Add ttas_spinlock.

app/invoke_parallel/main.cpp

@@ -34,7 +34,7 @@ long fib(long n) {
 
 
 int main() {
 int main() {
     PROFILE_ENABLE
     PROFILE_ENABLE
-    pls::scheduler scheduler{&my_scheduler_memory, 2};
+    pls::scheduler scheduler{&my_scheduler_memory, 8};
 
 
     long result;
     long result;
     scheduler.perform_work([&] {
     scheduler.perform_work([&] {

lib/pls/CMakeLists.txt

@@ -5,7 +5,9 @@ add_library(pls STATIC
         include/pls/algorithms/invoke_parallel.h
         include/pls/algorithms/invoke_parallel.h
         include/pls/algorithms/invoke_parallel_impl.h
         include/pls/algorithms/invoke_parallel_impl.h
 
 
-        include/pls/internal/base/spin_lock.h                   src/internal/base/spin_lock.cpp
+        include/pls/internal/base/spin_lock.h
+        include/pls/internal/base/tas_spin_lock.h               src/internal/base/tas_spin_lock.cpp
+        include/pls/internal/base/ttas_spin_lock.h              src/internal/base/ttas_spin_lock.cpp
         include/pls/internal/base/thread.h                      src/internal/base/thread.cpp
         include/pls/internal/base/thread.h                      src/internal/base/thread.cpp
         include/pls/internal/base/thread_impl.h
         include/pls/internal/base/thread_impl.h
         include/pls/internal/base/barrier.h                     src/internal/base/barrier.cpp
         include/pls/internal/base/barrier.h                     src/internal/base/barrier.cpp
@@ -20,6 +22,7 @@ add_library(pls STATIC
         include/pls/internal/helpers/prohibit_new.h
         include/pls/internal/helpers/prohibit_new.h
         include/pls/internal/helpers/profiler.h
         include/pls/internal/helpers/profiler.h
         include/pls/internal/helpers/mini_benchmark.h
         include/pls/internal/helpers/mini_benchmark.h
+        include/pls/internal/helpers/unique_id.h
 
 
         include/pls/internal/scheduling/root_task.h             src/internal/scheduling/root_task.cpp
         include/pls/internal/scheduling/root_task.h             src/internal/scheduling/root_task.cpp
         include/pls/internal/scheduling/thread_state.h          src/internal/scheduling/thread_state.cpp
         include/pls/internal/scheduling/thread_state.h          src/internal/scheduling/thread_state.cpp
@@ -29,7 +32,7 @@ add_library(pls STATIC
         include/pls/internal/scheduling/run_on_n_threads_task.h src/internal/scheduling/run_on_n_threads_task.cpp
         include/pls/internal/scheduling/run_on_n_threads_task.h src/internal/scheduling/run_on_n_threads_task.cpp
         include/pls/internal/scheduling/fork_join_task.h        src/internal/scheduling/fork_join_task.cpp
         include/pls/internal/scheduling/fork_join_task.h        src/internal/scheduling/fork_join_task.cpp
         include/pls/internal/scheduling/scheduler_memory.h      src/internal/scheduling/scheduler_memory.cpp
         include/pls/internal/scheduling/scheduler_memory.h      src/internal/scheduling/scheduler_memory.cpp
-        include/pls/internal/helpers/unique_id.h)
+)
 
 
 # Add everything in `./include` to be in the include path of this project
 # Add everything in `./include` to be in the include path of this project
 target_include_directories(pls
 target_include_directories(pls

lib/pls/include/pls/internal/base/spin_lock.h

@@ -2,32 +2,14 @@
 #ifndef PLS_SPINLOCK_H
 #ifndef PLS_SPINLOCK_H
 #define PLS_SPINLOCK_H
 #define PLS_SPINLOCK_H
 
 
-#include <atomic>
-#include <iostream>
-
-#include "pls/internal/base/thread.h"
+#include "tas_spin_lock.h"
+#include "ttas_spin_lock.h"
 
 
 namespace pls {
 namespace pls {
     namespace internal {
     namespace internal {
         namespace base {
         namespace base {
-            /**
-             * A simple set and test_and_set based spin lock implementation.
-             *
-             * PORTABILITY:
-             * Current implementation is based on C++ 11 atomic_flag.
-             */
-            class spin_lock {
-                std::atomic_flag flag_;
-                unsigned int yield_at_tries_;
-
-
-            public:
-                spin_lock(): flag_{ATOMIC_FLAG_INIT}, yield_at_tries_{1024} {};
-                spin_lock(const spin_lock& other): flag_{ATOMIC_FLAG_INIT}, yield_at_tries_{other.yield_at_tries_} {}
-
-                void lock();
-                void unlock();
-            };
+            // Default Spin-Lock implementation for this project.
+            using spin_lock = tas_spin_lock;
         }
         }
     }
     }
 }
 }

lib/pls/include/pls/internal/base/tas_spin_lock.h

+
+#ifndef PLS_TAS_SPIN_LOCK_H
+#define PLS_TAS_SPIN_LOCK_H
+
+#include "tas_spin_lock.h"
+
+#include <atomic>
+#include <iostream>
+
+#include "pls/internal/base/thread.h"
+
+namespace pls {
+    namespace internal {
+        namespace base {
+            /**
+             * A simple set and test_and_set based spin lock implementation.
+             *
+             * PORTABILITY:
+             * Current implementation is based on C++ 11 atomic_flag.
+             */
+            class tas_spin_lock {
+                std::atomic_flag flag_;
+                unsigned int yield_at_tries_;
+
+
+            public:
+                tas_spin_lock(): flag_{ATOMIC_FLAG_INIT}, yield_at_tries_{1024} {};
+                tas_spin_lock(const tas_spin_lock& other): flag_{ATOMIC_FLAG_INIT}, yield_at_tries_{other.yield_at_tries_} {}
+
+                void lock();
+                bool try_lock();
+                void unlock();
+            };
+        }
+    }
+}
+
+
+#endif //PLS_TAS_SPIN_LOCK_H

lib/pls/include/pls/internal/base/ttas_spin_lock.h

+
+#ifndef PLS_TTAS_SPIN_LOCK_H
+#define PLS_TTAS_SPIN_LOCK_H
+
+#include "tas_spin_lock.h"
+
+#include <atomic>
+#include <iostream>
+
+#include "pls/internal/base/thread.h"
+
+namespace pls {
+    namespace internal {
+        namespace base {
+            /**
+             * A simple set and test_and_set based spin lock implementation.
+             *
+             * PORTABILITY:
+             * Current implementation is based on C++ 11 atomic_flag.
+             */
+            class ttas_spin_lock {
+                std::atomic<int> flag_;
+                unsigned int yield_at_tries_;
+
+
+            public:
+                ttas_spin_lock(): flag_{0}, yield_at_tries_{1024} {};
+                ttas_spin_lock(const ttas_spin_lock& other): flag_{0}, yield_at_tries_{other.yield_at_tries_} {}
+
+                void lock();
+                bool try_lock();
+                void unlock();
+            };
+        }
+    }
+}
+
+
+#endif //PLS_TTAS_SPIN_LOCK_H

lib/pls/src/internal/base/spin_lock.cpp → lib/pls/src/internal/base/tas_spin_lock.cpp

 #include "pls/internal/helpers/profiler.h"
 #include "pls/internal/helpers/profiler.h"
-#include "pls/internal/base/spin_lock.h"
+#include "pls/internal/base/tas_spin_lock.h"
 
 
 namespace pls {
 namespace pls {
     namespace internal {
     namespace internal {
         namespace base {
         namespace base {
-            // TODO: Research/Measure how the memory fences/orders influence this.
-            //       For now we simply try to be safe by forcing this lock to
-            //       also act as a strict memory fence.
-            void spin_lock::lock() {
+            void tas_spin_lock::lock() {
                 PROFILE_LOCK("Acquire Lock")
                 PROFILE_LOCK("Acquire Lock")
                 int tries = 0;
                 int tries = 0;
-                while (flag_.test_and_set(std::memory_order_seq_cst)) {
+                while (flag_.test_and_set(std::memory_order_acquire)) {
                     tries++;
                     tries++;
                     if (tries % yield_at_tries_ == 0) {
                     if (tries % yield_at_tries_ == 0) {
                         this_thread::yield();
                         this_thread::yield();
@@ -18,8 +15,12 @@ namespace pls {
                 }
                 }
             }
             }
 
 
-            void spin_lock::unlock() {
-                flag_.clear(std::memory_order_seq_cst);
+            bool tas_spin_lock::try_lock() {
+                return flag_.test_and_set(std::memory_order_acquire) == 0;
+            }
+
+            void tas_spin_lock::unlock() {
+                flag_.clear(std::memory_order_release);
             }
             }
         }
         }
     }
     }

lib/pls/src/internal/base/ttas_spin_lock.cpp

+#include "pls/internal/helpers/profiler.h"
+#include "pls/internal/base/ttas_spin_lock.h"
+
+namespace pls {
+    namespace internal {
+        namespace base {
+            void ttas_spin_lock::lock() {
+                PROFILE_LOCK("Acquire Lock")
+                int tries = 0;
+                int expected = 0;
+
+                do {
+                    while (flag_.load(std::memory_order_relaxed) == 1) {
+                        tries++;
+                        if (tries % yield_at_tries_ == 0) {
+                            this_thread::yield();
+                        }
+                    }
+
+                    expected = 0;
+                } while (!flag_.compare_exchange_weak(expected, 1, std::memory_order_acquire));
+            }
+
+            bool ttas_spin_lock::try_lock() {
+                int expected = 0;
+                return flag_.load(std::memory_order_relaxed) == 0 && flag_.compare_exchange_weak(expected, 1, std::memory_order_acquire);
+            }
+
+            void ttas_spin_lock::unlock() {
+                flag_.store(0, std::memory_order_release);
+            }
+        }
+    }
+}