16 files changed, 726 insertions, 153 deletions

diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt
index 61ab68864..13ed68b3f 100644
--- a/src/common/CMakeLists.txt
+++ b/src/common/CMakeLists.txt
@@ -38,6 +38,7 @@ add_library(common STATIC
     common_precompiled_headers.h
     common_types.h
     concepts.h
+    container_hash.h
     demangle.cpp
     demangle.h
     div_ceil.h
@@ -132,6 +133,7 @@ add_library(common STATIC
     time_zone.h
     tiny_mt.h
     tree.h
+    typed_address.h
     uint128.h
     unique_function.h
     uuid.cpp
@@ -158,6 +160,8 @@ if(ARCHITECTURE_x86_64)
         PRIVATE
             x64/cpu_detect.cpp
             x64/cpu_detect.h
+            x64/cpu_wait.cpp
+            x64/cpu_wait.h
             x64/native_clock.cpp
             x64/native_clock.h
             x64/xbyak_abi.h
diff --git a/src/common/bounded_threadsafe_queue.h b/src/common/bounded_threadsafe_queue.h
index 21217801e..bd87aa09b 100644
--- a/src/common/bounded_threadsafe_queue.h
+++ b/src/common/bounded_threadsafe_queue.h
@@ -1,158 +1,249 @@
-// SPDX-FileCopyrightText: Copyright (c) 2020 Erik Rigtorp <erik@rigtorp.se>
-// SPDX-License-Identifier: MIT
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
 
 #pragma once
 
 #include <atomic>
-#include <bit>
 #include <condition_variable>
-#include <memory>
+#include <cstddef>
 #include <mutex>
 #include <new>
-#include <stop_token>
-#include <type_traits>
-#include <utility>
+
+#include "common/polyfill_thread.h"
 
 namespace Common {
 
-#if defined(__cpp_lib_hardware_interference_size)
-constexpr size_t hardware_interference_size = std::hardware_destructive_interference_size;
-#else
-constexpr size_t hardware_interference_size = 64;
-#endif
+namespace detail {
+constexpr size_t DefaultCapacity = 0x1000;
+} // namespace detail
+
+template <typename T, size_t Capacity = detail::DefaultCapacity>
+class SPSCQueue {
+    static_assert((Capacity & (Capacity - 1)) == 0, "Capacity must be a power of two.");
 
-template <typename T, size_t capacity = 0x400>
-class MPSCQueue {
 public:
-    explicit MPSCQueue() : allocator{std::allocator<Slot<T>>()} {
-        // Allocate one extra slot to prevent false sharing on the last slot
-        slots = allocator.allocate(capacity + 1);
-        // Allocators are not required to honor alignment for over-aligned types
-        // (see http://eel.is/c++draft/allocator.requirements#10) so we verify
-        // alignment here
-        if (reinterpret_cast<uintptr_t>(slots) % alignof(Slot<T>) != 0) {
-            allocator.deallocate(slots, capacity + 1);
-            throw std::bad_alloc();
-        }
-        for (size_t i = 0; i < capacity; ++i) {
-            std::construct_at(&slots[i]);
-        }
-        static_assert(std::has_single_bit(capacity), "capacity must be an integer power of 2");
-        static_assert(alignof(Slot<T>) == hardware_interference_size,
-                      "Slot must be aligned to cache line boundary to prevent false sharing");
-        static_assert(sizeof(Slot<T>) % hardware_interference_size == 0,
-                      "Slot size must be a multiple of cache line size to prevent "
-                      "false sharing between adjacent slots");
-        static_assert(sizeof(MPSCQueue) % hardware_interference_size == 0,
-                      "Queue size must be a multiple of cache line size to "
-                      "prevent false sharing between adjacent queues");
-    }
-
-    ~MPSCQueue() noexcept {
-        for (size_t i = 0; i < capacity; ++i) {
-            std::destroy_at(&slots[i]);
-        }
-        allocator.deallocate(slots, capacity + 1);
+    template <typename... Args>
+    bool TryEmplace(Args&&... args) {
+        return Emplace<PushMode::Try>(std::forward<Args>(args)...);
     }
 
-    // The queue must be both non-copyable and non-movable
-    MPSCQueue(const MPSCQueue&) = delete;
-    MPSCQueue& operator=(const MPSCQueue&) = delete;
+    template <typename... Args>
+    void EmplaceWait(Args&&... args) {
+        Emplace<PushMode::Wait>(std::forward<Args>(args)...);
+    }
 
-    MPSCQueue(MPSCQueue&&) = delete;
-    MPSCQueue& operator=(MPSCQueue&&) = delete;
+    bool TryPop(T& t) {
+        return Pop<PopMode::Try>(t);
+    }
 
-    void Push(const T& v) noexcept {
-        static_assert(std::is_nothrow_copy_constructible_v<T>,
-                      "T must be nothrow copy constructible");
-        emplace(v);
+    void PopWait(T& t) {
+        Pop<PopMode::Wait>(t);
     }
 
-    template <typename P, typename = std::enable_if_t<std::is_nothrow_constructible_v<T, P&&>>>
-    void Push(P&& v) noexcept {
-        emplace(std::forward<P>(v));
+    void PopWait(T& t, std::stop_token stop_token) {
+        Pop<PopMode::WaitWithStopToken>(t, stop_token);
     }
 
-    void Pop(T& v, std::stop_token stop) noexcept {
-        auto const tail = tail_.fetch_add(1);
-        auto& slot = slots[idx(tail)];
-        if (!slot.turn.test()) {
-            std::unique_lock lock{cv_mutex};
-            cv.wait(lock, stop, [&slot] { return slot.turn.test(); });
-        }
-        v = slot.move();
-        slot.destroy();
-        slot.turn.clear();
-        slot.turn.notify_one();
+    T PopWait() {
+        T t;
+        Pop<PopMode::Wait>(t);
+        return t;
+    }
+
+    T PopWait(std::stop_token stop_token) {
+        T t;
+        Pop<PopMode::WaitWithStopToken>(t, stop_token);
+        return t;
     }
 
 private:
-    template <typename U = T>
-    struct Slot {
-        ~Slot() noexcept {
-            if (turn.test()) {
-                destroy();
+    enum class PushMode {
+        Try,
+        Wait,
+        Count,
+    };
+
+    enum class PopMode {
+        Try,
+        Wait,
+        WaitWithStopToken,
+        Count,
+    };
+
+    template <PushMode Mode, typename... Args>
+    bool Emplace(Args&&... args) {
+        const size_t write_index = m_write_index.load(std::memory_order::relaxed);
+
+        if constexpr (Mode == PushMode::Try) {
+            // Check if we have free slots to write to.
+            if ((write_index - m_read_index.load(std::memory_order::acquire)) == Capacity) {
+                return false;
             }
+        } else if constexpr (Mode == PushMode::Wait) {
+            // Wait until we have free slots to write to.
+            std::unique_lock lock{producer_cv_mutex};
+            producer_cv.wait(lock, [this, write_index] {
+                return (write_index - m_read_index.load(std::memory_order::acquire)) < Capacity;
+            });
+        } else {
+            static_assert(Mode < PushMode::Count, "Invalid PushMode.");
         }
 
-        template <typename... Args>
-        void construct(Args&&... args) noexcept {
-            static_assert(std::is_nothrow_constructible_v<U, Args&&...>,
-                          "T must be nothrow constructible with Args&&...");
-            std::construct_at(reinterpret_cast<U*>(&storage), std::forward<Args>(args)...);
-        }
+        // Determine the position to write to.
+        const size_t pos = write_index % Capacity;
 
-        void destroy() noexcept {
-            static_assert(std::is_nothrow_destructible_v<U>, "T must be nothrow destructible");
-            std::destroy_at(reinterpret_cast<U*>(&storage));
-        }
+        // Emplace into the queue.
+        std::construct_at(std::addressof(m_data[pos]), std::forward<Args>(args)...);
+
+        // Increment the write index.
+        ++m_write_index;
+
+        // Notify the consumer that we have pushed into the queue.
+        std::scoped_lock lock{consumer_cv_mutex};
+        consumer_cv.notify_one();
+
+        return true;
+    }
+
+    template <PopMode Mode>
+    bool Pop(T& t, [[maybe_unused]] std::stop_token stop_token = {}) {
+        const size_t read_index = m_read_index.load(std::memory_order::relaxed);
 
-        U&& move() noexcept {
-            return reinterpret_cast<U&&>(storage);
+        if constexpr (Mode == PopMode::Try) {
+            // Check if the queue is empty.
+            if (read_index == m_write_index.load(std::memory_order::acquire)) {
+                return false;
+            }
+        } else if constexpr (Mode == PopMode::Wait) {
+            // Wait until the queue is not empty.
+            std::unique_lock lock{consumer_cv_mutex};
+            consumer_cv.wait(lock, [this, read_index] {
+                return read_index != m_write_index.load(std::memory_order::acquire);
+            });
+        } else if constexpr (Mode == PopMode::WaitWithStopToken) {
+            // Wait until the queue is not empty.
+            std::unique_lock lock{consumer_cv_mutex};
+            Common::CondvarWait(consumer_cv, lock, stop_token, [this, read_index] {
+                return read_index != m_write_index.load(std::memory_order::acquire);
+            });
+            if (stop_token.stop_requested()) {
+                return false;
+            }
+        } else {
+            static_assert(Mode < PopMode::Count, "Invalid PopMode.");
         }
 
-        // Align to avoid false sharing between adjacent slots
-        alignas(hardware_interference_size) std::atomic_flag turn{};
-        struct aligned_store {
-            struct type {
-                alignas(U) unsigned char data[sizeof(U)];
-            };
-        };
-        typename aligned_store::type storage;
-    };
+        // Determine the position to read from.
+        const size_t pos = read_index % Capacity;
+
+        // Pop the data off the queue, moving it.
+        t = std::move(m_data[pos]);
+
+        // Increment the read index.
+        ++m_read_index;
+
+        // Notify the producer that we have popped off the queue.
+        std::scoped_lock lock{producer_cv_mutex};
+        producer_cv.notify_one();
+
+        return true;
+    }
 
+    alignas(128) std::atomic_size_t m_read_index{0};
+    alignas(128) std::atomic_size_t m_write_index{0};
+
+    std::array<T, Capacity> m_data;
+
+    std::condition_variable_any producer_cv;
+    std::mutex producer_cv_mutex;
+    std::condition_variable_any consumer_cv;
+    std::mutex consumer_cv_mutex;
+};
+
+template <typename T, size_t Capacity = detail::DefaultCapacity>
+class MPSCQueue {
+public:
     template <typename... Args>
-    void emplace(Args&&... args) noexcept {
-        static_assert(std::is_nothrow_constructible_v<T, Args&&...>,
-                      "T must be nothrow constructible with Args&&...");
-        auto const head = head_.fetch_add(1);
-        auto& slot = slots[idx(head)];
-        slot.turn.wait(true);
-        slot.construct(std::forward<Args>(args)...);
-        slot.turn.test_and_set();
-        cv.notify_one();
+    bool TryEmplace(Args&&... args) {
+        std::scoped_lock lock{write_mutex};
+        return spsc_queue.TryEmplace(std::forward<Args>(args)...);
     }
 
-    constexpr size_t idx(size_t i) const noexcept {
-        return i & mask;
+    template <typename... Args>
+    void EmplaceWait(Args&&... args) {
+        std::scoped_lock lock{write_mutex};
+        spsc_queue.EmplaceWait(std::forward<Args>(args)...);
     }
 
-    static constexpr size_t mask = capacity - 1;
+    bool TryPop(T& t) {
+        return spsc_queue.TryPop(t);
+    }
 
-    // Align to avoid false sharing between head_ and tail_
-    alignas(hardware_interference_size) std::atomic<size_t> head_{0};
-    alignas(hardware_interference_size) std::atomic<size_t> tail_{0};
+    void PopWait(T& t) {
+        spsc_queue.PopWait(t);
+    }
 
-    std::mutex cv_mutex;
-    std::condition_variable_any cv;
+    void PopWait(T& t, std::stop_token stop_token) {
+        spsc_queue.PopWait(t, stop_token);
+    }
+
+    T PopWait() {
+        return spsc_queue.PopWait();
+    }
 
-    Slot<T>* slots;
-    [[no_unique_address]] std::allocator<Slot<T>> allocator;
+    T PopWait(std::stop_token stop_token) {
+        return spsc_queue.PopWait(stop_token);
+    }
 
-    static_assert(std::is_nothrow_copy_assignable_v<T> || std::is_nothrow_move_assignable_v<T>,
-                  "T must be nothrow copy or move assignable");
+private:
+    SPSCQueue<T, Capacity> spsc_queue;
+    std::mutex write_mutex;
+};
 
-    static_assert(std::is_nothrow_destructible_v<T>, "T must be nothrow destructible");
+template <typename T, size_t Capacity = detail::DefaultCapacity>
+class MPMCQueue {
+public:
+    template <typename... Args>
+    bool TryEmplace(Args&&... args) {
+        std::scoped_lock lock{write_mutex};
+        return spsc_queue.TryEmplace(std::forward<Args>(args)...);
+    }
+
+    template <typename... Args>
+    void EmplaceWait(Args&&... args) {
+        std::scoped_lock lock{write_mutex};
+        spsc_queue.EmplaceWait(std::forward<Args>(args)...);
+    }
+
+    bool TryPop(T& t) {
+        std::scoped_lock lock{read_mutex};
+        return spsc_queue.TryPop(t);
+    }
+
+    void PopWait(T& t) {
+        std::scoped_lock lock{read_mutex};
+        spsc_queue.PopWait(t);
+    }
+
+    void PopWait(T& t, std::stop_token stop_token) {
+        std::scoped_lock lock{read_mutex};
+        spsc_queue.PopWait(t, stop_token);
+    }
+
+    T PopWait() {
+        std::scoped_lock lock{read_mutex};
+        return spsc_queue.PopWait();
+    }
+
+    T PopWait(std::stop_token stop_token) {
+        std::scoped_lock lock{read_mutex};
+        return spsc_queue.PopWait(stop_token);
+    }
+
+private:
+    SPSCQueue<T, Capacity> spsc_queue;
+    std::mutex write_mutex;
+    std::mutex read_mutex;
 };
 
 } // namespace Common
diff --git a/src/common/container_hash.h b/src/common/container_hash.h
new file mode 100644
index 000000000..a5e357745
--- /dev/null
+++ b/src/common/container_hash.h
@@ -0,0 +1,92 @@
+// SPDX-FileCopyrightText: 2005-2014 Daniel James
+// SPDX-FileCopyrightText: 2016 Austin Appleby
+// SPDX-License-Identifier: BSL-1.0
+
+#include <array>
+#include <climits>
+#include <cstdint>
+#include <limits>
+#include <type_traits>
+#include <vector>
+
+namespace Common {
+
+namespace detail {
+
+template <typename T>
+    requires std::is_unsigned_v<T>
+inline std::size_t HashValue(T val) {
+    const unsigned int size_t_bits = std::numeric_limits<std::size_t>::digits;
+    const unsigned int length =
+        (std::numeric_limits<T>::digits - 1) / static_cast<unsigned int>(size_t_bits);
+
+    std::size_t seed = 0;
+
+    for (unsigned int i = length * size_t_bits; i > 0; i -= size_t_bits) {
+        seed ^= static_cast<size_t>(val >> i) + (seed << 6) + (seed >> 2);
+    }
+
+    seed ^= static_cast<size_t>(val) + (seed << 6) + (seed >> 2);
+
+    return seed;
+}
+
+template <size_t Bits>
+struct HashCombineImpl {
+    template <typename T>
+    static inline T fn(T seed, T value) {
+        seed ^= value + 0x9e3779b9 + (seed << 6) + (seed >> 2);
+        return seed;
+    }
+};
+
+template <>
+struct HashCombineImpl<64> {
+    static inline std::uint64_t fn(std::uint64_t h, std::uint64_t k) {
+        const std::uint64_t m = (std::uint64_t(0xc6a4a793) << 32) + 0x5bd1e995;
+        const int r = 47;
+
+        k *= m;
+        k ^= k >> r;
+        k *= m;
+
+        h ^= k;
+        h *= m;
+
+        // Completely arbitrary number, to prevent 0's
+        // from hashing to 0.
+        h += 0xe6546b64;
+
+        return h;
+    }
+};
+
+} // namespace detail
+
+template <typename T>
+inline void HashCombine(std::size_t& seed, const T& v) {
+    seed = detail::HashCombineImpl<sizeof(std::size_t) * CHAR_BIT>::fn(seed, detail::HashValue(v));
+}
+
+template <typename It>
+inline std::size_t HashRange(It first, It last) {
+    std::size_t seed = 0;
+
+    for (; first != last; ++first) {
+        HashCombine<typename std::iterator_traits<It>::value_type>(seed, *first);
+    }
+
+    return seed;
+}
+
+template <typename T, size_t Size>
+std::size_t HashValue(const std::array<T, Size>& v) {
+    return HashRange(v.cbegin(), v.cend());
+}
+
+template <typename T, typename Allocator>
+std::size_t HashValue(const std::vector<T, Allocator>& v) {
+    return HashRange(v.cbegin(), v.cend());
+}
+
+} // namespace Common
diff --git a/src/common/intrusive_red_black_tree.h b/src/common/intrusive_red_black_tree.h
index 5f6b34e82..bc2940fa0 100644
--- a/src/common/intrusive_red_black_tree.h
+++ b/src/common/intrusive_red_black_tree.h
@@ -96,10 +96,6 @@ public:
             return m_node == rhs.m_node;
         }
 
-        constexpr bool operator!=(const Iterator& rhs) const {
-            return !(*this == rhs);
-        }
-
         constexpr pointer operator->() const {
             return m_node;
         }
@@ -324,10 +320,6 @@ public:
             return m_impl == rhs.m_impl;
         }
 
-        constexpr bool operator!=(const Iterator& rhs) const {
-            return !(*this == rhs);
-        }
-
         constexpr pointer operator->() const {
             return Traits::GetParent(std::addressof(*m_impl));
         }
diff --git a/src/common/logging/backend.cpp b/src/common/logging/backend.cpp
index 2a3bded40..f96c7c222 100644
--- a/src/common/logging/backend.cpp
+++ b/src/common/logging/backend.cpp
@@ -28,7 +28,7 @@
 #ifdef _WIN32
 #include "common/string_util.h"
 #endif
-#include "common/threadsafe_queue.h"
+#include "common/bounded_threadsafe_queue.h"
 
 namespace Common::Log {
 
@@ -204,11 +204,11 @@ public:
 
     void PushEntry(Class log_class, Level log_level, const char* filename, unsigned int line_num,
                    const char* function, std::string&& message) {
-        if (!filter.CheckMessage(log_class, log_level))
+        if (!filter.CheckMessage(log_class, log_level)) {
             return;
-        const Entry& entry =
-            CreateEntry(log_class, log_level, filename, line_num, function, std::move(message));
-        message_queue.Push(entry);
+        }
+        message_queue.EmplaceWait(
+            CreateEntry(log_class, log_level, filename, line_num, function, std::move(message)));
     }
 
 private:
@@ -225,7 +225,7 @@ private:
                 ForEachBackend([&entry](Backend& backend) { backend.Write(entry); });
             };
             while (!stop_token.stop_requested()) {
-                entry = message_queue.PopWait(stop_token);
+                message_queue.PopWait(entry, stop_token);
                 if (entry.filename != nullptr) {
                     write_logs();
                 }
@@ -233,7 +233,7 @@ private:
             // Drain the logging queue. Only writes out up to MAX_LOGS_TO_WRITE to prevent a
             // case where a system is repeatedly spamming logs even on close.
             int max_logs_to_write = filter.IsDebug() ? INT_MAX : 100;
-            while (max_logs_to_write-- && message_queue.Pop(entry)) {
+            while (max_logs_to_write-- && message_queue.TryPop(entry)) {
                 write_logs();
             }
         });
@@ -273,7 +273,7 @@ private:
     ColorConsoleBackend color_console_backend{};
     FileBackend file_backend;
 
-    MPSCQueue<Entry, true> message_queue{};
+    MPSCQueue<Entry> message_queue{};
     std::chrono::steady_clock::time_point time_origin{std::chrono::steady_clock::now()};
     std::jthread backend_thread;
 };
diff --git a/src/common/range_map.h b/src/common/range_map.h
index 79c7ef547..ab73993e3 100644
--- a/src/common/range_map.h
+++ b/src/common/range_map.h
@@ -38,12 +38,12 @@ public:
         Map(address, address_end, null_value);
     }
 
-    [[nodiscard]] size_t GetContinousSizeFrom(KeyTBase address) const {
+    [[nodiscard]] size_t GetContinuousSizeFrom(KeyTBase address) const {
         const KeyT new_address = static_cast<KeyT>(address);
         if (new_address < 0) {
             return 0;
         }
-        return ContinousSizeInternal(new_address);
+        return ContinuousSizeInternal(new_address);
     }
 
     [[nodiscard]] ValueT GetValueAt(KeyT address) const {
@@ -59,7 +59,7 @@ private:
     using IteratorType = typename MapType::iterator;
     using ConstIteratorType = typename MapType::const_iterator;
 
-    size_t ContinousSizeInternal(KeyT address) const {
+    size_t ContinuousSizeInternal(KeyT address) const {
         const auto it = GetFirstElementBeforeOrOn(address);
         if (it == container.end() || it->second == null_value) {
             return 0;
diff --git a/src/common/string_util.cpp b/src/common/string_util.cpp
index e0b6180c5..feab1653d 100644
--- a/src/common/string_util.cpp
+++ b/src/common/string_util.cpp
@@ -125,18 +125,18 @@ std::string ReplaceAll(std::string result, const std::string& src, const std::st
     return result;
 }
 
-std::string UTF16ToUTF8(const std::u16string& input) {
+std::string UTF16ToUTF8(std::u16string_view input) {
     std::wstring_convert<std::codecvt_utf8_utf16<char16_t>, char16_t> convert;
-    return convert.to_bytes(input);
+    return convert.to_bytes(input.data(), input.data() + input.size());
 }
 
-std::u16string UTF8ToUTF16(const std::string& input) {
+std::u16string UTF8ToUTF16(std::string_view input) {
     std::wstring_convert<std::codecvt_utf8_utf16<char16_t>, char16_t> convert;
-    return convert.from_bytes(input);
+    return convert.from_bytes(input.data(), input.data() + input.size());
 }
 
 #ifdef _WIN32
-static std::wstring CPToUTF16(u32 code_page, const std::string& input) {
+static std::wstring CPToUTF16(u32 code_page, std::string_view input) {
     const auto size =
         MultiByteToWideChar(code_page, 0, input.data(), static_cast<int>(input.size()), nullptr, 0);
 
@@ -154,7 +154,7 @@ static std::wstring CPToUTF16(u32 code_page, const std::string& input) {
     return output;
 }
 
-std::string UTF16ToUTF8(const std::wstring& input) {
+std::string UTF16ToUTF8(std::wstring_view input) {
     const auto size = WideCharToMultiByte(CP_UTF8, 0, input.data(), static_cast<int>(input.size()),
                                           nullptr, 0, nullptr, nullptr);
     if (size == 0) {
@@ -172,7 +172,7 @@ std::string UTF16ToUTF8(const std::wstring& input) {
     return output;
 }
 
-std::wstring UTF8ToUTF16W(const std::string& input) {
+std::wstring UTF8ToUTF16W(std::string_view input) {
     return CPToUTF16(CP_UTF8, input);
 }
 
diff --git a/src/common/string_util.h b/src/common/string_util.h
index f8aecc875..c351f1a0c 100644
--- a/src/common/string_util.h
+++ b/src/common/string_util.h
@@ -36,12 +36,12 @@ bool SplitPath(const std::string& full_path, std::string* _pPath, std::string* _
 [[nodiscard]] std::string ReplaceAll(std::string result, const std::string& src,
                                      const std::string& dest);
 
-[[nodiscard]] std::string UTF16ToUTF8(const std::u16string& input);
-[[nodiscard]] std::u16string UTF8ToUTF16(const std::string& input);
+[[nodiscard]] std::string UTF16ToUTF8(std::u16string_view input);
+[[nodiscard]] std::u16string UTF8ToUTF16(std::string_view input);
 
 #ifdef _WIN32
-[[nodiscard]] std::string UTF16ToUTF8(const std::wstring& input);
-[[nodiscard]] std::wstring UTF8ToUTF16W(const std::string& str);
+[[nodiscard]] std::string UTF16ToUTF8(std::wstring_view input);
+[[nodiscard]] std::wstring UTF8ToUTF16W(std::string_view str);
 
 #endif
 
diff --git a/src/common/telemetry.cpp b/src/common/telemetry.cpp
index d26394359..91352912d 100644
--- a/src/common/telemetry.cpp
+++ b/src/common/telemetry.cpp
@@ -97,6 +97,7 @@ void AppendCPUInfo(FieldCollection& fc) {
     add_field("CPU_Extension_x64_PCLMULQDQ", caps.pclmulqdq);
     add_field("CPU_Extension_x64_POPCNT", caps.popcnt);
     add_field("CPU_Extension_x64_SHA", caps.sha);
+    add_field("CPU_Extension_x64_WAITPKG", caps.waitpkg);
 #else
     fc.AddField(FieldType::UserSystem, "CPU_Model", "Other");
 #endif
diff --git a/src/common/typed_address.h b/src/common/typed_address.h
new file mode 100644
index 000000000..64f4a07c2
--- /dev/null
+++ b/src/common/typed_address.h
@@ -0,0 +1,315 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include <compare>
+#include <type_traits>
+#include <fmt/format.h>
+
+#include "common/common_types.h"
+
+namespace Common {
+
+template <bool Virtual, typename T>
+class TypedAddress {
+public:
+    // Constructors.
+    constexpr inline TypedAddress() : m_address(0) {}
+    constexpr inline TypedAddress(uint64_t a) : m_address(a) {}
+
+    template <typename U>
+    constexpr inline explicit TypedAddress(const U* ptr)
+        : m_address(reinterpret_cast<uint64_t>(ptr)) {}
+
+    // Copy constructor.
+    constexpr inline TypedAddress(const TypedAddress& rhs) = default;
+
+    // Assignment operator.
+    constexpr inline TypedAddress& operator=(const TypedAddress& rhs) = default;
+
+    // Arithmetic operators.
+    template <typename I>
+    constexpr inline TypedAddress operator+(I rhs) const {
+        static_assert(std::is_integral_v<I>);
+        return m_address + rhs;
+    }
+
+    constexpr inline TypedAddress operator+(TypedAddress rhs) const {
+        return m_address + rhs.m_address;
+    }
+
+    constexpr inline TypedAddress operator++() {
+        return ++m_address;
+    }
+
+    constexpr inline TypedAddress operator++(int) {
+        return m_address++;
+    }
+
+    template <typename I>
+    constexpr inline TypedAddress operator-(I rhs) const {
+        static_assert(std::is_integral_v<I>);
+        return m_address - rhs;
+    }
+
+    constexpr inline ptrdiff_t operator-(TypedAddress rhs) const {
+        return m_address - rhs.m_address;
+    }
+
+    constexpr inline TypedAddress operator--() {
+        return --m_address;
+    }
+
+    constexpr inline TypedAddress operator--(int) {
+        return m_address--;
+    }
+
+    template <typename I>
+    constexpr inline TypedAddress operator+=(I rhs) {
+        static_assert(std::is_integral_v<I>);
+        m_address += rhs;
+        return *this;
+    }
+
+    template <typename I>
+    constexpr inline TypedAddress operator-=(I rhs) {
+        static_assert(std::is_integral_v<I>);
+        m_address -= rhs;
+        return *this;
+    }
+
+    // Logical operators.
+    constexpr inline uint64_t operator&(uint64_t mask) const {
+        return m_address & mask;
+    }
+
+    constexpr inline uint64_t operator|(uint64_t mask) const {
+        return m_address | mask;
+    }
+
+    template <typename I>
+    constexpr inline TypedAddress operator|=(I rhs) {
+        static_assert(std::is_integral_v<I>);
+        m_address |= rhs;
+        return *this;
+    }
+
+    constexpr inline uint64_t operator<<(int shift) const {
+        return m_address << shift;
+    }
+
+    constexpr inline uint64_t operator>>(int shift) const {
+        return m_address >> shift;
+    }
+
+    template <typename U>
+    constexpr inline size_t operator/(U size) const {
+        return m_address / size;
+    }
+
+    constexpr explicit operator bool() const {
+        return m_address != 0;
+    }
+
+    // constexpr inline uint64_t operator%(U align) const { return m_address % align; }
+
+    // Comparison operators.
+    constexpr bool operator==(const TypedAddress&) const = default;
+    constexpr auto operator<=>(const TypedAddress&) const = default;
+
+    // For convenience, also define comparison operators versus uint64_t.
+    constexpr inline bool operator==(uint64_t rhs) const {
+        return m_address == rhs;
+    }
+
+    // Allow getting the address explicitly, for use in accessors.
+    constexpr inline uint64_t GetValue() const {
+        return m_address;
+    }
+
+private:
+    uint64_t m_address{};
+};
+
+struct PhysicalAddressTag {};
+struct VirtualAddressTag {};
+struct ProcessAddressTag {};
+
+using PhysicalAddress = TypedAddress<false, PhysicalAddressTag>;
+using VirtualAddress = TypedAddress<true, VirtualAddressTag>;
+using ProcessAddress = TypedAddress<true, ProcessAddressTag>;
+
+// Define accessors.
+template <typename T>
+concept IsTypedAddress = std::same_as<T, PhysicalAddress> || std::same_as<T, VirtualAddress> ||
+                         std::same_as<T, ProcessAddress>;
+
+template <typename T>
+constexpr inline T Null = [] {
+    if constexpr (std::is_same<T, uint64_t>::value) {
+        return 0;
+    } else {
+        static_assert(std::is_same<T, PhysicalAddress>::value ||
+                      std::is_same<T, VirtualAddress>::value ||
+                      std::is_same<T, ProcessAddress>::value);
+        return T(0);
+    }
+}();
+
+// Basic type validations.
+static_assert(sizeof(PhysicalAddress) == sizeof(uint64_t));
+static_assert(sizeof(VirtualAddress) == sizeof(uint64_t));
+static_assert(sizeof(ProcessAddress) == sizeof(uint64_t));
+
+static_assert(std::is_trivially_copyable_v<PhysicalAddress>);
+static_assert(std::is_trivially_copyable_v<VirtualAddress>);
+static_assert(std::is_trivially_copyable_v<ProcessAddress>);
+
+static_assert(std::is_trivially_copy_constructible_v<PhysicalAddress>);
+static_assert(std::is_trivially_copy_constructible_v<VirtualAddress>);
+static_assert(std::is_trivially_copy_constructible_v<ProcessAddress>);
+
+static_assert(std::is_trivially_move_constructible_v<PhysicalAddress>);
+static_assert(std::is_trivially_move_constructible_v<VirtualAddress>);
+static_assert(std::is_trivially_move_constructible_v<ProcessAddress>);
+
+static_assert(std::is_trivially_copy_assignable_v<PhysicalAddress>);
+static_assert(std::is_trivially_copy_assignable_v<VirtualAddress>);
+static_assert(std::is_trivially_copy_assignable_v<ProcessAddress>);
+
+static_assert(std::is_trivially_move_assignable_v<PhysicalAddress>);
+static_assert(std::is_trivially_move_assignable_v<VirtualAddress>);
+static_assert(std::is_trivially_move_assignable_v<ProcessAddress>);
+
+static_assert(std::is_trivially_destructible_v<PhysicalAddress>);
+static_assert(std::is_trivially_destructible_v<VirtualAddress>);
+static_assert(std::is_trivially_destructible_v<ProcessAddress>);
+
+static_assert(Null<uint64_t> == 0);
+static_assert(Null<PhysicalAddress> == Null<uint64_t>);
+static_assert(Null<VirtualAddress> == Null<uint64_t>);
+static_assert(Null<ProcessAddress> == Null<uint64_t>);
+
+// Constructor/assignment validations.
+static_assert([] {
+    const PhysicalAddress a(5);
+    PhysicalAddress b(a);
+    return b;
+}() == PhysicalAddress(5));
+static_assert([] {
+    const PhysicalAddress a(5);
+    PhysicalAddress b(10);
+    b = a;
+    return b;
+}() == PhysicalAddress(5));
+
+// Arithmetic validations.
+static_assert(PhysicalAddress(10) + 5 == PhysicalAddress(15));
+static_assert(PhysicalAddress(10) - 5 == PhysicalAddress(5));
+static_assert([] {
+    PhysicalAddress v(10);
+    v += 5;
+    return v;
+}() == PhysicalAddress(15));
+static_assert([] {
+    PhysicalAddress v(10);
+    v -= 5;
+    return v;
+}() == PhysicalAddress(5));
+static_assert(PhysicalAddress(10)++ == PhysicalAddress(10));
+static_assert(++PhysicalAddress(10) == PhysicalAddress(11));
+static_assert(PhysicalAddress(10)-- == PhysicalAddress(10));
+static_assert(--PhysicalAddress(10) == PhysicalAddress(9));
+
+// Logical validations.
+static_assert((PhysicalAddress(0b11111111) >> 1) == 0b01111111);
+static_assert((PhysicalAddress(0b10101010) >> 1) == 0b01010101);
+static_assert((PhysicalAddress(0b11111111) << 1) == 0b111111110);
+static_assert((PhysicalAddress(0b01010101) << 1) == 0b10101010);
+static_assert((PhysicalAddress(0b11111111) & 0b01010101) == 0b01010101);
+static_assert((PhysicalAddress(0b11111111) & 0b10101010) == 0b10101010);
+static_assert((PhysicalAddress(0b01010101) & 0b10101010) == 0b00000000);
+static_assert((PhysicalAddress(0b00000000) | 0b01010101) == 0b01010101);
+static_assert((PhysicalAddress(0b11111111) | 0b01010101) == 0b11111111);
+static_assert((PhysicalAddress(0b10101010) | 0b01010101) == 0b11111111);
+
+// Comparisons.
+static_assert(PhysicalAddress(0) == PhysicalAddress(0));
+static_assert(PhysicalAddress(0) != PhysicalAddress(1));
+static_assert(PhysicalAddress(0) < PhysicalAddress(1));
+static_assert(PhysicalAddress(0) <= PhysicalAddress(1));
+static_assert(PhysicalAddress(1) > PhysicalAddress(0));
+static_assert(PhysicalAddress(1) >= PhysicalAddress(0));
+
+static_assert(!(PhysicalAddress(0) == PhysicalAddress(1)));
+static_assert(!(PhysicalAddress(0) != PhysicalAddress(0)));
+static_assert(!(PhysicalAddress(1) < PhysicalAddress(0)));
+static_assert(!(PhysicalAddress(1) <= PhysicalAddress(0)));
+static_assert(!(PhysicalAddress(0) > PhysicalAddress(1)));
+static_assert(!(PhysicalAddress(0) >= PhysicalAddress(1)));
+
+} // namespace Common
+
+template <bool Virtual, typename T>
+constexpr inline uint64_t GetInteger(Common::TypedAddress<Virtual, T> address) {
+    return address.GetValue();
+}
+
+template <>
+struct fmt::formatter<Common::PhysicalAddress> {
+    constexpr auto parse(fmt::format_parse_context& ctx) {
+        return ctx.begin();
+    }
+    template <typename FormatContext>
+    auto format(const Common::PhysicalAddress& addr, FormatContext& ctx) {
+        return fmt::format_to(ctx.out(), "{:#x}", static_cast<u64>(addr.GetValue()));
+    }
+};
+
+template <>
+struct fmt::formatter<Common::ProcessAddress> {
+    constexpr auto parse(fmt::format_parse_context& ctx) {
+        return ctx.begin();
+    }
+    template <typename FormatContext>
+    auto format(const Common::ProcessAddress& addr, FormatContext& ctx) {
+        return fmt::format_to(ctx.out(), "{:#x}", static_cast<u64>(addr.GetValue()));
+    }
+};
+
+template <>
+struct fmt::formatter<Common::VirtualAddress> {
+    constexpr auto parse(fmt::format_parse_context& ctx) {
+        return ctx.begin();
+    }
+    template <typename FormatContext>
+    auto format(const Common::VirtualAddress& addr, FormatContext& ctx) {
+        return fmt::format_to(ctx.out(), "{:#x}", static_cast<u64>(addr.GetValue()));
+    }
+};
+
+namespace std {
+
+template <>
+struct hash<Common::PhysicalAddress> {
+    size_t operator()(const Common::PhysicalAddress& k) const noexcept {
+        return k.GetValue();
+    }
+};
+
+template <>
+struct hash<Common::ProcessAddress> {
+    size_t operator()(const Common::ProcessAddress& k) const noexcept {
+        return k.GetValue();
+    }
+};
+
+template <>
+struct hash<Common::VirtualAddress> {
+    size_t operator()(const Common::VirtualAddress& k) const noexcept {
+        return k.GetValue();
+    }
+};
+
+} // namespace std
diff --git a/src/common/x64/cpu_detect.cpp b/src/common/x64/cpu_detect.cpp
index e54383a4a..72ed6e96c 100644
--- a/src/common/x64/cpu_detect.cpp
+++ b/src/common/x64/cpu_detect.cpp
@@ -144,6 +144,7 @@ static CPUCaps Detect() {
             caps.bmi2 = Common::Bit<8>(cpu_id[1]);
             caps.sha = Common::Bit<29>(cpu_id[1]);
 
+            caps.waitpkg = Common::Bit<5>(cpu_id[2]);
             caps.gfni = Common::Bit<8>(cpu_id[2]);
 
             __cpuidex(cpu_id, 0x00000007, 0x00000001);
diff --git a/src/common/x64/cpu_detect.h b/src/common/x64/cpu_detect.h
index ca8db19d6..8253944d6 100644
--- a/src/common/x64/cpu_detect.h
+++ b/src/common/x64/cpu_detect.h
@@ -67,6 +67,7 @@ struct CPUCaps {
     bool pclmulqdq : 1;
     bool popcnt : 1;
     bool sha : 1;
+    bool waitpkg : 1;
 };
 
 /**
diff --git a/src/common/x64/cpu_wait.cpp b/src/common/x64/cpu_wait.cpp
new file mode 100644
index 000000000..cfeef6a3d
--- /dev/null
+++ b/src/common/x64/cpu_wait.cpp
@@ -0,0 +1,69 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#include <thread>
+
+#ifdef _MSC_VER
+#include <intrin.h>
+#endif
+
+#include "common/x64/cpu_detect.h"
+#include "common/x64/cpu_wait.h"
+
+namespace Common::X64 {
+
+#ifdef _MSC_VER
+__forceinline static u64 FencedRDTSC() {
+    _mm_lfence();
+    _ReadWriteBarrier();
+    const u64 result = __rdtsc();
+    _mm_lfence();
+    _ReadWriteBarrier();
+    return result;
+}
+
+__forceinline static void TPAUSE() {
+    // 100,000 cycles is a reasonable amount of time to wait to save on CPU resources.
+    // For reference:
+    // At 1 GHz, 100K cycles is 100us
+    // At 2 GHz, 100K cycles is 50us
+    // At 4 GHz, 100K cycles is 25us
+    static constexpr auto PauseCycles = 100'000;
+    _tpause(0, FencedRDTSC() + PauseCycles);
+}
+#else
+static u64 FencedRDTSC() {
+    u64 eax;
+    u64 edx;
+    asm volatile("lfence\n\t"
+                 "rdtsc\n\t"
+                 "lfence\n\t"
+                 : "=a"(eax), "=d"(edx));
+    return (edx << 32) | eax;
+}
+
+static void TPAUSE() {
+    // 100,000 cycles is a reasonable amount of time to wait to save on CPU resources.
+    // For reference:
+    // At 1 GHz, 100K cycles is 100us
+    // At 2 GHz, 100K cycles is 50us
+    // At 4 GHz, 100K cycles is 25us
+    static constexpr auto PauseCycles = 100'000;
+    const auto tsc = FencedRDTSC() + PauseCycles;
+    const auto eax = static_cast<u32>(tsc & 0xFFFFFFFF);
+    const auto edx = static_cast<u32>(tsc >> 32);
+    asm volatile("tpause %0" : : "r"(0), "d"(edx), "a"(eax));
+}
+#endif
+
+void MicroSleep() {
+    static const bool has_waitpkg = GetCPUCaps().waitpkg;
+
+    if (has_waitpkg) {
+        TPAUSE();
+    } else {
+        std::this_thread::yield();
+    }
+}
+
+} // namespace Common::X64
diff --git a/src/common/x64/cpu_wait.h b/src/common/x64/cpu_wait.h
new file mode 100644
index 000000000..99d3757a7
--- /dev/null
+++ b/src/common/x64/cpu_wait.h
@@ -0,0 +1,10 @@
+// SPDX-FileCopyrightText: Copyright 2023 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+namespace Common::X64 {
+
+void MicroSleep();
+
+} // namespace Common::X64
diff --git a/src/common/x64/native_clock.cpp b/src/common/x64/native_clock.cpp
index 76c66e7ee..277b00662 100644
--- a/src/common/x64/native_clock.cpp
+++ b/src/common/x64/native_clock.cpp
@@ -27,16 +27,13 @@ __forceinline static u64 FencedRDTSC() {
 }
 #else
 static u64 FencedRDTSC() {
-    u64 result;
+    u64 eax;
+    u64 edx;
     asm volatile("lfence\n\t"
                  "rdtsc\n\t"
-                 "shl $32, %%rdx\n\t"
-                 "or %%rdx, %0\n\t"
-                 "lfence"
-                 : "=a"(result)
-                 :
-                 : "rdx", "memory", "cc");
-    return result;
+                 "lfence\n\t"
+                 : "=a"(eax), "=d"(edx));
+    return (edx << 32) | eax;
 }
 #endif
 
diff --git a/src/common/zstd_compression.cpp b/src/common/zstd_compression.cpp
index b71a41b78..cb6ec171b 100644
--- a/src/common/zstd_compression.cpp
+++ b/src/common/zstd_compression.cpp
@@ -33,7 +33,7 @@ std::vector<u8> CompressDataZSTDDefault(const u8* source, std::size_t source_siz
 
 std::vector<u8> DecompressDataZSTD(std::span<const u8> compressed) {
     const std::size_t decompressed_size =
-        ZSTD_getDecompressedSize(compressed.data(), compressed.size());
+        ZSTD_getFrameContentSize(compressed.data(), compressed.size());
     std::vector<u8> decompressed(decompressed_size);
 
     const std::size_t uncompressed_result_size = ZSTD_decompress(