| 1 | //===-- ThreadingTests.cpp --------------------------------------*- C++ -*-===// |
| 2 | // |
| 3 | // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions. |
| 4 | // See https://llvm.org/LICENSE.txt for license information. |
| 5 | // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception |
| 6 | // |
| 7 | //===----------------------------------------------------------------------===// |
| 8 | |
| 9 | #include "support/Threading.h" |
| 10 | #include "llvm/ADT/DenseMap.h" |
| 11 | #include "gmock/gmock.h" |
| 12 | #include "gtest/gtest.h" |
| 13 | #include <chrono> |
| 14 | #include <mutex> |
| 15 | |
| 16 | namespace clang { |
| 17 | namespace clangd { |
| 18 | class ThreadingTest : public ::testing::Test {}; |
| 19 | |
| 20 | TEST_F(ThreadingTest, TaskRunner) { |
| 21 | const int TasksCnt = 100; |
| 22 | // This should be const, but MSVC does not allow to use const vars in lambdas |
| 23 | // without capture. On the other hand, clang gives a warning that capture of |
| 24 | // const var is not required. |
| 25 | // Making it non-const makes both compilers happy. |
| 26 | int IncrementsPerTask = 1000; |
| 27 | |
| 28 | std::mutex Mutex; |
| 29 | int Counter(0); /* GUARDED_BY(Mutex) */ |
| 30 | { |
| 31 | AsyncTaskRunner Tasks; |
| 32 | auto ScheduleIncrements = [&]() { |
| 33 | for (int TaskI = 0; TaskI < TasksCnt; ++TaskI) { |
| 34 | Tasks.runAsync(Name: "task" , Action: [&Counter, &Mutex, IncrementsPerTask]() { |
| 35 | for (int Increment = 0; Increment < IncrementsPerTask; ++Increment) { |
| 36 | std::lock_guard<std::mutex> Lock(Mutex); |
| 37 | ++Counter; |
| 38 | } |
| 39 | }); |
| 40 | } |
| 41 | }; |
| 42 | |
| 43 | { |
| 44 | // Make sure runAsync is not running tasks synchronously on the same |
| 45 | // thread by locking the Mutex used for increments. |
| 46 | std::lock_guard<std::mutex> Lock(Mutex); |
| 47 | ScheduleIncrements(); |
| 48 | } |
| 49 | |
| 50 | Tasks.wait(); |
| 51 | { |
| 52 | std::lock_guard<std::mutex> Lock(Mutex); |
| 53 | ASSERT_EQ(Counter, TasksCnt * IncrementsPerTask); |
| 54 | } |
| 55 | |
| 56 | { |
| 57 | std::lock_guard<std::mutex> Lock(Mutex); |
| 58 | Counter = 0; |
| 59 | ScheduleIncrements(); |
| 60 | } |
| 61 | } |
| 62 | // Check that destructor has waited for tasks to finish. |
| 63 | std::lock_guard<std::mutex> Lock(Mutex); |
| 64 | ASSERT_EQ(Counter, TasksCnt * IncrementsPerTask); |
| 65 | } |
| 66 | |
| 67 | TEST_F(ThreadingTest, Memoize) { |
| 68 | const unsigned NumThreads = 5; |
| 69 | const unsigned NumKeys = 100; |
| 70 | const unsigned NumIterations = 100; |
| 71 | |
| 72 | Memoize<llvm::DenseMap<int, int>> Cache; |
| 73 | std::atomic<unsigned> ComputeCount(0); |
| 74 | std::atomic<int> ComputeResult[NumKeys]; |
| 75 | std::fill(first: std::begin(arr&: ComputeResult), last: std::end(arr&: ComputeResult), value: -1); |
| 76 | |
| 77 | AsyncTaskRunner Tasks; |
| 78 | for (unsigned I = 0; I < NumThreads; ++I) |
| 79 | Tasks.runAsync(Name: "worker" + std::to_string(val: I), Action: [&] { |
| 80 | for (unsigned J = 0; J < NumIterations; J++) |
| 81 | for (unsigned K = 0; K < NumKeys; K++) { |
| 82 | int Result = Cache.get(Key&: K, Compute: [&] { return ++ComputeCount; }); |
| 83 | EXPECT_THAT(ComputeResult[K].exchange(Result), |
| 84 | testing::AnyOf(-1, Result)) |
| 85 | << "Got inconsistent results from memoize" ; |
| 86 | } |
| 87 | }); |
| 88 | Tasks.wait(); |
| 89 | EXPECT_GE(ComputeCount, NumKeys) << "Computed each key once" ; |
| 90 | EXPECT_LE(ComputeCount, NumThreads * NumKeys) |
| 91 | << "Worst case, computed each key in every thread" ; |
| 92 | for (int Result : ComputeResult) |
| 93 | EXPECT_GT(Result, 0) << "All results in expected domain" ; |
| 94 | } |
| 95 | |
| 96 | TEST_F(ThreadingTest, MemoizeDeterministic) { |
| 97 | Memoize<llvm::DenseMap<int, char>> Cache; |
| 98 | |
| 99 | // Spawn two parallel computations, A and B. |
| 100 | // Force concurrency: neither can finish until both have started. |
| 101 | // Verify that cache returns consistent results. |
| 102 | AsyncTaskRunner Tasks; |
| 103 | std::atomic<char> ValueA(0), ValueB(0); |
| 104 | Notification ReleaseA, ReleaseB; |
| 105 | Tasks.runAsync(Name: "A" , Action: [&] { |
| 106 | ValueA = Cache.get(Key: 0, Compute: [&] { |
| 107 | ReleaseB.notify(); |
| 108 | ReleaseA.wait(); |
| 109 | return 'A'; |
| 110 | }); |
| 111 | }); |
| 112 | Tasks.runAsync(Name: "A" , Action: [&] { |
| 113 | ValueB = Cache.get(Key: 0, Compute: [&] { |
| 114 | ReleaseA.notify(); |
| 115 | ReleaseB.wait(); |
| 116 | return 'B'; |
| 117 | }); |
| 118 | }); |
| 119 | Tasks.wait(); |
| 120 | |
| 121 | ASSERT_EQ(ValueA, ValueB); |
| 122 | ASSERT_THAT(ValueA.load(), testing::AnyOf('A', 'B')); |
| 123 | } |
| 124 | |
| 125 | // It's hard to write a real test of this class, std::chrono is awkward to mock. |
| 126 | // But test some degenerate cases at least. |
| 127 | TEST(PeriodicThrottlerTest, Minimal) { |
| 128 | PeriodicThrottler Once(std::chrono::hours(24)); |
| 129 | EXPECT_TRUE(Once()); |
| 130 | EXPECT_FALSE(Once()); |
| 131 | EXPECT_FALSE(Once()); |
| 132 | |
| 133 | PeriodicThrottler Later(std::chrono::hours(24), |
| 134 | /*Delay=*/std::chrono::hours(24)); |
| 135 | EXPECT_FALSE(Later()); |
| 136 | EXPECT_FALSE(Later()); |
| 137 | EXPECT_FALSE(Later()); |
| 138 | |
| 139 | PeriodicThrottler Always(std::chrono::seconds(0)); |
| 140 | EXPECT_TRUE(Always()); |
| 141 | EXPECT_TRUE(Always()); |
| 142 | EXPECT_TRUE(Always()); |
| 143 | } |
| 144 | |
| 145 | } // namespace clangd |
| 146 | } // namespace clang |
| 147 | |