Split up lld Parallel.h a bit

[zturner]

Summary

This patch splits Parallel by introducing a new file called Executor.h and Executor.cpp.

Executor.h defines parallel infrastructure, but hides all concrete executor implementations in a cpp file inside of an anonymous namespaces so they can't leak out.

[ruiu]

lld/include/lld/Core/Executor.h

27–29

You want to remove the leading underscores and make them start with uppercase.

[krytarowski]

lld/lib/Core/Executor.cpp

19

#if !LLVM_ENABLE_THREADS

?

[zturner]

• Updated everything to use LLVM naming conventions.
• Hid the Executor stuff even more. Previously the abstract base class was in the Executor.h, now even that is hidden in an anonymous namespace in a cpp file. The only thing exposed now is TaskGroup.

That doesn't mean I'm giving up on trying to get this to parity with the C++ Parallel TS, but I'd like to get this ready for LLVM sooner rather than later, and this seems like the method with the least friction.

I plan to go check the Parallel TS as a follow up and see if there's anything else I can do to get more parity on the interfaces, but at least this way Executor is out of the picture for now.

[ruiu]

LGTM. This patch splits the files without changing any behavior and also fixes variable names. I want you to commit this before you make further changes, as I want to separate mechanical changes from non-mechanical ones.

lld/include/lld/Core/TaskGroup.h

28–31 ↗

(On Diff #97988)

Please fix these names.

56 ↗

(On Diff #97988)

And this.

[zturner]

Just to be clear, do you want me to commit this without the latest update? i.e. I should commit Diff 1 of this (without the variable name changes etc), then do all the renames and stuff in a separate patch (without review), then do further improvements?

[ruiu]

I don't have a strong preference, but as you already split the file and fixed the variable names, I'm fine if you check in both changes as one change.

lld/include/lld/Core/Executor.h

+//===- lld/Core/Executor.h - Thread Executor ------------------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLD_CORE_EXECUTOR_H
+#define LLD_CORE_EXECUTOR_H
+
+#include <atomic>
+#include <condition_variable>
+#include <mutex>
+#include <stack>
+
+namespace lld {
+/// \brief Allows one or more threads to wait on a potentially unknown number of
+///   events.
+///
+/// A latch starts at \p count. inc() increments this, and dec() decrements it.
+/// All calls to sync() will block while the count is not 0.
+///
+/// Calling dec() on a Latch with a count of 0 has undefined behaivor.
+class Latch {
+  uint32_t _count;
+  mutable std::mutex _condMut;
+  mutable std::condition_variable _cond;
+
+public:
+  explicit Latch(uint32_t count = 0) : _count(count) {}
+  ~Latch() { sync(); }
+
+  void inc() {
+    std::unique_lock<std::mutex> lock(_condMut);
+    ++_count;
+  }
+
+  void dec() {
+    std::unique_lock<std::mutex> lock(_condMut);
+    if (--_count == 0)
+      _cond.notify_all();
+  }
+
+  void sync() const {
+    std::unique_lock<std::mutex> lock(_condMut);
+    _cond.wait(lock, [&] { return _count == 0; });
+  }
+};
+
+/// \brief An abstract class that takes closures and runs them asynchronously.
+class Executor {
+public:
+  virtual ~Executor() = default;
+  virtual void add(std::function<void()> func) = 0;
+
+  static Executor *getDefaultExecutor();
+};
+
+/// \brief Allows launching a number of tasks and waiting for them to finish
+///   either explicitly via sync() or implicitly on destruction.
+class TaskGroup {
+  Latch _latch;
+
+public:
+  void spawn(std::function<void()> f) {
+    _latch.inc();
+    Executor::getDefaultExecutor()->add([&, f] {
+      f();
+      _latch.dec();
+    });
+  }
+
+  void sync() const { _latch.sync(); }
+};
+}
+
+#endif

lld/include/lld/Core/Parallel.h

 //===- lld/Core/Parallel.h - Parallel utilities ---------------------------===//
 //
 //                             The LLVM Linker
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 
 #ifndef LLD_CORE_PARALLEL_H
 #define LLD_CORE_PARALLEL_H
 
-#include "lld/Core/Instrumentation.h"
+#include "lld/Core/Executor.h"
 #include "lld/Core/LLVM.h"
 #include "llvm/Support/MathExtras.h"
-#include "llvm/Support/thread.h"
 
 #include <algorithm>
-#include <atomic>
-#include <condition_variable>
-#include <mutex>
-#include <stack>
 
 #if defined(_MSC_VER) && LLVM_ENABLE_THREADS
 #include <concrt.h>
 #include <ppl.h>
 #endif
 
 namespace lld {
-/// \brief Allows one or more threads to wait on a potentially unknown number of
-///   events.
-///
-/// A latch starts at \p count. inc() increments this, and dec() decrements it.
-/// All calls to sync() will block while the count is not 0.
-///
-/// Calling dec() on a Latch with a count of 0 has undefined behaivor.
-class Latch {
-  uint32_t _count;
-  mutable std::mutex _condMut;
-  mutable std::condition_variable _cond;
-
-public:
-  explicit Latch(uint32_t count = 0) : _count(count) {}
-  ~Latch() { sync(); }
-
-  void inc() {
-    std::unique_lock<std::mutex> lock(_condMut);
-    ++_count;
-  }
-
-  void dec() {
-    std::unique_lock<std::mutex> lock(_condMut);
-    if (--_count == 0)
-      _cond.notify_all();
-  }
-
-  void sync() const {
-    std::unique_lock<std::mutex> lock(_condMut);
-    _cond.wait(lock, [&] {
-      return _count == 0;
-    });
-  }
-};
-
-// Classes in this namespace are implementation details of this header.
-namespace internal {
-
-/// \brief An abstract class that takes closures and runs them asynchronously.
-class Executor {
-public:
-  virtual ~Executor() = default;
-  virtual void add(std::function<void()> func) = 0;
-};
-
-#if !defined(LLVM_ENABLE_THREADS) || LLVM_ENABLE_THREADS == 0
-class SyncExecutor : public Executor {
-public:
-  virtual void add(std::function<void()> func) {
-    func();
-  }
-};
-
-inline Executor *getDefaultExecutor() {
-  static SyncExecutor exec;
-  return &exec;
-}
-#elif defined(_MSC_VER)
-/// \brief An Executor that runs tasks via ConcRT.
-class ConcRTExecutor : public Executor {
-  struct Taskish {
-    Taskish(std::function<void()> task) : _task(task) {}
-
-    std::function<void()> _task;
-
-    static void run(void *p) {
-      Taskish *self = static_cast<Taskish *>(p);
-      self->_task();
-      concurrency::Free(self);
-    }
-  };
-
-public:
-  virtual void add(std::function<void()> func) {
-    Concurrency::CurrentScheduler::ScheduleTask(Taskish::run,
-        new (concurrency::Alloc(sizeof(Taskish))) Taskish(func));
-  }
-};
-
-inline Executor *getDefaultExecutor() {
-  static ConcRTExecutor exec;
-  return &exec;
-}
-#else
-/// \brief An implementation of an Executor that runs closures on a thread pool
-///   in filo order.
-class ThreadPoolExecutor : public Executor {
-public:
-  explicit ThreadPoolExecutor(unsigned threadCount =
-                                  std::thread::hardware_concurrency())
-      : _stop(false), _done(threadCount) {
-    // Spawn all but one of the threads in another thread as spawning threads
-    // can take a while.
-    std::thread([&, threadCount] {
-      for (size_t i = 1; i < threadCount; ++i) {
-        std::thread([=] {
-          work();
-        }).detach();
-      }
-      work();
-    }).detach();
-  }
-
-  ~ThreadPoolExecutor() override {
-    std::unique_lock<std::mutex> lock(_mutex);
-    _stop = true;
-    lock.unlock();
-    _cond.notify_all();
-    // Wait for ~Latch.
-  }
-
-  void add(std::function<void()> f) override {
-    std::unique_lock<std::mutex> lock(_mutex);
-    _workStack.push(f);
-    lock.unlock();
-    _cond.notify_one();
-  }
-
-private:
-  void work() {
-    while (true) {
-      std::unique_lock<std::mutex> lock(_mutex);
-      _cond.wait(lock, [&] {
-        return _stop || !_workStack.empty();
-      });
-      if (_stop)
-        break;
-      auto task = _workStack.top();
-      _workStack.pop();
-      lock.unlock();
-      task();
-    }
-    _done.dec();
-  }
-
-  std::atomic<bool> _stop;
-  std::stack<std::function<void()>> _workStack;
-  std::mutex _mutex;
-  std::condition_variable _cond;
-  Latch _done;
-};
-
-inline Executor *getDefaultExecutor() {
-  static ThreadPoolExecutor exec;
-  return &exec;
-}
-#endif
-
-}  // namespace internal
-
-/// \brief Allows launching a number of tasks and waiting for them to finish
-///   either explicitly via sync() or implicitly on destruction.
-class TaskGroup {
-  Latch _latch;
-
-public:
-  void spawn(std::function<void()> f) {
-    _latch.inc();
-    internal::getDefaultExecutor()->add([&, f] {
-      f();
-      _latch.dec();
-    });
-  }
-
-  void sync() const { _latch.sync(); }
-};
 
 #if !defined(LLVM_ENABLE_THREADS) || LLVM_ENABLE_THREADS == 0
 template <class RandomAccessIterator, class Comp>
 void parallel_sort(
     RandomAccessIterator start, RandomAccessIterator end,
     const Comp &comp = std::less<
         typename std::iterator_traits<RandomAccessIterator>::value_type>()) {
   std::sort(start, end, comp);

lld/lib/Core/CMakeLists.txt

 if(NOT LLD_BUILT_STANDALONE)
   set(tablegen_deps intrinsics_gen)
 endif()
 
 add_lld_library(lldCore
   DefinedAtom.cpp
   Error.cpp
+  Executor.cpp
   File.cpp
   LinkingContext.cpp
   Reader.cpp
   Reproduce.cpp
   Resolver.cpp
   SymbolTable.cpp
   TargetOptionsCommandFlags.cpp
   Writer.cpp

lld/lib/Core/Executor.cpp

+//===- lld/Core/Executor.cpp - Thread Executor ----------------------------===//
+//
+//                             The LLVM Linker
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+
+#include "lld/Core/Executor.h"
+
+#if defined(_MSC_VER) && LLVM_ENABLE_THREADS
+#include <concrt.h>
+#include <ppl.h>
+#endif
+
+using namespace lld;
+
+#if !defined(LLVM_ENABLE_THREADS) || LLVM_ENABLE_THREADS == 0
+namespace {
+class SyncExecutor : public Executor {
+public:
+  virtual void add(std::function<void()> func) { func(); }
+};
+}
+
+Executor *Executor::getDefaultExecutor() {
+  static SyncExecutor exec;
+  return &exec;
+}
+
+#elif defined(_MSC_VER)
+namespace {
+/// \brief An Executor that runs tasks via ConcRT.
+class ConcRTExecutor : public Executor {
+  struct Taskish {
+    Taskish(std::function<void()> task) : _task(task) {}
+
+    std::function<void()> _task;
+
+    static void run(void *p) {
+      Taskish *self = static_cast<Taskish *>(p);
+      self->_task();
+      concurrency::Free(self);
+    }
+  };
+
+public:
+  virtual void add(std::function<void()> func) {
+    Concurrency::CurrentScheduler::ScheduleTask(
+        Taskish::run, new (concurrency::Alloc(sizeof(Taskish))) Taskish(func));
+  }
+};
+}
+
+Executor *Executor::getDefaultExecutor() {
+  static ConcRTExecutor exec;
+  return &exec;
+}
+
+#else
+namespace {
+/// \brief An implementation of an Executor that runs closures on a thread pool
+///   in filo order.
+class ThreadPoolExecutor : public Executor {
+public:
+  explicit ThreadPoolExecutor(
+      unsigned threadCount = std::thread::hardware_concurrency())
+      : _stop(false), _done(threadCount) {
+    // Spawn all but one of the threads in another thread as spawning threads
+    // can take a while.
+    std::thread([&, threadCount] {
+      for (size_t i = 1; i < threadCount; ++i) {
+        std::thread([=] { work(); }).detach();
+      }
+      work();
+    }).detach();
+  }
+
+  ~ThreadPoolExecutor() override {
+    std::unique_lock<std::mutex> lock(_mutex);
+    _stop = true;
+    lock.unlock();
+    _cond.notify_all();
+    // Wait for ~Latch.
+  }
+
+  void add(std::function<void()> f) override {
+    std::unique_lock<std::mutex> lock(_mutex);
+    _workStack.push(f);
+    lock.unlock();
+    _cond.notify_one();
+  }
+
+private:
+  void work() {
+    while (true) {
+      std::unique_lock<std::mutex> lock(_mutex);
+      _cond.wait(lock, [&] { return _stop || !_workStack.empty(); });
+      if (_stop)
+        break;
+      auto task = _workStack.top();
+      _workStack.pop();
+      lock.unlock();
+      task();
+    }
+    _done.dec();
+  }
+
+  std::atomic<bool> _stop;
+  std::stack<std::function<void()>> _workStack;
+  std::mutex _mutex;
+  std::condition_variable _cond;
+  Latch _done;
+};
+}
+#endif