Add baseline performance and allocation tests for scheduling tasks and executing (#2009)

### Motivation:

In issue https://github.com/apple/swift-nio/issues/1316, we see a large number of allocations to happen when scheduling tasks. This can definitely be optimized. This PR adds a number of baseline allocation and performance tests for both `scheduleTask` and `execute`. In the next PRs, I am going to try a few optimizations to reduce the number of allocations.

### Modifications:

Added baseline performance and allocation tests for `scheduleTask` and `execute`

IntegrationTests/tests_04_performance/test_01_resources/test_schedule_10000_tasks.swift

@@ -16,20 +16,22 @@ import Dispatch
 import NIOPosix
 
 func run(identifier: String) {
-    let group = MultiThreadedEventLoopGroup(numberOfThreads: 1)
-    let loop = group.next()
-    let dg = DispatchGroup()
-
     measure(identifier: identifier) {
-        loop.execute {
-            for _ in 0..<10_000 {
-                dg.enter()
-                loop.scheduleTask(in: .nanoseconds(0)) { dg.leave() }
-            }
-        }
-        dg.wait()
-        return 10_000
-    }
+        let group = MultiThreadedEventLoopGroup(numberOfThreads: 1)
+        let loop = group.next()
+        let counter = try! loop.submit { () -> Int in
+            var counter: Int = 0
 
-    try! group.syncShutdownGracefully()
+            for _ in 0..<10000 {
+                loop.scheduleTask(in: .hours(1)) {
+                    counter &+= 1
+                }
+            }
+
+            return counter
+        }.wait()
+
+        try! group.syncShutdownGracefully()
+        return counter
+    }
 }

IntegrationTests/tests_04_performance/test_01_resources/test_schedule_and_run_10000_tasks.swift

@@ -0,0 +1,42 @@
+//===----------------------------------------------------------------------===//
+//
+// This source file is part of the SwiftNIO open source project
+//
+// Copyright (c) 2021 Apple Inc. and the SwiftNIO project authors
+// Licensed under Apache License v2.0
+//
+// See LICENSE.txt for license information
+// See CONTRIBUTORS.txt for the list of SwiftNIO project authors
+//
+// SPDX-License-Identifier: Apache-2.0
+//
+//===----------------------------------------------------------------------===//
+
+import Dispatch
+import NIOPosix
+
+func run(identifier: String) {
+    let group = MultiThreadedEventLoopGroup(numberOfThreads: 1)
+    let loop = group.next()
+    let dg = DispatchGroup()
+
+    measure(identifier: identifier) {
+        var counter = 0
+
+        try! loop.submit {
+            for _ in 0..<10000 {
+                dg.enter()
+
+                loop.scheduleTask(in: .nanoseconds(0)) {
+                    counter &+= 1
+                    dg.leave()
+                }
+            }
+        }.wait()
+        dg.wait()
+
+        return counter
+    }
+
+    try! group.syncShutdownGracefully()
+}

Sources/NIOPerformanceTester/ExecuteBenchmark.swift

@@ -0,0 +1,60 @@
+//===----------------------------------------------------------------------===//
+//
+// This source file is part of the SwiftNIO open source project
+//
+// Copyright (c) 2021 Apple Inc. and the SwiftNIO project authors
+// Licensed under Apache License v2.0
+//
+// See LICENSE.txt for license information
+// See CONTRIBUTORS.txt for the list of SwiftNIO project authors
+//
+// SPDX-License-Identifier: Apache-2.0
+//
+//===----------------------------------------------------------------------===//
+
+import Foundation
+import NIOCore
+import NIOPosix
+
+final class ExecuteBenchmark: Benchmark {
+    private var group: MultiThreadedEventLoopGroup!
+    private var loop: EventLoop!
+    private var dg: DispatchGroup!
+    private var counter = 0
+
+    func setUp() throws {
+        group = MultiThreadedEventLoopGroup(numberOfThreads: 1)
+        loop = group.next()
+        dg = DispatchGroup()
+
+        // We are preheating the EL to avoid growing the `ScheduledTask` `PriorityQueue`
+        // during the actual test
+        try! self.loop.submit {
+            var counter: Int = 0
+            for _ in 0..<100000 {
+                self.loop.scheduleTask(in: .nanoseconds(0)) {
+                    counter &+= 1
+                }
+            }
+        }.wait()
+    }
+
+    func tearDown() { }
+
+    func run() -> Int {
+        try! self.loop.submit {
+            for _ in 0..<10000 {
+                self.dg.enter()
+
+                self.loop.execute {
+                    self.counter &+= 1
+                    self.dg.leave()
+                }
+            }
+        }.wait()
+        self.dg.wait()
+
+        return self.counter
+    }
+
+}

Sources/NIOPerformanceTester/SchedulingAndRunningBenchmark.swift

@@ -0,0 +1,60 @@
+//===----------------------------------------------------------------------===//
+//
+// This source file is part of the SwiftNIO open source project
+//
+// Copyright (c) 2021 Apple Inc. and the SwiftNIO project authors
+// Licensed under Apache License v2.0
+//
+// See LICENSE.txt for license information
+// See CONTRIBUTORS.txt for the list of SwiftNIO project authors
+//
+// SPDX-License-Identifier: Apache-2.0
+//
+//===----------------------------------------------------------------------===//
+
+import Foundation
+import NIOCore
+import NIOPosix
+
+final class SchedulingAndRunningBenchmark: Benchmark {
+    private var group: MultiThreadedEventLoopGroup!
+    private var loop: EventLoop!
+    private var dg: DispatchGroup!
+    private var counter = 0
+
+    func setUp() throws {
+        group = MultiThreadedEventLoopGroup(numberOfThreads: 1)
+        loop = group.next()
+        dg = DispatchGroup()
+
+        // We are preheating the EL to avoid growing the `ScheduledTask` `PriorityQueue`
+        // during the actual test
+        try! self.loop.submit {
+            var counter: Int = 0
+            for _ in 0..<100000 {
+                self.loop.scheduleTask(in: .nanoseconds(0)) {
+                    counter &+= 1
+                }
+            }
+        }.wait()
+    }
+
+    func tearDown() { }
+
+    func run() -> Int {
+        try! self.loop.submit {
+            for _ in 0..<10000 {
+                self.dg.enter()
+
+                self.loop.scheduleTask(in: .nanoseconds(0)) {
+                    self.counter &+= 1
+                    self.dg.leave()
+                }
+            }
+        }.wait()
+        self.dg.wait()
+
+        return counter
+    }
+
+}

Sources/NIOPerformanceTester/SchedulingBenchmark.swift

@@ -0,0 +1,56 @@
+//===----------------------------------------------------------------------===//
+//
+// This source file is part of the SwiftNIO open source project
+//
+// Copyright (c) 2021 Apple Inc. and the SwiftNIO project authors
+// Licensed under Apache License v2.0
+//
+// See LICENSE.txt for license information
+// See CONTRIBUTORS.txt for the list of SwiftNIO project authors
+//
+// SPDX-License-Identifier: Apache-2.0
+//
+//===----------------------------------------------------------------------===//
+
+import Foundation
+import NIOCore
+import NIOPosix
+
+final class SchedulingBenchmark: Benchmark {
+    private var group: MultiThreadedEventLoopGroup!
+    private var loop: EventLoop!
+
+    func setUp() throws {
+        group = MultiThreadedEventLoopGroup(numberOfThreads: 1)
+        loop = group.next()
+
+        // We are preheating the EL to avoid growing the `ScheduledTask` `PriorityQueue`
+        // during the actual test
+        try! self.loop.submit {
+            var counter: Int = 0
+            for _ in 0..<100000 {
+                self.loop.scheduleTask(in: .nanoseconds(0)) {
+                    counter &+= 1
+                }
+            }
+        }.wait()
+    }
+
+    func tearDown() { }
+
+    func run() -> Int {
+        let counter = try! self.loop.submit { () -> Int in
+            var counter: Int = 0
+            for _ in 0..<10000 {
+                self.loop.scheduleTask(in: .hours(1)) {
+                    counter &+= 1
+                }
+            }
+
+            return counter
+        }.wait()
+
+        return counter
+    }
+
+}

Sources/NIOPerformanceTester/main.swift

@@ -833,4 +833,11 @@ try measureAndPrint(desc: "lock_4_threads_10M_ops",
 try measureAndPrint(desc: "lock_8_threads_10M_ops",
                     benchmark: LockBenchmark(numberOfThreads: 8, lockOperationsPerThread: 1_250_000))
 
+try measureAndPrint(desc: "schedule_10000_tasks",
+                    benchmark: SchedulingBenchmark())
 
+try measureAndPrint(desc: "schedule_and_run_10000_tasks",
+                    benchmark: SchedulingAndRunningBenchmark())
+
+try measureAndPrint(desc: "execute_10000",
+                    benchmark: ExecuteBenchmark())

docker/docker-compose.1604.53.yaml

@@ -51,7 +51,8 @@ services:
       - MAX_ALLOCS_ALLOWED_modifying_byte_buffer_view=2050
       - MAX_ALLOCS_ALLOWED_ping_pong_1000_reqs_1_conn=4400
       - MAX_ALLOCS_ALLOWED_read_10000_chunks_from_file=190050
-      - MAX_ALLOCS_ALLOWED_schedule_10000_tasks=90050
+      - MAX_ALLOCS_ALLOWED_schedule_10000_tasks=90150
+      - MAX_ALLOCS_ALLOWED_schedule_and_run_10000_tasks=100050
       - MAX_ALLOCS_ALLOWED_scheduling_10000_executions=20150
       - MAX_ALLOCS_ALLOWED_udp_1000_reqs_1_conn=12200
       - MAX_ALLOCS_ALLOWED_udp_1_reqs_1000_conn=188050

docker/docker-compose.1804.52.yaml

@@ -51,7 +51,8 @@ services:
       - MAX_ALLOCS_ALLOWED_modifying_byte_buffer_view=2050
       - MAX_ALLOCS_ALLOWED_ping_pong_1000_reqs_1_conn=4400
       - MAX_ALLOCS_ALLOWED_read_10000_chunks_from_file=200050
-      - MAX_ALLOCS_ALLOWED_schedule_10000_tasks=90050
+      - MAX_ALLOCS_ALLOWED_schedule_10000_tasks=90150
+      - MAX_ALLOCS_ALLOWED_schedule_and_run_10000_tasks=100050
       - MAX_ALLOCS_ALLOWED_scheduling_10000_executions=20150
       - MAX_ALLOCS_ALLOWED_udp_1000_reqs_1_conn=12200
       - MAX_ALLOCS_ALLOWED_udp_1_reqs_1000_conn=188050

docker/docker-compose.2004.54.yaml

@@ -51,7 +51,8 @@ services:
       - MAX_ALLOCS_ALLOWED_modifying_byte_buffer_view=2050
       - MAX_ALLOCS_ALLOWED_ping_pong_1000_reqs_1_conn=4400
       - MAX_ALLOCS_ALLOWED_read_10000_chunks_from_file=190050
-      - MAX_ALLOCS_ALLOWED_schedule_10000_tasks=90050
+      - MAX_ALLOCS_ALLOWED_schedule_10000_tasks=90150
+      - MAX_ALLOCS_ALLOWED_schedule_and_run_10000_tasks=100050
       - MAX_ALLOCS_ALLOWED_scheduling_10000_executions=20150
       - MAX_ALLOCS_ALLOWED_udp_1000_reqs_1_conn=12200
       - MAX_ALLOCS_ALLOWED_udp_1_reqs_1000_conn=190050

docker/docker-compose.2004.55.yaml

@@ -50,7 +50,8 @@ services:
       - MAX_ALLOCS_ALLOWED_modifying_byte_buffer_view=2050
       - MAX_ALLOCS_ALLOWED_ping_pong_1000_reqs_1_conn=4400
       - MAX_ALLOCS_ALLOWED_read_10000_chunks_from_file=190050
-      - MAX_ALLOCS_ALLOWED_schedule_10000_tasks=90050
+      - MAX_ALLOCS_ALLOWED_schedule_10000_tasks=90150
+      - MAX_ALLOCS_ALLOWED_schedule_and_run_10000_tasks=100050
       - MAX_ALLOCS_ALLOWED_scheduling_10000_executions=20150
       - MAX_ALLOCS_ALLOWED_udp_1000_reqs_1_conn=12200
       - MAX_ALLOCS_ALLOWED_udp_1_reqs_1000_conn=190050

docker/docker-compose.2004.main.yaml

@@ -51,7 +51,8 @@ services:
       - MAX_ALLOCS_ALLOWED_modifying_byte_buffer_view=2050
       - MAX_ALLOCS_ALLOWED_ping_pong_1000_reqs_1_conn=4400
       - MAX_ALLOCS_ALLOWED_read_10000_chunks_from_file=160050
-      - MAX_ALLOCS_ALLOWED_schedule_10000_tasks=90050
+      - MAX_ALLOCS_ALLOWED_schedule_10000_tasks=90150
+      - MAX_ALLOCS_ALLOWED_schedule_and_run_10000_tasks=100050
       - MAX_ALLOCS_ALLOWED_scheduling_10000_executions=20150
       - MAX_ALLOCS_ALLOWED_udp_1000_reqs_1_conn=12200
       - MAX_ALLOCS_ALLOWED_udp_1_reqs_1000_conn=188050