eliminate dependent computations.
*
* @param d object to consume.
*/
public final void consume(double d) {
if (COMPILER_BLACKHOLE) {
consumeCompiler(d);
} else {
consumeFull(d);
}
}
// Compiler blackholes block: let compilers figure out how to deal with it.
private static void consumeCompiler(boolean v) {}
private static void consumeCompiler(byte v) {}
private static void consumeCompiler(short v) {}
private static void consumeCompiler(char v) {}
private static void consumeCompiler(int v) {}
private static void consumeCompiler(float v) {}
private static void consumeCompiler(double v) {}
private static void consumeCompiler(long v) {}
private static void consumeCompiler(Object v) {}
// Full blackholes block: confuse compilers to get blackholing effects.
// See implementation comments at the top to understand what this code is doing.
private void consumeFull(byte b) {
byte b1 = this.b1; // volatile read
byte b2 = this.b2;
if ((b ^ b1) == (b ^ b2)) {
// SHOULD NEVER HAPPEN
nullBait.b1 = b; // implicit null pointer exception
}
}
private void consumeFull(boolean bool) {
boolean bool1 = this.bool1; // volatile read
boolean bool2 = this.bool2;
if ((bool ^ bool1) == (bool ^ bool2)) {
// SHOULD NEVER HAPPEN
nullBait.bool1 = bool; // implicit null pointer exception
}
}
private void consumeFull(char c) {
char c1 = this.c1; // volatile read
char c2 = this.c2;
if ((c ^ c1) == (c ^ c2)) {
// SHOULD NEVER HAPPEN
nullBait.c1 = c; // implicit null pointer exception
}
}
private void consumeFull(short s) {
short s1 = this.s1; // volatile read
short s2 = this.s2;
if ((s ^ s1) == (s ^ s2)) {
// SHOULD NEVER HAPPEN
nullBait.s1 = s; // implicit null pointer exception
}
}
private void consumeFull(int i) {
int i1 = this.i1; // volatile read
int i2 = this.i2;
if ((i ^ i1) == (i ^ i2)) {
// SHOULD NEVER HAPPEN
nullBait.i1 = i; // implicit null pointer exception
}
}
private void consumeFull(long l) {
long l1 = this.l1; // volatile read
long l2 = this.l2;
if ((l ^ l1) == (l ^ l2)) {
// SHOULD NEVER HAPPEN
nullBait.l1 = l; // implicit null pointer exception
}
}
private void consumeFull(float f) {
float f1 = this.f1; // volatile read
float f2 = this.f2;
if (f == f1 & f == f2) {
// SHOULD NEVER HAPPEN
nullBait.f1 = f; // implicit null pointer exception
}
}
private void consumeFull(double d) {
double d1 = this.d1; // volatile read
double d2 = this.d2;
if (d == d1 & d == d2) {
// SHOULD NEVER HAPPEN
nullBait.d1 = d; // implicit null pointer exception
}
}
private void consumeFull(Object obj) {
int tlrMask = this.tlrMask; // volatile read
int tlr = (this.tlr = (this.tlr * 1664525 + 1013904223));
if ((tlr & tlrMask) == 0) {
// SHOULD ALMOST NEVER HAPPEN IN MEASUREMENT
this.obj1 = new WeakReference<>(obj);
this.tlrMask = (tlrMask << 1) + 1;
}
}
private static volatile long consumedCPU = System.nanoTime();
/**
* Consume some amount of time tokens.
*
* This method does the CPU work almost linear to the number of tokens.
* The token cost may vary from system to system, and may change in
* future. (Translation: it is as reliable as we can get, but not absolutely
* reliable).
*
* See JMH samples for the complete demo, and core benchmarks for
* the performance assessments.
*
* @param tokens CPU tokens to consume
*/
public static void consumeCPU(long tokens) {
// If you are looking at this code trying to understand
// the non-linearity on low token counts, know this:
// we are pretty sure the generated assembly for almost all
// cases is the same, and the only explanation for the
// performance difference is hardware-specific effects.
// Be wary to waste more time on this. If you know more
// advanced and clever option to implement consumeCPU, let us
// know.
// Randomize start so that JIT could not memoize; this helps
// to break the loop optimizations if the method is called
// from the external loop body.
long t = consumedCPU;
// One of the rare cases when counting backwards is meaningful:
// for the forward loop HotSpot/x86 generates "cmp" with immediate
// on the hot path, while the backward loop tests against zero
// with "test". The immediate can have different lengths, which
// attribute to different machine code for different cases. We
// counter that with always counting backwards. We also mix the
// induction variable in, so that reversing the loop is the
// non-trivial optimization.
for (long i = tokens; i > 0; i--) {
t += (t * 0x5DEECE66DL + 0xBL + i) & (0xFFFFFFFFFFFFL);
}
// Need to guarantee side-effect on the result, but can't afford
// contention; make sure we update the shared state only in the
// unlikely case, so not to do the furious writes, but still
// dodge DCE.
if (t == 42) {
consumedCPU += t;
}
}
}