From 761ef78408d8682c04197cb3daaa77a2c65ecca3 Mon Sep 17 00:00:00 2001
From: MerryMage <MerryMage@users.noreply.github.com>
Date: Sat, 8 Sep 2018 20:42:25 +0100
Subject: [PATCH] common: Implement a ring buffer

---
 src/common/CMakeLists.txt |   1 +
 src/common/ring_buffer.h  | 111 ++++++++++++++++++++++++++++++++++++++
 2 files changed, 112 insertions(+)
 create mode 100644 src/common/ring_buffer.h

diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index dfc0ea0c16..eed86ae0dc 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -72,6 +72,7 @@ add_library(common STATIC
     param_package.cpp
     param_package.h
     quaternion.h
+    ring_buffer.h
     scm_rev.cpp
     scm_rev.h
     scope_exit.h
diff --git a/src/common/ring_buffer.h b/src/common/ring_buffer.h
new file mode 100644
index 0000000000..f44d9e8d35
--- /dev/null
+++ b/src/common/ring_buffer.h
@@ -0,0 +1,111 @@
+// Copyright 2018 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <algorithm>
+#include <array>
+#include <atomic>
+#include <cstddef>
+#include <cstring>
+#include <type_traits>
+#include <vector>
+#include "common/common_types.h"
+
+namespace Common {
+
+/// SPSC ring buffer
+/// @tparam T            Element type
+/// @tparam capacity     Number of slots in ring buffer
+/// @tparam granularity  Slot size in terms of number of elements
+template <typename T, std::size_t capacity, std::size_t granularity = 1>
+class RingBuffer {
+    /// A "slot" is made of `granularity` elements of `T`.
+    static constexpr std::size_t slot_size = granularity * sizeof(T);
+    // T must be safely memcpy-able and have a trivial default constructor.
+    static_assert(std::is_trivial_v<T>);
+    // Ensure capacity is sensible.
+    static_assert(capacity < std::numeric_limits<std::size_t>::max() / 2 / granularity);
+    static_assert((capacity & (capacity - 1)) == 0, "capacity must be a power of two");
+    // Ensure lock-free.
+    static_assert(std::atomic<std::size_t>::is_always_lock_free);
+
+public:
+    /// Pushes slots into the ring buffer
+    /// @param new_slots   Pointer to the slots to push
+    /// @param slot_count  Number of slots to push
+    /// @returns The number of slots actually pushed
+    std::size_t Push(const void* new_slots, std::size_t slot_count) {
+        const std::size_t write_index = m_write_index.load();
+        const std::size_t slots_free = capacity + m_read_index.load() - write_index;
+        const std::size_t push_count = std::min(slot_count, slots_free);
+
+        const std::size_t pos = write_index % capacity;
+        const std::size_t first_copy = std::min(capacity - pos, push_count);
+        const std::size_t second_copy = push_count - first_copy;
+
+        const char* in = static_cast<const char*>(new_slots);
+        std::memcpy(m_data.data() + pos * granularity, in, first_copy * slot_size);
+        in += first_copy * slot_size;
+        std::memcpy(m_data.data(), in, second_copy * slot_size);
+
+        m_write_index.store(write_index + push_count);
+
+        return push_count;
+    }
+
+    std::size_t Push(const std::vector<T>& input) {
+        return Push(input.data(), input.size());
+    }
+
+    /// Pops slots from the ring buffer
+    /// @param output     Where to store the popped slots
+    /// @param max_slots  Maximum number of slots to pop
+    /// @returns The number of slots actually popped
+    std::size_t Pop(void* output, std::size_t max_slots = ~std::size_t(0)) {
+        const std::size_t read_index = m_read_index.load();
+        const std::size_t slots_filled = m_write_index.load() - read_index;
+        const std::size_t pop_count = std::min(slots_filled, max_slots);
+
+        const std::size_t pos = read_index % capacity;
+        const std::size_t first_copy = std::min(capacity - pos, pop_count);
+        const std::size_t second_copy = pop_count - first_copy;
+
+        char* out = static_cast<char*>(output);
+        std::memcpy(out, m_data.data() + pos * granularity, first_copy * slot_size);
+        out += first_copy * slot_size;
+        std::memcpy(out, m_data.data(), second_copy * slot_size);
+
+        m_read_index.store(read_index + pop_count);
+
+        return pop_count;
+    }
+
+    std::vector<T> Pop(std::size_t max_slots = ~std::size_t(0)) {
+        std::vector<T> out(std::min(max_slots, capacity) * granularity);
+        const std::size_t count = Pop(out.data(), out.size() / granularity);
+        out.resize(count * granularity);
+        return out;
+    }
+
+    /// @returns Number of slots used
+    std::size_t Size() const {
+        return m_write_index.load() - m_read_index.load();
+    }
+
+    /// @returns Maximum size of ring buffer
+    constexpr std::size_t Capacity() const {
+        return capacity;
+    }
+
+private:
+    // It is important to align the below variables for performance reasons:
+    // Having them on the same cache-line would result in false-sharing between them.
+    alignas(128) std::atomic<std::size_t> m_read_index{0};
+    alignas(128) std::atomic<std::size_t> m_write_index{0};
+
+    std::array<T, granularity * capacity> m_data;
+};
+
+} // namespace Common