为什么“-XX:+UseLWPSynchronization”会对 Windows 操作系统上的印章锁定产生负面影响？

问题描述

我想在java中实现一个简单的速率限制器来学习如何使用jmh。在'https://github.com/William1104/rate-limiter'

创建了一个简单的 github 项目

有趣的是，当启用 '-XX:+UseLWPSynchronization' 选项时，某些实现（带有戳锁）的吞吐量会受到影响。基准测试是在 Windows 机器上执行的，我预计它对非 Solaris 系统没有影响。然而，测试结果显示不同。我知道有人可以帮助我了解到底发生了什么吗？

以下是我机器上的测试结果作为参考：

带选项： -server,-XX:+UnlockDiagnosticVMOptions,-XX:+UseNUMA

基准	(rateLimiterType)	模式	Cnt	得分	错误	单位
RaterLimiterBenchmark.thread_1	StampLockLongArrayRateLimiter	thrpt	90	21487.385	▒ 1082.163	操作数/毫秒
RaterLimiterBenchmark.thread_1	StampLockInstantArrayRateLimiter	thrpt	90	13162.330	▒ 1585.555	操作/毫秒
RaterLimiterBenchmark.thread_1	SynchronizedLongArrayRateLimiter	thrpt	90	15362.934	▒ 227.704	操作/毫秒
RaterLimiterBenchmark.thread_1	SynchronizedInstantArrayRateLimiter	thrpt	90	17281.675	▒ 2148.057	操作/毫秒
RaterLimiterBenchmark.thread_10	StampLockLongArrayRateLimiter	thrpt	90	6868.653	▒ 146.372	操作数/毫秒
RaterLimiterBenchmark.thread_10	StampLockInstantArrayRateLimiter	thrpt	90	8189.747	▒ 335.517	操作/毫秒
RaterLimiterBenchmark.thread_10	SynchronizedLongArrayRateLimiter	thrpt	90	6643.004	▒ 103.568	操作/毫秒
RaterLimiterBenchmark.thread_10	SynchronizedInstantArrayRateLimiter	thrpt	90	5252.975	▒ 190.363	操作/毫秒
RaterLimiterBenchmark.thread_100	StampLockLongArrayRateLimiter	thrpt	90	7352.890	▒ 2109.446	操作/毫秒
RaterLimiterBenchmark.thread_100	StampLockInstantArrayRateLimiter	thrpt	90	8675.814	▒ 922.653	操作/毫秒
RaterLimiterBenchmark.thread_100	SynchronizedLongArrayRateLimiter	thrpt	90	6509.368	▒ 157.212	操作/毫秒
RaterLimiterBenchmark.thread_100	SynchronizedInstantArrayRateLimiter	thrpt	90	5042.867	▒ 192.971	操作/毫秒

带选项：-server,-XX:+UseNUMA,-XX:+UseLWPSynchronization

基准	(rateLimiterType)	模式	Cnt	得分	错误	单位
RaterLimiterBenchmark.thread_1	StampLockLongArrayRateLimiter	thrpt	90	11383.198	▒ 353.921	操作/毫秒
RaterLimiterBenchmark.thread_1	StampLockInstantArrayRateLimiter	thrpt	90	11666.918	▒ 842.426	操作/毫秒
RaterLimiterBenchmark.thread_1	SynchronizedLongArrayRateLimiter	thrpt	90	15696.852	▒ 371.078	操作/毫秒
RaterLimiterBenchmark.thread_1	SynchronizedInstantArrayRateLimiter	thrpt	90	15357.617	▒ 650.846	操作/毫秒
RaterLimiterBenchmark.thread_10	StampLockLongArrayRateLimiter	thrpt	90	6937.050	▒ 130.727	操作/毫秒
RaterLimiterBenchmark.thread_10	StampLockInstantArrayRateLimiter	thrpt	90	8268.909	▒ 291.471	操作/毫秒
RaterLimiterBenchmark.thread_10	SynchronizedLongArrayRateLimiter	thrpt	90	9134.319	▒ 1208.998	操作/毫秒
RaterLimiterBenchmark.thread_10	SynchronizedInstantArrayRateLimiter	thrpt	90	5294.341	▒ 225.995	操作/毫秒
RaterLimiterBenchmark.thread_100	StampLockLongArrayRateLimiter	thrpt	90	8453.825	▒ 1075.312	操作/毫秒
RaterLimiterBenchmark.thread_100	StampLockInstantArrayRateLimiter	thrpt	90	16297.921	▒ 611.255	操作/毫秒
RaterLimiterBenchmark.thread_100	SynchronizedLongArrayRateLimiter	thrpt	90	12536.378	▒ 974.951	操作/毫秒
RaterLimiterBenchmark.thread_100	SynchronizedInstantArrayRateLimiter	thrpt	90	9051.560	▒ 1303.856	操作/毫秒

有 StampLockLongArrayRateLimiter 和 SynchronizedLongArrayRateLImiter 的实现：

package one.williamwong.ratelimiter;

import java.time.Duration;
import java.util.Arrays;
import java.util.concurrent.locks.StampedLock;

public class StampLockLongArrayRateLimiter implements IRateLimiter {

    private final long duration;
    private final long[] records;
    private final StampedLock lock;
    private int pointer;

    public StampLockLongArrayRateLimiter(int maxInvokes,Duration duration) {
        this.duration = duration.toNanos();
        this.records = new long[maxInvokes];
        this.lock = new StampedLock();
        this.pointer = 0;
    }

    @Override public void acquire() {
        final long stamp = lock.writeLock();
        try {
            final long now = System.nanoTime();
            if (records[pointer] != 0) {
                final long awayFromHead = now - records[pointer];
                if (awayFromHead < duration) {
                    handleExcessLimit(records.length,Duration.ofNanos(awayFromHead));
                }
            }
            records[pointer] = now;
            pointer = (pointer + 1) % records.length;
        } finally {
            lock.unlockWrite(stamp);
        }
    }

    @Override public void reset() {
        final long stamp = lock.writeLock();
        try {
            Arrays.fill(records,0);
            this.pointer = 0;
        } finally {
            lock.unlockWrite(stamp);
        }
    }

}

package one.williamwong.ratelimiter;

import java.time.Duration;
import java.util.Arrays;

public class SynchronizedLongArrayRateLimiter implements IRateLimiter {

    private final long duration;
    private final long[] records;
    private final Object lock;
    private int pointer;

    public SynchronizedLongArrayRateLimiter(int maxInvokes,Duration duration) {
        this.duration = duration.toNanos();
        this.records = new long[maxInvokes];
        this.lock = new Object();
        this.pointer = 0;
    }

    @Override
    public void acquire() {
        synchronized (lock) {
            final long now = System.nanoTime();
            if (records[pointer] != 0) {
                final long awayFromHead = now - records[pointer];
                if (awayFromHead < duration) {
                    handleExcessLimit(records.length,Duration.ofNanos(awayFromHead));
                }
            }
            records[pointer] = now;
            pointer = (pointer + 1) % records.length;
        }
    }

    @Override public void reset() {
        synchronized (lock) {
            Arrays.fill(records,0);
            this.pointer = 0;
        }
    }

}

解决方法

感谢您的评论。我使用不同的设置重新运行我的基准测试。如果我们每次迭代 1 秒再次执行 JMH，我得到以下结果：

带选项：-server,-XX:+UnlockDiagnosticVMOptions,-XX:+UseNUMA,-XX:-UseLWPSynchronization

基准	(rateLimiterType)	模式	Cnt	得分	错误	单位
RaterLimiterBenchmark.thread_1	StampLockLongArrayRateLimiter	thrpt	90	23573.282	▒ 364.739	操作/毫秒
RaterLimiterBenchmark.thread_1	StampLockInstantArrayRateLimiter	thrpt	90	23062.260	▒ 1035.395	操作/毫秒
RaterLimiterBenchmark.thread_1	SynchronizedLongArrayRateLimiter	thrpt	90	34667.411	▒ 246.003	操作/毫秒
RaterLimiterBenchmark.thread_1	SynchronizedInstantArrayRateLimiter	thrpt	90	36426.369	▒ 1248.360	操作/毫秒
RaterLimiterBenchmark.thread_10	StampLockLongArrayRateLimiter	thrpt	90	13592.158	▒ 76.319	操作/毫秒
RaterLimiterBenchmark.thread_10	StampLockInstantArrayRateLimiter	thrpt	90	14564.306	▒ 474.613	操作/毫秒
RaterLimiterBenchmark.thread_10	SynchronizedLongArrayRateLimiter	thrpt	90	13524.610	▒ 155.850	操作/毫秒
RaterLimiterBenchmark.thread_10	SynchronizedInstantArrayRateLimiter	thrpt	90	13080.967	▒ 309.736	操作/毫秒
RaterLimiterBenchmark.thread_100	StampLockLongArrayRateLimiter	thrpt	90	13224.529	▒ 459.035	操作/毫秒
RaterLimiterBenchmark.thread_100	StampLockInstantArrayRateLimiter	thrpt	90	13890.278	▒ 456.182	操作/毫秒
RaterLimiterBenchmark.thread_100	SynchronizedLongArrayRateLimiter	thrpt	90	12672.925	▒ 314.118	操作/毫秒
RaterLimiterBenchmark.thread_100	SynchronizedInstantArrayRateLimiter	thrpt	90	12245.120	▒ 296.395	操作/毫秒

带选项：-server,-XX:+UseLWPSynchronization

基准	(rateLimiterType)	模式	Cnt	得分	错误	单位
RaterLimiterBenchmark.thread_1	StampLockLongArrayRateLimiter	thrpt	90	24842.514	▒ 372.521	操作/毫秒
RaterLimiterBenchmark.thread_1	StampLockInstantArrayRateLimiter	thrpt	90	24327.864	▒ 322.659	操作/毫秒
RaterLimiterBenchmark.thread_1	SynchronizedLongArrayRateLimiter	thrpt	90	34490.411	▒ 330.288	操作/毫秒
RaterLimiterBenchmark.thread_1	SynchronizedInstantArrayRateLimiter	thrpt	90	38383.257	▒ 654.269	操作/毫秒
RaterLimiterBenchmark.thread_10	StampLockLongArrayRateLimiter	thrpt	90	13536.284	▒ 74.613	操作/毫秒
RaterLimiterBenchmark.thread_10	StampLockInstantArrayRateLimiter	thrpt	90	13702.022	▒ 289.616	操作/毫秒
RaterLimiterBenchmark.thread_10	SynchronizedLongArrayRateLimiter	thrpt	90	12530.107	▒ 243.471	操作/毫秒
RaterLimiterBenchmark.thread_10	SynchronizedInstantArrayRateLimiter	thrpt	90	10795.833	▒ 158.400	操作/毫秒
RaterLimiterBenchmark.thread_100	StampLockLongArrayRateLimiter	thrpt	90	13204.275	▒ 200.937	操作/毫秒
RaterLimiterBenchmark.thread_100	StampLockInstantArrayRateLimiter	thrpt	90	11606.823	▒ 224.213	操作/毫秒
RaterLimiterBenchmark.thread_100	SynchronizedLongArrayRateLimiter	thrpt	90	11504.124	▒ 107.543	操作/毫秒
RaterLimiterBenchmark.thread_100	SynchronizedInstantArrayRateLimiter	thrpt	90	10732.451	▒ 118.753	操作/毫秒

无论是否启用“UseLWPSynchronization”，我都没有观察到巨大的性能差异。我遇到的问题与不稳定的 JMH 设置有关。