src/mixed_arena.h

/*
 * Copyright 2015 WebAssembly Community Group participants
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */

#ifndef wasm_mixed_arena_h
#define wasm_mixed_arena_h

#include <atomic>
#include <cassert>
#include <memory>
#include <mutex>
#include <thread>
#include <type_traits>
#include <vector>

#include <support/alloc.h>

//
// Arena allocation for mixed-type data.
//
// Arena-style bump allocation is important for two reasons: First, so that
// allocation is quick, and second, so that allocated items are close together,
// which is cache-friendy. Arena allocation is also useful for a minor third
// reason which is to make freeing all the items in an arena very quick.
//
// Each WebAssembly Module has an arena allocator, which should be used
// for all of its AST nodes and so forth. When the Module is destroyed, the
// entire arena is cleaned up.
//
// When allocating an object in an arena, the object's proper constructor
// is called. Note that destructors are not called, because to make the
// arena simple and fast we do not track internal allocations inside it
// (and we can also avoid the need for virtual destructors).
//
// In general, optimization passes avoid allocation as much as possible.
// Many passes only remove or modify nodes anyhow, others can often
// reuse nodes that are being optimized out. This keeps things
// cache-friendly, and also makes the operations trivially thread-safe.
// In the rare case that a pass does need to allocate, and it is a
// parallel pass (so multiple threads might access the allocator),
// the MixedArena instance will notice if it is on a different thread
// than that arena's original thread, and will perform the allocation
// in a side arena for that other thread. This is done in a transparent
// way to the outside; as a result, it is always safe to allocate using
// a MixedArena, no matter which thread you are on. Allocations will
// of course be fastest on the original thread for the arena.
//

struct MixedArena {
  // fast bump allocation

  static const size_t CHUNK_SIZE = 32768;
  static const size_t MAX_ALIGN = 16; // allow 128bit SIMD

  // Each pointer in chunks is to a multiple of CHUNK_SIZE - typically 1,
  // but possibly more.
  std::vector<void*> chunks;

  size_t index = 0; // in last chunk

  std::thread::id threadId;

  // multithreaded allocation - each arena is valid on a specific thread.
  // if we are on the wrong thread, we atomically look in the linked
  // list of next, adding an allocator if necessary
  std::atomic<MixedArena*> next;

  MixedArena() {
    threadId = std::this_thread::get_id();
    next.store(nullptr);
  }

  // Allocate an amount of space with a guaranteed alignment
  void* allocSpace(size_t size, size_t align) {
    // the bump allocator data should not be modified by multiple threads at
    // once.
    auto myId = std::this_thread::get_id();
    if (myId != threadId) {
      MixedArena* curr = this;
      MixedArena* allocated = nullptr;
      while (myId != curr->threadId) {
        auto seen = curr->next.load();
        if (seen) {
          curr = seen;
          continue;
        }
        // there is a nullptr for next, so we may be able to place a new
        // allocator for us there. but carefully, as others may do so as
        // well. we may waste a few allocations here, but it doesn't matter
        // as this can only happen as the chain is built up, i.e.,
        // O(# of cores) per allocator, and our allocatrs are long-lived.
        if (!allocated) {
          allocated = new MixedArena(); // has our thread id
        }
        if (curr->next.compare_exchange_strong(seen, allocated)) {
          // we replaced it, so we are the next in the chain
          // we can forget about allocated, it is owned by the chain now
          allocated = nullptr;
          break;
        }
        // otherwise, the cmpxchg updated seen, and we continue to loop
        curr = seen;
      }
      if (allocated) {
        delete allocated;
      }
      return curr->allocSpace(size, align);
    }
    // First, move the current index in the last chunk to an aligned position.
    index = (index + align - 1) & (-align);
    if (index + size > CHUNK_SIZE || chunks.size() == 0) {
      // Allocate a new chunk.
      auto numChunks = (size + CHUNK_SIZE - 1) / CHUNK_SIZE;
      assert(size <= numChunks * CHUNK_SIZE);
      auto* allocation =
        wasm::aligned_malloc(MAX_ALIGN, numChunks * CHUNK_SIZE);
      if (!allocation) {
        abort();
      }
      chunks.push_back(allocation);
      index = 0;
    }
    uint8_t* ret = static_cast<uint8_t*>(chunks.back());
    ret += index;
    index += size; // TODO: if we allocated more than 1 chunk, reuse the
                   // remainder, right now we allocate another next time
    return static_cast<void*>(ret);
  }

  template<class T> T* alloc() {
    static_assert(alignof(T) <= MAX_ALIGN,
                  "maximum alignment not large enough");
    auto* ret = static_cast<T*>(allocSpace(sizeof(T), alignof(T)));
    new (ret) T(*this); // allocated objects receive the allocator, so they can
                        // allocate more later if necessary
    return ret;
  }

  void clear() {
    for (auto* chunk : chunks) {
      wasm::aligned_free(chunk);
    }
    chunks.clear();
  }

  ~MixedArena() {
    clear();
    if (next.load()) {
      delete next.load();
    }
  }
};

//
// A vector that allocates in an arena.
//
// TODO: specialize on the initial size of the array

template<typename SubType, typename T> class ArenaVectorBase {
protected:
  T* data = nullptr;
  size_t usedElements = 0, allocatedElements = 0;

  void reallocate(size_t size) {
    T* old = data;
    static_cast<SubType*>(this)->allocate(size);
    for (size_t i = 0; i < usedElements; i++) {
      data[i] = old[i];
    }
  }

public:
  struct Iterator;

  T& operator[](size_t index) const {
    assert(index < usedElements);
    return data[index];
  }

  size_t size() const { return usedElements; }

  bool empty() const { return size() == 0; }

  void resize(size_t size) {
    if (size > allocatedElements) {
      reallocate(size);
    }
    // construct new elements
    for (size_t i = usedElements; i < size; i++) {
      new (data + i) T();
    }
    usedElements = size;
  }

  T& back() const {
    assert(usedElements > 0);
    return data[usedElements - 1];
  }

  T& pop_back() {
    assert(usedElements > 0);
    usedElements--;
    return data[usedElements];
  }

  void push_back(T item) {
    if (usedElements == allocatedElements) {
      reallocate((allocatedElements + 1) * 2); // TODO: optimize
    }
    data[usedElements] = item;
    usedElements++;
  }

  T& front() const {
    assert(usedElements > 0);
    return data[0];
  }

  void erase(Iterator start_it, Iterator end_it) {
    assert(start_it.parent == end_it.parent && start_it.parent == this);
    assert(start_it.index <= end_it.index && end_it.index <= usedElements);
    size_t size = end_it.index - start_it.index;
    for (size_t cur = start_it.index; cur + size < usedElements; ++cur) {
      data[cur] = data[cur + size];
    }
    usedElements -= size;
  }

  void erase(Iterator it) { erase(it, it + 1); }

  void clear() { usedElements = 0; }

  void reserve(size_t size) {
    if (size > allocatedElements) {
      reallocate(size);
    }
  }

  template<typename ListType> void set(const ListType& list) {
    size_t size = list.size();
    if (allocatedElements < size) {
      static_cast<SubType*>(this)->allocate(size);
    }
    for (size_t i = 0; i < size; i++) {
      data[i] = list[i];
    }
    usedElements = size;
  }

  void operator=(SubType& other) { set(other); }

  void swap(SubType& other) {
    data = other.data;
    usedElements = other.usedElements;
    allocatedElements = other.allocatedElements;

    other.data = nullptr;
    other.usedElements = other.allocatedElements = 0;
  }

  // iteration

  struct Iterator {
    using iterator_category = std::random_access_iterator_tag;
    using value_type = T;
    using difference_type = std::ptrdiff_t;
    using pointer = T*;
    using reference = T&;

    const SubType* parent;
    size_t index;

    Iterator() : parent(nullptr), index(0) {}
    Iterator(const SubType* parent, size_t index)
      : parent(parent), index(index) {}

    bool operator==(const Iterator& other) const {
      return index == other.index && parent == other.parent;
    }

    bool operator!=(const Iterator& other) const { return !(*this == other); }

    bool operator<(const Iterator& other) const {
      assert(parent == other.parent);
      return index < other.index;
    }

    bool operator>(const Iterator& other) const { return other < *this; }

    bool operator<=(const Iterator& other) const { return !(other < *this); }

    bool operator>=(const Iterator& other) const { return !(*this < other); }

    Iterator& operator++() {
      index++;
      return *this;
    }

    Iterator& operator--() {
      index--;
      return *this;
    }

    Iterator operator++(int) {
      Iterator it = *this;
      ++*this;
      return it;
    }

    Iterator operator--(int) {
      Iterator it = *this;
      --*this;
      return it;
    }

    Iterator& operator+=(std::ptrdiff_t off) {
      index += off;
      return *this;
    }

    Iterator& operator-=(std::ptrdiff_t off) { return *this += -off; }

    Iterator operator+(std::ptrdiff_t off) const {
      return Iterator(*this) += off;
    }

    Iterator operator-(std::ptrdiff_t off) const { return *this + -off; }

    std::ptrdiff_t operator-(const Iterator& other) const {
      assert(parent == other.parent);
      return index - other.index;
    }

    friend Iterator operator+(std::ptrdiff_t off, const Iterator& it) {
      return it + off;
    }

    T& operator*() const { return (*parent)[index]; }

    T& operator[](std::ptrdiff_t off) const { return (*parent)[index + off]; }

    T* operator->() const { return &(*parent)[index]; }
  };

  Iterator begin() const {
    return Iterator(static_cast<const SubType*>(this), 0);
  }
  Iterator end() const {
    return Iterator(static_cast<const SubType*>(this), usedElements);
  }

  void allocate(size_t size) {
    abort(); // must be implemented in children
  }

  // C-API

  void insertAt(size_t index, T item) {
    assert(index <= size()); // appending is ok
    resize(size() + 1);
    for (auto i = size() - 1; i > index; --i) {
      data[i] = data[i - 1];
    }
    data[index] = item;
  }

  T removeAt(size_t index) {
    assert(index < size());
    auto item = data[index];
    for (auto i = index; i < size() - 1; ++i) {
      data[i] = data[i + 1];
    }
    resize(size() - 1);
    return item;
  }
};

// A vector that has an allocator for arena allocation
//
// TODO: consider not saving the allocator, but requiring it be
//       passed in when needed, would make this (and thus Blocks etc.
//       smaller)

template<typename T>
class ArenaVector : public ArenaVectorBase<ArenaVector<T>, T> {
private:
  MixedArena& allocator;

public:
  ArenaVector(MixedArena& allocator) : allocator(allocator) {}

  ArenaVector(ArenaVector<T>&& other) : allocator(other.allocator) {
    *this = other;
  }

  void allocate(size_t size) {
    this->allocatedElements = size;
    this->data = static_cast<T*>(
      allocator.allocSpace(sizeof(T) * this->allocatedElements, alignof(T)));
  }
};

#endif // wasm_mixed_arena_h