shadPS4/src/common/ringbuffer.h

// SPDX-FileCopyrightText: Copyright 2016 Mozilla Foundation
// SPDX-License-Identifier: ISC

#pragma once

#include <algorithm>
#include <atomic>
#include <cstdint>
#include <memory>
#include <thread>
#include "common/assert.h"

/**
 * Single producer single consumer lock-free and wait-free ring buffer.
 *
 * This data structure allows producing data from one thread, and consuming it
 * on another thread, safely and without explicit synchronization. If used on
 * two threads, this data structure uses atomics for thread safety. It is
 * possible to disable the use of atomics at compile time and only use this data
 * structure on one thread.
 *
 * The role for the producer and the consumer must be constant, i.e., the
 * producer should always be on one thread and the consumer should always be on
 * another thread.
 *
 * Some words about the inner workings of this class:
 * - Capacity is fixed. Only one allocation is performed, in the constructor.
 *   When reading and writing, the return value of the method allows checking if
 *   the ring buffer is empty or full.
 * - We always keep the read index at least one element ahead of the write
 *   index, so we can distinguish between an empty and a full ring buffer: an
 *   empty ring buffer is when the write index is at the same position as the
 *   read index. A full buffer is when the write index is exactly one position
 *   before the read index.
 * - We synchronize updates to the read index after having read the data, and
 *   the write index after having written the data. This means that the each
 *   thread can only touch a portion of the buffer that is not touched by the
 *   other thread.
 * - Callers are expected to provide buffers. When writing to the queue,
 *   elements are copied into the internal storage from the buffer passed in.
 *   When reading from the queue, the user is expected to provide a buffer.
 *   Because this is a ring buffer, data might not be contiguous in memory,
 *   providing an external buffer to copy into is an easy way to have linear
 *   data for further processing.
 */
template <typename T>
class RingBuffer {
public:
    /**
     * Constructor for a ring buffer.
     *
     * This performs an allocation, but is the only allocation that will happen
     * for the life time of a `RingBuffer`.
     *
     * @param capacity The maximum number of element this ring buffer will hold.
     */
    RingBuffer(int capacity)
        /* One more element to distinguish from empty and full buffer. */
        : capacity_(capacity + 1) {
        ASSERT(storage_capacity() < std::numeric_limits<int>::max() / 2 &&
               "buffer too large for the type of index used.");
        ASSERT(capacity_ > 0);

        data_.reset(new T[storage_capacity()]);
        /* If this queue is using atomics, initializing those members as the last
         * action in the constructor acts as a full barrier, and allow capacity() to
         * be thread-safe. */
        write_index_ = 0;
        read_index_ = 0;
    }
    /**
     * Push `count` zero or default constructed elements in the array.
     *
     * Only safely called on the producer thread.
     *
     * @param count The number of elements to enqueue.
     * @return The number of element enqueued.
     */
    int enqueue_default(int count) {
        return enqueue(nullptr, count);
    }
    /**
     * @brief Put an element in the queue
     *
     * Only safely called on the producer thread.
     *
     * @param element The element to put in the queue.
     *
     * @return 1 if the element was inserted, 0 otherwise.
     */
    int enqueue(T& element) {
        return enqueue(&element, 1);
    }
    /**
     * Push `count` elements in the ring buffer.
     *
     * Only safely called on the producer thread.
     *
     * @param elements a pointer to a buffer containing at least `count` elements.
     * If `elements` is nullptr, zero or default constructed elements are
     * enqueued.
     * @param count The number of elements to read from `elements`
     * @return The number of elements successfully coped from `elements` and
     * inserted into the ring buffer.
     */
    int enqueue(T* elements, int count) {
#ifndef NDEBUG
        assert_correct_thread(producer_id);
#endif

        int wr_idx = write_index_.load(std::memory_order_relaxed);
        int rd_idx = read_index_.load(std::memory_order_acquire);

        if (full_internal(rd_idx, wr_idx)) {
            return 0;
        }

        int to_write = std::min(available_write_internal(rd_idx, wr_idx), count);

        /* First part, from the write index to the end of the array. */
        int first_part = std::min(storage_capacity() - wr_idx, to_write);
        /* Second part, from the beginning of the array */
        int second_part = to_write - first_part;

        if (elements) {
            Copy(data_.get() + wr_idx, elements, first_part);
            Copy(data_.get(), elements + first_part, second_part);
        } else {
            ConstructDefault(data_.get() + wr_idx, first_part);
            ConstructDefault(data_.get(), second_part);
        }

        write_index_.store(increment_index(wr_idx, to_write), std::memory_order_release);

        return to_write;
    }
    /**
     * Retrieve at most `count` elements from the ring buffer, and copy them to
     * `elements`, if non-null.
     *
     * Only safely called on the consumer side.
     *
     * @param elements A pointer to a buffer with space for at least `count`
     * elements. If `elements` is `nullptr`, `count` element will be discarded.
     * @param count The maximum number of elements to dequeue.
     * @return The number of elements written to `elements`.
     */
    int dequeue(T* elements, int count) {
#ifndef NDEBUG
        assert_correct_thread(consumer_id);
#endif

        int rd_idx = read_index_.load(std::memory_order_relaxed);
        int wr_idx = write_index_.load(std::memory_order_acquire);

        if (empty_internal(rd_idx, wr_idx)) {
            return 0;
        }

        int to_read = std::min(available_read_internal(rd_idx, wr_idx), count);

        int first_part = std::min(storage_capacity() - rd_idx, to_read);
        int second_part = to_read - first_part;

        if (elements) {
            Copy(elements, data_.get() + rd_idx, first_part);
            Copy(elements + first_part, data_.get(), second_part);
        }

        read_index_.store(increment_index(rd_idx, to_read), std::memory_order_release);

        return to_read;
    }
    /**
     * Get the number of available element for consuming.
     *
     * Only safely called on the consumer thread.
     *
     * @return The number of available elements for reading.
     */
    int available_read() const {
#ifndef NDEBUG
        assert_correct_thread(consumer_id);
#endif
        return available_read_internal(read_index_.load(std::memory_order_relaxed),
                                       write_index_.load(std::memory_order_acquire));
    }
    /**
     * Get the number of available elements for consuming.
     *
     * Only safely called on the producer thread.
     *
     * @return The number of empty slots in the buffer, available for writing.
     */
    int available_write() const {
#ifndef NDEBUG
        assert_correct_thread(producer_id);
#endif
        return available_write_internal(read_index_.load(std::memory_order_acquire),
                                        write_index_.load(std::memory_order_relaxed));
    }
    /**
     * Get the total capacity, for this ring buffer.
     *
     * Can be called safely on any thread.
     *
     * @return The maximum capacity of this ring buffer.
     */
    int capacity() const {
        return storage_capacity() - 1;
    }
    /**
     * Reset the consumer and producer thread identifier, in case the thread are
     * being changed. This has to be externally synchronized. This is no-op when
     * asserts are disabled.
     */
    void reset_thread_ids() {
#ifndef NDEBUG
        consumer_id = producer_id = std::thread::id();
#endif
    }

private:
    /** Return true if the ring buffer is empty.
     *
     * @param read_index the read index to consider
     * @param write_index the write index to consider
     * @return true if the ring buffer is empty, false otherwise.
     **/
    bool empty_internal(int read_index, int write_index) const {
        return write_index == read_index;
    }
    /** Return true if the ring buffer is full.
     *
     * This happens if the write index is exactly one element behind the read
     * index.
     *
     * @param read_index the read index to consider
     * @param write_index the write index to consider
     * @return true if the ring buffer is full, false otherwise.
     **/
    bool full_internal(int read_index, int write_index) const {
        return (write_index + 1) % storage_capacity() == read_index;
    }
    /**
     * Return the size of the storage. It is one more than the number of elements
     * that can be stored in the buffer.
     *
     * @return the number of elements that can be stored in the buffer.
     */
    int storage_capacity() const {
        return capacity_;
    }
    /**
     * Returns the number of elements available for reading.
     *
     * @return the number of available elements for reading.
     */
    int available_read_internal(int read_index, int write_index) const {
        if (write_index >= read_index) {
            return write_index - read_index;
        } else {
            return write_index + storage_capacity() - read_index;
        }
    }
    /**
     * Returns the number of empty elements, available for writing.
     *
     * @return the number of elements that can be written into the array.
     */
    int available_write_internal(int read_index, int write_index) const {
        /* We substract one element here to always keep at least one sample
         * free in the buffer, to distinguish between full and empty array. */
        int rv = read_index - write_index - 1;
        if (write_index >= read_index) {
            rv += storage_capacity();
        }
        return rv;
    }
    /**
     * Increments an index, wrapping it around the storage.
     *
     * @param index a reference to the index to increment.
     * @param increment the number by which `index` is incremented.
     * @return the new index.
     */
    int increment_index(int index, int increment) const {
        ASSERT(increment >= 0);
        return (index + increment) % storage_capacity();
    }
    /**
     * @brief This allows checking that enqueue (resp. dequeue) are always called
     * by the right thread.
     *
     * @param id the id of the thread that has called the calling method first.
     */
#ifndef NDEBUG
    static void assert_correct_thread(std::thread::id& id) {
        if (id == std::thread::id()) {
            id = std::this_thread::get_id();
            return;
        }
        ASSERT(id == std::this_thread::get_id());
    }
#endif
    /** Similar to memcpy, but accounts for the size of an element. */
    template <typename CopyT>
    void PodCopy(CopyT* destination, const CopyT* source, size_t count) {
        static_assert(std::is_trivial<CopyT>::value, "Requires trivial type");
        ASSERT(destination && source);
        memcpy(destination, source, count * sizeof(CopyT));
    }
    /** Similar to a memset to zero, but accounts for the size of an element. */
    template <typename ZeroT>
    void PodZero(ZeroT* destination, size_t count) {
        static_assert(std::is_trivial<ZeroT>::value, "Requires trivial type");
        ASSERT(destination);
        memset(destination, 0, count * sizeof(ZeroT));
    }
    template <typename CopyT, typename Trait>
    void Copy(CopyT* destination, const CopyT* source, size_t count, Trait) {
        for (size_t i = 0; i < count; i++) {
            destination[i] = source[i];
        }
    }
    template <typename CopyT>
    void Copy(CopyT* destination, const CopyT* source, size_t count, std::true_type) {
        PodCopy(destination, source, count);
    }
    /**
     * This allows copying a number of elements from a `source` pointer to a
     * `destination` pointer, using `memcpy` if it is safe to do so, or a loop that
     * calls the constructors and destructors otherwise.
     */
    template <typename CopyT>
    void Copy(CopyT* destination, const T* source, size_t count) {
        ASSERT(destination && source);
        Copy(destination, source, count, typename std::is_trivial<CopyT>::type());
    }
    template <typename ConstructT, typename Trait>
    void ConstructDefault(ConstructT* destination, size_t count, Trait) {
        for (size_t i = 0; i < count; i++) {
            destination[i] = ConstructT();
        }
    }
    template <typename ConstructT>
    void ConstructDefault(ConstructT* destination, size_t count, std::true_type) {
        PodZero(destination, count);
    }
    /**
     * This allows zeroing (using memset) or default-constructing a number of
     * elements calling the constructors and destructors if necessary.
     */
    template <typename ConstructT>
    void ConstructDefault(ConstructT* destination, size_t count) {
        ASSERT(destination);
        ConstructDefault(destination, count, typename std::is_arithmetic<ConstructT>::type());
    }
    /** Index at which the oldest element is at, in samples. */
    std::atomic<int> read_index_;
    /** Index at which to write new elements. `write_index` is always at
     * least one element ahead of `read_index_`. */
    std::atomic<int> write_index_;
    /** Maximum number of elements that can be stored in the ring buffer. */
    const int capacity_;
    /** Data storage */
    std::unique_ptr<T[]> data_;
#ifndef NDEBUG
    /** The id of the only thread that is allowed to read from the queue. */
    mutable std::thread::id consumer_id;
    /** The id of the only thread that is allowed to write from the queue. */
    mutable std::thread::id producer_id;
#endif
};