Working version of our trading-deque

e34ea267 · FritzFlorian · 4cf3848f · e34ea267 · e34ea267 · e34ea267
Commit e34ea267 authored Dec 04, 2019 by FritzFlorian
15 changed files
--- a/app/benchmark_fft/main.cpp
+++ b/app/benchmark_fft/main.cpp
@@ -90,11 +90,10 @@ complex_vector prepare_input(int input_size) {
  return data;
 }
-static constexpr int NUM_ITERATIONS = 500;
+static constexpr int NUM_ITERATIONS = 1000;
-constexpr size_t NUM_THREADS = 8;
+constexpr size_t NUM_THREADS = 5;
 constexpr size_t NUM_TASKS = 128;
-constexpr size_t MAX_TASK_STACK_SIZE = 0;
 constexpr size_t NUM_CONTS = 128;
 constexpr size_t MAX_CONT_SIZE = 512;
@@ -104,7 +103,6 @@ int main() {
  static_scheduler_memory<NUM_THREADS,
                          NUM_TASKS,
-                          MAX_TASK_STACK_SIZE,
                          NUM_CONTS,
                          MAX_CONT_SIZE> static_scheduler_memory;
@@ -127,14 +125,14 @@ int main() {
  std::cout << "Framework:  " << std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count()
            << std::endl;
-//  start = std::chrono::steady_clock::now();
+  start = std::chrono::steady_clock::now();
-//  for (int i = 0; i < NUM_ITERATIONS; i++) {
+  for (int i = 0; i < NUM_ITERATIONS; i++) {
-//    complex_vector input_1(initial_input);
+    complex_vector input_1(initial_input);
-//    fft_normal(input_1.begin(), INPUT_SIZE);
+    fft_normal(input_1.begin(), INPUT_SIZE);
-//  }
+  }
-//  end = std::chrono::steady_clock::now();
+  end = std::chrono::steady_clock::now();
-//  std::cout << "Normal:     " << std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count()
+  std::cout << "Normal:     " << std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count()
-//            << std::endl;
+            << std::endl;
  return 0;
 }
--- a/app/playground/main.cpp
+++ b/app/playground/main.cpp
@@ -4,14 +4,14 @@
 #include "pls/internal/scheduling/scheduler.h"
 #include "pls/internal/scheduling/parallel_result.h"
 #include "pls/internal/scheduling/scheduler_memory.h"
-#include "pls/internal/base/thread.h"
+#include "pls/internal/data_structures/bounded_trading_deque.h"
 using namespace pls::internal;
 constexpr size_t NUM_THREADS = 1;
 constexpr size_t NUM_TASKS = 128;
-constexpr size_t MAX_TASK_STACK_SIZE = 0;
+static constexpr int NUM_ITERATIONS = 100;
 constexpr size_t NUM_CONTS = 128;
 constexpr size_t MAX_CONT_SIZE = 256;
@@ -45,33 +45,37 @@ scheduling::parallel_result<int> fib(int n) {
  });
 }
+static volatile int result;
 int main() {
  scheduling::static_scheduler_memory<NUM_THREADS,
                                      NUM_TASKS,
-                                      MAX_TASK_STACK_SIZE,
                                      NUM_CONTS,
                                      MAX_CONT_SIZE> static_scheduler_memory;
  scheduling::scheduler scheduler{static_scheduler_memory, NUM_THREADS};
  auto start = std::chrono::steady_clock::now();
-  std::cout << "fib = " << fib_normal(30) << std::endl;
+  for (int i = 0; i < NUM_ITERATIONS; i++) {
+    result = fib_normal(30);
+  }
  auto end = std::chrono::steady_clock::now();
  std::cout << "Normal:     " << std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count()
            << std::endl;
  start = std::chrono::steady_clock::now();
+  for (int i = 0; i < NUM_ITERATIONS; i++) {
    scheduler.perform_work([]() {
      return scheduling::scheduler::par([]() {
        return scheduling::parallel_result<int>(0);
      }, []() {
        return fib(30);
      }).then([](int, int b) {
-      std::cout << "fib = " << b << std::endl;
+        result = b;
        return scheduling::parallel_result<int>{0};
      });
    });
+  }
  end = std::chrono::steady_clock::now();
  std::cout << "Framework: " << std::chrono::duration_cast<std::chrono::milliseconds>(end - start).count() << std::endl;

--- a/lib/pls/include/pls/internal/base/error_handling.h
+++ b/lib/pls/include/pls/internal/base/error_handling.h
@@ -15,6 +15,8 @@
 void pls_error(const char *msg);
-#define PLS_ASSERT(cond, msg) if (!(cond)) { pls_error(msg); }
+// TODO: Distinguish between debug/internal asserts and production asserts.
+// TODO: Re-Enable Asserts
+#define PLS_ASSERT(cond, msg) //if (!(cond)) { pls_error(msg); }
 #endif //PLS_ERROR_HANDLING_H
--- a/lib/pls/include/pls/internal/data_structures/bounded_trading_deque.h
+++ b/lib/pls/include/pls/internal/data_structures/bounded_trading_deque.h
@@ -3,7 +3,7 @@
 #define PLS_INTERNAL_DATA_STRUCTURES_BOUNDED_TRADING_DEQUE_H_
 #include <atomic>
-#include <pls/internal/base/system_details.h>
+#include <tuple>
 #include "pls/internal/base/error_handling.h"
 #include "pls/internal/base/system_details.h"
@@ -17,31 +17,45 @@ namespace data_structures {
 template<typename TradedType>
 class traded_field {
+  static_assert(base::system_details::CACHE_LINE_SIZE >= 4,
+                "Traded objects must not use their last address bits, as we use them for status flags."
+                "As traded objects are usually cache aligned, we need big enough cache lines.");
  // TODO: Replace unsigned long with a portable sized integer
  //       (some systems might have different pointer sizes to long sizes).
+  static constexpr unsigned long SHIFT = 0x2lu;
+  static constexpr unsigned long TAG_BITS = 0x3lu;
+  static constexpr unsigned long RELEVANT_BITS = ~TAG_BITS;
+  static constexpr unsigned long EMPTY_TAG = 0x0lu;
+  static constexpr unsigned long STAMP_TAG = 0x1lu;
+  static constexpr unsigned long TRADE_TAG = 0x2lu;
-  void fill_with_tag(unsigned long tag) {
+ public:
-    pointer_ = (void *) ((tag << 1lu) | 0x1lu);
+  void fill_with_stamp(unsigned long stamp) {
+    pointer_ = (void *) ((stamp << SHIFT) | STAMP_TAG);
  }
-  unsigned long get_tag() {
+  unsigned long get_stamp() {
    PLS_ASSERT(is_filled_with_tag(), "Must only read out the tag when the traded field contains one.");
-    return ((unsigned long) (pointer_)) >> 1lu;
+    return ((unsigned long) pointer_) >> SHIFT;
  }
  bool is_filled_with_tag() {
-    return ((unsigned long) (pointer_)) & 0x1lu;
+    return (((unsigned long) pointer_) & TAG_BITS) == STAMP_TAG;
  }
-  void fill_with_object(TradedType *object) {
+  void fill_with_trade_object(TradedType *trade_object) {
-    PLS_ASSERT((object & 0x1lu) == 0,
+    PLS_ASSERT((((unsigned long) trade_object) & TAG_BITS) == 0,
               "Must only store aligned objects in this data structure (last bits are needed for tag bit)");
-    pointer_ = object;
+    pointer_ = reinterpret_cast<void *>(((unsigned long) trade_object) | TRADE_TAG);
  }
-  TradedType *get_object() {
+  TradedType *get_trade_object() {
    PLS_ASSERT(is_filled_with_object(), "Must only read out the object when the traded field contains one.");
-    return pointer_;
+    return reinterpret_cast<TradedType *>(((unsigned long) pointer_) & RELEVANT_BITS);
  }
  bool is_filled_with_object() {
-    return !is_filled_with_tag() && pointer_ != nullptr;
+    return (((unsigned long) pointer_) & TAG_BITS) == TRADE_TAG;
+  }
+  bool is_empty() {
+    return (((unsigned long) pointer_) & TAG_BITS) == EMPTY_TAG;
  }
 private:
@@ -50,19 +64,21 @@ class traded_field {
 template<typename EntryType, typename TradedType>
 class alignas(base::system_details::CACHE_LINE_SIZE) trading_deque_entry {
+ public:
  /*
   * Fill the slot with its initial values, making it ready for being stolen.
   * Performs no synchronization/memory ordering constraints.
   *
   * Method is called to init a field on pushBot.
   */
-  void fill_slots(EntryType *entry_item, unsigned long tag) {
+  void fill_slots(EntryType *entry_item, unsigned long expected_stamp) {
    entry_slot_.store(entry_item, std::memory_order_relaxed);
+    forwarding_stamp_.store(expected_stamp, std::memory_order_relaxed);
    // Relaxed is fine for this, as adding elements is synced over the bot pointer
    auto old = trade_slot_.load(std::memory_order_relaxed);
-    old.fill_with_tag(tag);
+    old.fill_with_stamp(expected_stamp);
-    trade_slot_.store(std::memory_order_relaxed);
+    trade_slot_.store(old, std::memory_order_relaxed);
  }
  /**
@@ -78,7 +94,7 @@ class alignas(base::system_details::CACHE_LINE_SIZE) trading_deque_entry {
    if (old_field_value.is_filled_with_tag()) {
      return optional<TradedType *>();
    } else {
-      return optional<TradedType *>(old_field_value.get_object());
+      return optional<TradedType *>(old_field_value.get_trade_object());
    }
  }
@@ -86,25 +102,34 @@ class alignas(base::system_details::CACHE_LINE_SIZE) trading_deque_entry {
    return entry_slot_;
  }
-  optional<EntryType *> trade_object(TradedType *offered_object, unsigned long expected_tag) {
+  bool is_empty() {
+    return trade_slot_.load(std::memory_order_seq_cst).is_empty();
+  }
+  optional<EntryType *> trade_object(TradedType *offered_object, unsigned long &expected_stamp) {
    // Read our potential result
    EntryType *result = entry_slot_.load(std::memory_order_relaxed);
+    unsigned long forwarding_stamp = forwarding_stamp_.load(std::memory_order_relaxed);
    // Try to get it by CAS with the expected field entry, giving up our offered_object for it
    traded_field<TradedType> expected_field;
-    expected_field.fill_with_tag(expected_tag);
+    expected_field.fill_with_stamp(expected_stamp);
    traded_field<TradedType> offered_field;
-    offered_field.fill_with_object(offered_object);
+    offered_field.fill_with_trade_object(offered_object);
    if (trade_slot_.compare_exchange_strong(expected_field, offered_field, std::memory_order_acq_rel)) {
-      return optional<EntryType *>(result);
+      return optional<EntryType *>{result};
    } else {
-      return optional<EntryType *>(nullptr);
+      if (expected_field.is_empty()) {
+        expected_stamp = forwarding_stamp;
+      }
+      return optional<EntryType *>{};
    }
  }
 private:
  std::atomic<EntryType *> entry_slot_{nullptr};
+  std::atomic<unsigned long> forwarding_stamp_{};
  std::atomic<traded_field<TradedType>> trade_slot_{};
 };
@@ -132,9 +157,9 @@ class bounded_trading_deque {
  void push_bot(EntryType *offered_object) {
    auto expected_stamp = bot_internal_.stamp;
-    auto *current_entry = entries_[bot_internal_.value];
+    auto &current_entry = entries_[bot_internal_.value];
-    current_entry->fill_slots(offered_object, expected_stamp);
+    current_entry.fill_slots(offered_object, expected_stamp);
    bot_internal_.stamp++;
    bot_internal_.value++;
@@ -157,28 +182,63 @@ class bounded_trading_deque {
    // Go one step back
    bot_internal_.value--;
-    auto *current_entry = entries_[bot_internal_.value];
+    auto &current_entry = entries_[bot_internal_.value];
-    optional<TradedType *> traded_object = current_entry->acquire_traded_type();
+    optional<TradedType *> traded_object = current_entry.acquire_traded_type();
    optional<EntryType *> queue_entry;
    if (traded_object) {
      // We do not return an entry, but the traded object
-      queue_entry = {};
+      queue_entry = optional<EntryType *>{};
    } else {
      // We still got it locally, grab the object
-      queue_entry = {current_entry->get_object()};
+      queue_entry = optional<EntryType *>{current_entry.get_object()};
-      // Keep the tag up to date (we must re-use it, as the head is just increasing by steps of one from the beginning)
-      bot_internal_.stamp--;
    }
    bot_.store(bot_internal_.value, std::memory_order_relaxed);
+    if (bot_internal_.value == 0) {
+      bot_internal_.stamp++;
+      top_.store({bot_internal_.stamp, 0}, std::memory_order_release);
+    }
    return pop_result{queue_entry, traded_object};
  }
+  std::tuple<optional<EntryType *>, stamped_integer> peek_top() {
+    auto local_top = top_.load();
+    auto local_bot = bot_.load();
+    if (local_top.value >= local_bot) {
+      return std::make_tuple(optional<EntryType *>{}, local_top);
+    } else {
+      return std::make_tuple(optional<EntryType *>{entries_[local_top.value].get_object()}, local_top);
+    }
+  }
  optional<EntryType *> pop_top(TradedType *trade_offer) {
    auto local_top = top_.load();
-    optional<EntryType *> entry = entries_[local_top.value].trade_object(trade_offer, local_top.stamp);
+    return pop_top(trade_offer, local_top);
+  }
+  optional<EntryType *> pop_top(TradedType *trade_offer, stamped_integer local_top) {
+    auto local_bot = bot_.load();
+    if (local_top.value >= local_bot) {
+      return optional<EntryType *>{};
+    }
+    unsigned long expected_top_stamp = local_top.stamp;
+    optional<EntryType *> entry = entries_[local_top.value].trade_object(trade_offer, expected_top_stamp);
+    if (entry) {
+      // We got it, for sure move the top pointer forward.
+      top_.compare_exchange_strong(local_top, {local_top.stamp + 1, local_top.value + 1});
+    } else {
+      // We did not get it....
+      if (entries_[local_top.value].is_empty()) {
+        // ...update the top stamp, so the next call can get it (we still make system progress, as the owner
+        // must have popped off the element)
+        top_.compare_exchange_strong(local_top, {expected_top_stamp, local_top.value});
+      } else {
+        // ...move the pointer forward if someone else put a valid trade object in there.
        top_.compare_exchange_strong(local_top, {local_top.stamp + 1, local_top.value + 1});
+      }
+    }
    return entry;
  }
@@ -193,6 +253,17 @@ class bounded_trading_deque {
  stamped_integer bot_internal_{0, 0};
 };
+template<typename EntryType, typename TradedType, size_t SIZE>
+class static_bounded_trading_deque {
+ public:
+  static_bounded_trading_deque() : items_{}, deque_{items_.data(), SIZE} {}
+  bounded_trading_deque<EntryType, TradedType> &get_deque() { return deque_; }
+ private:
+  std::array<trading_deque_entry<EntryType, TradedType>, SIZE> items_;
+  bounded_trading_deque<EntryType, TradedType> deque_;
+};
 }
 }
 }

--- a/lib/pls/include/pls/internal/data_structures/delayed_initialization.h
+++ b/lib/pls/include/pls/internal/data_structures/delayed_initialization.h
@@ -19,15 +19,38 @@ namespace data_structures {
 * Takes care of the de-construction the contained object if one is active.
 */
 template<typename T>
-class delayed_initialization {
+class alignas(alignof(T)) delayed_initialization {
 public:
  delayed_initialization() : memory_{}, initialized_{false} {}
+  delayed_initialization(const delayed_initialization &) = delete;
  delayed_initialization(delayed_initialization &&other) noexcept {
-    initialized_ = other.initialized_;
+    if (other.initialized()) {
-    if (other.initialized_) {
+      new((void *) memory_.data()) T(std::move(other.object()));
+      other.initialized_ = false;
+      initialized_ = true;
+    }
+  }
+  delayed_initialization &operator=(const delayed_initialization &) = delete;
+  delayed_initialization &operator=(delayed_initialization &&other) noexcept {
+    if (&other == this) {
+      return *this;
+    }
+    if (initialized() && other.initialized()) {
      object() = std::move(other.object());
      other.initialized_ = false;
+      initialized_ = true;
+      return *this;
    }
+    if (!initialized() && other.initialized_) {
+      new((void *) memory_.data()) T(std::move(other.object()));
+      other.initialized_ = false;
+      initialized_ = true;
+      return *this;
+    }
+    return *this;
  }
  template<typename ...ARGS>
@@ -62,14 +85,24 @@ class delayed_initialization {
    return *reinterpret_cast<T *>(memory_.data());
  }
+  const T &object() const {
+    PLS_ASSERT(initialized_, "Can not use an uninitialized delayed wrapper object!");
+    return *reinterpret_cast<const T *>(memory_.data());
+  }
  T &operator*() {
    return object();
  }
+  const T &operator*() const {
+    return object();
+  }
  bool initialized() const { return initialized_; }
 private:
-  std::array<char, sizeof(T)> memory_;
+  alignas(alignof(T)) std::array<char, sizeof(T)> memory_;
  bool initialized_;
 };

--- a/lib/pls/include/pls/internal/data_structures/optional.h
+++ b/lib/pls/include/pls/internal/data_structures/optional.h
@@ -3,6 +3,7 @@
 #define PLS_INTERNAL_DATA_STRUCTURES_OPTIONAL_H_
 #include <utility>
+#include <type_traits>
 #include "pls/internal/data_structures/delayed_initialization.h"
@@ -14,6 +15,38 @@ template<typename T>
 class optional {
 public:
  optional() = default;
+  optional(optional &other) noexcept : optional(const_cast<const optional &>(other)) {};
+  optional(const optional &other) noexcept {
+    if (other) {
+      data_.initialize(other.data_.object());
+    }
+  }
+  optional(optional &&other) noexcept {
+    data_ = std::move(other.data_);
+  }
+  optional &operator=(const optional &other) {
+    if (&other == this) {
+      return *this;
+    }
+    if (data_.initialized()) {
+      data_.destroy();
+    }
+    if (other) {
+      data_.initialize(other.data_.object());
+    }
+    return *this;
+  }
+  optional &operator=(optional &&other) noexcept {
+    if (&other == this) {
+      return *this;
+    }
+    data_ = std::move(other.data_);
+    return *this;
+  }
  template<typename ...ARGS>
  explicit optional(ARGS &&...args): data_{std::forward<ARGS>(args)...} {}

--- a/lib/pls/include/pls/internal/scheduling/cont.h
+++ b/lib/pls/include/pls/internal/scheduling/cont.h
@@ -46,6 +46,7 @@ class base_cont {
   * Will store the result in it's parent, but not mess with any counters.
   */
  virtual void execute_task() = 0;
+  virtual base_task *get_task() = 0;
  virtual void *get_right_result_pointer() = 0;
  virtual void *get_left_result_pointer() = 0;
@@ -120,14 +121,19 @@ class cont : public base_cont {
  void execute() override {
    using result_type = decltype(function_((*left_result_).value(), (*right_result_).value()));
    result_runner<result_type>::execute(*this);
-    this->get_memory_block()->free_buffer();
-    this->get_memory_block()->reset_state();
    this->~cont();
+    auto *memory_block = this->get_memory_block();
+    memory_block->free_buffer();
+    memory_block->reset_state();
  }
  void execute_task() override {
    task_.execute();
  }
+  base_task *get_task() override {
+    return &task_;
+  }
  void *get_left_result_pointer() override {
    return &left_result_;
@@ -136,10 +142,6 @@ class cont : public base_cont {
    return &right_result_;
  }
-  T2 *get_task() {
-    return &task_;
-  }
 private:
  // Initial data members. These slow down the fast path, try to init them lazy when possible.
  F function_;

--- a/lib/pls/include/pls/internal/scheduling/cont_manager.h
+++ b/lib/pls/include/pls/internal/scheduling/cont_manager.h
@@ -63,6 +63,25 @@ class cont_manager {
    return active_node_;
  }
+  bool is_clean() {
+    if (get_active_node()->get_depth() == 0) {
+      memory_block *current_node = active_node_;
+      for (size_t i = 1; i < num_conts_; i++) {
+        if (current_node->get_prev() != nullptr && current_node->get_prev()->get_next() != current_node) {
+          return false;
+        }
+        if (current_node->is_buffer_used()) {
+          return false;
+        }
+        current_node = current_node->get_next();
+      }
+    } else {
+      return false;
+    }
+    return true;
+  }
  // Manage the fall through behaviour/slow path behaviour
  bool falling_through() const {
    return fall_through_;
@@ -93,23 +112,24 @@ class cont_manager {
    fall_through_ = false;
    // Keep the target chain before we execute, as this potentially frees the memory
-    auto *target_chain = notified_cont->get_memory_block()->get_offered_chain().load();
+    auto *target_memory_block = notified_cont->get_memory_block();
+    auto *target_chain = target_memory_block->get_offered_chain().load();
    // Notify the next continuation of finishing a child...
-    if (notified_cont->get_memory_block()->get_results_missing().fetch_add(-1) == 1) {
+    if (target_memory_block->get_results_missing().fetch_add(-1) == 1) {
      // ... we finished the continuation.
      // We are now in charge continuing to execute the above continuation chain.
-      PLS_ASSERT(active_node_->get_prev()->get_depth() == notified_cont->get_memory_block()->get_depth(),
+      PLS_ASSERT(active_node_->get_prev()->get_depth() == target_memory_block->get_depth(),
                 "We must hold the system invariant to be in the correct depth.")
-      if (active_node_->get_prev() != notified_cont->get_memory_block()) {
+      if (active_node_->get_prev() != target_memory_block) {
        // We do not own the thing we will execute.
        // Own it by swapping the chain belonging to it in.
-        aquire_memory_chain(notified_cont->get_memory_block());
+        aquire_memory_chain(target_memory_block);
      }
      my_state.parent_cont_ = notified_cont->get_parent();
      my_state.right_spawn_ = notified_cont->is_right_child();
-      active_node_ = notified_cont->get_memory_block();
+      active_node_ = target_memory_block;
      notified_cont->execute();
      if (!falling_through() && notified_cont->get_parent() != nullptr) {
        fall_through_and_notify_cont(notified_cont->get_parent(), notified_cont->is_right_child());
@@ -119,9 +139,9 @@ class cont_manager {
      // ... we did not finish the last continuation.
      // We are no longer in charge of executing the above continuation chain.
-      PLS_ASSERT(active_node_->get_prev()->get_depth() == notified_cont->get_memory_block()->get_depth(),
+      PLS_ASSERT(active_node_->get_prev()->get_depth() == target_memory_block->get_depth(),
                 "We must hold the system invariant to be in the correct depth.")
-      if (active_node_->get_prev() == notified_cont->get_memory_block()) {
+      if (active_node_->get_prev() == target_memory_block) {
        // We own the thing we are not allowed to execute.
        // Get rid of the ownership by using the offered chain.
        aquire_memory_chain(target_chain);

--- a/lib/pls/include/pls/internal/scheduling/memory_block.h
+++ b/lib/pls/include/pls/internal/scheduling/memory_block.h
@@ -23,19 +23,22 @@ class memory_block {
      : prev_{prev},
        next_{nullptr},
        offered_chain_{nullptr},
-        state_{{initialized}},
        results_missing_{2},
        memory_buffer_{memory_buffer},
        memory_buffer_size_{memory_buffer_size},
        memory_buffer_used_{false},
-        depth_{depth} {};
+        depth_{depth},
+        owner_{0} {};
  template<typename T, typename ...ARGS>
  T *place_in_buffer(ARGS &&...args) {
    PLS_ASSERT(!memory_buffer_used_, "Must only allocate one continuation at once per node.");
    memory_buffer_used_ = true;
-    return new(memory_buffer_) T(std::forward<ARGS>(args)...);
+    auto *result = new(memory_buffer_) T(std::forward<ARGS>(args)...);
+    continuation_ = result;
+    return result;
  }
  void free_buffer() {
    PLS_ASSERT(memory_buffer_used_, "Can only free a memory spot when it is in use.")
@@ -44,14 +47,10 @@ class memory_block {
  bool is_buffer_used() {
    return memory_buffer_used_;
  }
+  base_cont *get_cont() {
-  // TODO: Fit the reset somewhere!!!
+    PLS_ASSERT(is_buffer_used(), "Can only read initialized buffer!");
-//    // Reset Associated counters
+    return continuation_;
-//    results_missing_.store(2);
+  }
-//    offered_chain_.store(nullptr);
-//    auto old_state = state_.load();
-//    state_.store({old_state.stamp + 1, initialized});
  memory_block *get_prev() {
    return prev_;
@@ -66,13 +65,6 @@ class memory_block {
    next_ = next;
  }
-  enum state { initialized, execute_local, stealing, stolen, invalid };
-  using stamped_state = data_structures::stamped_integer;
-  std::atomic<stamped_state> &get_state() {
-    return state_;
-  }
  std::atomic<memory_block *> &get_offered_chain() {
    return offered_chain_;
  }
@@ -87,11 +79,16 @@ class memory_block {
  void reset_state() {
    offered_chain_.store(nullptr);
-    auto old_state = state_.load();
-    state_.store({old_state.stamp + 1, initialized});
    results_missing_.store(2);
  }
+  void set_owner(int owner) {
+    owner_ = owner;
+  }
+  int get_owner() {
+    return owner_;
+  }
 private:
  // Linked list property of memory blocks (a complete list represents a threads currently owned memory).
  // Each block knows its chain start to allow stealing a whole chain in O(1)
@@ -103,13 +100,6 @@ class memory_block {
  // For this we need the offered chain's element up to the point we can steal.
  std::atomic<memory_block *> offered_chain_;
-  // The flag is needed for an ongoing stealing request.
-  // Stealing threads need to offer their memory block chain before the
-  // 'fully' own the stolen task. As long as that is not done the memory block
-  // chain can abort the steal request in order to be not blocked without a
-  // new, clean memory block chain to work with.
-  std::atomic<stamped_state> state_;
  // Management for coordinating concurrent result writing and stealing.
  // The result count decides atomically who gets to execute the continuation
  // and who therefore get's to own this memory block chain.
@@ -120,6 +110,8 @@ class memory_block {
  // This memory is managed explicitly by the continuation manager and runtime system
  // (they need to make sure to always call de-constructors and never allocate two continuations).
  char *memory_buffer_;
+  base_cont *continuation_;
  // These two are only helper properties helping with bugs during development.
  size_t memory_buffer_size_;
  bool memory_buffer_used_;
@@ -128,6 +120,9 @@ class memory_block {
  // Swapping parts of a memory chain will not reorder it, as always parts of
  // the same size are exchanged.
  const int depth_;
+  // TODO: Remove, debug only
+  int owner_;
 };
 }

--- a/lib/pls/include/pls/internal/scheduling/scheduler_impl.h
+++ b/lib/pls/include/pls/internal/scheduling/scheduler_impl.h
@@ -49,6 +49,7 @@ struct scheduler::starter {
    const bool is_right_cont = my_state.right_spawn_;
    base_cont *parent_cont = my_state.parent_cont_;
+    current_memory_block->set_owner(my_state.get_id());
    continuation_type *current_cont = current_memory_block->place_in_buffer<continuation_type>(parent_cont,
                                                                                               current_memory_block,
                                                                                               is_right_cont,
@@ -63,19 +64,32 @@ struct scheduler::starter {
    my_state.right_spawn_ = false;
    return_type_1 result_1 = function_1_();
    if (cont_manager.falling_through()) {
+      // Get our replacement from the task stack and store it for later use when we are actually blocked.
+      auto traded_memory = my_state.get_task_manager().try_pop_local();
+      current_cont->get_memory_block()->get_offered_chain().store(*traded_memory);
      // Unwind stack...
      return result_type{};
    }
    // Try to call second function on fast path
-    if (my_state.get_task_manager().steal_local_task()) {
+    auto traded_memory = my_state.get_task_manager().try_pop_local();
+    if (traded_memory) {
+      // The task got stolen...get_memory_block
+      // ...but we got a memory block that can be used if we block on this one.
+      current_cont->get_memory_block()->get_offered_chain().store(*traded_memory);
+      // Main scheduling loop is responsible for entering the result to the slow path...
+      current_cont->store_left_result(std::move(result_1));
+      cont_manager.fall_through_and_notify_cont(current_cont, false);
+      // Unwind stack...
+      return result_type{};
+    } else {
      my_state.right_spawn_ = true;
      return_type_2 result_2 = function_2_();
      if (cont_manager.falling_through()) {
        // Main scheduling loop is responsible for entering the result to the slow path...
        current_cont->store_left_result(std::move(result_1));
-        auto old_state = current_cont->get_memory_block()->get_state().load();
-        current_cont->get_memory_block()->get_state().store({old_state.stamp + 1, memory_block::invalid});
        current_cont->get_memory_block()->get_results_missing().fetch_add(-1);
        // Unwind stack...
        return result_type{};
@@ -101,12 +115,6 @@ struct scheduler::starter {
      }
      return cont_result;
    }
-    // Main scheduling loop is responsible for entering the result to the slow path...
-    current_cont->store_left_result(std::move(result_1));
-    cont_manager.fall_through_and_notify_cont(current_cont, false);
-    // Unwind stack...
-    return result_type{};
  };
 };

--- a/lib/pls/include/pls/internal/scheduling/scheduler_memory.h
+++ b/lib/pls/include/pls/internal/scheduling/scheduler_memory.h
@@ -28,7 +28,7 @@ class scheduler_memory {
  virtual thread_state &thread_state_for(size_t id) = 0;
 };
-template<size_t MAX_THREADS, size_t NUM_TASKS, size_t MAX_TASK_STACK_SIZE, size_t NUM_CONTS, size_t MAX_CONT_SIZE>
+template<size_t MAX_THREADS, size_t NUM_TASKS, size_t NUM_CONTS, size_t MAX_CONT_SIZE>
 class static_scheduler_memory : public scheduler_memory {
 public:
  size_t max_threads() const override {
@@ -44,7 +44,7 @@ class static_scheduler_memory : public scheduler_memory {
  }
 private:
-  using thread_state_type = thread_state_static<NUM_TASKS, MAX_TASK_STACK_SIZE, NUM_CONTS, MAX_CONT_SIZE>;
+  using thread_state_type = thread_state_static<NUM_TASKS, NUM_CONTS, MAX_CONT_SIZE>;
  alignas(base::system_details::CACHE_LINE_SIZE) std::array<base::thread, MAX_THREADS> threads_;
  alignas(base::system_details::CACHE_LINE_SIZE) std::array<thread_state_type, MAX_THREADS> thread_states_;
@@ -78,7 +78,7 @@ class heap_scheduler_memory : public scheduler_memory {
  }
 private:
-  using thread_state_type = thread_state_static<NUM_TASKS, MAX_TASK_STACK_SIZE, NUM_CONTS, MAX_CONT_SIZE>;
+  using thread_state_type = thread_state_static<NUM_TASKS, NUM_CONTS, MAX_CONT_SIZE>;
  // thread_state_type is aligned at the cache line and therefore overaligned (C++ 11 does not require
  // the new operator to obey alignments bigger than 16, cache lines are usually 64).
  // To allow this object to be allocated using 'new' (which the vector does internally),

--- a/lib/pls/include/pls/internal/scheduling/task_manager.h
+++ b/lib/pls/include/pls/internal/scheduling/task_manager.h
@@ -12,8 +12,9 @@
 #include "pls/internal/scheduling/cont_manager.h"
 #include "pls/internal/scheduling/memory_block.h"
-#include "pls/internal/data_structures/bounded_ws_deque.h"
+#include "pls/internal/data_structures/bounded_trading_deque.h"
 #include "pls/internal/data_structures/stamped_integer.h"
+#include "pls/internal/data_structures/optional.h"
 #include "pls/internal/base/spin_lock.h"
@@ -21,19 +22,6 @@ namespace pls {
 namespace internal {
 namespace scheduling {
-struct task_handle {
- public:
-  task_handle() : task_{nullptr}, task_memory_block_{nullptr} {};
-  explicit task_handle(base_task *task) : task_{task},
-                                          task_memory_block_{task->get_cont()->get_memory_block()} {};
-  base_task *task_;
-  // This seems redundant first, but is needed for a race-free steal.
-  // It could happen that the task's memory is overwritten and the pointer to it's memory block gets invalid.
-  // We can do this more elegantly in the future.
-  memory_block *task_memory_block_;
-};
 /**
 * Handles management of tasks in the system. Each thread has a local task manager,
 * responsible for allocating, freeing and publishing tasks for stealing.
@@ -42,60 +30,60 @@ class task_manager {
 public:
  // Publishes a task on the stack, i.e. makes it visible for other threads to steal.
  void publish_task(base_task *task) {
-    std::lock_guard<base::spin_lock> lock{lock_};
+//    std::lock_guard<base::spin_lock> lock{lock_};
-    task_deque_.push_bottom(task_handle{task});
+    task_deque_.push_bot(task->get_cont()->get_memory_block());
  }
  // Try to pop a local task from this task managers stack.
-  bool steal_local_task() {
+  data_structures::optional<memory_block *> try_pop_local() {
-    std::lock_guard<base::spin_lock> lock{lock_};
+//    std::lock_guard<base::spin_lock> lock{lock_};
-    return task_deque_.pop_bottom();
+    return task_deque_.pop_bot().traded_;
  }
  // Try to steal a task from a remote task_manager instance. The stolen task must be stored locally.
  // Returns a pair containing the actual task and if the steal was successful.
  base_task *steal_remote_task(cont_manager &stealing_cont_manager) {
-    std::lock_guard<base::spin_lock> lock{lock_};
+//    std::lock_guard<base::spin_lock> lock{lock_};
+    auto peek = task_deque_.peek_top();
-    // TODO: See if we can somehow make this trade lock free (and still be correct)
-    auto stolen_task_handle = task_deque_.pop_top();
+    if (std::get<0>(peek)) {
-    if (stolen_task_handle) {
+      memory_block *peeked_memory_block = (*std::get<0>(peek));
-      base_task *stolen_task = (*stolen_task_handle).task_;
+      auto peeked_depth = peeked_memory_block->get_depth();
-      memory_block *stolen_task_memory = (*stolen_task_handle).task_memory_block_;
-      auto stolen_task_depth = stolen_task_memory->get_depth();
-      auto &atomic_state = stolen_task_memory->get_state();
-      auto &atomic_offered_chain = stolen_task_memory->get_offered_chain();
-      // TODO: We ignore all we tried with lock free implementations here, just store the state how it is supposed to be
+      stealing_cont_manager.move_active_node(peeked_depth);
-      stealing_cont_manager.move_active_node(stolen_task_depth);
      auto offered_chain = stealing_cont_manager.get_active_node();
      stealing_cont_manager.move_active_node(1);
-      atomic_offered_chain.store(offered_chain);
-      atomic_state.store(memory_block::stolen);
-      return stolen_task;
+      auto stolen_memory_block = task_deque_.pop_top(offered_chain, std::get<1>(peek));
+      if (stolen_memory_block) {
+        PLS_ASSERT(*stolen_memory_block == peeked_memory_block, "Steal must only work if it is equal!");
+        return (*stolen_memory_block)->get_cont()->get_task();
      } else {
+        stealing_cont_manager.move_active_node(-(peeked_depth + 1));
        return nullptr;
      }
    }
-  explicit task_manager(data_structures::bounded_ws_deque<task_handle> &task_deque) : task_deque_{task_deque},
+    return nullptr;
-                                                                                      lock_{} {}
+  }
+  explicit task_manager(data_structures::bounded_trading_deque<memory_block, memory_block> &task_deque) :
+      task_deque_{task_deque} {}
 private:
-  data_structures::bounded_ws_deque<task_handle> &task_deque_;
+  data_structures::bounded_trading_deque<memory_block, memory_block> &task_deque_;
-  base::spin_lock lock_;
+  base::spin_lock lock_{};
 };
-template<size_t NUM_TASKS, size_t MAX_STACK_SIZE>
+template<size_t NUM_TASKS>
 class static_task_manager {
 public:
  static_task_manager() : task_deque_{}, task_manager_{task_deque_.get_deque()} {};
  task_manager &get_task_manager() { return task_manager_; }
 private:
-  data_structures::static_bounded_ws_deque<task_handle, NUM_TASKS> task_deque_;
+  data_structures::static_bounded_trading_deque<memory_block, memory_block, NUM_TASKS> task_deque_;
  task_manager task_manager_;
 };

--- a/lib/pls/include/pls/internal/scheduling/thread_state_static.h
+++ b/lib/pls/include/pls/internal/scheduling/thread_state_static.h
@@ -11,7 +11,7 @@ namespace pls {
 namespace internal {
 namespace scheduling {
-template<size_t NUM_TASKS, size_t MAX_TASK_STACK_SIZE, size_t NUM_CONTS, size_t MAX_CONT_SIZE>
+template<size_t NUM_TASKS, size_t NUM_CONTS, size_t MAX_CONT_SIZE>
 struct thread_state_static {
 public:
  thread_state_static()
@@ -21,7 +21,7 @@ struct thread_state_static {
  thread_state &get_thread_state() { return thread_state_; }
 private:
-  static_task_manager<NUM_TASKS, MAX_TASK_STACK_SIZE> static_task_manager_;
+  static_task_manager<NUM_TASKS> static_task_manager_;
  static_cont_manager<NUM_CONTS, MAX_CONT_SIZE> static_cont_manager_;
  thread_state thread_state_;
 };

--- a/lib/pls/src/internal/scheduling/scheduler.cpp
+++ b/lib/pls/src/internal/scheduling/scheduler.cpp
@@ -83,7 +83,7 @@ void scheduler::work_thread_work_section() {
      auto &target_state = my_state.get_scheduler().thread_state_for(target);
-      PLS_ASSERT(my_cont_manager.get_active_node()->get_depth() == 0, "Only steal with clean chain!");
+      PLS_ASSERT(my_cont_manager.is_clean(), "Only steal with clean chain!");
      auto *stolen_task = target_state.get_task_manager().steal_remote_task(my_cont_manager);
      if (stolen_task != nullptr) {
        my_state.parent_cont_ = stolen_task->get_cont();
@@ -99,6 +99,7 @@ void scheduler::work_thread_work_section() {
    }
  } while (!work_section_done_);
+  PLS_ASSERT(my_cont_manager.is_clean(), "Only finish work section with clean chain!");
 }
 void scheduler::terminate() {

--- a/test/data_structures_test.cpp
+++ b/test/data_structures_test.cpp
@@ -3,6 +3,7 @@
 #include "pls/internal/base/system_details.h"
 #include "pls/internal/data_structures/aligned_stack.h"
+#include "pls/internal/data_structures/bounded_trading_deque.h"
 using namespace pls::internal::data_structures;
 using namespace pls::internal::base;