diff --git a/akka-actor-tests/src/test/scala/akka/actor/ActorTimeoutSpec.scala b/akka-actor-tests/src/test/scala/akka/actor/ActorTimeoutSpec.scala
index c8df739b48..6800033d4c 100644
--- a/akka-actor-tests/src/test/scala/akka/actor/ActorTimeoutSpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/actor/ActorTimeoutSpec.scala
@@ -3,53 +3,30 @@
*/
package akka.actor
-import org.scalatest.BeforeAndAfterAll
import akka.util.duration._
-import akka.testkit.AkkaSpec
-import akka.testkit.DefaultTimeout
-import java.util.concurrent.TimeoutException
+import akka.testkit._
import akka.dispatch.Await
import akka.util.Timeout
import akka.pattern.{ ask, AskTimeoutException }
@org.junit.runner.RunWith(classOf[org.scalatest.junit.JUnitRunner])
-class ActorTimeoutSpec extends AkkaSpec with BeforeAndAfterAll with DefaultTimeout {
+class ActorTimeoutSpec extends AkkaSpec {
- val defaultTimeout = system.settings.ActorTimeout.duration
- val testTimeout = if (system.settings.ActorTimeout.duration < 400.millis) 500 millis else 100 millis
+ val testTimeout = 200.millis.dilated
"An Actor-based Future" must {
- "use the global default timeout if no implicit in scope" in {
- within(defaultTimeout - 100.millis, defaultTimeout + 400.millis) {
- val echo = system.actorOf(Props.empty)
- try {
- val d = system.settings.ActorTimeout.duration
- val f = echo ? "hallo"
- intercept[AskTimeoutException] { Await.result(f, d + d) }
- } finally { system.stop(echo) }
- }
- }
-
"use implicitly supplied timeout" in {
implicit val timeout = Timeout(testTimeout)
- within(testTimeout - 100.millis, testTimeout + 300.millis) {
- val echo = system.actorOf(Props.empty)
- try {
- val f = (echo ? "hallo").mapTo[String]
- intercept[AskTimeoutException] { Await.result(f, testTimeout + testTimeout) }
- } finally { system.stop(echo) }
- }
+ val echo = system.actorOf(Props.empty)
+ val f = (echo ? "hallo")
+ intercept[AskTimeoutException] { Await.result(f, testTimeout * 2) }
}
"use explicitly supplied timeout" in {
- within(testTimeout - 100.millis, testTimeout + 300.millis) {
- val echo = system.actorOf(Props.empty)
- val f = echo.?("hallo")(testTimeout)
- try {
- intercept[AskTimeoutException] { Await.result(f, testTimeout + 300.millis) }
- } finally { system.stop(echo) }
- }
+ val echo = system.actorOf(Props.empty)
+ val f = echo.?("hallo")(testTimeout)
+ intercept[AskTimeoutException] { Await.result(f, testTimeout * 2) }
}
}
}
diff --git a/akka-actor-tests/src/test/scala/akka/actor/ConsistencySpec.scala b/akka-actor-tests/src/test/scala/akka/actor/ConsistencySpec.scala
index 981ce89ef6..6f8639f4a4 100644
--- a/akka-actor-tests/src/test/scala/akka/actor/ConsistencySpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/actor/ConsistencySpec.scala
@@ -9,12 +9,15 @@ object ConsistencySpec {
consistency-dispatcher {
throughput = 1
keep-alive-time = 1 ms
- core-pool-size-min = 10
- core-pool-size-max = 10
- max-pool-size-min = 10
- max-pool-size-max = 10
- task-queue-type = array
- task-queue-size = 7
+ executor = "thread-pool-executor"
+ thread-pool-executor {
+ core-pool-size-min = 10
+ core-pool-size-max = 10
+ max-pool-size-min = 10
+ max-pool-size-max = 10
+ task-queue-type = array
+ task-queue-size = 7
+ }
}
"""
class CacheMisaligned(var value: Long, var padding1: Long, var padding2: Long, var padding3: Int) //Vars, no final fences
diff --git a/akka-actor-tests/src/test/scala/akka/actor/LocalActorRefProviderSpec.scala b/akka-actor-tests/src/test/scala/akka/actor/LocalActorRefProviderSpec.scala
index 82cd08fa77..5ebd8ff565 100644
--- a/akka-actor-tests/src/test/scala/akka/actor/LocalActorRefProviderSpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/actor/LocalActorRefProviderSpec.scala
@@ -14,8 +14,11 @@ object LocalActorRefProviderSpec {
akka {
actor {
default-dispatcher {
- core-pool-size-min = 16
- core-pool-size-max = 16
+ executor = "thread-pool-executor"
+ thread-pool-executor {
+ core-pool-size-min = 16
+ core-pool-size-max = 16
+ }
}
}
}
diff --git a/akka-actor-tests/src/test/scala/akka/actor/TypedActorSpec.scala b/akka-actor-tests/src/test/scala/akka/actor/TypedActorSpec.scala
index 49b37cc506..b83fe78338 100644
--- a/akka-actor-tests/src/test/scala/akka/actor/TypedActorSpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/actor/TypedActorSpec.scala
@@ -25,10 +25,13 @@ object TypedActorSpec {
val config = """
pooled-dispatcher {
type = BalancingDispatcher
- core-pool-size-min = 60
- core-pool-size-max = 60
- max-pool-size-min = 60
- max-pool-size-max = 60
+ executor = "thread-pool-executor"
+ thread-pool-executor {
+ core-pool-size-min = 60
+ core-pool-size-max = 60
+ max-pool-size-min = 60
+ max-pool-size-max = 60
+ }
}
"""
diff --git a/akka-actor-tests/src/test/scala/akka/actor/dispatch/ActorModelSpec.scala b/akka-actor-tests/src/test/scala/akka/actor/dispatch/ActorModelSpec.scala
index e2b697a08f..45e1954486 100644
--- a/akka-actor-tests/src/test/scala/akka/actor/dispatch/ActorModelSpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/actor/dispatch/ActorModelSpec.scala
@@ -448,16 +448,14 @@ object DispatcherModelSpec {
class MessageDispatcherInterceptorConfigurator(config: Config, prerequisites: DispatcherPrerequisites)
extends MessageDispatcherConfigurator(config, prerequisites) {
- private val instance: MessageDispatcher = {
- configureThreadPool(config,
- threadPoolConfig ⇒ new Dispatcher(prerequisites,
- config.getString("id"),
- config.getInt("throughput"),
- Duration(config.getNanoseconds("throughput-deadline-time"), TimeUnit.NANOSECONDS),
- mailboxType,
- threadPoolConfig,
- Duration(config.getMilliseconds("shutdown-timeout"), TimeUnit.MILLISECONDS)) with MessageDispatcherInterceptor).build
- }
+ private val instance: MessageDispatcher =
+ new Dispatcher(prerequisites,
+ config.getString("id"),
+ config.getInt("throughput"),
+ Duration(config.getNanoseconds("throughput-deadline-time"), TimeUnit.NANOSECONDS),
+ mailboxType,
+ configureExecutor(),
+ Duration(config.getMilliseconds("shutdown-timeout"), TimeUnit.MILLISECONDS)) with MessageDispatcherInterceptor
override def dispatcher(): MessageDispatcher = instance
}
@@ -522,16 +520,14 @@ object BalancingDispatcherModelSpec {
class BalancingMessageDispatcherInterceptorConfigurator(config: Config, prerequisites: DispatcherPrerequisites)
extends MessageDispatcherConfigurator(config, prerequisites) {
- private val instance: MessageDispatcher = {
- configureThreadPool(config,
- threadPoolConfig ⇒ new BalancingDispatcher(prerequisites,
- config.getString("id"),
- config.getInt("throughput"),
- Duration(config.getNanoseconds("throughput-deadline-time"), TimeUnit.NANOSECONDS),
- mailboxType,
- threadPoolConfig,
- Duration(config.getMilliseconds("shutdown-timeout"), TimeUnit.MILLISECONDS)) with MessageDispatcherInterceptor).build
- }
+ private val instance: MessageDispatcher =
+ new BalancingDispatcher(prerequisites,
+ config.getString("id"),
+ config.getInt("throughput"),
+ Duration(config.getNanoseconds("throughput-deadline-time"), TimeUnit.NANOSECONDS),
+ mailboxType,
+ configureExecutor(),
+ Duration(config.getMilliseconds("shutdown-timeout"), TimeUnit.MILLISECONDS)) with MessageDispatcherInterceptor
override def dispatcher(): MessageDispatcher = instance
}
diff --git a/akka-actor-tests/src/test/scala/akka/actor/dispatch/DispatcherActorSpec.scala b/akka-actor-tests/src/test/scala/akka/actor/dispatch/DispatcherActorSpec.scala
index 2dce8346db..4b3dd4a5b3 100644
--- a/akka-actor-tests/src/test/scala/akka/actor/dispatch/DispatcherActorSpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/actor/dispatch/DispatcherActorSpec.scala
@@ -16,14 +16,20 @@ object DispatcherActorSpec {
}
test-throughput-dispatcher {
throughput = 101
- core-pool-size-min = 1
- core-pool-size-max = 1
+ executor = "thread-pool-executor"
+ thread-pool-executor {
+ core-pool-size-min = 1
+ core-pool-size-max = 1
+ }
}
test-throughput-deadline-dispatcher {
throughput = 2
throughput-deadline-time = 100 milliseconds
- core-pool-size-min = 1
- core-pool-size-max = 1
+ executor = "thread-pool-executor"
+ thread-pool-executor {
+ core-pool-size-min = 1
+ core-pool-size-max = 1
+ }
}
"""
diff --git a/akka-actor-tests/src/test/scala/akka/config/ConfigSpec.scala b/akka-actor-tests/src/test/scala/akka/config/ConfigSpec.scala
index 67c7a51b60..ad39057d1d 100644
--- a/akka-actor-tests/src/test/scala/akka/config/ConfigSpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/config/ConfigSpec.scala
@@ -18,35 +18,62 @@ class ConfigSpec extends AkkaSpec(ConfigFactory.defaultReference) {
val settings = system.settings
val config = settings.config
- import config._
- getString("akka.version") must equal("2.0-SNAPSHOT")
- settings.ConfigVersion must equal("2.0-SNAPSHOT")
+ {
+ import config._
- getBoolean("akka.daemonic") must equal(false)
+ getString("akka.version") must equal("2.0-SNAPSHOT")
+ settings.ConfigVersion must equal("2.0-SNAPSHOT")
- getString("akka.actor.default-dispatcher.type") must equal("Dispatcher")
- getMilliseconds("akka.actor.default-dispatcher.keep-alive-time") must equal(60 * 1000)
- getDouble("akka.actor.default-dispatcher.core-pool-size-factor") must equal(3.0)
- getDouble("akka.actor.default-dispatcher.max-pool-size-factor") must equal(3.0)
- getInt("akka.actor.default-dispatcher.task-queue-size") must equal(-1)
- getString("akka.actor.default-dispatcher.task-queue-type") must equal("linked")
- getBoolean("akka.actor.default-dispatcher.allow-core-timeout") must equal(true)
- getInt("akka.actor.default-dispatcher.mailbox-capacity") must equal(-1)
- getMilliseconds("akka.actor.default-dispatcher.mailbox-push-timeout-time") must equal(10 * 1000)
- getString("akka.actor.default-dispatcher.mailboxType") must be("")
- getMilliseconds("akka.actor.default-dispatcher.shutdown-timeout") must equal(1 * 1000)
- getInt("akka.actor.default-dispatcher.throughput") must equal(5)
- getMilliseconds("akka.actor.default-dispatcher.throughput-deadline-time") must equal(0)
+ getBoolean("akka.daemonic") must equal(false)
+ getBoolean("akka.actor.serialize-messages") must equal(false)
+ settings.SerializeAllMessages must equal(false)
- getBoolean("akka.actor.serialize-messages") must equal(false)
- settings.SerializeAllMessages must equal(false)
+ getInt("akka.scheduler.ticksPerWheel") must equal(512)
+ settings.SchedulerTicksPerWheel must equal(512)
- getInt("akka.scheduler.ticksPerWheel") must equal(512)
- settings.SchedulerTicksPerWheel must equal(512)
+ getMilliseconds("akka.scheduler.tickDuration") must equal(100)
+ settings.SchedulerTickDuration must equal(100 millis)
+ }
- getMilliseconds("akka.scheduler.tickDuration") must equal(100)
- settings.SchedulerTickDuration must equal(100 millis)
+ {
+ val c = config.getConfig("akka.actor.default-dispatcher")
+
+ //General dispatcher config
+
+ {
+ c.getString("type") must equal("Dispatcher")
+ c.getString("executor") must equal("fork-join-executor")
+ c.getInt("mailbox-capacity") must equal(-1)
+ c.getMilliseconds("mailbox-push-timeout-time") must equal(10 * 1000)
+ c.getString("mailboxType") must be("")
+ c.getMilliseconds("shutdown-timeout") must equal(1 * 1000)
+ c.getInt("throughput") must equal(5)
+ c.getMilliseconds("throughput-deadline-time") must equal(0)
+ }
+
+ //Fork join executor config
+
+ {
+ val pool = c.getConfig("fork-join-executor")
+ pool.getInt("parallelism-min") must equal(8)
+ pool.getDouble("parallelism-factor") must equal(3.0)
+ pool.getInt("parallelism-max") must equal(64)
+ }
+
+ //Thread pool executor config
+
+ {
+ val pool = c.getConfig("thread-pool-executor")
+ import pool._
+ getMilliseconds("keep-alive-time") must equal(60 * 1000)
+ getDouble("core-pool-size-factor") must equal(3.0)
+ getDouble("max-pool-size-factor") must equal(3.0)
+ getInt("task-queue-size") must equal(-1)
+ getString("task-queue-type") must equal("linked")
+ getBoolean("allow-core-timeout") must equal(true)
+ }
+ }
}
}
}
diff --git a/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughput10000PerformanceSpec.scala b/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughput10000PerformanceSpec.scala
deleted file mode 100644
index 1ef92549c2..0000000000
--- a/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughput10000PerformanceSpec.scala
+++ /dev/null
@@ -1,169 +0,0 @@
-package akka.performance.microbench
-
-import akka.performance.workbench.PerformanceSpec
-import org.apache.commons.math.stat.descriptive.DescriptiveStatistics
-import akka.actor._
-import java.util.concurrent.{ ThreadPoolExecutor, CountDownLatch, TimeUnit }
-import akka.dispatch._
-import java.util.concurrent.ThreadPoolExecutor.AbortPolicy
-import java.util.concurrent.BlockingQueue
-import java.util.concurrent.LinkedBlockingQueue
-import akka.util.Duration
-import akka.util.duration._
-
-// -server -Xms512M -Xmx1024M -XX:+UseParallelGC -Dbenchmark=true -Dbenchmark.repeatFactor=500
-@org.junit.runner.RunWith(classOf[org.scalatest.junit.JUnitRunner])
-class TellThroughput10000PerformanceSpec extends PerformanceSpec {
- import TellThroughput10000PerformanceSpec._
-
- val repeat = 30000L * repeatFactor
-
- "Tell" must {
- "warmup" in {
- runScenario(4, warmup = true)
- }
- "warmup more" in {
- runScenario(4, warmup = true)
- }
- "perform with load 1" in {
- runScenario(1)
- }
- "perform with load 2" in {
- runScenario(2)
- }
- "perform with load 4" in {
- runScenario(4)
- }
- "perform with load 6" in {
- runScenario(6)
- }
- "perform with load 8" in {
- runScenario(8)
- }
- "perform with load 10" in {
- runScenario(10)
- }
- "perform with load 12" in {
- runScenario(12)
- }
- "perform with load 14" in {
- runScenario(14)
- }
- "perform with load 16" in {
- runScenario(16)
- }
- "perform with load 18" in {
- runScenario(18)
- }
- "perform with load 20" in {
- runScenario(20)
- }
- "perform with load 22" in {
- runScenario(22)
- }
- "perform with load 24" in {
- runScenario(24)
- }
- "perform with load 26" in {
- runScenario(26)
- }
- "perform with load 28" in {
- runScenario(28)
- }
- "perform with load 30" in {
- runScenario(30)
- }
- "perform with load 32" in {
- runScenario(32)
- }
- "perform with load 34" in {
- runScenario(34)
- }
- "perform with load 36" in {
- runScenario(36)
- }
- "perform with load 38" in {
- runScenario(38)
- }
- "perform with load 40" in {
- runScenario(40)
- }
- "perform with load 42" in {
- runScenario(42)
- }
- "perform with load 44" in {
- runScenario(44)
- }
- "perform with load 46" in {
- runScenario(46)
- }
- "perform with load 48" in {
- runScenario(48)
- }
-
- def runScenario(numberOfClients: Int, warmup: Boolean = false) {
- if (acceptClients(numberOfClients)) {
-
- val dispatcherKey = "benchmark.high-throughput-dispatcher"
- val latch = new CountDownLatch(numberOfClients)
- val repeatsPerClient = repeat / numberOfClients
- val destinations = for (i ← 0 until numberOfClients)
- yield system.actorOf(Props(new Destination).withDispatcher(dispatcherKey))
- val clients = for ((dest, j) ← destinations.zipWithIndex)
- yield system.actorOf(Props(new Client(dest, latch, repeatsPerClient)).withDispatcher(dispatcherKey))
-
- val start = System.nanoTime
- clients.foreach(_ ! Run)
- val ok = latch.await(maxRunDuration.toMillis, TimeUnit.MILLISECONDS)
- val durationNs = (System.nanoTime - start)
-
- if (!warmup) {
- ok must be(true)
- logMeasurement(numberOfClients, durationNs, repeat)
- }
- clients.foreach(system.stop(_))
- destinations.foreach(system.stop(_))
-
- }
- }
- }
-}
-
-object TellThroughput10000PerformanceSpec {
-
- case object Run
- case object Msg
-
- class Destination extends Actor {
- def receive = {
- case Msg ⇒ sender ! Msg
- }
- }
-
- class Client(
- actor: ActorRef,
- latch: CountDownLatch,
- repeat: Long) extends Actor {
-
- var sent = 0L
- var received = 0L
-
- def receive = {
- case Msg ⇒
- received += 1
- if (sent < repeat) {
- actor ! Msg
- sent += 1
- } else if (received >= repeat) {
- latch.countDown()
- }
- case Run ⇒
- for (i ← 0L until math.min(20000L, repeat)) {
- actor ! Msg
- sent += 1
- }
- }
-
- }
-
-}
diff --git a/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughputComputationPerformanceSpec.scala b/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughputComputationPerformanceSpec.scala
index 0b47a1f722..4bee0c8655 100644
--- a/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughputComputationPerformanceSpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughputComputationPerformanceSpec.scala
@@ -100,15 +100,14 @@ class TellThroughputComputationPerformanceSpec extends PerformanceSpec {
def runScenario(numberOfClients: Int, warmup: Boolean = false) {
if (acceptClients(numberOfClients)) {
- val clientDispatcher = "benchmark.client-dispatcher"
- val destinationDispatcher = "benchmark.destination-dispatcher"
+ val throughputDispatcher = "benchmark.throughput-dispatcher"
val latch = new CountDownLatch(numberOfClients)
val repeatsPerClient = repeat / numberOfClients
val destinations = for (i ← 0 until numberOfClients)
- yield system.actorOf(Props(new Destination).withDispatcher(destinationDispatcher))
+ yield system.actorOf(Props(new Destination).withDispatcher(throughputDispatcher))
val clients = for (dest ← destinations)
- yield system.actorOf(Props(new Client(dest, latch, repeatsPerClient)).withDispatcher(clientDispatcher))
+ yield system.actorOf(Props(new Client(dest, latch, repeatsPerClient)).withDispatcher(throughputDispatcher))
val start = System.nanoTime
clients.foreach(_ ! Run)
diff --git a/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughputPerformanceSpec.scala b/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughputPerformanceSpec.scala
index 552dbf62e9..f028fec6b0 100644
--- a/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughputPerformanceSpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughputPerformanceSpec.scala
@@ -16,10 +16,10 @@ class TellThroughputPerformanceSpec extends PerformanceSpec {
"Tell" must {
"warmup" in {
- runScenario(4, warmup = true)
+ runScenario(8, warmup = true)
}
"warmup more" in {
- runScenario(4, warmup = true)
+ runScenario(8, warmup = true)
}
"perform with load 1" in {
runScenario(1)
@@ -48,19 +48,66 @@ class TellThroughputPerformanceSpec extends PerformanceSpec {
"perform with load 16" in {
runScenario(16)
}
+ "perform with load 18" in {
+ runScenario(18)
+ }
+ "perform with load 20" in {
+ runScenario(20)
+ }
+ "perform with load 22" in {
+ runScenario(22)
+ }
+ "perform with load 24" in {
+ runScenario(24)
+ }
+ "perform with load 26" in {
+ runScenario(26)
+ }
+ "perform with load 28" in {
+ runScenario(28)
+ }
+ "perform with load 30" in {
+ runScenario(30)
+ }
+ "perform with load 32" in {
+ runScenario(32)
+ }
+ "perform with load 34" in {
+ runScenario(34)
+ }
+ "perform with load 36" in {
+ runScenario(36)
+ }
+ "perform with load 38" in {
+ runScenario(38)
+ }
+ "perform with load 40" in {
+ runScenario(40)
+ }
+ "perform with load 42" in {
+ runScenario(42)
+ }
+ "perform with load 44" in {
+ runScenario(44)
+ }
+ "perform with load 46" in {
+ runScenario(46)
+ }
+ "perform with load 48" in {
+ runScenario(48)
+ }
def runScenario(numberOfClients: Int, warmup: Boolean = false) {
if (acceptClients(numberOfClients)) {
- val clientDispatcher = "benchmark.client-dispatcher"
- val destinationDispatcher = "benchmark.destination-dispatcher"
+ val throughputDispatcher = "benchmark.throughput-dispatcher"
val latch = new CountDownLatch(numberOfClients)
val repeatsPerClient = repeat / numberOfClients
val destinations = for (i ← 0 until numberOfClients)
- yield system.actorOf(Props(new Destination).withDispatcher(destinationDispatcher))
+ yield system.actorOf(Props(new Destination).withDispatcher(throughputDispatcher))
val clients = for (dest ← destinations)
- yield system.actorOf(Props(new Client(dest, latch, repeatsPerClient)).withDispatcher(clientDispatcher))
+ yield system.actorOf(Props(new Client(dest, latch, repeatsPerClient)).withDispatcher(throughputDispatcher))
val start = System.nanoTime
clients.foreach(_ ! Run)
diff --git a/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughputPinnedDispatchersPerformanceSpec.scala b/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughputPinnedDispatchersPerformanceSpec.scala
deleted file mode 100644
index 4d9ad3eef1..0000000000
--- a/akka-actor-tests/src/test/scala/akka/performance/microbench/TellThroughputPinnedDispatchersPerformanceSpec.scala
+++ /dev/null
@@ -1,171 +0,0 @@
-package akka.performance.microbench
-
-import akka.performance.workbench.PerformanceSpec
-import org.apache.commons.math.stat.descriptive.DescriptiveStatistics
-import akka.actor._
-import java.util.concurrent.{ ThreadPoolExecutor, CountDownLatch, TimeUnit }
-import akka.dispatch._
-import java.util.concurrent.ThreadPoolExecutor.AbortPolicy
-import java.util.concurrent.BlockingQueue
-import java.util.concurrent.LinkedBlockingQueue
-import akka.util.Duration
-import akka.util.duration._
-
-// -server -Xms512M -Xmx1024M -XX:+UseParallelGC -Dbenchmark=true -Dbenchmark.repeatFactor=500
-@org.junit.runner.RunWith(classOf[org.scalatest.junit.JUnitRunner])
-class TellThroughputPinnedDispatchersPerformanceSpec extends PerformanceSpec {
- import TellThroughputPinnedDispatchersPerformanceSpec._
-
- val repeat = 30000L * repeatFactor
-
- "Tell" must {
- "warmup" in {
- runScenario(4, warmup = true)
- }
- "warmup more" in {
- runScenario(4, warmup = true)
- }
- "perform with load 1" in {
- runScenario(1)
- }
- "perform with load 2" in {
- runScenario(2)
- }
- "perform with load 4" in {
- runScenario(4)
- }
- "perform with load 6" in {
- runScenario(6)
- }
- "perform with load 8" in {
- runScenario(8)
- }
- "perform with load 10" in {
- runScenario(10)
- }
- "perform with load 12" in {
- runScenario(12)
- }
- "perform with load 14" in {
- runScenario(14)
- }
- "perform with load 16" in {
- runScenario(16)
- }
- "perform with load 18" in {
- runScenario(18)
- }
- "perform with load 20" in {
- runScenario(20)
- }
- "perform with load 22" in {
- runScenario(22)
- }
- "perform with load 24" in {
- runScenario(24)
- }
- "perform with load 26" in {
- runScenario(26)
- }
- "perform with load 28" in {
- runScenario(28)
- }
- "perform with load 30" in {
- runScenario(30)
- }
- "perform with load 32" in {
- runScenario(32)
- }
- "perform with load 34" in {
- runScenario(34)
- }
- "perform with load 36" in {
- runScenario(36)
- }
- "perform with load 38" in {
- runScenario(38)
- }
- "perform with load 40" in {
- runScenario(40)
- }
- "perform with load 42" in {
- runScenario(42)
- }
- "perform with load 44" in {
- runScenario(44)
- }
- "perform with load 46" in {
- runScenario(46)
- }
- "perform with load 48" in {
- runScenario(48)
- }
-
- def runScenario(numberOfClients: Int, warmup: Boolean = false) {
- if (acceptClients(numberOfClients)) {
-
- val pinnedDispatcher = "benchmark.pinned-dispatcher"
-
- val latch = new CountDownLatch(numberOfClients)
- val repeatsPerClient = repeat / numberOfClients
-
- val destinations = for (i ← 0 until numberOfClients)
- yield system.actorOf(Props(new Destination).withDispatcher(pinnedDispatcher))
- val clients = for ((dest, j) ← destinations.zipWithIndex)
- yield system.actorOf(Props(new Client(dest, latch, repeatsPerClient)).withDispatcher(pinnedDispatcher))
-
- val start = System.nanoTime
- clients.foreach(_ ! Run)
- val ok = latch.await(maxRunDuration.toMillis, TimeUnit.MILLISECONDS)
- val durationNs = (System.nanoTime - start)
-
- if (!warmup) {
- ok must be(true)
- logMeasurement(numberOfClients, durationNs, repeat)
- }
- clients.foreach(system.stop(_))
- destinations.foreach(system.stop(_))
-
- }
- }
- }
-}
-
-object TellThroughputPinnedDispatchersPerformanceSpec {
-
- case object Run
- case object Msg
-
- class Destination extends Actor {
- def receive = {
- case Msg ⇒ sender ! Msg
- }
- }
-
- class Client(
- actor: ActorRef,
- latch: CountDownLatch,
- repeat: Long) extends Actor {
-
- var sent = 0L
- var received = 0L
-
- def receive = {
- case Msg ⇒
- received += 1
- if (sent < repeat) {
- actor ! Msg
- sent += 1
- } else if (received >= repeat) {
- latch.countDown()
- }
- case Run ⇒
- for (i ← 0L until math.min(1000L, repeat)) {
- actor ! Msg
- sent += 1
- }
- }
-
- }
-
-}
diff --git a/akka-actor-tests/src/test/scala/akka/performance/trading/system/TradingLatencyPerformanceSpec.scala b/akka-actor-tests/src/test/scala/akka/performance/trading/system/TradingLatencyPerformanceSpec.scala
index 9ba77e71e8..58b2e7e315 100644
--- a/akka-actor-tests/src/test/scala/akka/performance/trading/system/TradingLatencyPerformanceSpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/performance/trading/system/TradingLatencyPerformanceSpec.scala
@@ -84,7 +84,7 @@ class TradingLatencyPerformanceSpec extends PerformanceSpec {
} yield Bid(s + i, 100 - i, 1000)
val orders = askOrders.zip(bidOrders).map(x ⇒ Seq(x._1, x._2)).flatten
- val clientDispatcher = "benchmark.client-dispatcher"
+ val latencyDispatcher = "benchmark.trading-dispatcher"
val ordersPerClient = repeat * orders.size / numberOfClients
val totalNumberOfOrders = ordersPerClient * numberOfClients
@@ -93,7 +93,7 @@ class TradingLatencyPerformanceSpec extends PerformanceSpec {
val start = System.nanoTime
val clients = (for (i ← 0 until numberOfClients) yield {
val receiver = receivers(i % receivers.size)
- val props = Props(new Client(receiver, orders, latch, ordersPerClient, clientDelay.toMicros.toInt)).withDispatcher(clientDispatcher)
+ val props = Props(new Client(receiver, orders, latch, ordersPerClient, clientDelay.toMicros.toInt)).withDispatcher(latencyDispatcher)
system.actorOf(props)
})
diff --git a/akka-actor-tests/src/test/scala/akka/performance/trading/system/TradingSystem.scala b/akka-actor-tests/src/test/scala/akka/performance/trading/system/TradingSystem.scala
index 7fe2783a9a..1adb2ecbc7 100644
--- a/akka-actor-tests/src/test/scala/akka/performance/trading/system/TradingSystem.scala
+++ b/akka-actor-tests/src/test/scala/akka/performance/trading/system/TradingSystem.scala
@@ -39,11 +39,9 @@ class AkkaTradingSystem(val system: ActorSystem) extends TradingSystem {
val orDispatcher = orderReceiverDispatcher
val meDispatcher = matchingEngineDispatcher
- // by default we use default-dispatcher
- def orderReceiverDispatcher: Option[String] = None
+ def orderReceiverDispatcher: Option[String] = Some("benchmark.trading-dispatcher")
- // by default we use default-dispatcher
- def matchingEngineDispatcher: Option[String] = None
+ def matchingEngineDispatcher: Option[String] = Some("benchmark.trading-dispatcher")
override val orderbooksGroupedByMatchingEngine: List[List[Orderbook]] =
for (groupOfSymbols: List[String] ← OrderbookRepository.orderbookSymbolsGroupedByMatchingEngine)
diff --git a/akka-actor-tests/src/test/scala/akka/performance/trading/system/TradingThroughputPerformanceSpec.scala b/akka-actor-tests/src/test/scala/akka/performance/trading/system/TradingThroughputPerformanceSpec.scala
index 7092f87666..a1033d7682 100644
--- a/akka-actor-tests/src/test/scala/akka/performance/trading/system/TradingThroughputPerformanceSpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/performance/trading/system/TradingThroughputPerformanceSpec.scala
@@ -81,7 +81,7 @@ class TradingThroughputPerformanceSpec extends PerformanceSpec {
} yield Bid(s + i, 100 - i, 1000)
val orders = askOrders.zip(bidOrders).map(x ⇒ Seq(x._1, x._2)).flatten
- val clientDispatcher = "benchmark.client-dispatcher"
+ val throughputDispatcher = "benchmark.trading-dispatcher"
val ordersPerClient = repeat * orders.size / numberOfClients
val totalNumberOfOrders = ordersPerClient * numberOfClients
@@ -90,7 +90,7 @@ class TradingThroughputPerformanceSpec extends PerformanceSpec {
val start = System.nanoTime
val clients = (for (i ← 0 until numberOfClients) yield {
val receiver = receivers(i % receivers.size)
- val props = Props(new Client(receiver, orders, latch, ordersPerClient)).withDispatcher(clientDispatcher)
+ val props = Props(new Client(receiver, orders, latch, ordersPerClient)).withDispatcher(throughputDispatcher)
system.actorOf(props)
})
diff --git a/akka-actor-tests/src/test/scala/akka/performance/workbench/BenchmarkConfig.scala b/akka-actor-tests/src/test/scala/akka/performance/workbench/BenchmarkConfig.scala
index 11ed21c9aa..e31e667678 100644
--- a/akka-actor-tests/src/test/scala/akka/performance/workbench/BenchmarkConfig.scala
+++ b/akka-actor-tests/src/test/scala/akka/performance/workbench/BenchmarkConfig.scala
@@ -20,38 +20,40 @@ object BenchmarkConfig {
resultDir = "target/benchmark"
useDummyOrderbook = false
- client-dispatcher {
- core-pool-size-min = ${benchmark.maxClients}
- core-pool-size-max = ${benchmark.maxClients}
- }
-
- destination-dispatcher {
- core-pool-size-min = ${benchmark.maxClients}
- core-pool-size-max = ${benchmark.maxClients}
- }
-
- high-throughput-dispatcher {
- throughput = 10000
- core-pool-size-min = ${benchmark.maxClients}
- core-pool-size-max = ${benchmark.maxClients}
- }
-
- pinned-dispatcher {
- type = PinnedDispatcher
+ throughput-dispatcher {
+ throughput = 5
+ executor = "fork-join-executor"
+ fork-join-executor {
+ parallelism-min = ${benchmark.maxClients}
+ parallelism-max = ${benchmark.maxClients}
+ }
}
latency-dispatcher {
throughput = 1
- core-pool-size-min = ${benchmark.maxClients}
- core-pool-size-max = ${benchmark.maxClients}
+ executor = "fork-join-executor"
+ fork-join-executor {
+ parallelism-min = ${benchmark.maxClients}
+ parallelism-max = ${benchmark.maxClients}
+ }
+ }
+
+ trading-dispatcher {
+ throughput = 5
+ executor = "fork-join-executor"
+ fork-join-executor {
+ parallelism-min = ${benchmark.maxClients}
+ parallelism-max = ${benchmark.maxClients}
+ }
}
}
""")
private val longRunningBenchmarkConfig = ConfigFactory.parseString("""
benchmark {
longRunning = true
+ minClients = 4
maxClients = 48
- repeatFactor = 150
+ repeatFactor = 2000
maxRunDuration = 120 seconds
useDummyOrderbook = true
}
diff --git a/akka-actor-tests/src/test/scala/akka/performance/workbench/PerformanceSpec.scala b/akka-actor-tests/src/test/scala/akka/performance/workbench/PerformanceSpec.scala
index 3d27f8a303..ca6e42d67f 100644
--- a/akka-actor-tests/src/test/scala/akka/performance/workbench/PerformanceSpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/performance/workbench/PerformanceSpec.scala
@@ -31,7 +31,8 @@ abstract class PerformanceSpec(cfg: Config = BenchmarkConfig.config) extends Akk
def compareResultWith: Option[String] = None
def acceptClients(numberOfClients: Int): Boolean = {
- (minClients <= numberOfClients && numberOfClients <= maxClients)
+ (minClients <= numberOfClients && numberOfClients <= maxClients &&
+ (maxClients <= 16 || numberOfClients % 4 == 0))
}
def logMeasurement(numberOfClients: Int, durationNs: Long, n: Long) {
diff --git a/akka-actor-tests/src/test/scala/akka/routing/ConfiguredLocalRoutingSpec.scala b/akka-actor-tests/src/test/scala/akka/routing/ConfiguredLocalRoutingSpec.scala
index f2707e042c..62800b8830 100644
--- a/akka-actor-tests/src/test/scala/akka/routing/ConfiguredLocalRoutingSpec.scala
+++ b/akka-actor-tests/src/test/scala/akka/routing/ConfiguredLocalRoutingSpec.scala
@@ -13,8 +13,11 @@ object ConfiguredLocalRoutingSpec {
akka {
actor {
default-dispatcher {
- core-pool-size-min = 8
- core-pool-size-max = 16
+ executor = "thread-pool-executor"
+ thread-pool-executor {
+ core-pool-size-min = 8
+ core-pool-size-max = 16
+ }
}
}
}
diff --git a/akka-actor/src/main/java/akka/jsr166y/ForkJoinPool.java b/akka-actor/src/main/java/akka/jsr166y/ForkJoinPool.java
new file mode 100644
index 0000000000..f92e5541f4
--- /dev/null
+++ b/akka-actor/src/main/java/akka/jsr166y/ForkJoinPool.java
@@ -0,0 +1,2674 @@
+/*
+ * Written by Doug Lea with assistance from members of JCP JSR-166
+ * Expert Group and released to the public domain, as explained at
+ * http://creativecommons.org/publicdomain/zero/1.0/
+ */
+
+package akka.jsr166y;
+
+import java.util.ArrayList;
+import java.util.Arrays;
+import java.util.Collection;
+import java.util.Collections;
+import java.util.List;
+import java.util.Random;
+import java.util.concurrent.AbstractExecutorService;
+import java.util.concurrent.Callable;
+import java.util.concurrent.ExecutorService;
+import java.util.concurrent.Future;
+import java.util.concurrent.RejectedExecutionException;
+import java.util.concurrent.RunnableFuture;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.atomic.AtomicInteger;
+import java.util.concurrent.atomic.AtomicLong;
+import java.util.concurrent.locks.AbstractQueuedSynchronizer;
+import java.util.concurrent.locks.Condition;
+import akka.util.Unsafe;
+
+/**
+ * An {@link ExecutorService} for running {@link ForkJoinTask}s.
+ * A {@code ForkJoinPool} provides the entry point for submissions
+ * from non-{@code ForkJoinTask} clients, as well as management and
+ * monitoring operations.
+ *
+ *
A {@code ForkJoinPool} differs from other kinds of {@link
+ * ExecutorService} mainly by virtue of employing
+ * work-stealing: all threads in the pool attempt to find and
+ * execute tasks submitted to the pool and/or created by other active
+ * tasks (eventually blocking waiting for work if none exist). This
+ * enables efficient processing when most tasks spawn other subtasks
+ * (as do most {@code ForkJoinTask}s), as well as when many small
+ * tasks are submitted to the pool from external clients. Especially
+ * when setting asyncMode to true in constructors, {@code
+ * ForkJoinPool}s may also be appropriate for use with event-style
+ * tasks that are never joined.
+ *
+ *
A {@code ForkJoinPool} is constructed with a given target
+ * parallelism level; by default, equal to the number of available
+ * processors. The pool attempts to maintain enough active (or
+ * available) threads by dynamically adding, suspending, or resuming
+ * internal worker threads, even if some tasks are stalled waiting to
+ * join others. However, no such adjustments are guaranteed in the
+ * face of blocked IO or other unmanaged synchronization. The nested
+ * {@link ManagedBlocker} interface enables extension of the kinds of
+ * synchronization accommodated.
+ *
+ *
In addition to execution and lifecycle control methods, this
+ * class provides status check methods (for example
+ * {@link #getStealCount}) that are intended to aid in developing,
+ * tuning, and monitoring fork/join applications. Also, method
+ * {@link #toString} returns indications of pool state in a
+ * convenient form for informal monitoring.
+ *
+ *
As is the case with other ExecutorServices, there are three
+ * main task execution methods summarized in the following table.
+ * These are designed to be used primarily by clients not already
+ * engaged in fork/join computations in the current pool. The main
+ * forms of these methods accept instances of {@code ForkJoinTask},
+ * but overloaded forms also allow mixed execution of plain {@code
+ * Runnable}- or {@code Callable}- based activities as well. However,
+ * tasks that are already executing in a pool should normally instead
+ * use the within-computation forms listed in the table unless using
+ * async event-style tasks that are not usually joined, in which case
+ * there is little difference among choice of methods.
+ *
+ *
+ *
+ *
+ *
Call from non-fork/join clients
+ *
Call from within fork/join computations
+ *
+ *
+ *
Arrange async execution
+ *
{@link #execute(ForkJoinTask)}
+ *
{@link ForkJoinTask#fork}
+ *
+ *
+ *
Await and obtain result
+ *
{@link #invoke(ForkJoinTask)}
+ *
{@link ForkJoinTask#invoke}
+ *
+ *
+ *
Arrange exec and obtain Future
+ *
{@link #submit(ForkJoinTask)}
+ *
{@link ForkJoinTask#fork} (ForkJoinTasks are Futures)
+ *
+ *
+ *
+ *
Sample Usage. Normally a single {@code ForkJoinPool} is
+ * used for all parallel task execution in a program or subsystem.
+ * Otherwise, use would not usually outweigh the construction and
+ * bookkeeping overhead of creating a large set of threads. For
+ * example, a common pool could be used for the {@code SortTasks}
+ * illustrated in {@link RecursiveAction}. Because {@code
+ * ForkJoinPool} uses threads in {@linkplain java.lang.Thread#isDaemon
+ * daemon} mode, there is typically no need to explicitly {@link
+ * #shutdown} such a pool upon program exit.
+ *
+ *
{@code
+ * static final ForkJoinPool mainPool = new ForkJoinPool();
+ * ...
+ * public void sort(long[] array) {
+ * mainPool.invoke(new SortTask(array, 0, array.length));
+ * }}
+ *
+ *
Implementation notes: This implementation restricts the
+ * maximum number of running threads to 32767. Attempts to create
+ * pools with greater than the maximum number result in
+ * {@code IllegalArgumentException}.
+ *
+ *
This implementation rejects submitted tasks (that is, by throwing
+ * {@link RejectedExecutionException}) only when the pool is shut down
+ * or internal resources have been exhausted.
+ *
+ * @since 1.7
+ * @author Doug Lea
+ */
+public class ForkJoinPool extends AbstractExecutorService {
+
+ /*
+ * Implementation Overview
+ *
+ * This class and its nested classes provide the main
+ * functionality and control for a set of worker threads:
+ * Submissions from non-FJ threads enter into submission queues.
+ * Workers take these tasks and typically split them into subtasks
+ * that may be stolen by other workers. Preference rules give
+ * first priority to processing tasks from their own queues (LIFO
+ * or FIFO, depending on mode), then to randomized FIFO steals of
+ * tasks in other queues.
+ *
+ * WorkQueues
+ * ==========
+ *
+ * Most operations occur within work-stealing queues (in nested
+ * class WorkQueue). These are special forms of Deques that
+ * support only three of the four possible end-operations -- push,
+ * pop, and poll (aka steal), under the further constraints that
+ * push and pop are called only from the owning thread (or, as
+ * extended here, under a lock), while poll may be called from
+ * other threads. (If you are unfamiliar with them, you probably
+ * want to read Herlihy and Shavit's book "The Art of
+ * Multiprocessor programming", chapter 16 describing these in
+ * more detail before proceeding.) The main work-stealing queue
+ * design is roughly similar to those in the papers "Dynamic
+ * Circular Work-Stealing Deque" by Chase and Lev, SPAA 2005
+ * (http://research.sun.com/scalable/pubs/index.html) and
+ * "Idempotent work stealing" by Michael, Saraswat, and Vechev,
+ * PPoPP 2009 (http://portal.acm.org/citation.cfm?id=1504186).
+ * The main differences ultimately stem from GC requirements that
+ * we null out taken slots as soon as we can, to maintain as small
+ * a footprint as possible even in programs generating huge
+ * numbers of tasks. To accomplish this, we shift the CAS
+ * arbitrating pop vs poll (steal) from being on the indices
+ * ("base" and "top") to the slots themselves. So, both a
+ * successful pop and poll mainly entail a CAS of a slot from
+ * non-null to null. Because we rely on CASes of references, we
+ * do not need tag bits on base or top. They are simple ints as
+ * used in any circular array-based queue (see for example
+ * ArrayDeque). Updates to the indices must still be ordered in a
+ * way that guarantees that top == base means the queue is empty,
+ * but otherwise may err on the side of possibly making the queue
+ * appear nonempty when a push, pop, or poll have not fully
+ * committed. Note that this means that the poll operation,
+ * considered individually, is not wait-free. One thief cannot
+ * successfully continue until another in-progress one (or, if
+ * previously empty, a push) completes. However, in the
+ * aggregate, we ensure at least probabilistic non-blockingness.
+ * If an attempted steal fails, a thief always chooses a different
+ * random victim target to try next. So, in order for one thief to
+ * progress, it suffices for any in-progress poll or new push on
+ * any empty queue to complete.
+ *
+ * This approach also enables support of a user mode in which local
+ * task processing is in FIFO, not LIFO order, simply by using
+ * poll rather than pop. This can be useful in message-passing
+ * frameworks in which tasks are never joined. However neither
+ * mode considers affinities, loads, cache localities, etc, so
+ * rarely provide the best possible performance on a given
+ * machine, but portably provide good throughput by averaging over
+ * these factors. (Further, even if we did try to use such
+ * information, we do not usually have a basis for exploiting it.
+ * For example, some sets of tasks profit from cache affinities,
+ * but others are harmed by cache pollution effects.)
+ *
+ * WorkQueues are also used in a similar way for tasks submitted
+ * to the pool. We cannot mix these tasks in the same queues used
+ * for work-stealing (this would contaminate lifo/fifo
+ * processing). Instead, we loosely associate submission queues
+ * with submitting threads, using a form of hashing. The
+ * ThreadLocal Submitter class contains a value initially used as
+ * a hash code for choosing existing queues, but may be randomly
+ * repositioned upon contention with other submitters. In
+ * essence, submitters act like workers except that they never
+ * take tasks, and they are multiplexed on to a finite number of
+ * shared work queues. However, classes are set up so that future
+ * extensions could allow submitters to optionally help perform
+ * tasks as well. Insertion of tasks in shared mode requires a
+ * lock (mainly to protect in the case of resizing) but we use
+ * only a simple spinlock (using bits in field runState), because
+ * submitters encountering a busy queue move on to try or create
+ * other queues, so never block.
+ *
+ * Management
+ * ==========
+ *
+ * The main throughput advantages of work-stealing stem from
+ * decentralized control -- workers mostly take tasks from
+ * themselves or each other. We cannot negate this in the
+ * implementation of other management responsibilities. The main
+ * tactic for avoiding bottlenecks is packing nearly all
+ * essentially atomic control state into two volatile variables
+ * that are by far most often read (not written) as status and
+ * consistency checks.
+ *
+ * Field "ctl" contains 64 bits holding all the information needed
+ * to atomically decide to add, inactivate, enqueue (on an event
+ * queue), dequeue, and/or re-activate workers. To enable this
+ * packing, we restrict maximum parallelism to (1<<15)-1 (which is
+ * far in excess of normal operating range) to allow ids, counts,
+ * and their negations (used for thresholding) to fit into 16bit
+ * fields.
+ *
+ * Field "runState" contains 32 bits needed to register and
+ * deregister WorkQueues, as well as to enable shutdown. It is
+ * only modified under a lock (normally briefly held, but
+ * occasionally protecting allocations and resizings) but even
+ * when locked remains available to check consistency. An
+ * auxiliary field "growHints", also only modified under lock,
+ * contains a candidate index for the next WorkQueue and
+ * a mask for submission queue indices.
+ *
+ * Recording WorkQueues. WorkQueues are recorded in the
+ * "workQueues" array that is created upon pool construction and
+ * expanded if necessary. Updates to the array while recording
+ * new workers and unrecording terminated ones are protected from
+ * each other by a lock but the array is otherwise concurrently
+ * readable, and accessed directly. To simplify index-based
+ * operations, the array size is always a power of two, and all
+ * readers must tolerate null slots. Shared (submission) queues
+ * are at even indices, worker queues at odd indices. Grouping
+ * them together in this way simplifies and speeds up task
+ * scanning. To avoid flailing during start-up, the array is
+ * presized to hold twice #parallelism workers (which is unlikely
+ * to need further resizing during execution). But to avoid
+ * dealing with so many null slots, variable runState includes a
+ * mask for the nearest power of two that contains all currently
+ * used indices.
+ *
+ * All worker thread creation is on-demand, triggered by task
+ * submissions, replacement of terminated workers, and/or
+ * compensation for blocked workers. However, all other support
+ * code is set up to work with other policies. To ensure that we
+ * do not hold on to worker references that would prevent GC, ALL
+ * accesses to workQueues are via indices into the workQueues
+ * array (which is one source of some of the messy code
+ * constructions here). In essence, the workQueues array serves as
+ * a weak reference mechanism. Thus for example the wait queue
+ * field of ctl stores indices, not references. Access to the
+ * workQueues in associated methods (for example signalWork) must
+ * both index-check and null-check the IDs. All such accesses
+ * ignore bad IDs by returning out early from what they are doing,
+ * since this can only be associated with termination, in which
+ * case it is OK to give up. All uses of the workQueues array
+ * also check that it is non-null (even if previously
+ * non-null). This allows nulling during termination, which is
+ * currently not necessary, but remains an option for
+ * resource-revocation-based shutdown schemes. It also helps
+ * reduce JIT issuance of uncommon-trap code, which tends to
+ * unnecessarily complicate control flow in some methods.
+ *
+ * Event Queuing. Unlike HPC work-stealing frameworks, we cannot
+ * let workers spin indefinitely scanning for tasks when none can
+ * be found immediately, and we cannot start/resume workers unless
+ * there appear to be tasks available. On the other hand, we must
+ * quickly prod them into action when new tasks are submitted or
+ * generated. In many usages, ramp-up time to activate workers is
+ * the main limiting factor in overall performance (this is
+ * compounded at program start-up by JIT compilation and
+ * allocation). So we try to streamline this as much as possible.
+ * We park/unpark workers after placing in an event wait queue
+ * when they cannot find work. This "queue" is actually a simple
+ * Treiber stack, headed by the "id" field of ctl, plus a 15bit
+ * counter value (that reflects the number of times a worker has
+ * been inactivated) to avoid ABA effects (we need only as many
+ * version numbers as worker threads). Successors are held in
+ * field WorkQueue.nextWait. Queuing deals with several intrinsic
+ * races, mainly that a task-producing thread can miss seeing (and
+ * signalling) another thread that gave up looking for work but
+ * has not yet entered the wait queue. We solve this by requiring
+ * a full sweep of all workers (via repeated calls to method
+ * scan()) both before and after a newly waiting worker is added
+ * to the wait queue. During a rescan, the worker might release
+ * some other queued worker rather than itself, which has the same
+ * net effect. Because enqueued workers may actually be rescanning
+ * rather than waiting, we set and clear the "parker" field of
+ * WorkQueues to reduce unnecessary calls to unpark. (This
+ * requires a secondary recheck to avoid missed signals.) Note
+ * the unusual conventions about Thread.interrupts surrounding
+ * parking and other blocking: Because interrupts are used solely
+ * to alert threads to check termination, which is checked anyway
+ * upon blocking, we clear status (using Thread.interrupted)
+ * before any call to park, so that park does not immediately
+ * return due to status being set via some other unrelated call to
+ * interrupt in user code.
+ *
+ * Signalling. We create or wake up workers only when there
+ * appears to be at least one task they might be able to find and
+ * execute. When a submission is added or another worker adds a
+ * task to a queue that previously had fewer than two tasks, they
+ * signal waiting workers (or trigger creation of new ones if
+ * fewer than the given parallelism level -- see signalWork).
+ * These primary signals are buttressed by signals during rescans;
+ * together these cover the signals needed in cases when more
+ * tasks are pushed but untaken, and improve performance compared
+ * to having one thread wake up all workers.
+ *
+ * Trimming workers. To release resources after periods of lack of
+ * use, a worker starting to wait when the pool is quiescent will
+ * time out and terminate if the pool has remained quiescent for
+ * SHRINK_RATE nanosecs. This will slowly propagate, eventually
+ * terminating all workers after long periods of non-use.
+ *
+ * Shutdown and Termination. A call to shutdownNow atomically sets
+ * a runState bit and then (non-atomically) sets each worker's
+ * runState status, cancels all unprocessed tasks, and wakes up
+ * all waiting workers. Detecting whether termination should
+ * commence after a non-abrupt shutdown() call requires more work
+ * and bookkeeping. We need consensus about quiescence (i.e., that
+ * there is no more work). The active count provides a primary
+ * indication but non-abrupt shutdown still requires a rechecking
+ * scan for any workers that are inactive but not queued.
+ *
+ * Joining Tasks
+ * =============
+ *
+ * Any of several actions may be taken when one worker is waiting
+ * to join a task stolen (or always held) by another. Because we
+ * are multiplexing many tasks on to a pool of workers, we can't
+ * just let them block (as in Thread.join). We also cannot just
+ * reassign the joiner's run-time stack with another and replace
+ * it later, which would be a form of "continuation", that even if
+ * possible is not necessarily a good idea since we sometimes need
+ * both an unblocked task and its continuation to progress.
+ * Instead we combine two tactics:
+ *
+ * Helping: Arranging for the joiner to execute some task that it
+ * would be running if the steal had not occurred.
+ *
+ * Compensating: Unless there are already enough live threads,
+ * method tryCompensate() may create or re-activate a spare
+ * thread to compensate for blocked joiners until they unblock.
+ *
+ * A third form (implemented in tryRemoveAndExec and
+ * tryPollForAndExec) amounts to helping a hypothetical
+ * compensator: If we can readily tell that a possible action of a
+ * compensator is to steal and execute the task being joined, the
+ * joining thread can do so directly, without the need for a
+ * compensation thread (although at the expense of larger run-time
+ * stacks, but the tradeoff is typically worthwhile).
+ *
+ * The ManagedBlocker extension API can't use helping so relies
+ * only on compensation in method awaitBlocker.
+ *
+ * The algorithm in tryHelpStealer entails a form of "linear"
+ * helping: Each worker records (in field currentSteal) the most
+ * recent task it stole from some other worker. Plus, it records
+ * (in field currentJoin) the task it is currently actively
+ * joining. Method tryHelpStealer uses these markers to try to
+ * find a worker to help (i.e., steal back a task from and execute
+ * it) that could hasten completion of the actively joined task.
+ * In essence, the joiner executes a task that would be on its own
+ * local deque had the to-be-joined task not been stolen. This may
+ * be seen as a conservative variant of the approach in Wagner &
+ * Calder "Leapfrogging: a portable technique for implementing
+ * efficient futures" SIGPLAN Notices, 1993
+ * (http://portal.acm.org/citation.cfm?id=155354). It differs in
+ * that: (1) We only maintain dependency links across workers upon
+ * steals, rather than use per-task bookkeeping. This sometimes
+ * requires a linear scan of workQueues array to locate stealers, but
+ * often doesn't because stealers leave hints (that may become
+ * stale/wrong) of where to locate them. A stealHint is only a
+ * hint because a worker might have had multiple steals and the
+ * hint records only one of them (usually the most current).
+ * Hinting isolates cost to when it is needed, rather than adding
+ * to per-task overhead. (2) It is "shallow", ignoring nesting
+ * and potentially cyclic mutual steals. (3) It is intentionally
+ * racy: field currentJoin is updated only while actively joining,
+ * which means that we miss links in the chain during long-lived
+ * tasks, GC stalls etc (which is OK since blocking in such cases
+ * is usually a good idea). (4) We bound the number of attempts
+ * to find work (see MAX_HELP_DEPTH) and fall back to suspending
+ * the worker and if necessary replacing it with another.
+ *
+ * It is impossible to keep exactly the target parallelism number
+ * of threads running at any given time. Determining the
+ * existence of conservatively safe helping targets, the
+ * availability of already-created spares, and the apparent need
+ * to create new spares are all racy, so we rely on multiple
+ * retries of each. Currently, in keeping with on-demand
+ * signalling policy, we compensate only if blocking would leave
+ * less than one active (non-waiting, non-blocked) worker.
+ * Additionally, to avoid some false alarms due to GC, lagging
+ * counters, system activity, etc, compensated blocking for joins
+ * is only attempted after rechecks stabilize in
+ * ForkJoinTask.awaitJoin. (Retries are interspersed with
+ * Thread.yield, for good citizenship.)
+ *
+ * Style notes: There is a lot of representation-level coupling
+ * among classes ForkJoinPool, ForkJoinWorkerThread, and
+ * ForkJoinTask. The fields of WorkQueue maintain data structures
+ * managed by ForkJoinPool, so are directly accessed. There is
+ * little point trying to reduce this, since any associated future
+ * changes in representations will need to be accompanied by
+ * algorithmic changes anyway. Several methods intrinsically
+ * sprawl because they must accumulate sets of consistent reads of
+ * volatiles held in local variables. Methods signalWork() and
+ * scan() are the main bottlenecks, so are especially heavily
+ * micro-optimized/mangled. There are lots of inline assignments
+ * (of form "while ((local = field) != 0)") which are usually the
+ * simplest way to ensure the required read orderings (which are
+ * sometimes critical). This leads to a "C"-like style of listing
+ * declarations of these locals at the heads of methods or blocks.
+ * There are several occurrences of the unusual "do {} while
+ * (!cas...)" which is the simplest way to force an update of a
+ * CAS'ed variable. There are also other coding oddities that help
+ * some methods perform reasonably even when interpreted (not
+ * compiled).
+ *
+ * The order of declarations in this file is:
+ * (1) Static utility functions
+ * (2) Nested (static) classes
+ * (3) Static fields
+ * (4) Fields, along with constants used when unpacking some of them
+ * (5) Internal control methods
+ * (6) Callbacks and other support for ForkJoinTask methods
+ * (7) Exported methods
+ * (8) Static block initializing statics in minimally dependent order
+ */
+
+ // Static utilities
+
+ /**
+ * Computes an initial hash code (also serving as a non-zero
+ * random seed) for a thread id. This method is expected to
+ * provide higher-quality hash codes than using method hashCode().
+ */
+ static final int hashId(long id) {
+ int h = (int)id ^ (int)(id >>> 32); // Use MurmurHash of thread id
+ h ^= h >>> 16; h *= 0x85ebca6b;
+ h ^= h >>> 13; h *= 0xc2b2ae35;
+ h ^= h >>> 16;
+ return (h == 0) ? 1 : h; // ensure nonzero
+ }
+
+ /**
+ * If there is a security manager, makes sure caller has
+ * permission to modify threads.
+ */
+ private static void checkPermission() {
+ SecurityManager security = System.getSecurityManager();
+ if (security != null)
+ security.checkPermission(modifyThreadPermission);
+ }
+
+ // Nested classes
+
+ /**
+ * Factory for creating new {@link ForkJoinWorkerThread}s.
+ * A {@code ForkJoinWorkerThreadFactory} must be defined and used
+ * for {@code ForkJoinWorkerThread} subclasses that extend base
+ * functionality or initialize threads with different contexts.
+ */
+ public static interface ForkJoinWorkerThreadFactory {
+ /**
+ * Returns a new worker thread operating in the given pool.
+ *
+ * @param pool the pool this thread works in
+ * @throws NullPointerException if the pool is null
+ */
+ public ForkJoinWorkerThread newThread(ForkJoinPool pool);
+ }
+
+ /**
+ * Default ForkJoinWorkerThreadFactory implementation; creates a
+ * new ForkJoinWorkerThread.
+ */
+ static class DefaultForkJoinWorkerThreadFactory
+ implements ForkJoinWorkerThreadFactory {
+ public ForkJoinWorkerThread newThread(ForkJoinPool pool) {
+ return new ForkJoinWorkerThread(pool);
+ }
+ }
+
+ /**
+ * A simple non-reentrant lock used for exclusion when managing
+ * queues and workers. We use a custom lock so that we can readily
+ * probe lock state in constructions that check among alternative
+ * actions. The lock is normally only very briefly held, and
+ * sometimes treated as a spinlock, but other usages block to
+ * reduce overall contention in those cases where locked code
+ * bodies perform allocation/resizing.
+ */
+ static final class Mutex extends AbstractQueuedSynchronizer {
+ public final boolean tryAcquire(int ignore) {
+ return compareAndSetState(0, 1);
+ }
+ public final boolean tryRelease(int ignore) {
+ setState(0);
+ return true;
+ }
+ public final void lock() { acquire(0); }
+ public final void unlock() { release(0); }
+ public final boolean isHeldExclusively() { return getState() == 1; }
+ public final Condition newCondition() { return new ConditionObject(); }
+ }
+
+ /**
+ * Class for artificial tasks that are used to replace the target
+ * of local joins if they are removed from an interior queue slot
+ * in WorkQueue.tryRemoveAndExec. We don't need the proxy to
+ * actually do anything beyond having a unique identity.
+ */
+ static final class EmptyTask extends ForkJoinTask {
+ EmptyTask() { status = ForkJoinTask.NORMAL; } // force done
+ public final Void getRawResult() { return null; }
+ public final void setRawResult(Void x) {}
+ public final boolean exec() { return true; }
+ }
+
+ /**
+ * Queues supporting work-stealing as well as external task
+ * submission. See above for main rationale and algorithms.
+ * Implementation relies heavily on "Unsafe" intrinsics
+ * and selective use of "volatile":
+ *
+ * Field "base" is the index (mod array.length) of the least valid
+ * queue slot, which is always the next position to steal (poll)
+ * from if nonempty. Reads and writes require volatile orderings
+ * but not CAS, because updates are only performed after slot
+ * CASes.
+ *
+ * Field "top" is the index (mod array.length) of the next queue
+ * slot to push to or pop from. It is written only by owner thread
+ * for push, or under lock for trySharedPush, and accessed by
+ * other threads only after reading (volatile) base. Both top and
+ * base are allowed to wrap around on overflow, but (top - base)
+ * (or more commonly -(base - top) to force volatile read of base
+ * before top) still estimates size.
+ *
+ * The array slots are read and written using the emulation of
+ * volatiles/atomics provided by Unsafe. Insertions must in
+ * general use putOrderedObject as a form of releasing store to
+ * ensure that all writes to the task object are ordered before
+ * its publication in the queue. (Although we can avoid one case
+ * of this when locked in trySharedPush.) All removals entail a
+ * CAS to null. The array is always a power of two. To ensure
+ * safety of Unsafe array operations, all accesses perform
+ * explicit null checks and implicit bounds checks via
+ * power-of-two masking.
+ *
+ * In addition to basic queuing support, this class contains
+ * fields described elsewhere to control execution. It turns out
+ * to work better memory-layout-wise to include them in this
+ * class rather than a separate class.
+ *
+ * Performance on most platforms is very sensitive to placement of
+ * instances of both WorkQueues and their arrays -- we absolutely
+ * do not want multiple WorkQueue instances or multiple queue
+ * arrays sharing cache lines. (It would be best for queue objects
+ * and their arrays to share, but there is nothing available to
+ * help arrange that). Unfortunately, because they are recorded
+ * in a common array, WorkQueue instances are often moved to be
+ * adjacent by garbage collectors. To reduce impact, we use field
+ * padding that works OK on common platforms; this effectively
+ * trades off slightly slower average field access for the sake of
+ * avoiding really bad worst-case access. (Until better JVM
+ * support is in place, this padding is dependent on transient
+ * properties of JVM field layout rules.) We also take care in
+ * allocating, sizing and resizing the array. Non-shared queue
+ * arrays are initialized (via method growArray) by workers before
+ * use. Others are allocated on first use.
+ */
+ static final class WorkQueue {
+ /**
+ * Capacity of work-stealing queue array upon initialization.
+ * Must be a power of two; at least 4, but set larger to
+ * reduce cacheline sharing among queues.
+ */
+ static final int INITIAL_QUEUE_CAPACITY = 1 << 8;
+
+ /**
+ * Maximum size for queue arrays. Must be a power of two less
+ * than or equal to 1 << (31 - width of array entry) to ensure
+ * lack of wraparound of index calculations, but defined to a
+ * value a bit less than this to help users trap runaway
+ * programs before saturating systems.
+ */
+ static final int MAXIMUM_QUEUE_CAPACITY = 1 << 26; // 64M
+
+ volatile long totalSteals; // cumulative number of steals
+ int seed; // for random scanning; initialize nonzero
+ volatile int eventCount; // encoded inactivation count; < 0 if inactive
+ int nextWait; // encoded record of next event waiter
+ int rescans; // remaining scans until block
+ int nsteals; // top-level task executions since last idle
+ final int mode; // lifo, fifo, or shared
+ int poolIndex; // index of this queue in pool (or 0)
+ int stealHint; // index of most recent known stealer
+ volatile int runState; // 1: locked, -1: terminate; else 0
+ volatile int base; // index of next slot for poll
+ int top; // index of next slot for push
+ ForkJoinTask[] array; // the elements (initially unallocated)
+ final ForkJoinWorkerThread owner; // owning thread or null if shared
+ volatile Thread parker; // == owner during call to park; else null
+ ForkJoinTask currentJoin; // task being joined in awaitJoin
+ ForkJoinTask currentSteal; // current non-local task being executed
+ // Heuristic padding to ameliorate unfortunate memory placements
+ Object p00, p01, p02, p03, p04, p05, p06, p07, p08, p09, p0a;
+
+ WorkQueue(ForkJoinWorkerThread owner, int mode) {
+ this.owner = owner;
+ this.mode = mode;
+ // Place indices in the center of array (that is not yet allocated)
+ base = top = INITIAL_QUEUE_CAPACITY >>> 1;
+ }
+
+ /**
+ * Returns number of tasks in the queue.
+ */
+ final int queueSize() {
+ int n = base - top; // non-owner callers must read base first
+ return (n >= 0) ? 0 : -n;
+ }
+
+ /**
+ * Pushes a task. Call only by owner in unshared queues.
+ *
+ * @param task the task. Caller must ensure non-null.
+ * @param p if non-null, pool to signal if necessary
+ * @throw RejectedExecutionException if array cannot be resized
+ */
+ final void push(ForkJoinTask task, ForkJoinPool p) {
+ ForkJoinTask[] a;
+ int s = top, m, n;
+ if ((a = array) != null) { // ignore if queue removed
+ U.putOrderedObject
+ (a, (((m = a.length - 1) & s) << ASHIFT) + ABASE, task);
+ if ((n = (top = s + 1) - base) <= 2) {
+ if (p != null)
+ p.signalWork();
+ }
+ else if (n >= m)
+ growArray(true);
+ }
+ }
+
+ /**
+ * Pushes a task if lock is free and array is either big
+ * enough or can be resized to be big enough.
+ *
+ * @param task the task. Caller must ensure non-null.
+ * @return true if submitted
+ */
+ final boolean trySharedPush(ForkJoinTask task) {
+ boolean submitted = false;
+ if (runState == 0 && U.compareAndSwapInt(this, RUNSTATE, 0, 1)) {
+ ForkJoinTask[] a = array;
+ int s = top;
+ try {
+ if ((a != null && a.length > s + 1 - base) ||
+ (a = growArray(false)) != null) { // must presize
+ int j = (((a.length - 1) & s) << ASHIFT) + ABASE;
+ U.putObject(a, (long)j, task); // don't need "ordered"
+ top = s + 1;
+ submitted = true;
+ }
+ } finally {
+ runState = 0; // unlock
+ }
+ }
+ return submitted;
+ }
+
+ /**
+ * Takes next task, if one exists, in FIFO order.
+ */
+ final ForkJoinTask poll() {
+ ForkJoinTask[] a; int b; ForkJoinTask t;
+ while ((b = base) - top < 0 && (a = array) != null) {
+ int j = (((a.length - 1) & b) << ASHIFT) + ABASE;
+ if ((t = (ForkJoinTask)U.getObjectVolatile(a, j)) != null &&
+ base == b &&
+ U.compareAndSwapObject(a, j, t, null)) {
+ base = b + 1;
+ return t;
+ }
+ }
+ return null;
+ }
+
+ /**
+ * Takes next task, if one exists, in LIFO order. Call only
+ * by owner in unshared queues. (We do not have a shared
+ * version of this method because it is never needed.)
+ */
+ final ForkJoinTask pop() {
+ ForkJoinTask t; int m;
+ ForkJoinTask[] a = array;
+ if (a != null && (m = a.length - 1) >= 0) {
+ for (int s; (s = top - 1) - base >= 0;) {
+ int j = ((m & s) << ASHIFT) + ABASE;
+ if ((t = (ForkJoinTask)U.getObjectVolatile(a, j)) == null)
+ break;
+ if (U.compareAndSwapObject(a, j, t, null)) {
+ top = s;
+ return t;
+ }
+ }
+ }
+ return null;
+ }
+
+ /**
+ * Takes next task, if one exists, in order specified by mode.
+ */
+ final ForkJoinTask nextLocalTask() {
+ return mode == 0 ? pop() : poll();
+ }
+
+ /**
+ * Returns next task, if one exists, in order specified by mode.
+ */
+ final ForkJoinTask peek() {
+ ForkJoinTask[] a = array; int m;
+ if (a == null || (m = a.length - 1) < 0)
+ return null;
+ int i = mode == 0 ? top - 1 : base;
+ int j = ((i & m) << ASHIFT) + ABASE;
+ return (ForkJoinTask)U.getObjectVolatile(a, j);
+ }
+
+ /**
+ * Returns task at index b if b is current base of queue.
+ */
+ final ForkJoinTask pollAt(int b) {
+ ForkJoinTask t; ForkJoinTask[] a;
+ if ((a = array) != null) {
+ int j = (((a.length - 1) & b) << ASHIFT) + ABASE;
+ if ((t = (ForkJoinTask)U.getObjectVolatile(a, j)) != null &&
+ base == b &&
+ U.compareAndSwapObject(a, j, t, null)) {
+ base = b + 1;
+ return t;
+ }
+ }
+ return null;
+ }
+
+ /**
+ * Pops the given task only if it is at the current top.
+ */
+ final boolean tryUnpush(ForkJoinTask t) {
+ ForkJoinTask[] a; int s;
+ if ((a = array) != null && (s = top) != base &&
+ U.compareAndSwapObject
+ (a, (((a.length - 1) & --s) << ASHIFT) + ABASE, t, null)) {
+ top = s;
+ return true;
+ }
+ return false;
+ }
+
+ /**
+ * Polls the given task only if it is at the current base.
+ */
+ final boolean pollFor(ForkJoinTask task) {
+ ForkJoinTask[] a; int b;
+ if ((b = base) - top < 0 && (a = array) != null) {
+ int j = (((a.length - 1) & b) << ASHIFT) + ABASE;
+ if (U.getObjectVolatile(a, j) == task && base == b &&
+ U.compareAndSwapObject(a, j, task, null)) {
+ base = b + 1;
+ return true;
+ }
+ }
+ return false;
+ }
+
+ /**
+ * If present, removes from queue and executes the given task, or
+ * any other cancelled task. Returns (true) immediately on any CAS
+ * or consistency check failure so caller can retry.
+ *
+ * @return false if no progress can be made
+ */
+ final boolean tryRemoveAndExec(ForkJoinTask task) {
+ boolean removed = false, empty = true, progress = true;
+ ForkJoinTask[] a; int m, s, b, n;
+ if ((a = array) != null && (m = a.length - 1) >= 0 &&
+ (n = (s = top) - (b = base)) > 0) {
+ for (ForkJoinTask t;;) { // traverse from s to b
+ int j = ((--s & m) << ASHIFT) + ABASE;
+ t = (ForkJoinTask)U.getObjectVolatile(a, j);
+ if (t == null) // inconsistent length
+ break;
+ else if (t == task) {
+ if (s + 1 == top) { // pop
+ if (!U.compareAndSwapObject(a, j, task, null))
+ break;
+ top = s;
+ removed = true;
+ }
+ else if (base == b) // replace with proxy
+ removed = U.compareAndSwapObject(a, j, task,
+ new EmptyTask());
+ break;
+ }
+ else if (t.status >= 0)
+ empty = false;
+ else if (s + 1 == top) { // pop and throw away
+ if (U.compareAndSwapObject(a, j, t, null))
+ top = s;
+ break;
+ }
+ if (--n == 0) {
+ if (!empty && base == b)
+ progress = false;
+ break;
+ }
+ }
+ }
+ if (removed)
+ task.doExec();
+ return progress;
+ }
+
+ /**
+ * Initializes or doubles the capacity of array. Call either
+ * by owner or with lock held -- it is OK for base, but not
+ * top, to move while resizings are in progress.
+ *
+ * @param rejectOnFailure if true, throw exception if capacity
+ * exceeded (relayed ultimately to user); else return null.
+ */
+ final ForkJoinTask[] growArray(boolean rejectOnFailure) {
+ ForkJoinTask[] oldA = array;
+ int size = oldA != null ? oldA.length << 1 : INITIAL_QUEUE_CAPACITY;
+ if (size <= MAXIMUM_QUEUE_CAPACITY) {
+ int oldMask, t, b;
+ ForkJoinTask[] a = array = new ForkJoinTask[size];
+ if (oldA != null && (oldMask = oldA.length - 1) >= 0 &&
+ (t = top) - (b = base) > 0) {
+ int mask = size - 1;
+ do {
+ ForkJoinTask x;
+ int oldj = ((b & oldMask) << ASHIFT) + ABASE;
+ int j = ((b & mask) << ASHIFT) + ABASE;
+ x = (ForkJoinTask)U.getObjectVolatile(oldA, oldj);
+ if (x != null &&
+ U.compareAndSwapObject(oldA, oldj, x, null))
+ U.putObjectVolatile(a, j, x);
+ } while (++b != t);
+ }
+ return a;
+ }
+ else if (!rejectOnFailure)
+ return null;
+ else
+ throw new RejectedExecutionException("Queue capacity exceeded");
+ }
+
+ /**
+ * Removes and cancels all known tasks, ignoring any exceptions.
+ */
+ final void cancelAll() {
+ ForkJoinTask.cancelIgnoringExceptions(currentJoin);
+ ForkJoinTask.cancelIgnoringExceptions(currentSteal);
+ for (ForkJoinTask t; (t = poll()) != null; )
+ ForkJoinTask.cancelIgnoringExceptions(t);
+ }
+
+ /**
+ * Computes next value for random probes. Scans don't require
+ * a very high quality generator, but also not a crummy one.
+ * Marsaglia xor-shift is cheap and works well enough. Note:
+ * This is manually inlined in several usages in ForkJoinPool
+ * to avoid writes inside busy scan loops.
+ */
+ final int nextSeed() {
+ int r = seed;
+ r ^= r << 13;
+ r ^= r >>> 17;
+ return seed = r ^= r << 5;
+ }
+
+ // Execution methods
+
+ /**
+ * Removes and runs tasks until empty, using local mode
+ * ordering.
+ */
+ final void runLocalTasks() {
+ if (base - top < 0) {
+ for (ForkJoinTask t; (t = nextLocalTask()) != null; )
+ t.doExec();
+ }
+ }
+
+ /**
+ * Executes a top-level task and any local tasks remaining
+ * after execution.
+ *
+ * @return true unless terminating
+ */
+ final boolean runTask(ForkJoinTask t) {
+ boolean alive = true;
+ if (t != null) {
+ currentSteal = t;
+ t.doExec();
+ runLocalTasks();
+ ++nsteals;
+ currentSteal = null;
+ }
+ else if (runState < 0) // terminating
+ alive = false;
+ return alive;
+ }
+
+ /**
+ * Executes a non-top-level (stolen) task.
+ */
+ final void runSubtask(ForkJoinTask t) {
+ if (t != null) {
+ ForkJoinTask ps = currentSteal;
+ currentSteal = t;
+ t.doExec();
+ currentSteal = ps;
+ }
+ }
+
+ /**
+ * Returns true if owned and not known to be blocked.
+ */
+ final boolean isApparentlyUnblocked() {
+ Thread wt; Thread.State s;
+ return (eventCount >= 0 &&
+ (wt = owner) != null &&
+ (s = wt.getState()) != Thread.State.BLOCKED &&
+ s != Thread.State.WAITING &&
+ s != Thread.State.TIMED_WAITING);
+ }
+
+ /**
+ * If this owned and is not already interrupted, try to
+ * interrupt and/or unpark, ignoring exceptions.
+ */
+ final void interruptOwner() {
+ Thread wt, p;
+ if ((wt = owner) != null && !wt.isInterrupted()) {
+ try {
+ wt.interrupt();
+ } catch (SecurityException ignore) {
+ }
+ }
+ if ((p = parker) != null)
+ U.unpark(p);
+ }
+
+ // Unsafe mechanics
+ private static final sun.misc.Unsafe U;
+ private static final long RUNSTATE;
+ private static final int ABASE;
+ private static final int ASHIFT;
+ static {
+ int s;
+ try {
+ U = getUnsafe();
+ Class k = WorkQueue.class;
+ Class ak = ForkJoinTask[].class;
+ RUNSTATE = U.objectFieldOffset
+ (k.getDeclaredField("runState"));
+ ABASE = U.arrayBaseOffset(ak);
+ s = U.arrayIndexScale(ak);
+ } catch (Exception e) {
+ throw new Error(e);
+ }
+ if ((s & (s-1)) != 0)
+ throw new Error("data type scale not a power of two");
+ ASHIFT = 31 - Integer.numberOfLeadingZeros(s);
+ }
+ }
+
+ /**
+ * Per-thread records for threads that submit to pools. Currently
+ * holds only pseudo-random seed / index that is used to choose
+ * submission queues in method doSubmit. In the future, this may
+ * also incorporate a means to implement different task rejection
+ * and resubmission policies.
+ */
+ static final class Submitter {
+ int seed;
+ Submitter() { seed = hashId(Thread.currentThread().getId()); }
+ }
+
+ /** ThreadLocal class for Submitters */
+ static final class ThreadSubmitter extends ThreadLocal {
+ public Submitter initialValue() { return new Submitter(); }
+ }
+
+ // static fields (initialized in static initializer below)
+
+ /**
+ * Creates a new ForkJoinWorkerThread. This factory is used unless
+ * overridden in ForkJoinPool constructors.
+ */
+ public static final ForkJoinWorkerThreadFactory
+ defaultForkJoinWorkerThreadFactory;
+
+ /**
+ * Generator for assigning sequence numbers as pool names.
+ */
+ private static final AtomicInteger poolNumberGenerator;
+
+ /**
+ * Permission required for callers of methods that may start or
+ * kill threads.
+ */
+ private static final RuntimePermission modifyThreadPermission;
+
+ /**
+ * Per-thread submission bookeeping. Shared across all pools
+ * to reduce ThreadLocal pollution and because random motion
+ * to avoid contention in one pool is likely to hold for others.
+ */
+ private static final ThreadSubmitter submitters;
+
+ // static constants
+
+ /**
+ * The wakeup interval (in nanoseconds) for a worker waiting for a
+ * task when the pool is quiescent to instead try to shrink the
+ * number of workers. The exact value does not matter too
+ * much. It must be short enough to release resources during
+ * sustained periods of idleness, but not so short that threads
+ * are continually re-created.
+ */
+ private static final long SHRINK_RATE =
+ 4L * 1000L * 1000L * 1000L; // 4 seconds
+
+ /**
+ * The timeout value for attempted shrinkage, includes
+ * some slop to cope with system timer imprecision.
+ */
+ private static final long SHRINK_TIMEOUT = SHRINK_RATE - (SHRINK_RATE / 10);
+
+ /**
+ * The maximum stolen->joining link depth allowed in tryHelpStealer.
+ * Depths for legitimate chains are unbounded, but we use a fixed
+ * constant to avoid (otherwise unchecked) cycles and to bound
+ * staleness of traversal parameters at the expense of sometimes
+ * blocking when we could be helping.
+ */
+ private static final int MAX_HELP_DEPTH = 16;
+
+ /**
+ * Bits and masks for control variables
+ *
+ * Field ctl is a long packed with:
+ * AC: Number of active running workers minus target parallelism (16 bits)
+ * TC: Number of total workers minus target parallelism (16 bits)
+ * ST: true if pool is terminating (1 bit)
+ * EC: the wait count of top waiting thread (15 bits)
+ * ID: poolIndex of top of Treiber stack of waiters (16 bits)
+ *
+ * When convenient, we can extract the upper 32 bits of counts and
+ * the lower 32 bits of queue state, u = (int)(ctl >>> 32) and e =
+ * (int)ctl. The ec field is never accessed alone, but always
+ * together with id and st. The offsets of counts by the target
+ * parallelism and the positionings of fields makes it possible to
+ * perform the most common checks via sign tests of fields: When
+ * ac is negative, there are not enough active workers, when tc is
+ * negative, there are not enough total workers, and when e is
+ * negative, the pool is terminating. To deal with these possibly
+ * negative fields, we use casts in and out of "short" and/or
+ * signed shifts to maintain signedness.
+ *
+ * When a thread is queued (inactivated), its eventCount field is
+ * set negative, which is the only way to tell if a worker is
+ * prevented from executing tasks, even though it must continue to
+ * scan for them to avoid queuing races. Note however that
+ * eventCount updates lag releases so usage requires care.
+ *
+ * Field runState is an int packed with:
+ * SHUTDOWN: true if shutdown is enabled (1 bit)
+ * SEQ: a sequence number updated upon (de)registering workers (15 bits)
+ * MASK: mask (power of 2 - 1) covering all registered poolIndexes (16 bits)
+ *
+ * The combination of mask and sequence number enables simple
+ * consistency checks: Staleness of read-only operations on the
+ * workQueues array can be checked by comparing runState before vs
+ * after the reads. The low 16 bits (i.e, anding with SMASK) hold
+ * the smallest power of two covering all indices, minus
+ * one.
+ */
+
+ // bit positions/shifts for fields
+ private static final int AC_SHIFT = 48;
+ private static final int TC_SHIFT = 32;
+ private static final int ST_SHIFT = 31;
+ private static final int EC_SHIFT = 16;
+
+ // bounds
+ private static final int POOL_MAX = 0x7fff; // max #workers - 1
+ private static final int SMASK = 0xffff; // short bits
+ private static final int SQMASK = 0xfffe; // even short bits
+ private static final int SHORT_SIGN = 1 << 15;
+ private static final int INT_SIGN = 1 << 31;
+
+ // masks
+ private static final long STOP_BIT = 0x0001L << ST_SHIFT;
+ private static final long AC_MASK = ((long)SMASK) << AC_SHIFT;
+ private static final long TC_MASK = ((long)SMASK) << TC_SHIFT;
+
+ // units for incrementing and decrementing
+ private static final long TC_UNIT = 1L << TC_SHIFT;
+ private static final long AC_UNIT = 1L << AC_SHIFT;
+
+ // masks and units for dealing with u = (int)(ctl >>> 32)
+ private static final int UAC_SHIFT = AC_SHIFT - 32;
+ private static final int UTC_SHIFT = TC_SHIFT - 32;
+ private static final int UAC_MASK = SMASK << UAC_SHIFT;
+ private static final int UTC_MASK = SMASK << UTC_SHIFT;
+ private static final int UAC_UNIT = 1 << UAC_SHIFT;
+ private static final int UTC_UNIT = 1 << UTC_SHIFT;
+
+ // masks and units for dealing with e = (int)ctl
+ private static final int E_MASK = 0x7fffffff; // no STOP_BIT
+ private static final int E_SEQ = 1 << EC_SHIFT;
+
+ // runState bits
+ private static final int SHUTDOWN = 1 << 31;
+ private static final int RS_SEQ = 1 << 16;
+ private static final int RS_SEQ_MASK = 0x7fff0000;
+
+ // access mode for WorkQueue
+ static final int LIFO_QUEUE = 0;
+ static final int FIFO_QUEUE = 1;
+ static final int SHARED_QUEUE = -1;
+
+ // Instance fields
+
+ /*
+ * Field layout order in this class tends to matter more than one
+ * would like. Runtime layout order is only loosely related to
+ * declaration order and may differ across JVMs, but the following
+ * empirically works OK on current JVMs.
+ */
+
+ volatile long ctl; // main pool control
+ final int parallelism; // parallelism level
+ final int localMode; // per-worker scheduling mode
+ int growHints; // for expanding indices/ranges
+ volatile int runState; // shutdown status, seq, and mask
+ WorkQueue[] workQueues; // main registry
+ final Mutex lock; // for registration
+ final Condition termination; // for awaitTermination
+ final ForkJoinWorkerThreadFactory factory; // factory for new workers
+ final Thread.UncaughtExceptionHandler ueh; // per-worker UEH
+ final AtomicLong stealCount; // collect counts when terminated
+ final AtomicInteger nextWorkerNumber; // to create worker name string
+ final String workerNamePrefix; // to create worker name string
+
+ // Creating, registering, deregistering and running workers
+
+ /**
+ * Tries to create and start a worker
+ */
+ private void addWorker() {
+ Throwable ex = null;
+ ForkJoinWorkerThread wt = null;
+ try {
+ if ((wt = factory.newThread(this)) != null) {
+ wt.start();
+ return;
+ }
+ } catch (Throwable e) {
+ ex = e;
+ }
+ deregisterWorker(wt, ex); // adjust counts etc on failure
+ }
+
+ /**
+ * Callback from ForkJoinWorkerThread constructor to assign a
+ * public name. This must be separate from registerWorker because
+ * it is called during the "super" constructor call in
+ * ForkJoinWorkerThread.
+ */
+ final String nextWorkerName() {
+ return workerNamePrefix.concat
+ (Integer.toString(nextWorkerNumber.addAndGet(1)));
+ }
+
+ /**
+ * Callback from ForkJoinWorkerThread constructor to establish and
+ * record its WorkQueue.
+ *
+ * @param wt the worker thread
+ */
+ final void registerWorker(ForkJoinWorkerThread wt) {
+ WorkQueue w = wt.workQueue;
+ Mutex lock = this.lock;
+ lock.lock();
+ try {
+ int g = growHints, k = g & SMASK;
+ WorkQueue[] ws = workQueues;
+ if (ws != null) { // ignore on shutdown
+ int n = ws.length;
+ if ((k & 1) == 0 || k >= n || ws[k] != null) {
+ for (k = 1; k < n && ws[k] != null; k += 2)
+ ; // workers are at odd indices
+ if (k >= n) // resize
+ workQueues = ws = Arrays.copyOf(ws, n << 1);
+ }
+ w.eventCount = w.poolIndex = k; // establish before recording
+ ws[k] = w;
+ growHints = (g & ~SMASK) | ((k + 2) & SMASK);
+ int rs = runState;
+ int m = rs & SMASK; // recalculate runState mask
+ if (k > m)
+ m = (m << 1) + 1;
+ runState = (rs & SHUTDOWN) | ((rs + RS_SEQ) & RS_SEQ_MASK) | m;
+ }
+ } finally {
+ lock.unlock();
+ }
+ }
+
+ /**
+ * Final callback from terminating worker, as well as upon failure
+ * to construct or start a worker in addWorker. Removes record of
+ * worker from array, and adjusts counts. If pool is shutting
+ * down, tries to complete termination.
+ *
+ * @param wt the worker thread or null if addWorker failed
+ * @param ex the exception causing failure, or null if none
+ */
+ final void deregisterWorker(ForkJoinWorkerThread wt, Throwable ex) {
+ WorkQueue w = null;
+ if (wt != null && (w = wt.workQueue) != null) {
+ w.runState = -1; // ensure runState is set
+ stealCount.getAndAdd(w.totalSteals + w.nsteals);
+ int idx = w.poolIndex;
+ Mutex lock = this.lock;
+ lock.lock();
+ try { // remove record from array
+ WorkQueue[] ws = workQueues;
+ if (ws != null && idx >= 0 && idx < ws.length && ws[idx] == w) {
+ ws[idx] = null;
+ growHints = (growHints & ~SMASK) | idx;
+ }
+ } finally {
+ lock.unlock();
+ }
+ }
+
+ long c; // adjust ctl counts
+ do {} while (!U.compareAndSwapLong
+ (this, CTL, c = ctl, (((c - AC_UNIT) & AC_MASK) |
+ ((c - TC_UNIT) & TC_MASK) |
+ (c & ~(AC_MASK|TC_MASK)))));
+
+ if (!tryTerminate(false, false) && w != null) {
+ w.cancelAll(); // cancel remaining tasks
+ if (w.array != null) // suppress signal if never ran
+ signalWork(); // wake up or create replacement
+ if (ex == null) // help clean refs on way out
+ ForkJoinTask.helpExpungeStaleExceptions();
+ }
+
+ if (ex != null) // rethrow
+ U.throwException(ex);
+ }
+
+ /**
+ * Top-level runloop for workers, called by ForkJoinWorkerThread.run.
+ */
+ final void runWorker(ForkJoinWorkerThread wt) {
+ // Initialize queue array and seed in this thread
+ WorkQueue w = wt.workQueue;
+ w.growArray(false);
+ w.seed = hashId(Thread.currentThread().getId());
+
+ do {} while (w.runTask(scan(w)));
+ }
+
+ // Submissions
+
+ /**
+ * Unless shutting down, adds the given task to a submission queue
+ * at submitter's current queue index (modulo submission
+ * range). If no queue exists at the index, one is created unless
+ * pool lock is busy. If the queue and/or lock are busy, another
+ * index is randomly chosen. The mask in growHints controls the
+ * effective index range of queues considered. The mask is
+ * expanded, up to the current workerQueue mask, upon any detected
+ * contention but otherwise remains small to avoid needlessly
+ * creating queues when there is no contention.
+ */
+ private void doSubmit(ForkJoinTask task) {
+ if (task == null)
+ throw new NullPointerException();
+ Submitter s = submitters.get();
+ for (int r = s.seed, m = growHints >>> 16;;) {
+ WorkQueue[] ws; WorkQueue q; Mutex lk;
+ int k = r & m & SQMASK; // use only even indices
+ if (runState < 0 || (ws = workQueues) == null || ws.length <= k)
+ throw new RejectedExecutionException(); // shutting down
+ if ((q = ws[k]) == null && (lk = lock).tryAcquire(0)) {
+ try { // try to create new queue
+ if (ws == workQueues && (q = ws[k]) == null) {
+ int rs; // update runState seq
+ ws[k] = q = new WorkQueue(null, SHARED_QUEUE);
+ runState = (((rs = runState) & SHUTDOWN) |
+ ((rs + RS_SEQ) & ~SHUTDOWN));
+ }
+ } finally {
+ lk.unlock();
+ }
+ }
+ if (q != null) {
+ if (q.trySharedPush(task)) {
+ signalWork();
+ return;
+ }
+ else if (m < parallelism - 1 && m < (runState & SMASK)) {
+ Mutex lock = this.lock;
+ lock.lock(); // block until lock free
+ int g = growHints;
+ if (g >>> 16 == m) // expand range
+ growHints = (((m << 1) + 1) << 16) | (g & SMASK);
+ lock.unlock(); // no need for try/finally
+ }
+ else if ((r & m) == 0)
+ Thread.yield(); // occasionally yield if busy
+ }
+ if (m == (m = growHints >>> 16)) {
+ r ^= r << 13; // update seed unless new range
+ r ^= r >>> 17; // same xorshift as WorkQueues
+ s.seed = r ^= r << 5;
+ }
+ }
+ }
+
+ // Maintaining ctl counts
+
+ /**
+ * Increments active count; mainly called upon return from blocking.
+ */
+ final void incrementActiveCount() {
+ long c;
+ do {} while (!U.compareAndSwapLong(this, CTL, c = ctl, c + AC_UNIT));
+ }
+
+ /**
+ * Tries to activate or create a worker if too few are active.
+ */
+ final void signalWork() {
+ long c; int u;
+ while ((u = (int)((c = ctl) >>> 32)) < 0) { // too few active
+ WorkQueue[] ws = workQueues; int e, i; WorkQueue w; Thread p;
+ if ((e = (int)c) > 0) { // at least one waiting
+ if (ws != null && (i = e & SMASK) < ws.length &&
+ (w = ws[i]) != null && w.eventCount == (e | INT_SIGN)) {
+ long nc = (((long)(w.nextWait & E_MASK)) |
+ ((long)(u + UAC_UNIT) << 32));
+ if (U.compareAndSwapLong(this, CTL, c, nc)) {
+ w.eventCount = (e + E_SEQ) & E_MASK;
+ if ((p = w.parker) != null)
+ U.unpark(p); // activate and release
+ break;
+ }
+ }
+ else
+ break;
+ }
+ else if (e == 0 && (u & SHORT_SIGN) != 0) { // too few total
+ long nc = (long)(((u + UTC_UNIT) & UTC_MASK) |
+ ((u + UAC_UNIT) & UAC_MASK)) << 32;
+ if (U.compareAndSwapLong(this, CTL, c, nc)) {
+ addWorker();
+ break;
+ }
+ }
+ else
+ break;
+ }
+ }
+
+ /**
+ * Tries to decrement active count (sometimes implicitly) and
+ * possibly release or create a compensating worker in preparation
+ * for blocking. Fails on contention or termination.
+ *
+ * @return true if the caller can block, else should recheck and retry
+ */
+ final boolean tryCompensate() {
+ WorkQueue w; Thread p;
+ int pc = parallelism, e, u, ac, tc, i;
+ long c = ctl;
+ WorkQueue[] ws = workQueues;
+ if ((e = (int)c) >= 0) {
+ if ((ac = ((u = (int)(c >>> 32)) >> UAC_SHIFT)) <= 0 &&
+ e != 0 && ws != null && (i = e & SMASK) < ws.length &&
+ (w = ws[i]) != null) {
+ long nc = (long)(w.nextWait & E_MASK) | (c & (AC_MASK|TC_MASK));
+ if (w.eventCount == (e | INT_SIGN) &&
+ U.compareAndSwapLong(this, CTL, c, nc)) {
+ w.eventCount = (e + E_SEQ) & E_MASK;
+ if ((p = w.parker) != null)
+ U.unpark(p);
+ return true; // release an idle worker
+ }
+ }
+ else if ((tc = (short)(u >>> UTC_SHIFT)) >= 0 && ac + pc > 1) {
+ long nc = ((c - AC_UNIT) & AC_MASK) | (c & ~AC_MASK);
+ if (U.compareAndSwapLong(this, CTL, c, nc))
+ return true; // no compensation needed
+ }
+ else if (tc + pc < POOL_MAX) {
+ long nc = ((c + TC_UNIT) & TC_MASK) | (c & ~TC_MASK);
+ if (U.compareAndSwapLong(this, CTL, c, nc)) {
+ addWorker();
+ return true; // create replacement
+ }
+ }
+ }
+ return false;
+ }
+
+ // Scanning for tasks
+
+ /**
+ * Scans for and, if found, returns one task, else possibly
+ * inactivates the worker. This method operates on single reads of
+ * volatile state and is designed to be re-invoked continuously in
+ * part because it returns upon detecting inconsistencies,
+ * contention, or state changes that indicate possible success on
+ * re-invocation.
+ *
+ * The scan searches for tasks across queues, randomly selecting
+ * the first #queues probes, favoring steals over submissions
+ * (by exploiting even/odd indexing), and then performing a
+ * circular sweep of all queues. The scan terminates upon either
+ * finding a non-empty queue, or completing a full sweep. If the
+ * worker is not inactivated, it takes and returns a task from
+ * this queue. On failure to find a task, we take one of the
+ * following actions, after which the caller will retry calling
+ * this method unless terminated.
+ *
+ * * If pool is terminating, terminate the worker.
+ *
+ * * If not a complete sweep, try to release a waiting worker. If
+ * the scan terminated because the worker is inactivated, then the
+ * released worker will often be the calling worker, and it can
+ * succeed obtaining a task on the next call. Or maybe it is
+ * another worker, but with same net effect. Releasing in other
+ * cases as well ensures that we have enough workers running.
+ *
+ * * If the caller has run a task since the last empty scan,
+ * return (to allow rescan) if other workers are not also yet
+ * enqueued. Field WorkQueue.rescans counts down on each scan to
+ * ensure eventual inactivation and blocking.
+ *
+ * * If not already enqueued, try to inactivate and enqueue the
+ * worker on wait queue.
+ *
+ * * If already enqueued and none of the above apply, either park
+ * awaiting signal, or if this is the most recent waiter and pool
+ * is quiescent, relay to idleAwaitWork to check for termination
+ * and possibly shrink pool.
+ *
+ * @param w the worker (via its WorkQueue)
+ * @return a task or null of none found
+ */
+ private final ForkJoinTask scan(WorkQueue w) {
+ boolean swept = false; // true after full empty scan
+ WorkQueue[] ws; // volatile read order matters
+ int r = w.seed, ec = w.eventCount; // ec is negative if inactive
+ int rs = runState, m = rs & SMASK;
+ if ((ws = workQueues) != null && ws.length > m) { // consistency check
+ for (int k = 0, j = -1 - m; ; ++j) {
+ WorkQueue q; int b;
+ if (j < 0) { // random probes while j negative
+ r ^= r << 13; r ^= r >>> 17; k = (r ^= r << 5) | (j & 1);
+ } // worker (not submit) for odd j
+ else // cyclic scan when j >= 0
+ k += 7; // step 7 reduces array packing bias
+ if ((q = ws[k & m]) != null && (b = q.base) - q.top < 0) {
+ ForkJoinTask t = (ec >= 0) ? q.pollAt(b) : null;
+ w.seed = r; // save seed for next scan
+ if (t != null)
+ return t;
+ break;
+ }
+ else if (j - m > m) {
+ if (rs == runState) // staleness check
+ swept = true;
+ break;
+ }
+ }
+
+ // Decode ctl on empty scan
+ long c = ctl; int e = (int)c, a = (int)(c >> AC_SHIFT), nr, ns;
+ if (e < 0) // pool is terminating
+ w.runState = -1;
+ else if (!swept) { // try to release a waiter
+ WorkQueue v; Thread p;
+ if (e > 0 && a < 0 && (v = ws[e & m]) != null &&
+ v.eventCount == (e | INT_SIGN)) {
+ long nc = ((long)(v.nextWait & E_MASK) |
+ ((c + AC_UNIT) & (AC_MASK|TC_MASK)));
+ if (U.compareAndSwapLong(this, CTL, c, nc)) {
+ v.eventCount = (e + E_SEQ) & E_MASK;
+ if ((p = v.parker) != null)
+ U.unpark(p);
+ }
+ }
+ }
+ else if ((nr = w.rescans) > 0) { // continue rescanning
+ int ac = a + parallelism;
+ if (((w.rescans = (ac < nr) ? ac : nr - 1) & 3) == 0 &&
+ w.eventCount == ec)
+ Thread.yield(); // occasionally yield
+ }
+ else if (ec >= 0) { // try to enqueue
+ long nc = (long)ec | ((c - AC_UNIT) & (AC_MASK|TC_MASK));
+ w.nextWait = e;
+ w.eventCount = ec | INT_SIGN;// mark as inactive
+ if (!U.compareAndSwapLong(this, CTL, c, nc))
+ w.eventCount = ec; // unmark on CAS failure
+ else if ((ns = w.nsteals) != 0) {
+ w.nsteals = 0; // set rescans if ran task
+ w.rescans = a + parallelism;
+ w.totalSteals += ns;
+ }
+ }
+ else { // already queued
+ if (parallelism == -a)
+ idleAwaitWork(w); // quiescent
+ if (w.eventCount == ec) {
+ Thread.interrupted(); // clear status
+ ForkJoinWorkerThread wt = w.owner;
+ U.putObject(wt, PARKBLOCKER, this);
+ w.parker = wt; // emulate LockSupport.park
+ if (w.eventCount == ec) // recheck
+ U.park(false, 0L); // block
+ w.parker = null;
+ U.putObject(wt, PARKBLOCKER, null);
+ }
+ }
+ }
+ return null;
+ }
+
+ /**
+ * If inactivating worker w has caused pool to become quiescent,
+ * checks for pool termination, and, so long as this is not the
+ * only worker, waits for event for up to SHRINK_RATE nanosecs.
+ * On timeout, if ctl has not changed, terminates the worker,
+ * which will in turn wake up another worker to possibly repeat
+ * this process.
+ *
+ * @param w the calling worker
+ */
+ private void idleAwaitWork(WorkQueue w) {
+ long c; int nw, ec;
+ if (!tryTerminate(false, false) &&
+ (int)((c = ctl) >> AC_SHIFT) + parallelism == 0 &&
+ (ec = w.eventCount) == ((int)c | INT_SIGN) &&
+ (nw = w.nextWait) != 0) {
+ long nc = ((long)(nw & E_MASK) | // ctl to restore on timeout
+ ((c + AC_UNIT) & AC_MASK) | (c & TC_MASK));
+ ForkJoinWorkerThread wt = w.owner;
+ while (ctl == c) {
+ long startTime = System.nanoTime();
+ Thread.interrupted(); // timed variant of version in scan()
+ U.putObject(wt, PARKBLOCKER, this);
+ w.parker = wt;
+ if (ctl == c)
+ U.park(false, SHRINK_RATE);
+ w.parker = null;
+ U.putObject(wt, PARKBLOCKER, null);
+ if (ctl != c)
+ break;
+ if (System.nanoTime() - startTime >= SHRINK_TIMEOUT &&
+ U.compareAndSwapLong(this, CTL, c, nc)) {
+ w.eventCount = (ec + E_SEQ) | E_MASK;
+ w.runState = -1; // shrink
+ break;
+ }
+ }
+ }
+ }
+
+ /**
+ * Tries to locate and execute tasks for a stealer of the given
+ * task, or in turn one of its stealers, Traces currentSteal ->
+ * currentJoin links looking for a thread working on a descendant
+ * of the given task and with a non-empty queue to steal back and
+ * execute tasks from. The first call to this method upon a
+ * waiting join will often entail scanning/search, (which is OK
+ * because the joiner has nothing better to do), but this method
+ * leaves hints in workers to speed up subsequent calls. The
+ * implementation is very branchy to cope with potential
+ * inconsistencies or loops encountering chains that are stale,
+ * unknown, or of length greater than MAX_HELP_DEPTH links. All
+ * of these cases are dealt with by just retrying by caller.
+ *
+ * @param joiner the joining worker
+ * @param task the task to join
+ * @return true if found or ran a task (and so is immediately retryable)
+ */
+ final boolean tryHelpStealer(WorkQueue joiner, ForkJoinTask task) {
+ ForkJoinTask subtask; // current target
+ boolean progress = false;
+ int depth = 0; // current chain depth
+ int m = runState & SMASK;
+ WorkQueue[] ws = workQueues;
+
+ if (ws != null && ws.length > m && (subtask = task).status >= 0) {
+ outer:for (WorkQueue j = joiner;;) {
+ // Try to find the stealer of subtask, by first using hint
+ WorkQueue stealer = null;
+ WorkQueue v = ws[j.stealHint & m];
+ if (v != null && v.currentSteal == subtask)
+ stealer = v;
+ else {
+ for (int i = 1; i <= m; i += 2) {
+ if ((v = ws[i]) != null && v.currentSteal == subtask) {
+ stealer = v;
+ j.stealHint = i; // save hint
+ break;
+ }
+ }
+ if (stealer == null)
+ break;
+ }
+
+ for (WorkQueue q = stealer;;) { // Try to help stealer
+ ForkJoinTask t; int b;
+ if (task.status < 0)
+ break outer;
+ if ((b = q.base) - q.top < 0) {
+ progress = true;
+ if (subtask.status < 0)
+ break outer; // stale
+ if ((t = q.pollAt(b)) != null) {
+ stealer.stealHint = joiner.poolIndex;
+ joiner.runSubtask(t);
+ }
+ }
+ else { // empty - try to descend to find stealer's stealer
+ ForkJoinTask next = stealer.currentJoin;
+ if (++depth == MAX_HELP_DEPTH || subtask.status < 0 ||
+ next == null || next == subtask)
+ break outer; // max depth, stale, dead-end, cyclic
+ subtask = next;
+ j = stealer;
+ break;
+ }
+ }
+ }
+ }
+ return progress;
+ }
+
+ /**
+ * If task is at base of some steal queue, steals and executes it.
+ *
+ * @param joiner the joining worker
+ * @param task the task
+ */
+ final void tryPollForAndExec(WorkQueue joiner, ForkJoinTask task) {
+ WorkQueue[] ws;
+ int m = runState & SMASK;
+ if ((ws = workQueues) != null && ws.length > m) {
+ for (int j = 1; j <= m && task.status >= 0; j += 2) {
+ WorkQueue q = ws[j];
+ if (q != null && q.pollFor(task)) {
+ joiner.runSubtask(task);
+ break;
+ }
+ }
+ }
+ }
+
+ /**
+ * Returns a non-empty steal queue, if one is found during a random,
+ * then cyclic scan, else null. This method must be retried by
+ * caller if, by the time it tries to use the queue, it is empty.
+ */
+ private WorkQueue findNonEmptyStealQueue(WorkQueue w) {
+ int r = w.seed; // Same idea as scan(), but ignoring submissions
+ for (WorkQueue[] ws;;) {
+ int m = runState & SMASK;
+ if ((ws = workQueues) == null)
+ return null;
+ if (ws.length > m) {
+ WorkQueue q;
+ for (int k = 0, j = -1 - m;; ++j) {
+ if (j < 0) {
+ r ^= r << 13; r ^= r >>> 17; k = r ^= r << 5;
+ }
+ else
+ k += 7;
+ if ((q = ws[(k | 1) & m]) != null && q.base - q.top < 0) {
+ w.seed = r;
+ return q;
+ }
+ else if (j - m > m)
+ return null;
+ }
+ }
+ }
+ }
+
+ /**
+ * Runs tasks until {@code isQuiescent()}. We piggyback on
+ * active count ctl maintenance, but rather than blocking
+ * when tasks cannot be found, we rescan until all others cannot
+ * find tasks either.
+ */
+ final void helpQuiescePool(WorkQueue w) {
+ for (boolean active = true;;) {
+ w.runLocalTasks(); // exhaust local queue
+ WorkQueue q = findNonEmptyStealQueue(w);
+ if (q != null) {
+ ForkJoinTask t;
+ if (!active) { // re-establish active count
+ long c;
+ active = true;
+ do {} while (!U.compareAndSwapLong
+ (this, CTL, c = ctl, c + AC_UNIT));
+ }
+ if ((t = q.poll()) != null)
+ w.runSubtask(t);
+ }
+ else {
+ long c;
+ if (active) { // decrement active count without queuing
+ active = false;
+ do {} while (!U.compareAndSwapLong
+ (this, CTL, c = ctl, c -= AC_UNIT));
+ }
+ else
+ c = ctl; // re-increment on exit
+ if ((int)(c >> AC_SHIFT) + parallelism == 0) {
+ do {} while (!U.compareAndSwapLong
+ (this, CTL, c = ctl, c + AC_UNIT));
+ break;
+ }
+ }
+ }
+ }
+
+ /**
+ * Gets and removes a local or stolen task for the given worker.
+ *
+ * @return a task, if available
+ */
+ final ForkJoinTask nextTaskFor(WorkQueue w) {
+ for (ForkJoinTask t;;) {
+ WorkQueue q;
+ if ((t = w.nextLocalTask()) != null)
+ return t;
+ if ((q = findNonEmptyStealQueue(w)) == null)
+ return null;
+ if ((t = q.poll()) != null)
+ return t;
+ }
+ }
+
+ /**
+ * Returns the approximate (non-atomic) number of idle threads per
+ * active thread to offset steal queue size for method
+ * ForkJoinTask.getSurplusQueuedTaskCount().
+ */
+ final int idlePerActive() {
+ // Approximate at powers of two for small values, saturate past 4
+ int p = parallelism;
+ int a = p + (int)(ctl >> AC_SHIFT);
+ return (a > (p >>>= 1) ? 0 :
+ a > (p >>>= 1) ? 1 :
+ a > (p >>>= 1) ? 2 :
+ a > (p >>>= 1) ? 4 :
+ 8);
+ }
+
+ // Termination
+
+ /**
+ * Possibly initiates and/or completes termination. The caller
+ * triggering termination runs three passes through workQueues:
+ * (0) Setting termination status, followed by wakeups of queued
+ * workers; (1) cancelling all tasks; (2) interrupting lagging
+ * threads (likely in external tasks, but possibly also blocked in
+ * joins). Each pass repeats previous steps because of potential
+ * lagging thread creation.
+ *
+ * @param now if true, unconditionally terminate, else only
+ * if no work and no active workers
+ * @param enable if true, enable shutdown when next possible
+ * @return true if now terminating or terminated
+ */
+ private boolean tryTerminate(boolean now, boolean enable) {
+ Mutex lock = this.lock;
+ for (long c;;) {
+ if (((c = ctl) & STOP_BIT) != 0) { // already terminating
+ if ((short)(c >>> TC_SHIFT) == -parallelism) {
+ lock.lock(); // don't need try/finally
+ termination.signalAll(); // signal when 0 workers
+ lock.unlock();
+ }
+ return true;
+ }
+ if (runState >= 0) { // not yet enabled
+ if (!enable)
+ return false;
+ lock.lock();
+ runState |= SHUTDOWN;
+ lock.unlock();
+ }
+ if (!now) { // check if idle & no tasks
+ if ((int)(c >> AC_SHIFT) != -parallelism ||
+ hasQueuedSubmissions())
+ return false;
+ // Check for unqueued inactive workers. One pass suffices.
+ WorkQueue[] ws = workQueues; WorkQueue w;
+ if (ws != null) {
+ for (int i = 1; i < ws.length; i += 2) {
+ if ((w = ws[i]) != null && w.eventCount >= 0)
+ return false;
+ }
+ }
+ }
+ if (U.compareAndSwapLong(this, CTL, c, c | STOP_BIT)) {
+ for (int pass = 0; pass < 3; ++pass) {
+ WorkQueue[] ws = workQueues;
+ if (ws != null) {
+ WorkQueue w;
+ int n = ws.length;
+ for (int i = 0; i < n; ++i) {
+ if ((w = ws[i]) != null) {
+ w.runState = -1;
+ if (pass > 0) {
+ w.cancelAll();
+ if (pass > 1)
+ w.interruptOwner();
+ }
+ }
+ }
+ // Wake up workers parked on event queue
+ int i, e; long cc; Thread p;
+ while ((e = (int)(cc = ctl) & E_MASK) != 0 &&
+ (i = e & SMASK) < n &&
+ (w = ws[i]) != null) {
+ long nc = ((long)(w.nextWait & E_MASK) |
+ ((cc + AC_UNIT) & AC_MASK) |
+ (cc & (TC_MASK|STOP_BIT)));
+ if (w.eventCount == (e | INT_SIGN) &&
+ U.compareAndSwapLong(this, CTL, cc, nc)) {
+ w.eventCount = (e + E_SEQ) & E_MASK;
+ w.runState = -1;
+ if ((p = w.parker) != null)
+ U.unpark(p);
+ }
+ }
+ }
+ }
+ }
+ }
+ }
+
+ // Exported methods
+
+ // Constructors
+
+ /**
+ * Creates a {@code ForkJoinPool} with parallelism equal to {@link
+ * java.lang.Runtime#availableProcessors}, using the {@linkplain
+ * #defaultForkJoinWorkerThreadFactory default thread factory},
+ * no UncaughtExceptionHandler, and non-async LIFO processing mode.
+ *
+ * @throws SecurityException if a security manager exists and
+ * the caller is not permitted to modify threads
+ * because it does not hold {@link
+ * java.lang.RuntimePermission}{@code ("modifyThread")}
+ */
+ public ForkJoinPool() {
+ this(Runtime.getRuntime().availableProcessors(),
+ defaultForkJoinWorkerThreadFactory, null, false);
+ }
+
+ /**
+ * Creates a {@code ForkJoinPool} with the indicated parallelism
+ * level, the {@linkplain
+ * #defaultForkJoinWorkerThreadFactory default thread factory},
+ * no UncaughtExceptionHandler, and non-async LIFO processing mode.
+ *
+ * @param parallelism the parallelism level
+ * @throws IllegalArgumentException if parallelism less than or
+ * equal to zero, or greater than implementation limit
+ * @throws SecurityException if a security manager exists and
+ * the caller is not permitted to modify threads
+ * because it does not hold {@link
+ * java.lang.RuntimePermission}{@code ("modifyThread")}
+ */
+ public ForkJoinPool(int parallelism) {
+ this(parallelism, defaultForkJoinWorkerThreadFactory, null, false);
+ }
+
+ /**
+ * Creates a {@code ForkJoinPool} with the given parameters.
+ *
+ * @param parallelism the parallelism level. For default value,
+ * use {@link java.lang.Runtime#availableProcessors}.
+ * @param factory the factory for creating new threads. For default value,
+ * use {@link #defaultForkJoinWorkerThreadFactory}.
+ * @param handler the handler for internal worker threads that
+ * terminate due to unrecoverable errors encountered while executing
+ * tasks. For default value, use {@code null}.
+ * @param asyncMode if true,
+ * establishes local first-in-first-out scheduling mode for forked
+ * tasks that are never joined. This mode may be more appropriate
+ * than default locally stack-based mode in applications in which
+ * worker threads only process event-style asynchronous tasks.
+ * For default value, use {@code false}.
+ * @throws IllegalArgumentException if parallelism less than or
+ * equal to zero, or greater than implementation limit
+ * @throws NullPointerException if the factory is null
+ * @throws SecurityException if a security manager exists and
+ * the caller is not permitted to modify threads
+ * because it does not hold {@link
+ * java.lang.RuntimePermission}{@code ("modifyThread")}
+ */
+ public ForkJoinPool(int parallelism,
+ ForkJoinWorkerThreadFactory factory,
+ Thread.UncaughtExceptionHandler handler,
+ boolean asyncMode) {
+ checkPermission();
+ if (factory == null)
+ throw new NullPointerException();
+ if (parallelism <= 0 || parallelism > POOL_MAX)
+ throw new IllegalArgumentException();
+ this.parallelism = parallelism;
+ this.factory = factory;
+ this.ueh = handler;
+ this.localMode = asyncMode ? FIFO_QUEUE : LIFO_QUEUE;
+ this.growHints = 1;
+ long np = (long)(-parallelism); // offset ctl counts
+ this.ctl = ((np << AC_SHIFT) & AC_MASK) | ((np << TC_SHIFT) & TC_MASK);
+ // initialize workQueues array with room for 2*parallelism if possible
+ int n = parallelism << 1;
+ if (n >= POOL_MAX)
+ n = POOL_MAX;
+ else { // See Hackers Delight, sec 3.2, where n < (1 << 16)
+ n |= n >>> 1; n |= n >>> 2; n |= n >>> 4; n |= n >>> 8;
+ }
+ this.workQueues = new WorkQueue[(n + 1) << 1]; // #slots = 2 * #workers
+ this.termination = (this.lock = new Mutex()).newCondition();
+ this.stealCount = new AtomicLong();
+ this.nextWorkerNumber = new AtomicInteger();
+ StringBuilder sb = new StringBuilder("ForkJoinPool-");
+ sb.append(poolNumberGenerator.incrementAndGet());
+ sb.append("-worker-");
+ this.workerNamePrefix = sb.toString();
+ }
+
+ // Execution methods
+
+ /**
+ * Performs the given task, returning its result upon completion.
+ * If the computation encounters an unchecked Exception or Error,
+ * it is rethrown as the outcome of this invocation. Rethrown
+ * exceptions behave in the same way as regular exceptions, but,
+ * when possible, contain stack traces (as displayed for example
+ * using {@code ex.printStackTrace()}) of both the current thread
+ * as well as the thread actually encountering the exception;
+ * minimally only the latter.
+ *
+ * @param task the task
+ * @return the task's result
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
+ */
+ public T invoke(ForkJoinTask task) {
+ doSubmit(task);
+ return task.join();
+ }
+
+ /**
+ * Arranges for (asynchronous) execution of the given task.
+ *
+ * @param task the task
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
+ */
+ public void execute(ForkJoinTask task) {
+ doSubmit(task);
+ }
+
+ // AbstractExecutorService methods
+
+ /**
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
+ */
+ public void execute(Runnable task) {
+ if (task == null)
+ throw new NullPointerException();
+ ForkJoinTask job;
+ if (task instanceof ForkJoinTask) // avoid re-wrap
+ job = (ForkJoinTask) task;
+ else
+ job = ForkJoinTask.adapt(task, null);
+ doSubmit(job);
+ }
+
+ /**
+ * Submits a ForkJoinTask for execution.
+ *
+ * @param task the task to submit
+ * @return the task
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
+ */
+ public ForkJoinTask submit(ForkJoinTask task) {
+ doSubmit(task);
+ return task;
+ }
+
+ /**
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
+ */
+ public ForkJoinTask submit(Callable task) {
+ if (task == null)
+ throw new NullPointerException();
+ ForkJoinTask job = ForkJoinTask.adapt(task);
+ doSubmit(job);
+ return job;
+ }
+
+ /**
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
+ */
+ public ForkJoinTask submit(Runnable task, T result) {
+ if (task == null)
+ throw new NullPointerException();
+ ForkJoinTask job = ForkJoinTask.adapt(task, result);
+ doSubmit(job);
+ return job;
+ }
+
+ /**
+ * @throws NullPointerException if the task is null
+ * @throws RejectedExecutionException if the task cannot be
+ * scheduled for execution
+ */
+ public ForkJoinTask submit(Runnable task) {
+ if (task == null)
+ throw new NullPointerException();
+ ForkJoinTask job;
+ if (task instanceof ForkJoinTask) // avoid re-wrap
+ job = (ForkJoinTask) task;
+ else
+ job = ForkJoinTask.adapt(task, null);
+ doSubmit(job);
+ return job;
+ }
+
+ /**
+ * @throws NullPointerException {@inheritDoc}
+ * @throws RejectedExecutionException {@inheritDoc}
+ */
+ public List> invokeAll(Collection> tasks) {
+ // In previous versions of this class, this method constructed
+ // a task to run ForkJoinTask.invokeAll, but now external
+ // invocation of multiple tasks is at least as efficient.
+ List> fs = new ArrayList>(tasks.size());
+ // Workaround needed because method wasn't declared with
+ // wildcards in return type but should have been.
+ @SuppressWarnings({"unchecked", "rawtypes"})
+ List> futures = (List>) (List) fs;
+
+ boolean done = false;
+ try {
+ for (Callable t : tasks) {
+ ForkJoinTask f = ForkJoinTask.adapt(t);
+ doSubmit(f);
+ fs.add(f);
+ }
+ for (ForkJoinTask f : fs)
+ f.quietlyJoin();
+ done = true;
+ return futures;
+ } finally {
+ if (!done)
+ for (ForkJoinTask f : fs)
+ f.cancel(false);
+ }
+ }
+
+ /**
+ * Returns the factory used for constructing new workers.
+ *
+ * @return the factory used for constructing new workers
+ */
+ public ForkJoinWorkerThreadFactory getFactory() {
+ return factory;
+ }
+
+ /**
+ * Returns the handler for internal worker threads that terminate
+ * due to unrecoverable errors encountered while executing tasks.
+ *
+ * @return the handler, or {@code null} if none
+ */
+ public Thread.UncaughtExceptionHandler getUncaughtExceptionHandler() {
+ return ueh;
+ }
+
+ /**
+ * Returns the targeted parallelism level of this pool.
+ *
+ * @return the targeted parallelism level of this pool
+ */
+ public int getParallelism() {
+ return parallelism;
+ }
+
+ /**
+ * Returns the number of worker threads that have started but not
+ * yet terminated. The result returned by this method may differ
+ * from {@link #getParallelism} when threads are created to
+ * maintain parallelism when others are cooperatively blocked.
+ *
+ * @return the number of worker threads
+ */
+ public int getPoolSize() {
+ return parallelism + (short)(ctl >>> TC_SHIFT);
+ }
+
+ /**
+ * Returns {@code true} if this pool uses local first-in-first-out
+ * scheduling mode for forked tasks that are never joined.
+ *
+ * @return {@code true} if this pool uses async mode
+ */
+ public boolean getAsyncMode() {
+ return localMode != 0;
+ }
+
+ /**
+ * Returns an estimate of the number of worker threads that are
+ * not blocked waiting to join tasks or for other managed
+ * synchronization. This method may overestimate the
+ * number of running threads.
+ *
+ * @return the number of worker threads
+ */
+ public int getRunningThreadCount() {
+ int rc = 0;
+ WorkQueue[] ws; WorkQueue w;
+ if ((ws = workQueues) != null) {
+ for (int i = 1; i < ws.length; i += 2) {
+ if ((w = ws[i]) != null && w.isApparentlyUnblocked())
+ ++rc;
+ }
+ }
+ return rc;
+ }
+
+ /**
+ * Returns an estimate of the number of threads that are currently
+ * stealing or executing tasks. This method may overestimate the
+ * number of active threads.
+ *
+ * @return the number of active threads
+ */
+ public int getActiveThreadCount() {
+ int r = parallelism + (int)(ctl >> AC_SHIFT);
+ return (r <= 0) ? 0 : r; // suppress momentarily negative values
+ }
+
+ /**
+ * Returns {@code true} if all worker threads are currently idle.
+ * An idle worker is one that cannot obtain a task to execute
+ * because none are available to steal from other threads, and
+ * there are no pending submissions to the pool. This method is
+ * conservative; it might not return {@code true} immediately upon
+ * idleness of all threads, but will eventually become true if
+ * threads remain inactive.
+ *
+ * @return {@code true} if all threads are currently idle
+ */
+ public boolean isQuiescent() {
+ return (int)(ctl >> AC_SHIFT) + parallelism == 0;
+ }
+
+ /**
+ * Returns an estimate of the total number of tasks stolen from
+ * one thread's work queue by another. The reported value
+ * underestimates the actual total number of steals when the pool
+ * is not quiescent. This value may be useful for monitoring and
+ * tuning fork/join programs: in general, steal counts should be
+ * high enough to keep threads busy, but low enough to avoid
+ * overhead and contention across threads.
+ *
+ * @return the number of steals
+ */
+ public long getStealCount() {
+ long count = stealCount.get();
+ WorkQueue[] ws; WorkQueue w;
+ if ((ws = workQueues) != null) {
+ for (int i = 1; i < ws.length; i += 2) {
+ if ((w = ws[i]) != null)
+ count += w.totalSteals;
+ }
+ }
+ return count;
+ }
+
+ /**
+ * Returns an estimate of the total number of tasks currently held
+ * in queues by worker threads (but not including tasks submitted
+ * to the pool that have not begun executing). This value is only
+ * an approximation, obtained by iterating across all threads in
+ * the pool. This method may be useful for tuning task
+ * granularities.
+ *
+ * @return the number of queued tasks
+ */
+ public long getQueuedTaskCount() {
+ long count = 0;
+ WorkQueue[] ws; WorkQueue w;
+ if ((ws = workQueues) != null) {
+ for (int i = 1; i < ws.length; i += 2) {
+ if ((w = ws[i]) != null)
+ count += w.queueSize();
+ }
+ }
+ return count;
+ }
+
+ /**
+ * Returns an estimate of the number of tasks submitted to this
+ * pool that have not yet begun executing. This method may take
+ * time proportional to the number of submissions.
+ *
+ * @return the number of queued submissions
+ */
+ public int getQueuedSubmissionCount() {
+ int count = 0;
+ WorkQueue[] ws; WorkQueue w;
+ if ((ws = workQueues) != null) {
+ for (int i = 0; i < ws.length; i += 2) {
+ if ((w = ws[i]) != null)
+ count += w.queueSize();
+ }
+ }
+ return count;
+ }
+
+ /**
+ * Returns {@code true} if there are any tasks submitted to this
+ * pool that have not yet begun executing.
+ *
+ * @return {@code true} if there are any queued submissions
+ */
+ public boolean hasQueuedSubmissions() {
+ WorkQueue[] ws; WorkQueue w;
+ if ((ws = workQueues) != null) {
+ for (int i = 0; i < ws.length; i += 2) {
+ if ((w = ws[i]) != null && w.queueSize() != 0)
+ return true;
+ }
+ }
+ return false;
+ }
+
+ /**
+ * Removes and returns the next unexecuted submission if one is
+ * available. This method may be useful in extensions to this
+ * class that re-assign work in systems with multiple pools.
+ *
+ * @return the next submission, or {@code null} if none
+ */
+ protected ForkJoinTask pollSubmission() {
+ WorkQueue[] ws; WorkQueue w; ForkJoinTask t;
+ if ((ws = workQueues) != null) {
+ for (int i = 0; i < ws.length; i += 2) {
+ if ((w = ws[i]) != null && (t = w.poll()) != null)
+ return t;
+ }
+ }
+ return null;
+ }
+
+ /**
+ * Removes all available unexecuted submitted and forked tasks
+ * from scheduling queues and adds them to the given collection,
+ * without altering their execution status. These may include
+ * artificially generated or wrapped tasks. This method is
+ * designed to be invoked only when the pool is known to be
+ * quiescent. Invocations at other times may not remove all
+ * tasks. A failure encountered while attempting to add elements
+ * to collection {@code c} may result in elements being in
+ * neither, either or both collections when the associated
+ * exception is thrown. The behavior of this operation is
+ * undefined if the specified collection is modified while the
+ * operation is in progress.
+ *
+ * @param c the collection to transfer elements into
+ * @return the number of elements transferred
+ */
+ protected int drainTasksTo(Collection> c) {
+ int count = 0;
+ WorkQueue[] ws; WorkQueue w; ForkJoinTask t;
+ if ((ws = workQueues) != null) {
+ for (int i = 0; i < ws.length; ++i) {
+ if ((w = ws[i]) != null) {
+ while ((t = w.poll()) != null) {
+ c.add(t);
+ ++count;
+ }
+ }
+ }
+ }
+ return count;
+ }
+
+ /**
+ * Returns a string identifying this pool, as well as its state,
+ * including indications of run state, parallelism level, and
+ * worker and task counts.
+ *
+ * @return a string identifying this pool, as well as its state
+ */
+ public String toString() {
+ // Use a single pass through workQueues to collect counts
+ long qt = 0L, qs = 0L; int rc = 0;
+ long st = stealCount.get();
+ long c = ctl;
+ WorkQueue[] ws; WorkQueue w;
+ if ((ws = workQueues) != null) {
+ for (int i = 0; i < ws.length; ++i) {
+ if ((w = ws[i]) != null) {
+ int size = w.queueSize();
+ if ((i & 1) == 0)
+ qs += size;
+ else {
+ qt += size;
+ st += w.totalSteals;
+ if (w.isApparentlyUnblocked())
+ ++rc;
+ }
+ }
+ }
+ }
+ int pc = parallelism;
+ int tc = pc + (short)(c >>> TC_SHIFT);
+ int ac = pc + (int)(c >> AC_SHIFT);
+ if (ac < 0) // ignore transient negative
+ ac = 0;
+ String level;
+ if ((c & STOP_BIT) != 0)
+ level = (tc == 0) ? "Terminated" : "Terminating";
+ else
+ level = runState < 0 ? "Shutting down" : "Running";
+ return super.toString() +
+ "[" + level +
+ ", parallelism = " + pc +
+ ", size = " + tc +
+ ", active = " + ac +
+ ", running = " + rc +
+ ", steals = " + st +
+ ", tasks = " + qt +
+ ", submissions = " + qs +
+ "]";
+ }
+
+ /**
+ * Initiates an orderly shutdown in which previously submitted
+ * tasks are executed, but no new tasks will be accepted.
+ * Invocation has no additional effect if already shut down.
+ * Tasks that are in the process of being submitted concurrently
+ * during the course of this method may or may not be rejected.
+ *
+ * @throws SecurityException if a security manager exists and
+ * the caller is not permitted to modify threads
+ * because it does not hold {@link
+ * java.lang.RuntimePermission}{@code ("modifyThread")}
+ */
+ public void shutdown() {
+ checkPermission();
+ tryTerminate(false, true);
+ }
+
+ /**
+ * Attempts to cancel and/or stop all tasks, and reject all
+ * subsequently submitted tasks. Tasks that are in the process of
+ * being submitted or executed concurrently during the course of
+ * this method may or may not be rejected. This method cancels
+ * both existing and unexecuted tasks, in order to permit
+ * termination in the presence of task dependencies. So the method
+ * always returns an empty list (unlike the case for some other
+ * Executors).
+ *
+ * @return an empty list
+ * @throws SecurityException if a security manager exists and
+ * the caller is not permitted to modify threads
+ * because it does not hold {@link
+ * java.lang.RuntimePermission}{@code ("modifyThread")}
+ */
+ public List shutdownNow() {
+ checkPermission();
+ tryTerminate(true, true);
+ return Collections.emptyList();
+ }
+
+ /**
+ * Returns {@code true} if all tasks have completed following shut down.
+ *
+ * @return {@code true} if all tasks have completed following shut down
+ */
+ public boolean isTerminated() {
+ long c = ctl;
+ return ((c & STOP_BIT) != 0L &&
+ (short)(c >>> TC_SHIFT) == -parallelism);
+ }
+
+ /**
+ * Returns {@code true} if the process of termination has
+ * commenced but not yet completed. This method may be useful for
+ * debugging. A return of {@code true} reported a sufficient
+ * period after shutdown may indicate that submitted tasks have
+ * ignored or suppressed interruption, or are waiting for IO,
+ * causing this executor not to properly terminate. (See the
+ * advisory notes for class {@link ForkJoinTask} stating that
+ * tasks should not normally entail blocking operations. But if
+ * they do, they must abort them on interrupt.)
+ *
+ * @return {@code true} if terminating but not yet terminated
+ */
+ public boolean isTerminating() {
+ long c = ctl;
+ return ((c & STOP_BIT) != 0L &&
+ (short)(c >>> TC_SHIFT) != -parallelism);
+ }
+
+ /**
+ * Returns {@code true} if this pool has been shut down.
+ *
+ * @return {@code true} if this pool has been shut down
+ */
+ public boolean isShutdown() {
+ return runState < 0;
+ }
+
+ /**
+ * Blocks until all tasks have completed execution after a shutdown
+ * request, or the timeout occurs, or the current thread is
+ * interrupted, whichever happens first.
+ *
+ * @param timeout the maximum time to wait
+ * @param unit the time unit of the timeout argument
+ * @return {@code true} if this executor terminated and
+ * {@code false} if the timeout elapsed before termination
+ * @throws InterruptedException if interrupted while waiting
+ */
+ public boolean awaitTermination(long timeout, TimeUnit unit)
+ throws InterruptedException {
+ long nanos = unit.toNanos(timeout);
+ final Mutex lock = this.lock;
+ lock.lock();
+ try {
+ for (;;) {
+ if (isTerminated())
+ return true;
+ if (nanos <= 0)
+ return false;
+ nanos = termination.awaitNanos(nanos);
+ }
+ } finally {
+ lock.unlock();
+ }
+ }
+
+ /**
+ * Interface for extending managed parallelism for tasks running
+ * in {@link ForkJoinPool}s.
+ *
+ *
A {@code ManagedBlocker} provides two methods. Method
+ * {@code isReleasable} must return {@code true} if blocking is
+ * not necessary. Method {@code block} blocks the current thread
+ * if necessary (perhaps internally invoking {@code isReleasable}
+ * before actually blocking). These actions are performed by any
+ * thread invoking {@link ForkJoinPool#managedBlock}. The
+ * unusual methods in this API accommodate synchronizers that may,
+ * but don't usually, block for long periods. Similarly, they
+ * allow more efficient internal handling of cases in which
+ * additional workers may be, but usually are not, needed to
+ * ensure sufficient parallelism. Toward this end,
+ * implementations of method {@code isReleasable} must be amenable
+ * to repeated invocation.
+ *
+ *
For example, here is a ManagedBlocker based on a
+ * ReentrantLock:
+ *
+ */
+ public static interface ManagedBlocker {
+ /**
+ * Possibly blocks the current thread, for example waiting for
+ * a lock or condition.
+ *
+ * @return {@code true} if no additional blocking is necessary
+ * (i.e., if isReleasable would return true)
+ * @throws InterruptedException if interrupted while waiting
+ * (the method is not required to do so, but is allowed to)
+ */
+ boolean block() throws InterruptedException;
+
+ /**
+ * Returns {@code true} if blocking is unnecessary.
+ */
+ boolean isReleasable();
+ }
+
+ /**
+ * Blocks in accord with the given blocker. If the current thread
+ * is a {@link ForkJoinWorkerThread}, this method possibly
+ * arranges for a spare thread to be activated if necessary to
+ * ensure sufficient parallelism while the current thread is blocked.
+ *
+ *
If the caller is not a {@link ForkJoinTask}, this method is
+ * behaviorally equivalent to
+ *
{@code
+ * while (!blocker.isReleasable())
+ * if (blocker.block())
+ * return;
+ * }
+ *
+ * If the caller is a {@code ForkJoinTask}, then the pool may
+ * first be expanded to ensure parallelism, and later adjusted.
+ *
+ * @param blocker the blocker
+ * @throws InterruptedException if blocker.block did so
+ */
+ public static void managedBlock(ManagedBlocker blocker)
+ throws InterruptedException {
+ Thread t = Thread.currentThread();
+ ForkJoinPool p = ((t instanceof ForkJoinWorkerThread) ?
+ ((ForkJoinWorkerThread)t).pool : null);
+ while (!blocker.isReleasable()) {
+ if (p == null || p.tryCompensate()) {
+ try {
+ do {} while (!blocker.isReleasable() && !blocker.block());
+ } finally {
+ if (p != null)
+ p.incrementActiveCount();
+ }
+ break;
+ }
+ }
+ }
+
+ // AbstractExecutorService overrides. These rely on undocumented
+ // fact that ForkJoinTask.adapt returns ForkJoinTasks that also
+ // implement RunnableFuture.
+
+ protected RunnableFuture newTaskFor(Runnable runnable, T value) {
+ return (RunnableFuture) ForkJoinTask.adapt(runnable, value);
+ }
+
+ protected RunnableFuture newTaskFor(Callable callable) {
+ return (RunnableFuture) ForkJoinTask.adapt(callable);
+ }
+
+ // Unsafe mechanics
+ private static final sun.misc.Unsafe U;
+ private static final long CTL;
+ private static final long PARKBLOCKER;
+
+ static {
+ poolNumberGenerator = new AtomicInteger();
+ modifyThreadPermission = new RuntimePermission("modifyThread");
+ defaultForkJoinWorkerThreadFactory =
+ new DefaultForkJoinWorkerThreadFactory();
+ submitters = new ThreadSubmitter();
+ try {
+ U = getUnsafe();
+ Class k = ForkJoinPool.class;
+ CTL = U.objectFieldOffset
+ (k.getDeclaredField("ctl"));
+ Class tk = Thread.class;
+ PARKBLOCKER = U.objectFieldOffset
+ (tk.getDeclaredField("parkBlocker"));
+ } catch (Exception e) {
+ throw new Error(e);
+ }
+ }
+
+ /**
+ * Returns a sun.misc.Unsafe. Suitable for use in a 3rd party package.
+ * Replace with a simple call to Unsafe.getUnsafe when integrating
+ * into a jdk.
+ *
+ * @return a sun.misc.Unsafe
+ */
+ private static sun.misc.Unsafe getUnsafe() {
+ return Unsafe.instance;
+ }
+}
diff --git a/akka-actor/src/main/java/akka/jsr166y/ForkJoinTask.java b/akka-actor/src/main/java/akka/jsr166y/ForkJoinTask.java
new file mode 100644
index 0000000000..996d05e647
--- /dev/null
+++ b/akka-actor/src/main/java/akka/jsr166y/ForkJoinTask.java
@@ -0,0 +1,1543 @@
+/*
+ * Written by Doug Lea with assistance from members of JCP JSR-166
+ * Expert Group and released to the public domain, as explained at
+ * http://creativecommons.org/publicdomain/zero/1.0/
+ */
+
+package akka.jsr166y;
+import java.io.Serializable;
+import java.util.Collection;
+import java.util.List;
+import java.util.RandomAccess;
+import java.lang.ref.WeakReference;
+import java.lang.ref.ReferenceQueue;
+import java.util.concurrent.Callable;
+import java.util.concurrent.CancellationException;
+import java.util.concurrent.ExecutionException;
+import java.util.concurrent.Future;
+import java.util.concurrent.RejectedExecutionException;
+import java.util.concurrent.RunnableFuture;
+import java.util.concurrent.TimeUnit;
+import java.util.concurrent.TimeoutException;
+import java.util.concurrent.locks.ReentrantLock;
+import java.lang.reflect.Constructor;
+
+/**
+ * Abstract base class for tasks that run within a {@link ForkJoinPool}.
+ * A {@code ForkJoinTask} is a thread-like entity that is much
+ * lighter weight than a normal thread. Huge numbers of tasks and
+ * subtasks may be hosted by a small number of actual threads in a
+ * ForkJoinPool, at the price of some usage limitations.
+ *
+ *
A "main" {@code ForkJoinTask} begins execution when submitted
+ * to a {@link ForkJoinPool}. Once started, it will usually in turn
+ * start other subtasks. As indicated by the name of this class,
+ * many programs using {@code ForkJoinTask} employ only methods
+ * {@link #fork} and {@link #join}, or derivatives such as {@link
+ * #invokeAll(ForkJoinTask...) invokeAll}. However, this class also
+ * provides a number of other methods that can come into play in
+ * advanced usages, as well as extension mechanics that allow
+ * support of new forms of fork/join processing.
+ *
+ *
A {@code ForkJoinTask} is a lightweight form of {@link Future}.
+ * The efficiency of {@code ForkJoinTask}s stems from a set of
+ * restrictions (that are only partially statically enforceable)
+ * reflecting their main use as computational tasks calculating pure
+ * functions or operating on purely isolated objects. The primary
+ * coordination mechanisms are {@link #fork}, that arranges
+ * asynchronous execution, and {@link #join}, that doesn't proceed
+ * until the task's result has been computed. Computations should
+ * ideally avoid {@code synchronized} methods or blocks, and should
+ * minimize other blocking synchronization apart from joining other
+ * tasks or using synchronizers such as Phasers that are advertised to
+ * cooperate with fork/join scheduling. Subdividable tasks should also
+ * not perform blocking IO, and should ideally access variables that
+ * are completely independent of those accessed by other running
+ * tasks. These guidelines are loosely enforced by not permitting
+ * checked exceptions such as {@code IOExceptions} to be
+ * thrown. However, computations may still encounter unchecked
+ * exceptions, that are rethrown to callers attempting to join
+ * them. These exceptions may additionally include {@link
+ * RejectedExecutionException} stemming from internal resource
+ * exhaustion, such as failure to allocate internal task
+ * queues. Rethrown exceptions behave in the same way as regular
+ * exceptions, but, when possible, contain stack traces (as displayed
+ * for example using {@code ex.printStackTrace()}) of both the thread
+ * that initiated the computation as well as the thread actually
+ * encountering the exception; minimally only the latter.
+ *
+ *
It is possible to define and use ForkJoinTasks that may block,
+ * but doing do requires three further considerations: (1) Completion
+ * of few if any other tasks should be dependent on a task
+ * that blocks on external synchronization or IO. Event-style async
+ * tasks that are never joined often fall into this category. (2) To
+ * minimize resource impact, tasks should be small; ideally performing
+ * only the (possibly) blocking action. (3) Unless the {@link
+ * ForkJoinPool.ManagedBlocker} API is used, or the number of possibly
+ * blocked tasks is known to be less than the pool's {@link
+ * ForkJoinPool#getParallelism} level, the pool cannot guarantee that
+ * enough threads will be available to ensure progress or good
+ * performance.
+ *
+ *
The primary method for awaiting completion and extracting
+ * results of a task is {@link #join}, but there are several variants:
+ * The {@link Future#get} methods support interruptible and/or timed
+ * waits for completion and report results using {@code Future}
+ * conventions. Method {@link #invoke} is semantically
+ * equivalent to {@code fork(); join()} but always attempts to begin
+ * execution in the current thread. The "quiet" forms of
+ * these methods do not extract results or report exceptions. These
+ * may be useful when a set of tasks are being executed, and you need
+ * to delay processing of results or exceptions until all complete.
+ * Method {@code invokeAll} (available in multiple versions)
+ * performs the most common form of parallel invocation: forking a set
+ * of tasks and joining them all.
+ *
+ *
In the most typical usages, a fork-join pair act like a call
+ * (fork) and return (join) from a parallel recursive function. As is
+ * the case with other forms of recursive calls, returns (joins)
+ * should be performed innermost-first. For example, {@code a.fork();
+ * b.fork(); b.join(); a.join();} is likely to be substantially more
+ * efficient than joining {@code a} before {@code b}.
+ *
+ *
The execution status of tasks may be queried at several levels
+ * of detail: {@link #isDone} is true if a task completed in any way
+ * (including the case where a task was cancelled without executing);
+ * {@link #isCompletedNormally} is true if a task completed without
+ * cancellation or encountering an exception; {@link #isCancelled} is
+ * true if the task was cancelled (in which case {@link #getException}
+ * returns a {@link java.util.concurrent.CancellationException}); and
+ * {@link #isCompletedAbnormally} is true if a task was either
+ * cancelled or encountered an exception, in which case {@link
+ * #getException} will return either the encountered exception or
+ * {@link java.util.concurrent.CancellationException}.
+ *
+ *
The ForkJoinTask class is not usually directly subclassed.
+ * Instead, you subclass one of the abstract classes that support a
+ * particular style of fork/join processing, typically {@link
+ * RecursiveAction} for computations that do not return results, or
+ * {@link RecursiveTask} for those that do. Normally, a concrete
+ * ForkJoinTask subclass declares fields comprising its parameters,
+ * established in a constructor, and then defines a {@code compute}
+ * method that somehow uses the control methods supplied by this base
+ * class. While these methods have {@code public} access (to allow
+ * instances of different task subclasses to call each other's
+ * methods), some of them may only be called from within other
+ * ForkJoinTasks (as may be determined using method {@link
+ * #inForkJoinPool}). Attempts to invoke them in other contexts
+ * result in exceptions or errors, possibly including
+ * {@code ClassCastException}.
+ *
+ *
Method {@link #join} and its variants are appropriate for use
+ * only when completion dependencies are acyclic; that is, the
+ * parallel computation can be described as a directed acyclic graph
+ * (DAG). Otherwise, executions may encounter a form of deadlock as
+ * tasks cyclically wait for each other. However, this framework
+ * supports other methods and techniques (for example the use of
+ * {@link Phaser}, {@link #helpQuiesce}, and {@link #complete}) that
+ * may be of use in constructing custom subclasses for problems that
+ * are not statically structured as DAGs. To support such usages a
+ * ForkJoinTask may be atomically marked using {@link
+ * #markForkJoinTask} and checked for marking using {@link
+ * #isMarkedForkJoinTask}. The ForkJoinTask implementation does not
+ * use these {@code protected} methods or marks for any purpose, but
+ * they may be of use in the construction of specialized subclasses.
+ * For example, parallel graph traversals can use the supplied methods
+ * to avoid revisiting nodes/tasks that have already been processed.
+ * Also, completion based designs can use them to record that one
+ * subtask has completed. (Method names for marking are bulky in part
+ * to encourage definition of methods that reflect their usage
+ * patterns.)
+ *
+ *
Most base support methods are {@code final}, to prevent
+ * overriding of implementations that are intrinsically tied to the
+ * underlying lightweight task scheduling framework. Developers
+ * creating new basic styles of fork/join processing should minimally
+ * implement {@code protected} methods {@link #exec}, {@link
+ * #setRawResult}, and {@link #getRawResult}, while also introducing
+ * an abstract computational method that can be implemented in its
+ * subclasses, possibly relying on other {@code protected} methods
+ * provided by this class.
+ *
+ *
ForkJoinTasks should perform relatively small amounts of
+ * computation. Large tasks should be split into smaller subtasks,
+ * usually via recursive decomposition. As a very rough rule of thumb,
+ * a task should perform more than 100 and less than 10000 basic
+ * computational steps, and should avoid indefinite looping. If tasks
+ * are too big, then parallelism cannot improve throughput. If too
+ * small, then memory and internal task maintenance overhead may
+ * overwhelm processing.
+ *
+ *
This class provides {@code adapt} methods for {@link Runnable}
+ * and {@link Callable}, that may be of use when mixing execution of
+ * {@code ForkJoinTasks} with other kinds of tasks. When all tasks are
+ * of this form, consider using a pool constructed in asyncMode.
+ *
+ *
ForkJoinTasks are {@code Serializable}, which enables them to be
+ * used in extensions such as remote execution frameworks. It is
+ * sensible to serialize tasks only before or after, but not during,
+ * execution. Serialization is not relied on during execution itself.
+ *
+ * @since 1.7
+ * @author Doug Lea
+ */
+public abstract class ForkJoinTask implements Future, Serializable {
+
+ /*
+ * See the internal documentation of class ForkJoinPool for a
+ * general implementation overview. ForkJoinTasks are mainly
+ * responsible for maintaining their "status" field amidst relays
+ * to methods in ForkJoinWorkerThread and ForkJoinPool.
+ *
+ * The methods of this class are more-or-less layered into
+ * (1) basic status maintenance
+ * (2) execution and awaiting completion
+ * (3) user-level methods that additionally report results.
+ * This is sometimes hard to see because this file orders exported
+ * methods in a way that flows well in javadocs.
+ */
+
+ /**
+ * The number of times to try to help join a task without any
+ * apparent progress before giving up and blocking. The value is
+ * arbitrary but should be large enough to cope with transient
+ * stalls (due to GC etc) that can cause helping methods not to be
+ * able to proceed because other workers have not progressed to
+ * the point where subtasks can be found or taken.
+ */
+ private static final int HELP_RETRIES = 32;
+
+ /*
+ * The status field holds run control status bits packed into a
+ * single int to minimize footprint and to ensure atomicity (via
+ * CAS). Status is initially zero, and takes on nonnegative
+ * values until completed, upon which status holds value
+ * NORMAL, CANCELLED, or EXCEPTIONAL. Tasks undergoing blocking
+ * waits by other threads have the SIGNAL bit set. Completion of
+ * a stolen task with SIGNAL set awakens any waiters via
+ * notifyAll. Even though suboptimal for some purposes, we use
+ * basic builtin wait/notify to take advantage of "monitor
+ * inflation" in JVMs that we would otherwise need to emulate to
+ * avoid adding further per-task bookkeeping overhead. We want
+ * these monitors to be "fat", i.e., not use biasing or thin-lock
+ * techniques, so use some odd coding idioms that tend to avoid
+ * them.
+ */
+
+ /** The run status of this task */
+ volatile int status; // accessed directly by pool and workers
+ static final int NORMAL = 0xfffffffc; // negative with low 2 bits 0
+ static final int CANCELLED = 0xfffffff8; // must be < NORMAL
+ static final int EXCEPTIONAL = 0xfffffff4; // must be < CANCELLED
+ static final int SIGNAL = 0x00000001;
+ static final int MARKED = 0x00000002;
+
+ /**
+ * Marks completion and wakes up threads waiting to join this
+ * task, also clearing signal request bits. A specialization for
+ * NORMAL completion is in method doExec.
+ *
+ * @param completion one of NORMAL, CANCELLED, EXCEPTIONAL
+ * @return completion status on exit
+ */
+ private int setCompletion(int completion) {
+ for (int s;;) {
+ if ((s = status) < 0)
+ return s;
+ if (U.compareAndSwapInt(this, STATUS, s, (s & ~SIGNAL)|completion)) {
+ if ((s & SIGNAL) != 0)
+ synchronized (this) { notifyAll(); }
+ return completion;
+ }
+ }
+ }
+
+ /**
+ * Primary execution method for stolen tasks. Unless done, calls
+ * exec and records status if completed, but doesn't wait for
+ * completion otherwise.
+ *
+ * @return status on exit from this method
+ */
+ final int doExec() {
+ int s; boolean completed;
+ if ((s = status) >= 0) {
+ try {
+ completed = exec();
+ } catch (Throwable rex) {
+ return setExceptionalCompletion(rex);
+ }
+ while ((s = status) >= 0 && completed) {
+ if (U.compareAndSwapInt(this, STATUS, s, (s & ~SIGNAL)|NORMAL)) {
+ if ((s & SIGNAL) != 0)
+ synchronized (this) { notifyAll(); }
+ return NORMAL;
+ }
+ }
+ }
+ return s;
+ }
+
+ /**
+ * Blocks a non-worker-thread until completion.
+ * @return status upon completion
+ */
+ private int externalAwaitDone() {
+ int s;
+ if ((s = status) >= 0) {
+ boolean interrupted = false;
+ synchronized (this) {
+ while ((s = status) >= 0) {
+ if (U.compareAndSwapInt(this, STATUS, s, s | SIGNAL)) {
+ try {
+ wait();
+ } catch (InterruptedException ie) {
+ interrupted = true;
+ }
+ }
+ }
+ }
+ if (interrupted)
+ Thread.currentThread().interrupt();
+ }
+ return s;
+ }
+
+ /**
+ * Blocks a non-worker-thread until completion or interruption or timeout.
+ */
+ private int externalInterruptibleAwaitDone(long millis)
+ throws InterruptedException {
+ int s;
+ if (Thread.interrupted())
+ throw new InterruptedException();
+ if ((s = status) >= 0) {
+ synchronized (this) {
+ while ((s = status) >= 0) {
+ if (U.compareAndSwapInt(this, STATUS, s, s | SIGNAL)) {
+ wait(millis);
+ if (millis > 0L)
+ break;
+ }
+ }
+ }
+ }
+ return s;
+ }
+
+
+ /**
+ * Implementation for join, get, quietlyJoin. Directly handles
+ * only cases of already-completed, external wait, and
+ * unfork+exec. Others are relayed to awaitJoin.
+ *
+ * @return status upon completion
+ */
+ private int doJoin() {
+ int s; Thread t; ForkJoinWorkerThread wt; ForkJoinPool.WorkQueue w;
+ if ((s = status) >= 0) {
+ if (!((t = Thread.currentThread()) instanceof ForkJoinWorkerThread))
+ s = externalAwaitDone();
+ else if (!(w = (wt = (ForkJoinWorkerThread)t).workQueue).
+ tryUnpush(this) || (s = doExec()) >= 0)
+ s = awaitJoin(w, wt.pool);
+ }
+ return s;
+ }
+
+ /**
+ * Helps and/or blocks until joined.
+ *
+ * @param w the joiner
+ * @param p the pool
+ * @return status upon completion
+ */
+ private int awaitJoin(ForkJoinPool.WorkQueue w, ForkJoinPool p) {
+ int s;
+ ForkJoinTask prevJoin = w.currentJoin;
+ w.currentJoin = this;
+ for (int k = HELP_RETRIES; (s = status) >= 0;) {
+ if ((w.queueSize() > 0) ?
+ w.tryRemoveAndExec(this) : // self-help
+ p.tryHelpStealer(w, this)) // help process tasks
+ k = HELP_RETRIES; // reset if made progress
+ else if ((s = status) < 0) // recheck
+ break;
+ else if (--k > 0) {
+ if ((k & 3) == 1)
+ Thread.yield(); // occasionally yield
+ }
+ else if (k == 0)
+ p.tryPollForAndExec(w, this); // uncommon self-help case
+ else if (p.tryCompensate()) { // true if can block
+ try {
+ int ss = status;
+ if (ss >= 0 && // assert need signal
+ U.compareAndSwapInt(this, STATUS, ss, ss | SIGNAL)) {
+ synchronized (this) {
+ if (status >= 0) // block
+ wait();
+ }
+ }
+ } catch (InterruptedException ignore) {
+ } finally {
+ p.incrementActiveCount(); // re-activate
+ }
+ }
+ }
+ w.currentJoin = prevJoin;
+ return s;
+ }
+
+ /**
+ * Implementation for invoke, quietlyInvoke.
+ *
+ * @return status upon completion
+ */
+ private int doInvoke() {
+ int s; Thread t;
+ if ((s = doExec()) >= 0) {
+ if (!((t = Thread.currentThread()) instanceof ForkJoinWorkerThread))
+ s = externalAwaitDone();
+ else {
+ ForkJoinWorkerThread wt = (ForkJoinWorkerThread)t;
+ s = awaitJoin(wt.workQueue, wt.pool);
+ }
+ }
+ return s;
+ }
+
+ // Exception table support
+
+ /**
+ * Table of exceptions thrown by tasks, to enable reporting by
+ * callers. Because exceptions are rare, we don't directly keep
+ * them with task objects, but instead use a weak ref table. Note
+ * that cancellation exceptions don't appear in the table, but are
+ * instead recorded as status values.
+ *
+ * Note: These statics are initialized below in static block.
+ */
+ private static final ExceptionNode[] exceptionTable;
+ private static final ReentrantLock exceptionTableLock;
+ private static final ReferenceQueue