Fix locking bug in high-level stealing.

The lock was not released when an split_task was executed. We did a dirty fix by allowing the method to unlock the lock itself, but this has to be done cleaner in the future.

lib/pls/include/pls/internal/scheduling/abstract_task.h

@@ -28,7 +28,7 @@ namespace pls {
                 int depth() { return depth_; }
             protected:
                 virtual bool internal_stealing(abstract_task* other_task) = 0;
-                virtual bool split_task() = 0;
+                virtual bool split_task(base::spin_lock* lock) = 0;
 
                 bool steal_work();
             };

lib/pls/include/pls/internal/scheduling/root_task.h

@@ -40,7 +40,7 @@ namespace pls {
                     return false;
                 }
 
-                bool split_task() override {
+                bool split_task(base::spin_lock* /*lock*/) override {
                     return false;
                 }
             };
@@ -64,7 +64,7 @@ namespace pls {
                     return false;
                 }
 
-                bool split_task() override {
+                bool split_task(base::spin_lock* /*lock*/) override {
                     return false;
                 }
             };

lib/pls/include/pls/internal/scheduling/run_on_n_threads_task.h

@@ -57,7 +57,7 @@ namespace pls {
                     return false;
                 }
 
-                bool split_task() override;
+                bool split_task(base::spin_lock* lock) override;
             };
 
             template<typename Function>
@@ -83,16 +83,19 @@ namespace pls {
                     return false;
                 }
 
-                bool split_task() override {
+                bool split_task(base::spin_lock* /*lock*/) override {
                     return false;
                 }
             };
 
             template<typename Function>
-            bool run_on_n_threads_task<Function>::split_task() {
+            bool run_on_n_threads_task<Function>::split_task(base::spin_lock* lock) {
                 if (get_counter() <= 0) {
                     return false;
                 }
+                // In success case, unlock.
+                // TODO: this locking is complicated and error prone.
+                lock->unlock();
 
                 auto scheduler = base::this_thread::state<thread_state>()->scheduler_;
                 auto task = run_on_n_threads_task_worker<Function>{function_, this};

lib/pls/include/pls/internal/scheduling/tbb_task.h

@@ -60,6 +60,7 @@ namespace pls {
                 base::aligned_stack* my_stack_;
 
                 // Double-Ended Queue management
+                base::spin_lock lock_;
                 tbb_sub_task* top_;
                 tbb_sub_task* bottom_;
 
@@ -70,7 +71,7 @@ namespace pls {
                 tbb_sub_task* get_stolen_sub_task();
 
                 bool internal_stealing(abstract_task* other_task) override;
-                bool split_task() override;
+                bool split_task(base::spin_lock* /*lock*/) override;
 
             public:
                 explicit tbb_task(tbb_sub_task* root_task):

lib/pls/src/internal/scheduling/abstract_task.cpp

@@ -13,7 +13,9 @@ namespace pls {
                 for (size_t i = 1; i < my_scheduler->num_threads(); i++) {
                     size_t target = (my_id + i) % my_scheduler->num_threads();
                     auto target_state = my_scheduler->thread_state_for(target);
-                    std::lock_guard<base::spin_lock> lock{target_state->lock_};
+
+                    // TODO: Cleaner Locking Using std::guarded_lock
+                    target_state->lock_.lock();
 
                     // Dig down to our level
                     abstract_task* current_task = target_state->root_task_;
@@ -27,6 +29,7 @@ namespace pls {
                             current_task->depth_ == depth_) {
                             if (internal_stealing(current_task)) {
                                 // internal steal was a success, hand it back to the internal scheduler
+                                target_state->lock_.unlock();
                                 return true;
                             }
 
@@ -39,13 +42,14 @@ namespace pls {
                     // Execute 'top level task steal' if possible
                     // (only try deeper tasks to keep depth restricted stealing)
                     while (current_task != nullptr) {
-                        if (current_task->split_task()) {
+                        if (current_task->split_task(&target_state->lock_)) {
                             // internal steal was no success (we did a top level task steal)
                             return false;
                         }
 
                         current_task = current_task->child_task_;
                     }
+                    target_state->lock_.unlock();
                 }
 
                 // internal steal was no success

lib/pls/src/internal/scheduling/tbb_task.cpp

@@ -11,7 +11,7 @@ namespace pls {
                 below_{nullptr},
                 above_{nullptr} {}
 
-            tbb_sub_task::tbb_sub_task(const tbb_sub_task& other) {
+            tbb_sub_task::tbb_sub_task(const tbb_sub_task& /*other*/) {
                 // Do Nothing, will be inited after this anyways
             }
 
@@ -34,14 +34,17 @@ namespace pls {
                 sub_task->stack_state_ = tbb_task_->my_stack_->save_state();
 
                 // Put sub_task into stealing queue
-                if (tbb_task_->bottom_ != nullptr) {
-                    tbb_task_->bottom_->below_ = sub_task;
-                } else {
-                    tbb_task_->top_ = sub_task;
+                {
+                    std::lock_guard<base::spin_lock> lock{tbb_task_->lock_};
+                    if (tbb_task_->bottom_ != nullptr) {
+                        tbb_task_->bottom_->below_ = sub_task;
+                    } else {
+                        tbb_task_->top_ = sub_task;
+                    }
+                    sub_task->above_ = tbb_task_->bottom_;
+                    sub_task->below_ = nullptr;
+                    tbb_task_->bottom_ = sub_task;
                 }
-                sub_task->above_ = tbb_task_->bottom_;
-                sub_task->below_ = nullptr;
-                tbb_task_->bottom_ = sub_task;
             }
 
             void tbb_sub_task::wait_for_all() {
@@ -61,12 +64,14 @@ namespace pls {
             }
 
             tbb_sub_task* tbb_task::get_local_sub_task() {
+                // Remove from bottom of queue
+                std::lock_guard<base::spin_lock> lock{lock_};
+
                 if (bottom_ == nullptr) {
                     return nullptr;
                 }
 
-                // Remove from bottom of queue
-                tbb_sub_task* result = bottom_;
+                tbb_sub_task *result = bottom_;
                 bottom_ = bottom_->above_;
                 if (bottom_ == nullptr) {
                     top_ = nullptr;
@@ -78,11 +83,14 @@ namespace pls {
             }
 
             tbb_sub_task* tbb_task::get_stolen_sub_task() {
+                // Remove from top of queue
+                std::lock_guard<base::spin_lock> lock{lock_};
+
                 if (top_ == nullptr) {
                     return nullptr;
                 }
 
-                tbb_sub_task* result = top_;
+                tbb_sub_task *result = top_;
                 top_ = top_->below_;
                 if (top_ == nullptr) {
                     bottom_ = nullptr;
@@ -110,11 +118,14 @@ namespace pls {
                 }
             }
 
-            bool tbb_task::split_task() {
+            bool tbb_task::split_task(base::spin_lock* lock) {
                 tbb_sub_task* stolen_sub_task = get_stolen_sub_task();
                 if (stolen_sub_task == nullptr) {
                     return false;
                 }
+                // In success case, unlock.
+                // TODO: this locking is complicated and error prone.
+                lock->unlock();
 
                 tbb_task task{stolen_sub_task};
                 scheduler::execute_task(task, depth());