65 files changed, 3687 insertions, 998 deletions

diff --git a/CMakeLists.txt b/CMakeLists.txt
index 44ed4196d..467d769a2 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -157,8 +157,14 @@ if (ENABLE_SDL2)
         target_include_directories(SDL2 INTERFACE "${SDL2_INCLUDE_DIR}")
     else()
         find_package(SDL2 REQUIRED)
-        include_directories(${SDL2_INCLUDE_DIRS})
 
+        # Some installations don't set SDL2_LIBRARIES
+        if("${SDL2_LIBRARIES}" STREQUAL "")
+            message(WARNING "SDL2_LIBRARIES wasn't set, manually setting to SDL2::SDL2")
+            set(SDL2_LIBRARIES "SDL2::SDL2")
+        endif()
+
+        include_directories(${SDL2_INCLUDE_DIRS})
         add_library(SDL2 INTERFACE)
         target_link_libraries(SDL2 INTERFACE "${SDL2_LIBRARIES}")
     endif()
diff --git a/externals/httplib/README.md b/externals/httplib/README.md
index 0e26522b5..73037d297 100644
--- a/externals/httplib/README.md
+++ b/externals/httplib/README.md
@@ -1,4 +1,4 @@
-From https://github.com/yhirose/cpp-httplib/commit/d9479bc0b12e8a1e8bce2d34da4feeef488581f3
+From https://github.com/yhirose/cpp-httplib/tree/fce8e6fefdab4ad48bc5b25c98e5ebfda4f3cf53
 
 MIT License
 
diff --git a/externals/httplib/httplib.h b/externals/httplib/httplib.h
index fa2edcc94..e03842e6d 100644
--- a/externals/httplib/httplib.h
+++ b/externals/httplib/httplib.h
@@ -1,7 +1,7 @@
 //
 //  httplib.h
 //
-//  Copyright (c) 2019 Yuji Hirose. All rights reserved.
+//  Copyright (c) 2020 Yuji Hirose. All rights reserved.
 //  MIT License
 //
 
@@ -11,6 +11,7 @@
 /*
  * Configuration
  */
+
 #ifndef CPPHTTPLIB_KEEPALIVE_TIMEOUT_SECOND
 #define CPPHTTPLIB_KEEPALIVE_TIMEOUT_SECOND 5
 #endif
@@ -40,7 +41,7 @@
 #endif
 
 #ifndef CPPHTTPLIB_PAYLOAD_MAX_LENGTH
-#define CPPHTTPLIB_PAYLOAD_MAX_LENGTH (std::numeric_limits<size_t>::max)()
+#define CPPHTTPLIB_PAYLOAD_MAX_LENGTH (std::numeric_limits<size_t>::max())
 #endif
 
 #ifndef CPPHTTPLIB_RECV_BUFSIZ
@@ -48,9 +49,14 @@
 #endif
 
 #ifndef CPPHTTPLIB_THREAD_POOL_COUNT
-#define CPPHTTPLIB_THREAD_POOL_COUNT 8
+#define CPPHTTPLIB_THREAD_POOL_COUNT                                           \
+  (std::max(1u, std::thread::hardware_concurrency() - 1))
 #endif
 
+/*
+ * Headers
+ */
+
 #ifdef _WIN32
 #ifndef _CRT_SECURE_NO_WARNINGS
 #define _CRT_SECURE_NO_WARNINGS
@@ -62,9 +68,9 @@
 
 #if defined(_MSC_VER)
 #ifdef _WIN64
-typedef __int64 ssize_t;
+using ssize_t = __int64;
 #else
-typedef int ssize_t;
+using ssize_t = int;
 #endif
 
 #if _MSC_VER < 1900
@@ -100,7 +106,7 @@ typedef int ssize_t;
 #define strcasecmp _stricmp
 #endif // strcasecmp
 
-typedef SOCKET socket_t;
+using socket_t = SOCKET;
 #ifdef CPPHTTPLIB_USE_POLL
 #define poll(fds, nfds, timeout) WSAPoll(fds, nfds, timeout)
 #endif
@@ -109,23 +115,25 @@ typedef SOCKET socket_t;
 
 #include <arpa/inet.h>
 #include <cstring>
+#include <ifaddrs.h>
 #include <netdb.h>
 #include <netinet/in.h>
 #ifdef CPPHTTPLIB_USE_POLL
 #include <poll.h>
 #endif
+#include <csignal>
 #include <pthread.h>
-#include <signal.h>
 #include <sys/select.h>
 #include <sys/socket.h>
 #include <unistd.h>
 
-typedef int socket_t;
+using socket_t = int;
 #define INVALID_SOCKET (-1)
 #endif //_WIN32
 
-#include <assert.h>
+#include <array>
 #include <atomic>
+#include <cassert>
 #include <condition_variable>
 #include <errno.h>
 #include <fcntl.h>
@@ -143,9 +151,13 @@ typedef int socket_t;
 
 #ifdef CPPHTTPLIB_OPENSSL_SUPPORT
 #include <openssl/err.h>
+#include <openssl/md5.h>
 #include <openssl/ssl.h>
 #include <openssl/x509v3.h>
 
+#include <iomanip>
+#include <sstream>
+
 // #if OPENSSL_VERSION_NUMBER < 0x1010100fL
 // #error Sorry, OpenSSL versions prior to 1.1.1 are not supported
 // #endif
@@ -162,6 +174,9 @@ inline const unsigned char *ASN1_STRING_get0_data(const ASN1_STRING *asn1) {
 #include <zlib.h>
 #endif
 
+/*
+ * Declaration
+ */
 namespace httplib {
 
 namespace detail {
@@ -176,37 +191,15 @@ struct ci {
 
 } // namespace detail
 
-enum class HttpVersion { v1_0 = 0, v1_1 };
-
-typedef std::multimap<std::string, std::string, detail::ci> Headers;
-
-typedef std::multimap<std::string, std::string> Params;
-typedef std::smatch Match;
-
-typedef std::function<void(const char *data, size_t data_len)> DataSink;
-
-typedef std::function<void()> Done;
-
-typedef std::function<void(size_t offset, size_t length, DataSink sink,
-                           Done done)>
-    ContentProvider;
+using Headers = std::multimap<std::string, std::string, detail::ci>;
 
-typedef std::function<bool(const char *data, size_t data_length, size_t offset,
-                           uint64_t content_length)>
-    ContentReceiver;
+using Params = std::multimap<std::string, std::string>;
+using Match = std::smatch;
 
-typedef std::function<bool(uint64_t current, uint64_t total)> Progress;
+using Progress = std::function<bool(uint64_t current, uint64_t total)>;
 
 struct Response;
-typedef std::function<bool(const Response &response)> ResponseHandler;
-
-struct MultipartFile {
-  std::string filename;
-  std::string content_type;
-  size_t offset = 0;
-  size_t length = 0;
-};
-typedef std::multimap<std::string, MultipartFile> MultipartFiles;
+using ResponseHandler = std::function<bool(const Response &response)>;
 
 struct MultipartFormData {
   std::string name;
@@ -214,10 +207,53 @@ struct MultipartFormData {
   std::string filename;
   std::string content_type;
 };
-typedef std::vector<MultipartFormData> MultipartFormDataItems;
+using MultipartFormDataItems = std::vector<MultipartFormData>;
+using MultipartFormDataMap = std::multimap<std::string, MultipartFormData>;
+
+class DataSink {
+public:
+  DataSink() = default;
+  DataSink(const DataSink &) = delete;
+  DataSink &operator=(const DataSink &) = delete;
+  DataSink(DataSink &&) = delete;
+  DataSink &operator=(DataSink &&) = delete;
+
+  std::function<void(const char *data, size_t data_len)> write;
+  std::function<void()> done;
+  std::function<bool()> is_writable;
+};
+
+using ContentProvider =
+    std::function<void(size_t offset, size_t length, DataSink &sink)>;
+
+using ContentReceiver =
+    std::function<bool(const char *data, size_t data_length)>;
+
+using MultipartContentHeader =
+    std::function<bool(const MultipartFormData &file)>;
+
+class ContentReader {
+public:
+  using Reader = std::function<bool(ContentReceiver receiver)>;
+  using MultipartReader = std::function<bool(MultipartContentHeader header,
+                                             ContentReceiver receiver)>;
+
+  ContentReader(Reader reader, MultipartReader muitlpart_reader)
+      : reader_(reader), muitlpart_reader_(muitlpart_reader) {}
+
+  bool operator()(MultipartContentHeader header,
+                  ContentReceiver receiver) const {
+    return muitlpart_reader_(header, receiver);
+  }
+
+  bool operator()(ContentReceiver receiver) const { return reader_(receiver); }
+
+  Reader reader_;
+  MultipartReader muitlpart_reader_;
+};
 
-typedef std::pair<ssize_t, ssize_t> Range;
-typedef std::vector<Range> Ranges;
+using Range = std::pair<ssize_t, ssize_t>;
+using Ranges = std::vector<Range>;
 
 struct Request {
   std::string method;
@@ -229,7 +265,7 @@ struct Request {
   std::string version;
   std::string target;
   Params params;
-  MultipartFiles files;
+  MultipartFormDataMap files;
   Ranges ranges;
   Match matches;
 
@@ -253,13 +289,19 @@ struct Request {
   std::string get_param_value(const char *key, size_t id = 0) const;
   size_t get_param_value_count(const char *key) const;
 
+  bool is_multipart_form_data() const;
+
   bool has_file(const char *key) const;
-  MultipartFile get_file_value(const char *key) const;
+  MultipartFormData get_file_value(const char *key) const;
+
+  // private members...
+  size_t content_length;
+  ContentProvider content_provider;
 };
 
 struct Response {
   std::string version;
-  int status;
+  int status = -1;
   Headers headers;
   std::string body;
 
@@ -269,106 +311,81 @@ struct Response {
   void set_header(const char *key, const char *val);
   void set_header(const char *key, const std::string &val);
 
-  void set_redirect(const char *uri);
+  void set_redirect(const char *url);
   void set_content(const char *s, size_t n, const char *content_type);
   void set_content(const std::string &s, const char *content_type);
 
   void set_content_provider(
       size_t length,
-      std::function<void(size_t offset, size_t length, DataSink sink)> provider,
+      std::function<void(size_t offset, size_t length, DataSink &sink)>
+          provider,
       std::function<void()> resource_releaser = [] {});
 
   void set_chunked_content_provider(
-      std::function<void(size_t offset, DataSink sink, Done done)> provider,
+      std::function<void(size_t offset, DataSink &sink)> provider,
       std::function<void()> resource_releaser = [] {});
 
-  Response() : status(-1), content_provider_resource_length(0) {}
-
+  Response() = default;
+  Response(const Response &) = default;
+  Response &operator=(const Response &) = default;
+  Response(Response &&) = default;
+  Response &operator=(Response &&) = default;
   ~Response() {
     if (content_provider_resource_releaser) {
       content_provider_resource_releaser();
     }
   }
 
-  size_t content_provider_resource_length;
+  // private members...
+  size_t content_length = 0;
   ContentProvider content_provider;
   std::function<void()> content_provider_resource_releaser;
 };
 
 class Stream {
 public:
-  virtual ~Stream() {}
+  virtual ~Stream() = default;
+
+  virtual bool is_readable() const = 0;
+  virtual bool is_writable() const = 0;
+
   virtual int read(char *ptr, size_t size) = 0;
-  virtual int write(const char *ptr, size_t size1) = 0;
-  virtual int write(const char *ptr) = 0;
-  virtual int write(const std::string &s) = 0;
+  virtual int write(const char *ptr, size_t size) = 0;
   virtual std::string get_remote_addr() const = 0;
 
   template <typename... Args>
   int write_format(const char *fmt, const Args &... args);
-};
-
-class SocketStream : public Stream {
-public:
-  SocketStream(socket_t sock);
-  virtual ~SocketStream();
-
-  virtual int read(char *ptr, size_t size);
-  virtual int write(const char *ptr, size_t size);
-  virtual int write(const char *ptr);
-  virtual int write(const std::string &s);
-  virtual std::string get_remote_addr() const;
-
-private:
-  socket_t sock_;
-};
-
-class BufferStream : public Stream {
-public:
-  BufferStream() {}
-  virtual ~BufferStream() {}
-
-  virtual int read(char *ptr, size_t size);
-  virtual int write(const char *ptr, size_t size);
-  virtual int write(const char *ptr);
-  virtual int write(const std::string &s);
-  virtual std::string get_remote_addr() const;
-
-  const std::string &get_buffer() const;
-
-private:
-  std::string buffer;
+  int write(const char *ptr);
+  int write(const std::string &s);
 };
 
 class TaskQueue {
 public:
-  TaskQueue() {}
-  virtual ~TaskQueue() {}
+  TaskQueue() = default;
+  virtual ~TaskQueue() = default;
   virtual void enqueue(std::function<void()> fn) = 0;
   virtual void shutdown() = 0;
 };
 
-#if CPPHTTPLIB_THREAD_POOL_COUNT > 0
 class ThreadPool : public TaskQueue {
 public:
-  ThreadPool(size_t n) : shutdown_(false) {
+  explicit ThreadPool(size_t n) : shutdown_(false) {
     while (n) {
-      auto t = std::make_shared<std::thread>(worker(*this));
-      threads_.push_back(t);
+      threads_.emplace_back(worker(*this));
       n--;
     }
   }
 
   ThreadPool(const ThreadPool &) = delete;
-  virtual ~ThreadPool() {}
+  ~ThreadPool() override = default;
 
-  virtual void enqueue(std::function<void()> fn) override {
+  void enqueue(std::function<void()> fn) override {
     std::unique_lock<std::mutex> lock(mutex_);
     jobs_.push_back(fn);
     cond_.notify_one();
   }
 
-  virtual void shutdown() override {
+  void shutdown() override {
     // Stop all worker threads...
     {
       std::unique_lock<std::mutex> lock(mutex_);
@@ -378,14 +395,14 @@ public:
     cond_.notify_all();
 
     // Join...
-    for (auto t : threads_) {
-      t->join();
+    for (auto &t : threads_) {
+      t.join();
     }
   }
 
 private:
   struct worker {
-    worker(ThreadPool &pool) : pool_(pool) {}
+    explicit worker(ThreadPool &pool) : pool_(pool) {}
 
     void operator()() {
       for (;;) {
@@ -411,7 +428,7 @@ private:
   };
   friend struct worker;
 
-  std::vector<std::shared_ptr<std::thread>> threads_;
+  std::vector<std::thread> threads_;
   std::list<std::function<void()>> jobs_;
 
   bool shutdown_;
@@ -419,46 +436,16 @@ private:
   std::condition_variable cond_;
   std::mutex mutex_;
 };
-#else
-class Threads : public TaskQueue {
-public:
-  Threads() : running_threads_(0) {}
-  virtual ~Threads() {}
-
-  virtual void enqueue(std::function<void()> fn) override {
-    std::thread([=]() {
-      {
-        std::lock_guard<std::mutex> guard(running_threads_mutex_);
-        running_threads_++;
-      }
-
-      fn();
-
-      {
-        std::lock_guard<std::mutex> guard(running_threads_mutex_);
-        running_threads_--;
-      }
-    }).detach();
-  }
-
-  virtual void shutdown() override {
-    for (;;) {
-      std::this_thread::sleep_for(std::chrono::milliseconds(10));
-      std::lock_guard<std::mutex> guard(running_threads_mutex_);
-      if (!running_threads_) { break; }
-    }
-  }
 
-private:
-  std::mutex running_threads_mutex_;
-  int running_threads_;
-};
-#endif
+using Logger = std::function<void(const Request &, const Response &)>;
 
 class Server {
 public:
-  typedef std::function<void(const Request &, Response &)> Handler;
-  typedef std::function<void(const Request &, const Response &)> Logger;
+  using Handler = std::function<void(const Request &, Response &)>;
+  using HandlerWithContentReader = std::function<void(
+      const Request &, Response &, const ContentReader &content_reader)>;
+  using Expect100ContinueHandler =
+      std::function<int(const Request &, Response &)>;
 
   Server();
 
@@ -468,21 +455,32 @@ public:
 
   Server &Get(const char *pattern, Handler handler);
   Server &Post(const char *pattern, Handler handler);
-
+  Server &Post(const char *pattern, HandlerWithContentReader handler);
   Server &Put(const char *pattern, Handler handler);
+  Server &Put(const char *pattern, HandlerWithContentReader handler);
   Server &Patch(const char *pattern, Handler handler);
+  Server &Patch(const char *pattern, HandlerWithContentReader handler);
   Server &Delete(const char *pattern, Handler handler);
   Server &Options(const char *pattern, Handler handler);
 
-  bool set_base_dir(const char *path);
+  [[deprecated]] bool set_base_dir(const char *dir,
+                                   const char *mount_point = nullptr);
+  bool set_mount_point(const char *mount_point, const char *dir);
+  bool remove_mount_point(const char *mount_point);
+  void set_file_extension_and_mimetype_mapping(const char *ext,
+                                               const char *mime);
   void set_file_request_handler(Handler handler);
 
   void set_error_handler(Handler handler);
   void set_logger(Logger logger);
 
+  void set_expect_100_continue_handler(Expect100ContinueHandler handler);
+
   void set_keep_alive_max_count(size_t count);
+  void set_read_timeout(time_t sec, time_t usec);
   void set_payload_max_length(size_t length);
 
+  bool bind_to_port(const char *host, int port, int socket_flags = 0);
   int bind_to_any_port(const char *host, int socket_flags = 0);
   bool listen_after_bind();
 
@@ -496,22 +494,29 @@ public:
 protected:
   bool process_request(Stream &strm, bool last_connection,
                        bool &connection_close,
-                       std::function<void(Request &)> setup_request);
+                       const std::function<void(Request &)> &setup_request);
 
   size_t keep_alive_max_count_;
+  time_t read_timeout_sec_;
+  time_t read_timeout_usec_;
   size_t payload_max_length_;
 
 private:
-  typedef std::vector<std::pair<std::regex, Handler>> Handlers;
+  using Handlers = std::vector<std::pair<std::regex, Handler>>;
+  using HandlersForContentReader =
+      std::vector<std::pair<std::regex, HandlerWithContentReader>>;
 
   socket_t create_server_socket(const char *host, int port,
                                 int socket_flags) const;
   int bind_internal(const char *host, int port, int socket_flags);
   bool listen_internal();
 
-  bool routing(Request &req, Response &res);
-  bool handle_file_request(Request &req, Response &res);
+  bool routing(Request &req, Response &res, Stream &strm, bool last_connection);
+  bool handle_file_request(Request &req, Response &res, bool head = false);
   bool dispatch_request(Request &req, Response &res, Handlers &handlers);
+  bool dispatch_request_for_content_reader(Request &req, Response &res,
+                                           ContentReader content_reader,
+                                           HandlersForContentReader &handlers);
 
   bool parse_request_line(const char *s, Request &req);
   bool write_response(Stream &strm, bool last_connection, const Request &req,
@@ -519,26 +524,43 @@ private:
   bool write_content_with_provider(Stream &strm, const Request &req,
                                    Response &res, const std::string &boundary,
                                    const std::string &content_type);
+  bool read_content(Stream &strm, bool last_connection, Request &req,
+                    Response &res);
+  bool read_content_with_content_receiver(
+      Stream &strm, bool last_connection, Request &req, Response &res,
+      ContentReceiver receiver, MultipartContentHeader multipart_header,
+      ContentReceiver multipart_receiver);
+  bool read_content_core(Stream &strm, bool last_connection, Request &req,
+                         Response &res, ContentReceiver receiver,
+                         MultipartContentHeader mulitpart_header,
+                         ContentReceiver multipart_receiver);
 
   virtual bool process_and_close_socket(socket_t sock);
 
   std::atomic<bool> is_running_;
   std::atomic<socket_t> svr_sock_;
-  std::string base_dir_;
+  std::vector<std::pair<std::string, std::string>> base_dirs_;
+  std::map<std::string, std::string> file_extension_and_mimetype_map_;
   Handler file_request_handler_;
   Handlers get_handlers_;
   Handlers post_handlers_;
+  HandlersForContentReader post_handlers_for_content_reader_;
   Handlers put_handlers_;
+  HandlersForContentReader put_handlers_for_content_reader_;
   Handlers patch_handlers_;
+  HandlersForContentReader patch_handlers_for_content_reader_;
   Handlers delete_handlers_;
   Handlers options_handlers_;
   Handler error_handler_;
   Logger logger_;
+  Expect100ContinueHandler expect_100_continue_handler_;
 };
 
 class Client {
 public:
-  Client(const char *host, int port = 80, time_t timeout_sec = 300);
+  explicit Client(const std::string &host, int port = 80,
+                  const std::string &client_cert_path = std::string(),
+                  const std::string &client_key_path = std::string());
 
   virtual ~Client();
 
@@ -586,6 +608,15 @@ public:
                                  const std::string &body,
                                  const char *content_type);
 
+  std::shared_ptr<Response> Post(const char *path, size_t content_length,
+                                 ContentProvider content_provider,
+                                 const char *content_type);
+
+  std::shared_ptr<Response> Post(const char *path, const Headers &headers,
+                                 size_t content_length,
+                                 ContentProvider content_provider,
+                                 const char *content_type);
+
   std::shared_ptr<Response> Post(const char *path, const Params &params);
 
   std::shared_ptr<Response> Post(const char *path, const Headers &headers,
@@ -604,6 +635,20 @@ public:
                                 const std::string &body,
                                 const char *content_type);
 
+  std::shared_ptr<Response> Put(const char *path, size_t content_length,
+                                ContentProvider content_provider,
+                                const char *content_type);
+
+  std::shared_ptr<Response> Put(const char *path, const Headers &headers,
+                                size_t content_length,
+                                ContentProvider content_provider,
+                                const char *content_type);
+
+  std::shared_ptr<Response> Put(const char *path, const Params &params);
+
+  std::shared_ptr<Response> Put(const char *path, const Headers &headers,
+                                const Params &params);
+
   std::shared_ptr<Response> Patch(const char *path, const std::string &body,
                                   const char *content_type);
 
@@ -611,6 +656,15 @@ public:
                                   const std::string &body,
                                   const char *content_type);
 
+  std::shared_ptr<Response> Patch(const char *path, size_t content_length,
+                                  ContentProvider content_provider,
+                                  const char *content_type);
+
+  std::shared_ptr<Response> Patch(const char *path, const Headers &headers,
+                                  size_t content_length,
+                                  ContentProvider content_provider,
+                                  const char *content_type);
+
   std::shared_ptr<Response> Delete(const char *path);
 
   std::shared_ptr<Response> Delete(const char *path, const std::string &body,
@@ -631,9 +685,33 @@ public:
   bool send(const std::vector<Request> &requests,
             std::vector<Response> &responses);
 
+  void set_timeout_sec(time_t timeout_sec);
+
+  void set_read_timeout(time_t sec, time_t usec);
+
   void set_keep_alive_max_count(size_t count);
 
-  void follow_location(bool on);
+  void set_basic_auth(const char *username, const char *password);
+
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+  void set_digest_auth(const char *username, const char *password);
+#endif
+
+  void set_follow_location(bool on);
+
+  void set_compress(bool on);
+
+  void set_interface(const char *intf);
+
+  void set_proxy(const char *host, int port);
+
+  void set_proxy_basic_auth(const char *username, const char *password);
+
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+  void set_proxy_digest_auth(const char *username, const char *password);
+#endif
+
+  void set_logger(Logger logger);
 
 protected:
   bool process_request(Stream &strm, const Request &req, Response &res,
@@ -641,16 +719,85 @@ protected:
 
   const std::string host_;
   const int port_;
-  time_t timeout_sec_;
   const std::string host_and_port_;
-  size_t keep_alive_max_count_;
-  size_t follow_location_;
+
+  // Settings
+  std::string client_cert_path_;
+  std::string client_key_path_;
+
+  time_t timeout_sec_ = 300;
+  time_t read_timeout_sec_ = CPPHTTPLIB_READ_TIMEOUT_SECOND;
+  time_t read_timeout_usec_ = CPPHTTPLIB_READ_TIMEOUT_USECOND;
+
+  size_t keep_alive_max_count_ = CPPHTTPLIB_KEEPALIVE_MAX_COUNT;
+
+  std::string basic_auth_username_;
+  std::string basic_auth_password_;
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+  std::string digest_auth_username_;
+  std::string digest_auth_password_;
+#endif
+
+  bool follow_location_ = false;
+
+  bool compress_ = false;
+
+  std::string interface_;
+
+  std::string proxy_host_;
+  int proxy_port_;
+
+  std::string proxy_basic_auth_username_;
+  std::string proxy_basic_auth_password_;
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+  std::string proxy_digest_auth_username_;
+  std::string proxy_digest_auth_password_;
+#endif
+
+  Logger logger_;
+
+  void copy_settings(const Client &rhs) {
+    client_cert_path_ = rhs.client_cert_path_;
+    client_key_path_ = rhs.client_key_path_;
+    timeout_sec_ = rhs.timeout_sec_;
+    read_timeout_sec_ = rhs.read_timeout_sec_;
+    read_timeout_usec_ = rhs.read_timeout_usec_;
+    keep_alive_max_count_ = rhs.keep_alive_max_count_;
+    basic_auth_username_ = rhs.basic_auth_username_;
+    basic_auth_password_ = rhs.basic_auth_password_;
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+    digest_auth_username_ = rhs.digest_auth_username_;
+    digest_auth_password_ = rhs.digest_auth_password_;
+#endif
+    follow_location_ = rhs.follow_location_;
+    compress_ = rhs.compress_;
+    interface_ = rhs.interface_;
+    proxy_host_ = rhs.proxy_host_;
+    proxy_port_ = rhs.proxy_port_;
+    proxy_basic_auth_username_ = rhs.proxy_basic_auth_username_;
+    proxy_basic_auth_password_ = rhs.proxy_basic_auth_password_;
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+    proxy_digest_auth_username_ = rhs.proxy_digest_auth_username_;
+    proxy_digest_auth_password_ = rhs.proxy_digest_auth_password_;
+#endif
+    logger_ = rhs.logger_;
+  }
 
 private:
   socket_t create_client_socket() const;
   bool read_response_line(Stream &strm, Response &res);
-  void write_request(Stream &strm, const Request &req, bool last_connection);
+  bool write_request(Stream &strm, const Request &req, bool last_connection);
   bool redirect(const Request &req, Response &res);
+  bool handle_request(Stream &strm, const Request &req, Response &res,
+                      bool last_connection, bool &connection_close);
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+  bool connect(socket_t sock, Response &res, bool &error);
+#endif
+
+  std::shared_ptr<Response> send_with_content_provider(
+      const char *method, const char *path, const Headers &headers,
+      const std::string &body, size_t content_length,
+      ContentProvider content_provider, const char *content_type);
 
   virtual bool process_and_close_socket(
       socket_t sock, size_t request_count,
@@ -692,22 +839,6 @@ inline void Post(std::vector<Request> &requests, const char *path,
 }
 
 #ifdef CPPHTTPLIB_OPENSSL_SUPPORT
-class SSLSocketStream : public Stream {
-public:
-  SSLSocketStream(socket_t sock, SSL *ssl);
-  virtual ~SSLSocketStream();
-
-  virtual int read(char *ptr, size_t size);
-  virtual int write(const char *ptr, size_t size);
-  virtual int write(const char *ptr);
-  virtual int write(const std::string &s);
-  virtual std::string get_remote_addr() const;
-
-private:
-  socket_t sock_;
-  SSL *ssl_;
-};
-
 class SSLServer : public Server {
 public:
   SSLServer(const char *cert_path, const char *private_key_path,
@@ -727,9 +858,9 @@ private:
 
 class SSLClient : public Client {
 public:
-  SSLClient(const char *host, int port = 443, time_t timeout_sec = 300,
-            const char *client_cert_path = nullptr,
-            const char *client_key_path = nullptr);
+  SSLClient(const std::string &host, int port = 443,
+            const std::string &client_cert_path = std::string(),
+            const std::string &client_key_path = std::string());
 
   virtual ~SSLClient();
 
@@ -737,11 +868,12 @@ public:
 
   void set_ca_cert_path(const char *ca_ceert_file_path,
                         const char *ca_cert_dir_path = nullptr);
+
   void enable_server_certificate_verification(bool enabled);
 
   long get_openssl_verify_result() const;
 
-  SSL_CTX* ssl_context() const noexcept;
+  SSL_CTX *ssl_context() const noexcept;
 
 private:
   virtual bool process_and_close_socket(
@@ -759,6 +891,7 @@ private:
   SSL_CTX *ctx_;
   std::mutex ctx_mutex_;
   std::vector<std::string> host_components_;
+
   std::string ca_cert_file_path_;
   std::string ca_cert_dir_path_;
   bool server_certificate_verification_ = false;
@@ -766,9 +899,12 @@ private:
 };
 #endif
 
+// ----------------------------------------------------------------------------
+
 /*
  * Implementation
  */
+
 namespace detail {
 
 inline bool is_hex(char c, int &v) {
@@ -932,8 +1068,8 @@ inline void read_file(const std::string &path, std::string &out) {
 
 inline std::string file_extension(const std::string &path) {
   std::smatch m;
-  auto pat = std::regex("\\.([a-zA-Z0-9]+)$");
-  if (std::regex_search(path, m, pat)) { return m[1].str(); }
+  static auto re = std::regex("\\.([a-zA-Z0-9]+)$");
+  if (std::regex_search(path, m, re)) { return m[1].str(); }
   return std::string();
 }
 
@@ -976,6 +1112,11 @@ public:
     }
   }
 
+  bool end_with_crlf() const {
+    auto end = ptr() + size();
+    return size() >= 2 && end[-2] == '\r' && end[-1] == '\n';
+  }
+
   bool getline() {
     fixed_buffer_used_size_ = 0;
     glowable_buffer_.clear();
@@ -1019,7 +1160,7 @@ private:
   Stream &strm_;
   char *fixed_buffer_;
   const size_t fixed_buffer_size_;
-  size_t fixed_buffer_used_size_;
+  size_t fixed_buffer_used_size_ = 0;
   std::string glowable_buffer_;
 };
 
@@ -1053,6 +1194,28 @@ inline int select_read(socket_t sock, time_t sec, time_t usec) {
 #endif
 }
 
+inline int select_write(socket_t sock, time_t sec, time_t usec) {
+#ifdef CPPHTTPLIB_USE_POLL
+  struct pollfd pfd_read;
+  pfd_read.fd = sock;
+  pfd_read.events = POLLOUT;
+
+  auto timeout = static_cast<int>(sec * 1000 + usec / 1000);
+
+  return poll(&pfd_read, 1, timeout);
+#else
+  fd_set fds;
+  FD_ZERO(&fds);
+  FD_SET(sock, &fds);
+
+  timeval tv;
+  tv.tv_sec = static_cast<long>(sec);
+  tv.tv_usec = static_cast<long>(usec);
+
+  return select(static_cast<int>(sock + 1), nullptr, &fds, nullptr, &tv);
+#endif
+}
+
 inline bool wait_until_socket_is_ready(socket_t sock, time_t sec, time_t usec) {
 #ifdef CPPHTTPLIB_USE_POLL
   struct pollfd pfd_read;
@@ -1065,7 +1228,8 @@ inline bool wait_until_socket_is_ready(socket_t sock, time_t sec, time_t usec) {
       pfd_read.revents & (POLLIN | POLLOUT)) {
     int error = 0;
     socklen_t len = sizeof(error);
-    return getsockopt(sock, SOL_SOCKET, SO_ERROR, reinterpret_cast<char*>(&error), &len) >= 0 &&
+    return getsockopt(sock, SOL_SOCKET, SO_ERROR,
+                      reinterpret_cast<char *>(&error), &len) >= 0 &&
            !error;
   }
   return false;
@@ -1085,27 +1249,86 @@ inline bool wait_until_socket_is_ready(socket_t sock, time_t sec, time_t usec) {
       (FD_ISSET(sock, &fdsr) || FD_ISSET(sock, &fdsw))) {
     int error = 0;
     socklen_t len = sizeof(error);
-    return getsockopt(sock, SOL_SOCKET, SO_ERROR, (char *)&error, &len) >= 0 &&
+    return getsockopt(sock, SOL_SOCKET, SO_ERROR,
+                      reinterpret_cast<char *>(&error), &len) >= 0 &&
            !error;
   }
   return false;
 #endif
 }
 
+class SocketStream : public Stream {
+public:
+  SocketStream(socket_t sock, time_t read_timeout_sec,
+               time_t read_timeout_usec);
+  ~SocketStream() override;
+
+  bool is_readable() const override;
+  bool is_writable() const override;
+  int read(char *ptr, size_t size) override;
+  int write(const char *ptr, size_t size) override;
+  std::string get_remote_addr() const override;
+
+private:
+  socket_t sock_;
+  time_t read_timeout_sec_;
+  time_t read_timeout_usec_;
+};
+
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+class SSLSocketStream : public Stream {
+public:
+  SSLSocketStream(socket_t sock, SSL *ssl, time_t read_timeout_sec,
+                  time_t read_timeout_usec);
+  virtual ~SSLSocketStream();
+
+  bool is_readable() const override;
+  bool is_writable() const override;
+  int read(char *ptr, size_t size) override;
+  int write(const char *ptr, size_t size) override;
+  std::string get_remote_addr() const override;
+
+private:
+  socket_t sock_;
+  SSL *ssl_;
+  time_t read_timeout_sec_;
+  time_t read_timeout_usec_;
+};
+#endif
+
+class BufferStream : public Stream {
+public:
+  BufferStream() = default;
+  ~BufferStream() override = default;
+
+  bool is_readable() const override;
+  bool is_writable() const override;
+  int read(char *ptr, size_t size) override;
+  int write(const char *ptr, size_t size) override;
+  std::string get_remote_addr() const override;
+
+  const std::string &get_buffer() const;
+
+private:
+  std::string buffer;
+  int position = 0;
+};
+
 template <typename T>
-inline bool process_and_close_socket(bool is_client_request, socket_t sock,
-                                     size_t keep_alive_max_count, T callback) {
+inline bool process_socket(bool is_client_request, socket_t sock,
+                           size_t keep_alive_max_count, time_t read_timeout_sec,
+                           time_t read_timeout_usec, T callback) {
   assert(keep_alive_max_count > 0);
 
-  bool ret = false;
+  auto ret = false;
 
   if (keep_alive_max_count > 1) {
     auto count = keep_alive_max_count;
     while (count > 0 &&
            (is_client_request ||
-            detail::select_read(sock, CPPHTTPLIB_KEEPALIVE_TIMEOUT_SECOND,
-                                CPPHTTPLIB_KEEPALIVE_TIMEOUT_USECOND) > 0)) {
-      SocketStream strm(sock);
+            select_read(sock, CPPHTTPLIB_KEEPALIVE_TIMEOUT_SECOND,
+                        CPPHTTPLIB_KEEPALIVE_TIMEOUT_USECOND) > 0)) {
+      SocketStream strm(sock, read_timeout_sec, read_timeout_usec);
       auto last_connection = count == 1;
       auto connection_close = false;
 
@@ -1114,12 +1337,22 @@ inline bool process_and_close_socket(bool is_client_request, socket_t sock,
 
       count--;
     }
-  } else {
-    SocketStream strm(sock);
+  } else { // keep_alive_max_count  is 0 or 1
+    SocketStream strm(sock, read_timeout_sec, read_timeout_usec);
     auto dummy_connection_close = false;
     ret = callback(strm, true, dummy_connection_close);
   }
 
+  return ret;
+}
+
+template <typename T>
+inline bool process_and_close_socket(bool is_client_request, socket_t sock,
+                                     size_t keep_alive_max_count,
+                                     time_t read_timeout_sec,
+                                     time_t read_timeout_usec, T callback) {
+  auto ret = process_socket(is_client_request, sock, keep_alive_max_count,
+                            read_timeout_sec, read_timeout_usec, callback);
   close_socket(sock);
   return ret;
 }
@@ -1165,6 +1398,23 @@ socket_t create_socket(const char *host, int port, Fn fn,
 #ifdef _WIN32
     auto sock = WSASocketW(rp->ai_family, rp->ai_socktype, rp->ai_protocol,
                            nullptr, 0, WSA_FLAG_NO_HANDLE_INHERIT);
+    /**
+     * Since the WSA_FLAG_NO_HANDLE_INHERIT is only supported on Windows 7 SP1
+     * and above the socket creation fails on older Windows Systems.
+     *
+     * Let's try to create a socket the old way in this case.
+     *
+     * Reference:
+     * https://docs.microsoft.com/en-us/windows/win32/api/winsock2/nf-winsock2-wsasocketa
+     *
+     * WSA_FLAG_NO_HANDLE_INHERIT:
+     * This flag is supported on Windows 7 with SP1, Windows Server 2008 R2 with
+     * SP1, and later
+     *
+     */
+    if (sock == INVALID_SOCKET) {
+      sock = socket(rp->ai_family, rp->ai_socktype, rp->ai_protocol);
+    }
 #else
     auto sock = socket(rp->ai_family, rp->ai_socktype, rp->ai_protocol);
 #endif
@@ -1176,9 +1426,11 @@ socket_t create_socket(const char *host, int port, Fn fn,
 
     // Make 'reuse address' option available
     int yes = 1;
-    setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, reinterpret_cast<char*>(&yes), sizeof(yes));
+    setsockopt(sock, SOL_SOCKET, SO_REUSEADDR, reinterpret_cast<char *>(&yes),
+               sizeof(yes));
 #ifdef SO_REUSEPORT
-    setsockopt(sock, SOL_SOCKET, SO_REUSEPORT, reinterpret_cast<char*>(&yes), sizeof(yes));
+    setsockopt(sock, SOL_SOCKET, SO_REUSEPORT, reinterpret_cast<char *>(&yes),
+               sizeof(yes));
 #endif
 
     // bind or connect
@@ -1213,27 +1465,105 @@ inline bool is_connection_error() {
 #endif
 }
 
+inline bool bind_ip_address(socket_t sock, const char *host) {
+  struct addrinfo hints;
+  struct addrinfo *result;
+
+  memset(&hints, 0, sizeof(struct addrinfo));
+  hints.ai_family = AF_UNSPEC;
+  hints.ai_socktype = SOCK_STREAM;
+  hints.ai_protocol = 0;
+
+  if (getaddrinfo(host, "0", &hints, &result)) { return false; }
+
+  auto ret = false;
+  for (auto rp = result; rp; rp = rp->ai_next) {
+    const auto &ai = *rp;
+    if (!::bind(sock, ai.ai_addr, static_cast<int>(ai.ai_addrlen))) {
+      ret = true;
+      break;
+    }
+  }
+
+  freeaddrinfo(result);
+  return ret;
+}
+
+inline std::string if2ip(const std::string &ifn) {
+#ifndef _WIN32
+  struct ifaddrs *ifap;
+  getifaddrs(&ifap);
+  for (auto ifa = ifap; ifa; ifa = ifa->ifa_next) {
+    if (ifa->ifa_addr && ifn == ifa->ifa_name) {
+      if (ifa->ifa_addr->sa_family == AF_INET) {
+        auto sa = reinterpret_cast<struct sockaddr_in *>(ifa->ifa_addr);
+        char buf[INET_ADDRSTRLEN];
+        if (inet_ntop(AF_INET, &sa->sin_addr, buf, INET_ADDRSTRLEN)) {
+          freeifaddrs(ifap);
+          return std::string(buf, INET_ADDRSTRLEN);
+        }
+      }
+    }
+  }
+  freeifaddrs(ifap);
+#endif
+  return std::string();
+}
+
+inline socket_t create_client_socket(const char *host, int port,
+                                     time_t timeout_sec,
+                                     const std::string &intf) {
+  return create_socket(
+      host, port, [&](socket_t sock, struct addrinfo &ai) -> bool {
+        if (!intf.empty()) {
+          auto ip = if2ip(intf);
+          if (ip.empty()) { ip = intf; }
+          if (!bind_ip_address(sock, ip.c_str())) { return false; }
+        }
+
+        set_nonblocking(sock, true);
+
+        auto ret = ::connect(sock, ai.ai_addr, static_cast<int>(ai.ai_addrlen));
+        if (ret < 0) {
+          if (is_connection_error() ||
+              !wait_until_socket_is_ready(sock, timeout_sec, 0)) {
+            close_socket(sock);
+            return false;
+          }
+        }
+
+        set_nonblocking(sock, false);
+        return true;
+      });
+}
+
 inline std::string get_remote_addr(socket_t sock) {
   struct sockaddr_storage addr;
   socklen_t len = sizeof(addr);
 
   if (!getpeername(sock, reinterpret_cast<struct sockaddr *>(&addr), &len)) {
-    char ipstr[NI_MAXHOST];
+    std::array<char, NI_MAXHOST> ipstr{};
 
-    if (!getnameinfo(reinterpret_cast<struct sockaddr *>(&addr), len, ipstr, sizeof(ipstr),
-                     nullptr, 0, NI_NUMERICHOST)) {
-      return ipstr;
+    if (!getnameinfo(reinterpret_cast<struct sockaddr *>(&addr), len,
+                     ipstr.data(), static_cast<unsigned int>(ipstr.size()), nullptr, 0, NI_NUMERICHOST)) {
+      return ipstr.data();
     }
   }
 
   return std::string();
 }
 
-inline const char *find_content_type(const std::string &path) {
+inline const char *
+find_content_type(const std::string &path,
+                  const std::map<std::string, std::string> &user_data) {
   auto ext = file_extension(path);
+
+  auto it = user_data.find(ext);
+  if (it != user_data.end()) { return it->second.c_str(); }
+
   if (ext == "txt") {
     return "text/plain";
-  } else if (ext == "html") {
+  } else if (ext == "html" || ext == "htm") {
     return "text/html";
   } else if (ext == "css") {
     return "text/css";
@@ -1253,6 +1583,8 @@ inline const char *find_content_type(const std::string &path) {
     return "application/pdf";
   } else if (ext == "js") {
     return "application/javascript";
+  } else if (ext == "wasm") {
+    return "application/wasm";
   } else if (ext == "xml") {
     return "application/xml";
   } else if (ext == "xhtml") {
@@ -1263,19 +1595,25 @@ inline const char *find_content_type(const std::string &path) {
 
 inline const char *status_message(int status) {
   switch (status) {
+  case 100: return "Continue";
   case 200: return "OK";
+  case 202: return "Accepted";
+  case 204: return "No Content";
   case 206: return "Partial Content";
   case 301: return "Moved Permanently";
   case 302: return "Found";
   case 303: return "See Other";
   case 304: return "Not Modified";
   case 400: return "Bad Request";
+  case 401: return "Unauthorized";
   case 403: return "Forbidden";
   case 404: return "Not Found";
   case 413: return "Payload Too Large";
   case 414: return "Request-URI Too Long";
   case 415: return "Unsupported Media Type";
   case 416: return "Range Not Satisfiable";
+  case 417: return "Expectation Failed";
+  case 503: return "Service Unavailable";
 
   default:
   case 500: return "Internal Server Error";
@@ -1302,18 +1640,18 @@ inline bool compress(std::string &content) {
   if (ret != Z_OK) { return false; }
 
   strm.avail_in = content.size();
-  strm.next_in = const_cast<Bytef*>(reinterpret_cast<const Bytef*>(content.data()));
+  strm.next_in =
+      const_cast<Bytef *>(reinterpret_cast<const Bytef *>(content.data()));
 
   std::string compressed;
 
-  const auto bufsiz = 16384;
-  char buff[bufsiz];
+  std::array<char, 16384> buff{};
   do {
-    strm.avail_out = bufsiz;
-    strm.next_out = reinterpret_cast<Bytef*>(buff);
+    strm.avail_out = buff.size();
+    strm.next_out = reinterpret_cast<Bytef *>(buff.data());
     ret = deflate(&strm, Z_FINISH);
     assert(ret != Z_STREAM_ERROR);
-    compressed.append(buff, bufsiz - strm.avail_out);
+    compressed.append(buff.data(), buff.size() - strm.avail_out);
   } while (strm.avail_out == 0);
 
   assert(ret == Z_STREAM_END);
@@ -1347,13 +1685,12 @@ public:
     int ret = Z_OK;
 
     strm.avail_in = data_length;
-    strm.next_in = const_cast<Bytef*>(reinterpret_cast<const Bytef *>(data));
+    strm.next_in = const_cast<Bytef *>(reinterpret_cast<const Bytef *>(data));
 
-    const auto bufsiz = 16384;
-    char buff[bufsiz];
+    std::array<char, 16384> buff{};
     do {
-      strm.avail_out = bufsiz;
-      strm.next_out = reinterpret_cast<Bytef*>(buff);
+      strm.avail_out = buff.size();
+      strm.next_out = reinterpret_cast<Bytef *>(buff.data());
 
       ret = inflate(&strm, Z_NO_FLUSH);
       assert(ret != Z_STREAM_ERROR);
@@ -1363,10 +1700,12 @@ public:
       case Z_MEM_ERROR: inflateEnd(&strm); return false;
       }
 
-      if (!callback(buff, bufsiz - strm.avail_out)) { return false; }
+      if (!callback(buff.data(), buff.size() - strm.avail_out)) {
+        return false;
+      }
     } while (strm.avail_out == 0);
 
-    return ret == Z_STREAM_END;
+    return ret == Z_OK || ret == Z_STREAM_END;
   }
 
 private:
@@ -1397,18 +1736,35 @@ inline uint64_t get_header_value_uint64(const Headers &headers, const char *key,
 }
 
 inline bool read_headers(Stream &strm, Headers &headers) {
-  static std::regex re(R"((.+?):\s*(.+?)\s*\r\n)");
-
   const auto bufsiz = 2048;
   char buf[bufsiz];
-
-  stream_line_reader reader(strm, buf, bufsiz);
+  stream_line_reader line_reader(strm, buf, bufsiz);
 
   for (;;) {
-    if (!reader.getline()) { return false; }
-    if (!strcmp(reader.ptr(), "\r\n")) { break; }
+    if (!line_reader.getline()) { return false; }
+
+    // Check if the line ends with CRLF.
+    if (line_reader.end_with_crlf()) {
+      // Blank line indicates end of headers.
+      if (line_reader.size() == 2) { break; }
+    } else {
+      continue; // Skip invalid line.
+    }
+
+    // Skip trailing spaces and tabs.
+    auto end = line_reader.ptr() + line_reader.size() - 2;
+    while (line_reader.ptr() < end && (end[-1] == ' ' || end[-1] == '\t')) {
+      end--;
+    }
+
+    // Horizontal tab and ' ' are considered whitespace and are ignored when on
+    // the left or right side of the header value:
+    //  - https://stackoverflow.com/questions/50179659/
+    //  - https://www.w3.org/Protocols/rfc2616/rfc2616-sec4.html
+    static const std::regex re(R"((.+?):[\t ]*(.+))");
+
     std::cmatch m;
-    if (std::regex_match(reader.ptr(), m, re)) {
+    if (std::regex_match(line_reader.ptr(), end, m, re)) {
       auto key = std::string(m[1]);
       auto val = std::string(m[2]);
       headers.emplace(key, val);
@@ -1418,12 +1774,8 @@ inline bool read_headers(Stream &strm, Headers &headers) {
   return true;
 }
 
-typedef std::function<bool(const char *data, size_t data_length)>
-    ContentReceiverCore;
-
 inline bool read_content_with_length(Stream &strm, uint64_t len,
-                                     Progress progress,
-                                     ContentReceiverCore out) {
+                                     Progress progress, ContentReceiver out) {
   char buf[CPPHTTPLIB_RECV_BUFSIZ];
 
   uint64_t r = 0;
@@ -1455,7 +1807,7 @@ inline void skip_content_with_length(Stream &strm, uint64_t len) {
   }
 }
 
-inline bool read_content_without_length(Stream &strm, ContentReceiverCore out) {
+inline bool read_content_without_length(Stream &strm, ContentReceiver out) {
   char buf[CPPHTTPLIB_RECV_BUFSIZ];
   for (;;) {
     auto n = strm.read(buf, CPPHTTPLIB_RECV_BUFSIZ);
@@ -1470,33 +1822,34 @@ inline bool read_content_without_length(Stream &strm, ContentReceiverCore out) {
   return true;
 }
 
-inline bool read_content_chunked(Stream &strm, ContentReceiverCore out) {
+inline bool read_content_chunked(Stream &strm, ContentReceiver out) {
   const auto bufsiz = 16;
   char buf[bufsiz];
 
-  stream_line_reader reader(strm, buf, bufsiz);
+  stream_line_reader line_reader(strm, buf, bufsiz);
 
-  if (!reader.getline()) { return false; }
+  if (!line_reader.getline()) { return false; }
 
-  auto chunk_len = std::stoi(reader.ptr(), 0, 16);
+  auto chunk_len = std::stoi(line_reader.ptr(), 0, 16);
 
   while (chunk_len > 0) {
     if (!read_content_with_length(strm, chunk_len, nullptr, out)) {
       return false;
     }
 
-    if (!reader.getline()) { return false; }
+    if (!line_reader.getline()) { return false; }
 
-    if (strcmp(reader.ptr(), "\r\n")) { break; }
+    if (strcmp(line_reader.ptr(), "\r\n")) { break; }
 
-    if (!reader.getline()) { return false; }
+    if (!line_reader.getline()) { return false; }
 
-    chunk_len = std::stoi(reader.ptr(), 0, 16);
+    chunk_len = std::stoi(line_reader.ptr(), 0, 16);
   }
 
   if (chunk_len == 0) {
     // Reader terminator after chunks
-    if (!reader.getline() || strcmp(reader.ptr(), "\r\n")) return false;
+    if (!line_reader.getline() || strcmp(line_reader.ptr(), "\r\n"))
+      return false;
   }
 
   return true;
@@ -1509,14 +1862,14 @@ inline bool is_chunked_transfer_encoding(const Headers &headers) {
 
 template <typename T>
 bool read_content(Stream &strm, T &x, size_t payload_max_length, int &status,
-                  Progress progress, ContentReceiverCore receiver) {
+                  Progress progress, ContentReceiver receiver) {
 
-  ContentReceiverCore out = [&](const char *buf, size_t n) {
+  ContentReceiver out = [&](const char *buf, size_t n) {
     return receiver(buf, n);
   };
 
 #ifdef CPPHTTPLIB_ZLIB_SUPPORT
-  detail::decompressor decompressor;
+  decompressor decompressor;
 
   if (!decompressor.is_valid()) {
     status = 500;
@@ -1586,39 +1939,47 @@ inline ssize_t write_content(Stream &strm, ContentProvider content_provider,
   size_t end_offset = offset + length;
   while (offset < end_offset) {
     ssize_t written_length = 0;
-    content_provider(
-        offset, end_offset - offset,
-        [&](const char *d, size_t l) {
-          offset += l;
-          written_length = strm.write(d, l);
-        },
-        [&](void) { written_length = -1; });
+
+    DataSink data_sink;
+    data_sink.write = [&](const char *d, size_t l) {
+      offset += l;
+      written_length = strm.write(d, l);
+    };
+    data_sink.done = [&](void) { written_length = -1; };
+    data_sink.is_writable = [&](void) { return strm.is_writable(); };
+
+    content_provider(offset, end_offset - offset, data_sink);
     if (written_length < 0) { return written_length; }
   }
   return static_cast<ssize_t>(offset - begin_offset);
 }
 
+template <typename T>
 inline ssize_t write_content_chunked(Stream &strm,
-                                     ContentProvider content_provider) {
+                                     ContentProvider content_provider,
+                                     T is_shutting_down) {
   size_t offset = 0;
   auto data_available = true;
   ssize_t total_written_length = 0;
-  while (data_available) {
+  while (data_available && !is_shutting_down()) {
     ssize_t written_length = 0;
-    content_provider(
-        offset, 0,
-        [&](const char *d, size_t l) {
-          data_available = l > 0;
-          offset += l;
-
-          // Emit chunked response header and footer for each chunk
-          auto chunk = from_i_to_hex(l) + "\r\n" + std::string(d, l) + "\r\n";
-          written_length = strm.write(chunk);
-        },
-        [&](void) {
-          data_available = false;
-          written_length = strm.write("0\r\n\r\n");
-        });
+
+    DataSink data_sink;
+    data_sink.write = [&](const char *d, size_t l) {
+      data_available = l > 0;
+      offset += l;
+
+      // Emit chunked response header and footer for each chunk
+      auto chunk = from_i_to_hex(l) + "\r\n" + std::string(d, l) + "\r\n";
+      written_length = strm.write(chunk);
+    };
+    data_sink.done = [&](void) {
+      data_available = false;
+      written_length = strm.write("0\r\n\r\n");
+    };
+    data_sink.is_writable = [&](void) { return strm.is_writable(); };
+
+    content_provider(offset, 0, data_sink);
 
     if (written_length < 0) { return written_length; }
     total_written_length += written_length;
@@ -1629,17 +1990,12 @@ inline ssize_t write_content_chunked(Stream &strm,
 template <typename T>
 inline bool redirect(T &cli, const Request &req, Response &res,
                      const std::string &path) {
-  Request new_req;
-  new_req.method = req.method;
+  Request new_req = req;
   new_req.path = path;
-  new_req.headers = req.headers;
-  new_req.body = req.body;
-  new_req.redirect_count = req.redirect_count - 1;
-  new_req.response_handler = req.response_handler;
-  new_req.content_receiver = req.content_receiver;
-  new_req.progress = req.progress;
+  new_req.redirect_count -= 1;
 
   Response new_res;
+
   auto ret = cli.send(new_req, new_res);
   if (ret) { res = new_res; }
   return ret;
@@ -1656,7 +2012,7 @@ inline std::string encode_url(const std::string &s) {
     case '\n': result += "%0A"; break;
     case '\'': result += "%27"; break;
     case ',': result += "%2C"; break;
-    case ':': result += "%3A"; break;
+    // case ':': result += "%3A"; break; // ok? probably...
     case ';': result += "%3B"; break;
     default:
       auto c = static_cast<uint8_t>(s[i]);
@@ -1716,11 +2072,11 @@ inline void parse_query_text(const std::string &s, Params &params) {
   split(&s[0], &s[s.size()], '&', [&](const char *b, const char *e) {
     std::string key;
     std::string val;
-    split(b, e, '=', [&](const char *b, const char *e) {
+    split(b, e, '=', [&](const char *b2, const char *e2) {
       if (key.empty()) {
-        key.assign(b, e);
+        key.assign(b2, e2);
       } else {
-        val.assign(b, e);
+        val.assign(b2, e2);
       }
     });
     params.emplace(key, decode_url(val));
@@ -1736,112 +2092,207 @@ inline bool parse_multipart_boundary(const std::string &content_type,
   return true;
 }
 
-inline bool parse_multipart_formdata(const std::string &boundary,
-                                     const std::string &body,
-                                     MultipartFiles &files) {
-  static std::string dash = "--";
-  static std::string crlf = "\r\n";
-
-  static std::regex re_content_type("Content-Type: (.*?)",
-                                    std::regex_constants::icase);
-
-  static std::regex re_content_disposition(
-      "Content-Disposition: form-data; name=\"(.*?)\"(?:; filename=\"(.*?)\")?",
-      std::regex_constants::icase);
-
-  auto dash_boundary = dash + boundary;
-
-  auto pos = body.find(dash_boundary);
-  if (pos != 0) { return false; }
-
-  pos += dash_boundary.size();
+inline bool parse_range_header(const std::string &s, Ranges &ranges) {
+  static auto re_first_range = std::regex(R"(bytes=(\d*-\d*(?:,\s*\d*-\d*)*))");
+  std::smatch m;
+  if (std::regex_match(s, m, re_first_range)) {
+    auto pos = m.position(1);
+    auto len = m.length(1);
+    bool all_valid_ranges = true;
+    split(&s[pos], &s[pos + len], ',', [&](const char *b, const char *e) {
+      if (!all_valid_ranges) return;
+      static auto re_another_range = std::regex(R"(\s*(\d*)-(\d*))");
+      std::cmatch cm;
+      if (std::regex_match(b, e, cm, re_another_range)) {
+        ssize_t first = -1;
+        if (!cm.str(1).empty()) {
+          first = static_cast<ssize_t>(std::stoll(cm.str(1)));
+        }
 
-  auto next_pos = body.find(crlf, pos);
-  if (next_pos == std::string::npos) { return false; }
+        ssize_t last = -1;
+        if (!cm.str(2).empty()) {
+          last = static_cast<ssize_t>(std::stoll(cm.str(2)));
+        }
 
-  pos = next_pos + crlf.size();
+        if (first != -1 && last != -1 && first > last) {
+          all_valid_ranges = false;
+          return;
+        }
+        ranges.emplace_back(std::make_pair(first, last));
+      }
+    });
+    return all_valid_ranges;
+  }
+  return false;
+}
 
-  while (pos < body.size()) {
-    next_pos = body.find(crlf, pos);
-    if (next_pos == std::string::npos) { return false; }
+class MultipartFormDataParser {
+public:
+  MultipartFormDataParser() {}
 
-    std::string name;
-    MultipartFile file;
+  void set_boundary(const std::string &boundary) { boundary_ = boundary; }
 
-    auto header = body.substr(pos, (next_pos - pos));
+  bool is_valid() const { return is_valid_; }
 
-    while (pos != next_pos) {
-      std::smatch m;
-      if (std::regex_match(header, m, re_content_type)) {
-        file.content_type = m[1];
-      } else if (std::regex_match(header, m, re_content_disposition)) {
-        name = m[1];
-        file.filename = m[2];
+  template <typename T, typename U>
+  bool parse(const char *buf, size_t n, T content_callback, U header_callback) {
+    static const std::regex re_content_type(R"(^Content-Type:\s*(.*?)\s*$)",
+                                            std::regex_constants::icase);
+
+    static const std::regex re_content_disposition(
+        "^Content-Disposition:\\s*form-data;\\s*name=\"(.*?)\"(?:;\\s*filename="
+        "\"(.*?)\")?\\s*$",
+        std::regex_constants::icase);
+
+    buf_.append(buf, n); // TODO: performance improvement
+
+    while (!buf_.empty()) {
+      switch (state_) {
+      case 0: { // Initial boundary
+        auto pattern = dash_ + boundary_ + crlf_;
+        if (pattern.size() > buf_.size()) { return true; }
+        auto pos = buf_.find(pattern);
+        if (pos != 0) {
+          is_done_ = true;
+          return false;
+        }
+        buf_.erase(0, pattern.size());
+        off_ += pattern.size();
+        state_ = 1;
+        break;
       }
+      case 1: { // New entry
+        clear_file_info();
+        state_ = 2;
+        break;
+      }
+      case 2: { // Headers
+        auto pos = buf_.find(crlf_);
+        while (pos != std::string::npos) {
+          // Empty line
+          if (pos == 0) {
+            if (!header_callback(file_)) {
+              is_valid_ = false;
+              is_done_ = false;
+              return false;
+            }
+            buf_.erase(0, crlf_.size());
+            off_ += crlf_.size();
+            state_ = 3;
+            break;
+          }
 
-      pos = next_pos + crlf.size();
-
-      next_pos = body.find(crlf, pos);
-      if (next_pos == std::string::npos) { return false; }
-
-      header = body.substr(pos, (next_pos - pos));
-    }
-
-    pos = next_pos + crlf.size();
+          auto header = buf_.substr(0, pos);
+          {
+            std::smatch m;
+            if (std::regex_match(header, m, re_content_type)) {
+              file_.content_type = m[1];
+            } else if (std::regex_match(header, m, re_content_disposition)) {
+              file_.name = m[1];
+              file_.filename = m[2];
+            }
+          }
 
-    next_pos = body.find(crlf + dash_boundary, pos);
+          buf_.erase(0, pos + crlf_.size());
+          off_ += pos + crlf_.size();
+          pos = buf_.find(crlf_);
+        }
+        break;
+      }
+      case 3: { // Body
+        {
+          auto pattern = crlf_ + dash_;
+          if (pattern.size() > buf_.size()) { return true; }
+
+          auto pos = buf_.find(pattern);
+          if (pos == std::string::npos) { pos = buf_.size(); }
+          if (!content_callback(buf_.data(), pos)) {
+            is_valid_ = false;
+            is_done_ = false;
+            return false;
+          }
 
-    if (next_pos == std::string::npos) { return false; }
+          off_ += pos;
+          buf_.erase(0, pos);
+        }
 
-    file.offset = pos;
-    file.length = next_pos - pos;
+        {
+          auto pattern = crlf_ + dash_ + boundary_;
+          if (pattern.size() > buf_.size()) { return true; }
+
+          auto pos = buf_.find(pattern);
+          if (pos != std::string::npos) {
+            if (!content_callback(buf_.data(), pos)) {
+              is_valid_ = false;
+              is_done_ = false;
+              return false;
+            }
 
-    pos = next_pos + crlf.size() + dash_boundary.size();
+            off_ += pos + pattern.size();
+            buf_.erase(0, pos + pattern.size());
+            state_ = 4;
+          } else {
+            if (!content_callback(buf_.data(), pattern.size())) {
+              is_valid_ = false;
+              is_done_ = false;
+              return false;
+            }
 
-    next_pos = body.find(crlf, pos);
-    if (next_pos == std::string::npos) { return false; }
+            off_ += pattern.size();
+            buf_.erase(0, pattern.size());
+          }
+        }
+        break;
+      }
+      case 4: { // Boundary
+        if (crlf_.size() > buf_.size()) { return true; }
+        if (buf_.find(crlf_) == 0) {
+          buf_.erase(0, crlf_.size());
+          off_ += crlf_.size();
+          state_ = 1;
+        } else {
+          auto pattern = dash_ + crlf_;
+          if (pattern.size() > buf_.size()) { return true; }
+          if (buf_.find(pattern) == 0) {
+            buf_.erase(0, pattern.size());
+            off_ += pattern.size();
+            is_valid_ = true;
+            state_ = 5;
+          } else {
+            is_done_ = true;
+            return true;
+          }
+        }
+        break;
+      }
+      case 5: { // Done
+        is_valid_ = false;
+        return false;
+      }
+      }
+    }
 
-    files.emplace(name, file);
+    return true;
+  }
 
-    pos = next_pos + crlf.size();
+private:
+  void clear_file_info() {
+    file_.name.clear();
+    file_.filename.clear();
+    file_.content_type.clear();
   }
 
-  return true;
-}
+  const std::string dash_ = "--";
+  const std::string crlf_ = "\r\n";
+  std::string boundary_;
 
-inline bool parse_range_header(const std::string &s, Ranges &ranges) {
-  try {
-    static auto re = std::regex(R"(bytes=(\d*-\d*(?:,\s*\d*-\d*)*))");
-    std::smatch m;
-    if (std::regex_match(s, m, re)) {
-      auto pos = m.position(1);
-      auto len = m.length(1);
-      detail::split(&s[pos], &s[pos + len], ',',
-                    [&](const char *b, const char *e) {
-                      static auto re = std::regex(R"(\s*(\d*)-(\d*))");
-                      std::cmatch m;
-                      if (std::regex_match(b, e, m, re)) {
-                        ssize_t first = -1;
-                        if (!m.str(1).empty()) {
-                          first = static_cast<ssize_t>(std::stoll(m.str(1)));
-                        }
-
-                        ssize_t last = -1;
-                        if (!m.str(2).empty()) {
-                          last = static_cast<ssize_t>(std::stoll(m.str(2)));
-                        }
-
-                        if (first != -1 && last != -1 && first > last) {
-                          throw std::runtime_error("invalid range error");
-                        }
-                        ranges.emplace_back(std::make_pair(first, last));
-                      }
-                    });
-      return true;
-    }
-    return false;
-  } catch (...) { return false; }
-}
+  std::string buf_;
+  size_t state_ = 0;
+  size_t is_valid_ = false;
+  size_t is_done_ = false;
+  size_t off_ = 0;
+  MultipartFormData file_;
+};
 
 inline std::string to_lower(const char *beg, const char *end) {
   std::string out;
@@ -1915,7 +2366,7 @@ bool process_multipart_ranges_data(const Request &req, Response &res,
       ctoken("\r\n");
     }
 
-    auto offsets = detail::get_range_offset_and_length(req, res.body.size(), i);
+    auto offsets = get_range_offset_and_length(req, res.body.size(), i);
     auto offset = offsets.first;
     auto length = offsets.second;
 
@@ -1978,8 +2429,7 @@ inline bool write_multipart_ranges_data(Stream &strm, const Request &req,
       [&](const std::string &token) { strm.write(token); },
       [&](const char *token) { strm.write(token); },
       [&](size_t offset, size_t length) {
-        return detail::write_content(strm, res.content_provider, offset,
-                                     length) >= 0;
+        return write_content(strm, res.content_provider, offset, length) >= 0;
       });
 }
 
@@ -1988,11 +2438,56 @@ get_range_offset_and_length(const Request &req, const Response &res,
                             size_t index) {
   auto r = req.ranges[index];
 
-  if (r.second == -1) { r.second = res.content_provider_resource_length - 1; }
+  if (r.second == -1) { r.second = res.content_length - 1; }
 
   return std::make_pair(r.first, r.second - r.first + 1);
 }
 
+inline bool expect_content(const Request &req) {
+  if (req.method == "POST" || req.method == "PUT" || req.method == "PATCH" ||
+      req.method == "PRI") {
+    return true;
+  }
+  // TODO: check if Content-Length is set
+  return false;
+}
+
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+template <typename CTX, typename Init, typename Update, typename Final>
+inline std::string message_digest(const std::string &s, Init init,
+                                  Update update, Final final,
+                                  size_t digest_length) {
+  using namespace std;
+
+  std::vector<unsigned char> md(digest_length, 0);
+  CTX ctx;
+  init(&ctx);
+  update(&ctx, s.data(), s.size());
+  final(md.data(), &ctx);
+
+  stringstream ss;
+  for (auto c : md) {
+    ss << setfill('0') << setw(2) << hex << (unsigned int)c;
+  }
+  return ss.str();
+}
+
+inline std::string MD5(const std::string &s) {
+  return message_digest<MD5_CTX>(s, MD5_Init, MD5_Update, MD5_Final,
+                                 MD5_DIGEST_LENGTH);
+}
+
+inline std::string SHA_256(const std::string &s) {
+  return message_digest<SHA256_CTX>(s, SHA256_Init, SHA256_Update, SHA256_Final,
+                                    SHA256_DIGEST_LENGTH);
+}
+
+inline std::string SHA_512(const std::string &s) {
+  return message_digest<SHA512_CTX>(s, SHA512_Init, SHA512_Update, SHA512_Final,
+                                    SHA512_DIGEST_LENGTH);
+}
+#endif
+
 #ifdef _WIN32
 class WSInit {
 public:
@@ -2025,9 +2520,103 @@ inline std::pair<std::string, std::string> make_range_header(Ranges ranges) {
 
 inline std::pair<std::string, std::string>
 make_basic_authentication_header(const std::string &username,
-                                 const std::string &password) {
+                                 const std::string &password,
+                                 bool is_proxy = false) {
   auto field = "Basic " + detail::base64_encode(username + ":" + password);
-  return std::make_pair("Authorization", field);
+  auto key = is_proxy ? "Proxy-Authorization" : "Authorization";
+  return std::make_pair(key, field);
+}
+
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+inline std::pair<std::string, std::string> make_digest_authentication_header(
+    const Request &req, const std::map<std::string, std::string> &auth,
+    size_t cnonce_count, const std::string &cnonce, const std::string &username,
+    const std::string &password, bool is_proxy = false) {
+  using namespace std;
+
+  string nc;
+  {
+    stringstream ss;
+    ss << setfill('0') << setw(8) << hex << cnonce_count;
+    nc = ss.str();
+  }
+
+  auto qop = auth.at("qop");
+  if (qop.find("auth-int") != std::string::npos) {
+    qop = "auth-int";
+  } else {
+    qop = "auth";
+  }
+
+  std::string algo = "MD5";
+  if (auth.find("algorithm") != auth.end()) { algo = auth.at("algorithm"); }
+
+  string response;
+  {
+    auto H = algo == "SHA-256"
+                 ? detail::SHA_256
+                 : algo == "SHA-512" ? detail::SHA_512 : detail::MD5;
+
+    auto A1 = username + ":" + auth.at("realm") + ":" + password;
+
+    auto A2 = req.method + ":" + req.path;
+    if (qop == "auth-int") { A2 += ":" + H(req.body); }
+
+    response = H(H(A1) + ":" + auth.at("nonce") + ":" + nc + ":" + cnonce +
+                 ":" + qop + ":" + H(A2));
+  }
+
+  auto field = "Digest username=\"hello\", realm=\"" + auth.at("realm") +
+               "\", nonce=\"" + auth.at("nonce") + "\", uri=\"" + req.path +
+               "\", algorithm=" + algo + ", qop=" + qop + ", nc=\"" + nc +
+               "\", cnonce=\"" + cnonce + "\", response=\"" + response + "\"";
+
+  auto key = is_proxy ? "Proxy-Authorization" : "Authorization";
+  return std::make_pair(key, field);
+}
+#endif
+
+inline bool parse_www_authenticate(const httplib::Response &res,
+                                   std::map<std::string, std::string> &auth,
+                                   bool is_proxy) {
+  auto auth_key = is_proxy ? "Proxy-Authenticate" : "WWW-Authenticate";
+  if (res.has_header(auth_key)) {
+    static auto re = std::regex(R"~((?:(?:,\s*)?(.+?)=(?:"(.*?)"|([^,]*))))~");
+    auto s = res.get_header_value(auth_key);
+    auto pos = s.find(' ');
+    if (pos != std::string::npos) {
+      auto type = s.substr(0, pos);
+      if (type == "Basic") {
+        return false;
+      } else if (type == "Digest") {
+        s = s.substr(pos + 1);
+        auto beg = std::sregex_iterator(s.begin(), s.end(), re);
+        for (auto i = beg; i != std::sregex_iterator(); ++i) {
+          auto m = *i;
+          auto key = s.substr(m.position(1), m.length(1));
+          auto val = m.length(2) > 0 ? s.substr(m.position(2), m.length(2))
+                                     : s.substr(m.position(3), m.length(3));
+          auth[key] = val;
+        }
+        return true;
+      }
+    }
+  }
+  return false;
+}
+
+// https://stackoverflow.com/questions/440133/how-do-i-create-a-random-alpha-numeric-string-in-c/440240#answer-440240
+inline std::string random_string(size_t length) {
+  auto randchar = []() -> char {
+    const char charset[] = "0123456789"
+                           "ABCDEFGHIJKLMNOPQRSTUVWXYZ"
+                           "abcdefghijklmnopqrstuvwxyz";
+    const size_t max_index = (sizeof(charset) - 1);
+    return charset[rand() % max_index];
+  };
+  std::string str(length, 0);
+  std::generate_n(str.begin(), length, randchar);
+  return str;
 }
 
 // Request implementation
@@ -2068,14 +2657,19 @@ inline size_t Request::get_param_value_count(const char *key) const {
   return std::distance(r.first, r.second);
 }
 
+inline bool Request::is_multipart_form_data() const {
+  const auto &content_type = get_header_value("Content-Type");
+  return !content_type.find("multipart/form-data");
+}
+
 inline bool Request::has_file(const char *key) const {
   return files.find(key) != files.end();
 }
 
-inline MultipartFile Request::get_file_value(const char *key) const {
+inline MultipartFormData Request::get_file_value(const char *key) const {
   auto it = files.find(key);
   if (it != files.end()) { return it->second; }
-  return MultipartFile();
+  return MultipartFormData();
 }
 
 // Response implementation
@@ -2119,40 +2713,47 @@ inline void Response::set_content(const std::string &s,
 }
 
 inline void Response::set_content_provider(
-    size_t length,
-    std::function<void(size_t offset, size_t length, DataSink sink)> provider,
+    size_t in_length,
+    std::function<void(size_t offset, size_t length, DataSink &sink)> provider,
     std::function<void()> resource_releaser) {
-  assert(length > 0);
-  content_provider_resource_length = length;
-  content_provider = [provider](size_t offset, size_t length, DataSink sink,
-                                Done) { provider(offset, length, sink); };
+  assert(in_length > 0);
+  content_length = in_length;
+  content_provider = [provider](size_t offset, size_t length, DataSink &sink) {
+    provider(offset, length, sink);
+  };
   content_provider_resource_releaser = resource_releaser;
 }
 
 inline void Response::set_chunked_content_provider(
-    std::function<void(size_t offset, DataSink sink, Done done)> provider,
+    std::function<void(size_t offset, DataSink &sink)> provider,
     std::function<void()> resource_releaser) {
-  content_provider_resource_length = 0;
-  content_provider = [provider](size_t offset, size_t, DataSink sink,
-                                Done done) { provider(offset, sink, done); };
+  content_length = 0;
+  content_provider = [provider](size_t offset, size_t, DataSink &sink) {
+    provider(offset, sink);
+  };
   content_provider_resource_releaser = resource_releaser;
 }
 
 // Rstream implementation
+inline int Stream::write(const char *ptr) { return write(ptr, strlen(ptr)); }
+
+inline int Stream::write(const std::string &s) {
+  return write(s.data(), s.size());
+}
+
 template <typename... Args>
 inline int Stream::write_format(const char *fmt, const Args &... args) {
-  const auto bufsiz = 2048;
-  char buf[bufsiz];
+  std::array<char, 2048> buf;
 
 #if defined(_MSC_VER) && _MSC_VER < 1900
-  auto n = _snprintf_s(buf, bufsiz, bufsiz - 1, fmt, args...);
+  auto n = _snprintf_s(buf, bufsiz, buf.size() - 1, fmt, args...);
 #else
-  auto n = snprintf(buf, bufsiz - 1, fmt, args...);
+  auto n = snprintf(buf.data(), buf.size() - 1, fmt, args...);
 #endif
   if (n <= 0) { return n; }
 
-  if (n >= bufsiz - 1) {
-    std::vector<char> glowable_buf(bufsiz);
+  if (n >= static_cast<int>(buf.size()) - 1) {
+    std::vector<char> glowable_buf(buf.size());
 
     while (n >= static_cast<int>(glowable_buf.size() - 1)) {
       glowable_buf.resize(glowable_buf.size() * 2);
@@ -2165,33 +2766,36 @@ inline int Stream::write_format(const char *fmt, const Args &... args) {
     }
     return write(&glowable_buf[0], n);
   } else {
-    return write(buf, n);
+    return write(buf.data(), n);
   }
 }
 
+namespace detail {
+
 // Socket stream implementation
-inline SocketStream::SocketStream(socket_t sock) : sock_(sock) {}
+inline SocketStream::SocketStream(socket_t sock, time_t read_timeout_sec,
+                                  time_t read_timeout_usec)
+    : sock_(sock), read_timeout_sec_(read_timeout_sec),
+      read_timeout_usec_(read_timeout_usec) {}
 
 inline SocketStream::~SocketStream() {}
 
-inline int SocketStream::read(char *ptr, size_t size) {
-  if (detail::select_read(sock_, CPPHTTPLIB_READ_TIMEOUT_SECOND,
-                          CPPHTTPLIB_READ_TIMEOUT_USECOND) > 0) {
-    return recv(sock_, ptr, static_cast<int>(size), 0);
-  }
-  return -1;
+inline bool SocketStream::is_readable() const {
+  return detail::select_read(sock_, read_timeout_sec_, read_timeout_usec_) > 0;
 }
 
-inline int SocketStream::write(const char *ptr, size_t size) {
-  return send(sock_, ptr, static_cast<int>(size), 0);
+inline bool SocketStream::is_writable() const {
+  return detail::select_write(sock_, 0, 0) > 0;
 }
 
-inline int SocketStream::write(const char *ptr) {
-  return write(ptr, strlen(ptr));
+inline int SocketStream::read(char *ptr, size_t size) {
+  if (is_readable()) { return recv(sock_, ptr, static_cast<int>(size), 0); }
+  return -1;
 }
 
-inline int SocketStream::write(const std::string &s) {
-  return write(s.data(), s.size());
+inline int SocketStream::write(const char *ptr, size_t size) {
+  if (is_writable()) { return send(sock_, ptr, static_cast<int>(size), 0); }
+  return -1;
 }
 
 inline std::string SocketStream::get_remote_addr() const {
@@ -2199,12 +2803,18 @@ inline std::string SocketStream::get_remote_addr() const {
 }
 
 // Buffer stream implementation
+inline bool BufferStream::is_readable() const { return true; }
+
+inline bool BufferStream::is_writable() const { return true; }
+
 inline int BufferStream::read(char *ptr, size_t size) {
 #if defined(_MSC_VER) && _MSC_VER < 1900
-  return static_cast<int>(buffer._Copy_s(ptr, size, size));
+  int len_read = static_cast<int>(buffer._Copy_s(ptr, size, size, position));
 #else
-  return static_cast<int>(buffer.copy(ptr, size));
+  int len_read = static_cast<int>(buffer.copy(ptr, size, position));
 #endif
+  position += len_read;
+  return len_read;
 }
 
 inline int BufferStream::write(const char *ptr, size_t size) {
@@ -2212,33 +2822,23 @@ inline int BufferStream::write(const char *ptr, size_t size) {
   return static_cast<int>(size);
 }
 
-inline int BufferStream::write(const char *ptr) {
-  return write(ptr, strlen(ptr));
-}
-
-inline int BufferStream::write(const std::string &s) {
-  return write(s.data(), s.size());
-}
-
 inline std::string BufferStream::get_remote_addr() const { return ""; }
 
 inline const std::string &BufferStream::get_buffer() const { return buffer; }
 
+} // namespace detail
+
 // HTTP server implementation
 inline Server::Server()
     : keep_alive_max_count_(CPPHTTPLIB_KEEPALIVE_MAX_COUNT),
+      read_timeout_sec_(CPPHTTPLIB_READ_TIMEOUT_SECOND),
+      read_timeout_usec_(CPPHTTPLIB_READ_TIMEOUT_USECOND),
       payload_max_length_(CPPHTTPLIB_PAYLOAD_MAX_LENGTH), is_running_(false),
       svr_sock_(INVALID_SOCKET) {
 #ifndef _WIN32
   signal(SIGPIPE, SIG_IGN);
 #endif
-  new_task_queue = [] {
-#if CPPHTTPLIB_THREAD_POOL_COUNT > 0
-    return new ThreadPool(CPPHTTPLIB_THREAD_POOL_COUNT);
-#else
-    return new Threads();
-#endif
-  };
+  new_task_queue = [] { return new ThreadPool(CPPHTTPLIB_THREAD_POOL_COUNT); };
 }
 
 inline Server::~Server() {}
@@ -2253,16 +2853,37 @@ inline Server &Server::Post(const char *pattern, Handler handler) {
   return *this;
 }
 
+inline Server &Server::Post(const char *pattern,
+                            HandlerWithContentReader handler) {
+  post_handlers_for_content_reader_.push_back(
+      std::make_pair(std::regex(pattern), handler));
+  return *this;
+}
+
 inline Server &Server::Put(const char *pattern, Handler handler) {
   put_handlers_.push_back(std::make_pair(std::regex(pattern), handler));
   return *this;
 }
 
+inline Server &Server::Put(const char *pattern,
+                           HandlerWithContentReader handler) {
+  put_handlers_for_content_reader_.push_back(
+      std::make_pair(std::regex(pattern), handler));
+  return *this;
+}
+
 inline Server &Server::Patch(const char *pattern, Handler handler) {
   patch_handlers_.push_back(std::make_pair(std::regex(pattern), handler));
   return *this;
 }
 
+inline Server &Server::Patch(const char *pattern,
+                             HandlerWithContentReader handler) {
+  patch_handlers_for_content_reader_.push_back(
+      std::make_pair(std::regex(pattern), handler));
+  return *this;
+}
+
 inline Server &Server::Delete(const char *pattern, Handler handler) {
   delete_handlers_.push_back(std::make_pair(std::regex(pattern), handler));
   return *this;
@@ -2273,32 +2894,68 @@ inline Server &Server::Options(const char *pattern, Handler handler) {
   return *this;
 }
 
-inline bool Server::set_base_dir(const char *path) {
-  if (detail::is_dir(path)) {
-    base_dir_ = path;
-    return true;
+inline bool Server::set_base_dir(const char *dir, const char *mount_point) {
+  return set_mount_point(mount_point, dir);
+}
+
+inline bool Server::set_mount_point(const char *mount_point, const char *dir) {
+  if (detail::is_dir(dir)) {
+    std::string mnt = mount_point ? mount_point : "/";
+    if (!mnt.empty() && mnt[0] == '/') {
+      base_dirs_.emplace_back(mnt, dir);
+      return true;
+    }
+  }
+  return false;
+}
+
+inline bool Server::remove_mount_point(const char *mount_point) {
+  for (auto it = base_dirs_.begin(); it != base_dirs_.end(); ++it) {
+    if (it->first == mount_point) {
+      base_dirs_.erase(it);
+      return true;
+    }
   }
   return false;
 }
 
+inline void Server::set_file_extension_and_mimetype_mapping(const char *ext,
+                                                            const char *mime) {
+  file_extension_and_mimetype_map_[ext] = mime;
+}
+
 inline void Server::set_file_request_handler(Handler handler) {
-  file_request_handler_ = handler;
+  file_request_handler_ = std::move(handler);
 }
 
 inline void Server::set_error_handler(Handler handler) {
-  error_handler_ = handler;
+  error_handler_ = std::move(handler);
 }
 
-inline void Server::set_logger(Logger logger) { logger_ = logger; }
+inline void Server::set_logger(Logger logger) { logger_ = std::move(logger); }
+
+inline void
+Server::set_expect_100_continue_handler(Expect100ContinueHandler handler) {
+  expect_100_continue_handler_ = std::move(handler);
+}
 
 inline void Server::set_keep_alive_max_count(size_t count) {
   keep_alive_max_count_ = count;
 }
 
+inline void Server::set_read_timeout(time_t sec, time_t usec) {
+  read_timeout_sec_ = sec;
+  read_timeout_usec_ = usec;
+}
+
 inline void Server::set_payload_max_length(size_t length) {
   payload_max_length_ = length;
 }
 
+inline bool Server::bind_to_port(const char *host, int port, int socket_flags) {
+  if (bind_internal(host, port, socket_flags) < 0) return false;
+  return true;
+}
 inline int Server::bind_to_any_port(const char *host, int socket_flags) {
   return bind_internal(host, 0, socket_flags);
 }
@@ -2306,8 +2963,7 @@ inline int Server::bind_to_any_port(const char *host, int socket_flags) {
 inline bool Server::listen_after_bind() { return listen_internal(); }
 
 inline bool Server::listen(const char *host, int port, int socket_flags) {
-  if (bind_internal(host, port, socket_flags) < 0) return false;
-  return listen_internal();
+  return bind_to_port(host, port, socket_flags) && listen_internal();
 }
 
 inline bool Server::is_running() const { return is_running_; }
@@ -2322,8 +2978,9 @@ inline void Server::stop() {
 }
 
 inline bool Server::parse_request_line(const char *s, Request &req) {
-  static std::regex re("(GET|HEAD|POST|PUT|DELETE|CONNECT|OPTIONS|TRACE|PATCH|PRI) "
-                       "(([^?]+)(?:\\?(.+?))?) (HTTP/1\\.[01])\r\n");
+  const static std::regex re(
+      "(GET|HEAD|POST|PUT|DELETE|CONNECT|OPTIONS|TRACE|PATCH|PRI) "
+      "(([^?]+)(?:\\?(.*?))?) (HTTP/1\\.[01])\r\n");
 
   std::cmatch m;
   if (std::regex_match(s, m, re)) {
@@ -2348,9 +3005,11 @@ inline bool Server::write_response(Stream &strm, bool last_connection,
 
   if (400 <= res.status && error_handler_) { error_handler_(req, res); }
 
+  detail::BufferStream bstrm;
+
   // Response line
-  if (!strm.write_format("HTTP/1.1 %d %s\r\n", res.status,
-                         detail::status_message(res.status))) {
+  if (!bstrm.write_format("HTTP/1.1 %d %s\r\n", res.status,
+                          detail::status_message(res.status))) {
     return false;
   }
 
@@ -2363,11 +3022,12 @@ inline bool Server::write_response(Stream &strm, bool last_connection,
     res.set_header("Connection", "Keep-Alive");
   }
 
-  if (!res.has_header("Content-Type")) {
+  if (!res.has_header("Content-Type") &&
+      (!res.body.empty() || res.content_length > 0)) {
     res.set_header("Content-Type", "text/plain");
   }
 
-  if (!res.has_header("Accept-Ranges")) {
+  if (!res.has_header("Accept-Ranges") && req.method == "HEAD") {
     res.set_header("Accept-Ranges", "bytes");
   }
 
@@ -2388,17 +3048,17 @@ inline bool Server::write_response(Stream &strm, bool last_connection,
   }
 
   if (res.body.empty()) {
-    if (res.content_provider_resource_length > 0) {
+    if (res.content_length > 0) {
       size_t length = 0;
       if (req.ranges.empty()) {
-        length = res.content_provider_resource_length;
+        length = res.content_length;
       } else if (req.ranges.size() == 1) {
-        auto offsets = detail::get_range_offset_and_length(
-            req, res.content_provider_resource_length, 0);
+        auto offsets =
+            detail::get_range_offset_and_length(req, res.content_length, 0);
         auto offset = offsets.first;
         length = offsets.second;
         auto content_range = detail::make_content_range_header_field(
-            offset, length, res.content_provider_resource_length);
+            offset, length, res.content_length);
         res.set_header("Content-Range", content_range);
       } else {
         length = detail::get_multipart_ranges_data_length(req, res, boundary,
@@ -2430,7 +3090,7 @@ inline bool Server::write_response(Stream &strm, bool last_connection,
     }
 
 #ifdef CPPHTTPLIB_ZLIB_SUPPORT
-    // TODO: 'Accpet-Encoding' has gzip, not gzip;q=0
+    // TODO: 'Accept-Encoding' has gzip, not gzip;q=0
     const auto &encodings = req.get_header_value("Accept-Encoding");
     if (encodings.find("gzip") != std::string::npos &&
         detail::can_compress(res.get_header_value("Content-Type"))) {
@@ -2444,7 +3104,11 @@ inline bool Server::write_response(Stream &strm, bool last_connection,
     res.set_header("Content-Length", length);
   }
 
-  if (!detail::write_headers(strm, res, Headers())) { return false; }
+  if (!detail::write_headers(bstrm, res, Headers())) { return false; }
+
+  // Flush buffer
+  auto &data = bstrm.get_buffer();
+  strm.write(data.data(), data.size());
 
   // Body
   if (req.method != "HEAD") {
@@ -2468,15 +3132,15 @@ inline bool
 Server::write_content_with_provider(Stream &strm, const Request &req,
                                     Response &res, const std::string &boundary,
                                     const std::string &content_type) {
-  if (res.content_provider_resource_length) {
+  if (res.content_length) {
     if (req.ranges.empty()) {
       if (detail::write_content(strm, res.content_provider, 0,
-                                res.content_provider_resource_length) < 0) {
+                                res.content_length) < 0) {
         return false;
       }
     } else if (req.ranges.size() == 1) {
-      auto offsets = detail::get_range_offset_and_length(
-          req, res.content_provider_resource_length, 0);
+      auto offsets =
+          detail::get_range_offset_and_length(req, res.content_length, 0);
       auto offset = offsets.first;
       auto length = offsets.second;
       if (detail::write_content(strm, res.content_provider, offset, length) <
@@ -2490,29 +3154,123 @@ Server::write_content_with_provider(Stream &strm, const Request &req,
       }
     }
   } else {
-    if (detail::write_content_chunked(strm, res.content_provider) < 0) {
+    auto is_shutting_down = [this]() {
+      return this->svr_sock_ == INVALID_SOCKET;
+    };
+    if (detail::write_content_chunked(strm, res.content_provider,
+                                      is_shutting_down) < 0) {
       return false;
     }
   }
   return true;
 }
 
-inline bool Server::handle_file_request(Request &req, Response &res) {
-  if (!base_dir_.empty() && detail::is_valid_path(req.path)) {
-    std::string path = base_dir_ + req.path;
+inline bool Server::read_content(Stream &strm, bool last_connection,
+                                 Request &req, Response &res) {
+  MultipartFormDataMap::iterator cur;
+  auto ret = read_content_core(
+      strm, last_connection, req, res,
+      // Regular
+      [&](const char *buf, size_t n) {
+        if (req.body.size() + n > req.body.max_size()) { return false; }
+        req.body.append(buf, n);
+        return true;
+      },
+      // Multipart
+      [&](const MultipartFormData &file) {
+        cur = req.files.emplace(file.name, file);
+        return true;
+      },
+      [&](const char *buf, size_t n) {
+        auto &content = cur->second.content;
+        if (content.size() + n > content.max_size()) { return false; }
+        content.append(buf, n);
+        return true;
+      });
 
-    if (!path.empty() && path.back() == '/') { path += "index.html"; }
+  const auto &content_type = req.get_header_value("Content-Type");
+  if (!content_type.find("application/x-www-form-urlencoded")) {
+    detail::parse_query_text(req.body, req.params);
+  }
 
-    if (detail::is_file(path)) {
-      detail::read_file(path, res.body);
-      auto type = detail::find_content_type(path);
-      if (type) { res.set_header("Content-Type", type); }
-      res.status = 200;
-      if (file_request_handler_) { file_request_handler_(req, res); }
-      return true;
+  return ret;
+}
+
+inline bool Server::read_content_with_content_receiver(
+    Stream &strm, bool last_connection, Request &req, Response &res,
+    ContentReceiver receiver, MultipartContentHeader multipart_header,
+    ContentReceiver multipart_receiver) {
+  return read_content_core(strm, last_connection, req, res, receiver,
+                           multipart_header, multipart_receiver);
+}
+
+inline bool Server::read_content_core(Stream &strm, bool last_connection,
+                                      Request &req, Response &res,
+                                      ContentReceiver receiver,
+                                      MultipartContentHeader mulitpart_header,
+                                      ContentReceiver multipart_receiver) {
+  detail::MultipartFormDataParser multipart_form_data_parser;
+  ContentReceiver out;
+
+  if (req.is_multipart_form_data()) {
+    const auto &content_type = req.get_header_value("Content-Type");
+    std::string boundary;
+    if (!detail::parse_multipart_boundary(content_type, boundary)) {
+      res.status = 400;
+      return write_response(strm, last_connection, req, res);
     }
+
+    multipart_form_data_parser.set_boundary(boundary);
+    out = [&](const char *buf, size_t n) {
+      return multipart_form_data_parser.parse(buf, n, multipart_receiver,
+                                              mulitpart_header);
+    };
+  } else {
+    out = receiver;
+  }
+
+  if (!detail::read_content(strm, req, payload_max_length_, res.status,
+                            Progress(), out)) {
+    return write_response(strm, last_connection, req, res);
   }
 
+  if (req.is_multipart_form_data()) {
+    if (!multipart_form_data_parser.is_valid()) {
+      res.status = 400;
+      return write_response(strm, last_connection, req, res);
+    }
+  }
+
+  return true;
+}
+
+inline bool Server::handle_file_request(Request &req, Response &res,
+                                        bool head) {
+  for (const auto &kv : base_dirs_) {
+    const auto &mount_point = kv.first;
+    const auto &base_dir = kv.second;
+
+    // Prefix match
+    if (!req.path.find(mount_point)) {
+      std::string sub_path = "/" + req.path.substr(mount_point.size());
+      if (detail::is_valid_path(sub_path)) {
+        auto path = base_dir + sub_path;
+        if (path.back() == '/') { path += "index.html"; }
+
+        if (detail::is_file(path)) {
+          detail::read_file(path, res.body);
+          auto type =
+              detail::find_content_type(path, file_extension_and_mimetype_map_);
+          if (type) { res.set_header("Content-Type", type); }
+          res.status = 200;
+          if (!head && file_request_handler_) {
+            file_request_handler_(req, res);
+          }
+          return true;
+        }
+      }
+    }
+  }
   return false;
 }
 
@@ -2605,9 +3363,51 @@ inline bool Server::listen_internal() {
   return ret;
 }
 
-inline bool Server::routing(Request &req, Response &res) {
-  if (req.method == "GET" && handle_file_request(req, res)) { return true; }
+inline bool Server::routing(Request &req, Response &res, Stream &strm,
+                            bool last_connection) {
+  // File handler
+  bool is_head_request = req.method == "HEAD";
+  if ((req.method == "GET" || is_head_request) &&
+      handle_file_request(req, res, is_head_request)) {
+    return true;
+  }
+
+  if (detail::expect_content(req)) {
+    // Content reader handler
+    {
+      ContentReader reader(
+          [&](ContentReceiver receiver) {
+            return read_content_with_content_receiver(
+                strm, last_connection, req, res, receiver, nullptr, nullptr);
+          },
+          [&](MultipartContentHeader header, ContentReceiver receiver) {
+            return read_content_with_content_receiver(
+                strm, last_connection, req, res, nullptr, header, receiver);
+          });
+
+      if (req.method == "POST") {
+        if (dispatch_request_for_content_reader(
+                req, res, reader, post_handlers_for_content_reader_)) {
+          return true;
+        }
+      } else if (req.method == "PUT") {
+        if (dispatch_request_for_content_reader(
+                req, res, reader, put_handlers_for_content_reader_)) {
+          return true;
+        }
+      } else if (req.method == "PATCH") {
+        if (dispatch_request_for_content_reader(
+                req, res, reader, patch_handlers_for_content_reader_)) {
+          return true;
+        }
+      }
+    }
 
+    // Read content into `req.body`
+    if (!read_content(strm, last_connection, req, res)) { return false; }
+  }
+
+  // Regular handler
   if (req.method == "GET" || req.method == "HEAD") {
     return dispatch_request(req, res, get_handlers_);
   } else if (req.method == "POST") {
@@ -2640,17 +3440,31 @@ inline bool Server::dispatch_request(Request &req, Response &res,
   return false;
 }
 
+inline bool Server::dispatch_request_for_content_reader(
+    Request &req, Response &res, ContentReader content_reader,
+    HandlersForContentReader &handlers) {
+  for (const auto &x : handlers) {
+    const auto &pattern = x.first;
+    const auto &handler = x.second;
+
+    if (std::regex_match(req.path, req.matches, pattern)) {
+      handler(req, res, content_reader);
+      return true;
+    }
+  }
+  return false;
+}
+
 inline bool
 Server::process_request(Stream &strm, bool last_connection,
                         bool &connection_close,
-                        std::function<void(Request &)> setup_request) {
-  const auto bufsiz = 2048;
-  char buf[bufsiz];
+                        const std::function<void(Request &)> &setup_request) {
+  std::array<char, 2048> buf{};
 
-  detail::stream_line_reader reader(strm, buf, bufsiz);
+  detail::stream_line_reader line_reader(strm, buf.data(), buf.size());
 
   // Connection has been closed on client
-  if (!reader.getline()) { return false; }
+  if (!line_reader.getline()) { return false; }
 
   Request req;
   Response res;
@@ -2658,7 +3472,7 @@ Server::process_request(Stream &strm, bool last_connection,
   res.version = "HTTP/1.1";
 
   // Check if the request URI doesn't exceed the limit
-  if (reader.size() > CPPHTTPLIB_REQUEST_URI_MAX_LENGTH) {
+  if (line_reader.size() > CPPHTTPLIB_REQUEST_URI_MAX_LENGTH) {
     Headers dummy;
     detail::read_headers(strm, dummy);
     res.status = 414;
@@ -2666,7 +3480,7 @@ Server::process_request(Stream &strm, bool last_connection,
   }
 
   // Request line and headers
-  if (!parse_request_line(reader.ptr(), req) ||
+  if (!parse_request_line(line_reader.ptr(), req) ||
       !detail::read_headers(strm, req.headers)) {
     res.status = 400;
     return write_response(strm, last_connection, req, res);
@@ -2683,33 +3497,6 @@ Server::process_request(Stream &strm, bool last_connection,
 
   req.set_header("REMOTE_ADDR", strm.get_remote_addr());
 
-  // Body
-  if (req.method == "POST" || req.method == "PUT" || req.method == "PATCH" || req.method == "PRI") {
-    if (!detail::read_content(strm, req, payload_max_length_, res.status,
-                              Progress(), [&](const char *buf, size_t n) {
-                                if (req.body.size() + n > req.body.max_size()) {
-                                  return false;
-                                }
-                                req.body.append(buf, n);
-                                return true;
-                              })) {
-      return write_response(strm, last_connection, req, res);
-    }
-
-    const auto &content_type = req.get_header_value("Content-Type");
-
-    if (!content_type.find("application/x-www-form-urlencoded")) {
-      detail::parse_query_text(req.body, req.params);
-    } else if (!content_type.find("multipart/form-data")) {
-      std::string boundary;
-      if (!detail::parse_multipart_boundary(content_type, boundary) ||
-          !detail::parse_multipart_formdata(boundary, req.body, req.files)) {
-        res.status = 400;
-        return write_response(strm, last_connection, req, res);
-      }
-    }
-  }
-
   if (req.has_header("Range")) {
     const auto &range_header_value = req.get_header_value("Range");
     if (!detail::parse_range_header(range_header_value, req.ranges)) {
@@ -2719,7 +3506,23 @@ Server::process_request(Stream &strm, bool last_connection,
 
   if (setup_request) { setup_request(req); }
 
-  if (routing(req, res)) {
+  if (req.get_header_value("Expect") == "100-continue") {
+    auto status = 100;
+    if (expect_100_continue_handler_) {
+      status = expect_100_continue_handler_(req, res);
+    }
+    switch (status) {
+    case 100:
+    case 417:
+      strm.write_format("HTTP/1.1 %d %s\r\n\r\n", status,
+                        detail::status_message(status));
+      break;
+    default: return write_response(strm, last_connection, req, res);
+    }
+  }
+
+  // Rounting
+  if (routing(req, res, strm, last_connection)) {
     if (res.status == -1) { res.status = req.ranges.empty() ? 200 : 206; }
   } else {
     if (res.status == -1) { res.status = 404; }
@@ -2732,7 +3535,7 @@ inline bool Server::is_valid() const { return true; }
 
 inline bool Server::process_and_close_socket(socket_t sock) {
   return detail::process_and_close_socket(
-      false, sock, keep_alive_max_count_,
+      false, sock, keep_alive_max_count_, read_timeout_sec_, read_timeout_usec_,
       [this](Stream &strm, bool last_connection, bool &connection_close) {
         return process_request(strm, last_connection, connection_close,
                                nullptr);
@@ -2740,47 +3543,37 @@ inline bool Server::process_and_close_socket(socket_t sock) {
 }
 
 // HTTP client implementation
-inline Client::Client(const char *host, int port, time_t timeout_sec)
-    : host_(host), port_(port), timeout_sec_(timeout_sec),
+inline Client::Client(const std::string &host, int port,
+                      const std::string &client_cert_path,
+                      const std::string &client_key_path)
+    : host_(host), port_(port),
       host_and_port_(host_ + ":" + std::to_string(port_)),
-      keep_alive_max_count_(CPPHTTPLIB_KEEPALIVE_MAX_COUNT),
-      follow_location_(false) {}
+      client_cert_path_(client_cert_path), client_key_path_(client_key_path) {}
 
 inline Client::~Client() {}
 
 inline bool Client::is_valid() const { return true; }
 
 inline socket_t Client::create_client_socket() const {
-  return detail::create_socket(
-      host_.c_str(), port_, [=](socket_t sock, struct addrinfo &ai) -> bool {
-        detail::set_nonblocking(sock, true);
-
-        auto ret = connect(sock, ai.ai_addr, static_cast<int>(ai.ai_addrlen));
-        if (ret < 0) {
-          if (detail::is_connection_error() ||
-              !detail::wait_until_socket_is_ready(sock, timeout_sec_, 0)) {
-            detail::close_socket(sock);
-            return false;
-          }
-        }
-
-        detail::set_nonblocking(sock, false);
-        return true;
-      });
+  if (!proxy_host_.empty()) {
+    return detail::create_client_socket(proxy_host_.c_str(), proxy_port_,
+                                        timeout_sec_, interface_);
+  }
+  return detail::create_client_socket(host_.c_str(), port_, timeout_sec_,
+                                      interface_);
 }
 
 inline bool Client::read_response_line(Stream &strm, Response &res) {
-  const auto bufsiz = 2048;
-  char buf[bufsiz];
+  std::array<char, 2048> buf;
 
-  detail::stream_line_reader reader(strm, buf, bufsiz);
+  detail::stream_line_reader line_reader(strm, buf.data(), buf.size());
 
-  if (!reader.getline()) { return false; }
+  if (!line_reader.getline()) { return false; }
 
   const static std::regex re("(HTTP/1\\.[01]) (\\d+?) .*\r\n");
 
   std::cmatch m;
-  if (std::regex_match(reader.ptr(), m, re)) {
+  if (std::regex_match(line_reader.ptr(), m, re)) {
     res.version = std::string(m[1]);
     res.status = std::stoi(std::string(m[2]));
   }
@@ -2789,22 +3582,21 @@ inline bool Client::read_response_line(Stream &strm, Response &res) {
 }
 
 inline bool Client::send(const Request &req, Response &res) {
-  if (req.path.empty()) { return false; }
-
   auto sock = create_client_socket();
   if (sock == INVALID_SOCKET) { return false; }
 
-  auto ret = process_and_close_socket(
-      sock, 1, [&](Stream &strm, bool last_connection, bool &connection_close) {
-        return process_request(strm, req, res, last_connection,
-                               connection_close);
-      });
-
-  if (ret && follow_location_ && (300 < res.status && res.status < 400)) {
-    ret = redirect(req, res);
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+  if (is_ssl() && !proxy_host_.empty()) {
+    bool error;
+    if (!connect(sock, res, error)) { return error; }
   }
+#endif
 
-  return ret;
+  return process_and_close_socket(
+      sock, 1, [&](Stream &strm, bool last_connection, bool &connection_close) {
+        return handle_request(strm, req, res, last_connection,
+                              connection_close);
+      });
 }
 
 inline bool Client::send(const std::vector<Request> &requests,
@@ -2814,32 +3606,136 @@ inline bool Client::send(const std::vector<Request> &requests,
     auto sock = create_client_socket();
     if (sock == INVALID_SOCKET) { return false; }
 
-    if (!process_and_close_socket(
-            sock, requests.size() - i,
-            [&](Stream &strm, bool last_connection, bool &connection_close) -> bool {
-              auto &req = requests[i];
-              auto res = Response();
-              i++;
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+    if (is_ssl() && !proxy_host_.empty()) {
+      Response res;
+      bool error;
+      if (!connect(sock, res, error)) { return false; }
+    }
+#endif
+
+    if (!process_and_close_socket(sock, requests.size() - i,
+                                  [&](Stream &strm, bool last_connection,
+                                      bool &connection_close) -> bool {
+                                    auto &req = requests[i++];
+                                    auto res = Response();
+                                    auto ret = handle_request(strm, req, res,
+                                                              last_connection,
+                                                              connection_close);
+                                    if (ret) {
+                                      responses.emplace_back(std::move(res));
+                                    }
+                                    return ret;
+                                  })) {
+      return false;
+    }
+  }
 
-              if (req.path.empty()) { return false; }
-              auto ret = process_request(strm, req, res, last_connection,
-                                         connection_close);
+  return true;
+}
+
+inline bool Client::handle_request(Stream &strm, const Request &req,
+                                   Response &res, bool last_connection,
+                                   bool &connection_close) {
+  if (req.path.empty()) { return false; }
+
+  bool ret;
+
+  if (!is_ssl() && !proxy_host_.empty()) {
+    auto req2 = req;
+    req2.path = "http://" + host_and_port_ + req.path;
+    ret = process_request(strm, req2, res, last_connection, connection_close);
+  } else {
+    ret = process_request(strm, req, res, last_connection, connection_close);
+  }
+
+  if (!ret) { return false; }
+
+  if (300 < res.status && res.status < 400 && follow_location_) {
+    ret = redirect(req, res);
+  }
 
-              if (ret && follow_location_ &&
-                  (300 < res.status && res.status < 400)) {
-                ret = redirect(req, res);
-              }
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+  if (res.status == 401 || res.status == 407) {
+    auto is_proxy = res.status == 407;
+    const auto &username =
+        is_proxy ? proxy_digest_auth_username_ : digest_auth_username_;
+    const auto &password =
+        is_proxy ? proxy_digest_auth_password_ : digest_auth_password_;
+
+    if (!username.empty() && !password.empty()) {
+      std::map<std::string, std::string> auth;
+      if (parse_www_authenticate(res, auth, is_proxy)) {
+        Request new_req = req;
+        auto key = is_proxy ? "Proxy-Authorization" : "WWW-Authorization";
+        new_req.headers.erase(key);
+        new_req.headers.insert(make_digest_authentication_header(
+            req, auth, 1, random_string(10), username, password, is_proxy));
+
+        Response new_res;
+
+        ret = send(new_req, new_res);
+        if (ret) { res = new_res; }
+      }
+    }
+  }
+#endif
+
+  return ret;
+}
 
-              if (ret) { responses.emplace_back(std::move(res)); }
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+inline bool Client::connect(socket_t sock, Response &res, bool &error) {
+  error = true;
+  Response res2;
+
+  if (!detail::process_socket(
+          true, sock, 1, read_timeout_sec_, read_timeout_usec_,
+          [&](Stream &strm, bool /*last_connection*/, bool &connection_close) {
+            Request req2;
+            req2.method = "CONNECT";
+            req2.path = host_and_port_;
+            return process_request(strm, req2, res2, false, connection_close);
+          })) {
+    detail::close_socket(sock);
+    error = false;
+    return false;
+  }
 
-              return ret;
-            })) {
+  if (res2.status == 407) {
+    if (!proxy_digest_auth_username_.empty() &&
+        !proxy_digest_auth_password_.empty()) {
+      std::map<std::string, std::string> auth;
+      if (parse_www_authenticate(res2, auth, true)) {
+        Response res3;
+        if (!detail::process_socket(
+                true, sock, 1, read_timeout_sec_, read_timeout_usec_,
+                [&](Stream &strm, bool /*last_connection*/,
+                    bool &connection_close) {
+                  Request req3;
+                  req3.method = "CONNECT";
+                  req3.path = host_and_port_;
+                  req3.headers.insert(make_digest_authentication_header(
+                      req3, auth, 1, random_string(10),
+                      proxy_digest_auth_username_, proxy_digest_auth_password_,
+                      true));
+                  return process_request(strm, req3, res3, false,
+                                         connection_close);
+                })) {
+          detail::close_socket(sock);
+          error = false;
+          return false;
+        }
+      }
+    } else {
+      res = res2;
       return false;
     }
   }
 
   return true;
 }
+#endif
 
 inline bool Client::redirect(const Request &req, Response &res) {
   if (req.redirect_count == 0) { return false; }
@@ -2847,46 +3743,47 @@ inline bool Client::redirect(const Request &req, Response &res) {
   auto location = res.get_header_value("location");
   if (location.empty()) { return false; }
 
-  std::regex re(
+  const static std::regex re(
       R"(^(?:([^:/?#]+):)?(?://([^/?#]*))?([^?#]*(?:\?[^#]*)?)(?:#.*)?)");
 
+  std::smatch m;
+  if (!regex_match(location, m, re)) { return false; }
+
   auto scheme = is_ssl() ? "https" : "http";
 
-  std::smatch m;
-  if (regex_match(location, m, re)) {
-    auto next_scheme = m[1].str();
-    auto next_host = m[2].str();
-    auto next_path = m[3].str();
-    if (next_host.empty()) { next_host = host_; }
-    if (next_path.empty()) { next_path = "/"; }
-
-    if (next_scheme == scheme && next_host == host_) {
-      return detail::redirect(*this, req, res, next_path);
-    } else {
-      if (next_scheme == "https") {
+  auto next_scheme = m[1].str();
+  auto next_host = m[2].str();
+  auto next_path = m[3].str();
+  if (next_scheme.empty()) { next_scheme = scheme; }
+  if (next_scheme.empty()) { next_scheme = scheme; }
+  if (next_host.empty()) { next_host = host_; }
+  if (next_path.empty()) { next_path = "/"; }
+
+  if (next_scheme == scheme && next_host == host_) {
+    return detail::redirect(*this, req, res, next_path);
+  } else {
+    if (next_scheme == "https") {
 #ifdef CPPHTTPLIB_OPENSSL_SUPPORT
-        SSLClient cli(next_host.c_str());
-        cli.follow_location(true);
-        return detail::redirect(cli, req, res, next_path);
+      SSLClient cli(next_host.c_str());
+      cli.copy_settings(*this);
+      return detail::redirect(cli, req, res, next_path);
 #else
-        return false;
+      return false;
 #endif
-      } else {
-        Client cli(next_host.c_str());
-        cli.follow_location(true);
-        return detail::redirect(cli, req, res, next_path);
-      }
+    } else {
+      Client cli(next_host.c_str());
+      cli.copy_settings(*this);
+      return detail::redirect(cli, req, res, next_path);
     }
   }
-  return false;
 }
 
-inline void Client::write_request(Stream &strm, const Request &req,
+inline bool Client::write_request(Stream &strm, const Request &req,
                                   bool last_connection) {
-  BufferStream bstrm;
+  detail::BufferStream bstrm;
 
   // Request line
-  auto path = detail::encode_url(req.path);
+  const auto &path = detail::encode_url(req.path);
 
   bstrm.write_format("%s %s HTTP/1.1\r\n", req.method.c_str(), path.c_str());
 
@@ -2913,11 +3810,14 @@ inline void Client::write_request(Stream &strm, const Request &req,
   if (!req.has_header("Accept")) { headers.emplace("Accept", "*/*"); }
 
   if (!req.has_header("User-Agent")) {
-    headers.emplace("User-Agent", "cpp-httplib/0.2");
+    headers.emplace("User-Agent", "cpp-httplib/0.5");
   }
 
   if (req.body.empty()) {
-    if (req.method == "POST" || req.method == "PUT" || req.method == "PATCH") {
+    if (req.content_provider) {
+      auto length = std::to_string(req.content_length);
+      headers.emplace("Content-Length", length);
+    } else {
       headers.emplace("Content-Length", "0");
     }
   } else {
@@ -2931,21 +3831,100 @@ inline void Client::write_request(Stream &strm, const Request &req,
     }
   }
 
-  detail::write_headers(bstrm, req, headers);
+  if (!basic_auth_username_.empty() && !basic_auth_password_.empty()) {
+    headers.insert(make_basic_authentication_header(
+        basic_auth_username_, basic_auth_password_, false));
+  }
 
-  // Body
-  if (!req.body.empty()) { bstrm.write(req.body); }
+  if (!proxy_basic_auth_username_.empty() &&
+      !proxy_basic_auth_password_.empty()) {
+    headers.insert(make_basic_authentication_header(
+        proxy_basic_auth_username_, proxy_basic_auth_password_, true));
+  }
+
+  detail::write_headers(bstrm, req, headers);
 
   // Flush buffer
   auto &data = bstrm.get_buffer();
   strm.write(data.data(), data.size());
+
+  // Body
+  if (req.body.empty()) {
+    if (req.content_provider) {
+      size_t offset = 0;
+      size_t end_offset = req.content_length;
+
+      DataSink data_sink;
+      data_sink.write = [&](const char *d, size_t l) {
+        auto written_length = strm.write(d, l);
+        offset += written_length;
+      };
+      data_sink.is_writable = [&](void) { return strm.is_writable(); };
+
+      while (offset < end_offset) {
+        req.content_provider(offset, end_offset - offset, data_sink);
+      }
+    }
+  } else {
+    strm.write(req.body);
+  }
+
+  return true;
+}
+
+inline std::shared_ptr<Response> Client::send_with_content_provider(
+    const char *method, const char *path, const Headers &headers,
+    const std::string &body, size_t content_length,
+    ContentProvider content_provider, const char *content_type) {
+  Request req;
+  req.method = method;
+  req.headers = headers;
+  req.path = path;
+
+  req.headers.emplace("Content-Type", content_type);
+
+#ifdef CPPHTTPLIB_ZLIB_SUPPORT
+  if (compress_) {
+    if (content_provider) {
+      size_t offset = 0;
+
+      DataSink data_sink;
+      data_sink.write = [&](const char *data, size_t data_len) {
+        req.body.append(data, data_len);
+        offset += data_len;
+      };
+      data_sink.is_writable = [&](void) { return true; };
+
+      while (offset < content_length) {
+        content_provider(offset, content_length - offset, data_sink);
+      }
+    } else {
+      req.body = body;
+    }
+
+    if (!detail::compress(req.body)) { return nullptr; }
+    req.headers.emplace("Content-Encoding", "gzip");
+  } else
+#endif
+  {
+    if (content_provider) {
+      req.content_length = content_length;
+      req.content_provider = content_provider;
+    } else {
+      req.body = body;
+    }
+  }
+
+  auto res = std::make_shared<Response>();
+
+  return send(req, *res) ? res : nullptr;
 }
 
 inline bool Client::process_request(Stream &strm, const Request &req,
                                     Response &res, bool last_connection,
                                     bool &connection_close) {
   // Send request
-  write_request(strm, req, last_connection);
+  if (!write_request(strm, req, last_connection)) { return false; }
 
   // Receive response and headers
   if (!read_response_line(strm, res) ||
@@ -2963,21 +3942,16 @@ inline bool Client::process_request(Stream &strm, const Request &req,
   }
 
   // Body
-  if (req.method != "HEAD") {
-    detail::ContentReceiverCore out = [&](const char *buf, size_t n) {
+  if (req.method != "HEAD" && req.method != "CONNECT") {
+    ContentReceiver out = [&](const char *buf, size_t n) {
       if (res.body.size() + n > res.body.max_size()) { return false; }
       res.body.append(buf, n);
       return true;
     };
 
     if (req.content_receiver) {
-      auto offset = std::make_shared<size_t>();
-      auto length = get_header_value_uint64(res.headers, "Content-Length", 0);
-      auto receiver = req.content_receiver;
-      out = [offset, length, receiver](const char *buf, size_t n) {
-        auto ret = receiver(buf, n, *offset, length);
-        (*offset) += n;
-        return ret;
+      out = [&](const char *buf, size_t n) {
+        return req.content_receiver(buf, n);
       };
     }
 
@@ -2988,6 +3962,9 @@ inline bool Client::process_request(Stream &strm, const Request &req,
     }
   }
 
+  // Log
+  if (logger_) { logger_(req, res); }
+
   return true;
 }
 
@@ -2997,25 +3974,25 @@ inline bool Client::process_and_close_socket(
                        bool &connection_close)>
         callback) {
   request_count = std::min(request_count, keep_alive_max_count_);
-  return detail::process_and_close_socket(true, sock, request_count, callback);
+  return detail::process_and_close_socket(true, sock, request_count,
+                                          read_timeout_sec_, read_timeout_usec_,
+                                          callback);
 }
 
 inline bool Client::is_ssl() const { return false; }
 
 inline std::shared_ptr<Response> Client::Get(const char *path) {
-  Progress dummy;
-  return Get(path, Headers(), dummy);
+  return Get(path, Headers(), Progress());
 }
 
 inline std::shared_ptr<Response> Client::Get(const char *path,
                                              Progress progress) {
-  return Get(path, Headers(), progress);
+  return Get(path, Headers(), std::move(progress));
 }
 
 inline std::shared_ptr<Response> Client::Get(const char *path,
                                              const Headers &headers) {
-  Progress dummy;
-  return Get(path, headers, dummy);
+  return Get(path, headers, Progress());
 }
 
 inline std::shared_ptr<Response>
@@ -3024,7 +4001,7 @@ Client::Get(const char *path, const Headers &headers, Progress progress) {
   req.method = "GET";
   req.path = path;
   req.headers = headers;
-  req.progress = progress;
+  req.progress = std::move(progress);
 
   auto res = std::make_shared<Response>();
   return send(req, *res) ? res : nullptr;
@@ -3032,36 +4009,36 @@ Client::Get(const char *path, const Headers &headers, Progress progress) {
 
 inline std::shared_ptr<Response> Client::Get(const char *path,
                                              ContentReceiver content_receiver) {
-  Progress dummy;
-  return Get(path, Headers(), nullptr, content_receiver, dummy);
+  return Get(path, Headers(), nullptr, std::move(content_receiver), Progress());
 }
 
 inline std::shared_ptr<Response> Client::Get(const char *path,
                                              ContentReceiver content_receiver,
                                              Progress progress) {
-  return Get(path, Headers(), nullptr, content_receiver, progress);
+  return Get(path, Headers(), nullptr, std::move(content_receiver),
+             std::move(progress));
 }
 
 inline std::shared_ptr<Response> Client::Get(const char *path,
                                              const Headers &headers,
                                              ContentReceiver content_receiver) {
-  Progress dummy;
-  return Get(path, headers, nullptr, content_receiver, dummy);
+  return Get(path, headers, nullptr, std::move(content_receiver), Progress());
 }
 
 inline std::shared_ptr<Response> Client::Get(const char *path,
                                              const Headers &headers,
                                              ContentReceiver content_receiver,
                                              Progress progress) {
-  return Get(path, headers, nullptr, content_receiver, progress);
+  return Get(path, headers, nullptr, std::move(content_receiver),
+             std::move(progress));
 }
 
 inline std::shared_ptr<Response> Client::Get(const char *path,
                                              const Headers &headers,
                                              ResponseHandler response_handler,
                                              ContentReceiver content_receiver) {
-  Progress dummy;
-  return Get(path, headers, response_handler, content_receiver, dummy);
+  return Get(path, headers, std::move(response_handler), content_receiver,
+             Progress());
 }
 
 inline std::shared_ptr<Response> Client::Get(const char *path,
@@ -3073,9 +4050,9 @@ inline std::shared_ptr<Response> Client::Get(const char *path,
   req.method = "GET";
   req.path = path;
   req.headers = headers;
-  req.response_handler = response_handler;
-  req.content_receiver = content_receiver;
-  req.progress = progress;
+  req.response_handler = std::move(response_handler);
+  req.content_receiver = std::move(content_receiver);
+  req.progress = std::move(progress);
 
   auto res = std::make_shared<Response>();
   return send(req, *res) ? res : nullptr;
@@ -3107,17 +4084,8 @@ inline std::shared_ptr<Response> Client::Post(const char *path,
                                               const Headers &headers,
                                               const std::string &body,
                                               const char *content_type) {
-  Request req;
-  req.method = "POST";
-  req.headers = headers;
-  req.path = path;
-
-  req.headers.emplace("Content-Type", content_type);
-  req.body = body;
-
-  auto res = std::make_shared<Response>();
-
-  return send(req, *res) ? res : nullptr;
+  return send_with_content_provider("POST", path, headers, body, 0, nullptr,
+                                    content_type);
 }
 
 inline std::shared_ptr<Response> Client::Post(const char *path,
@@ -3125,6 +4093,21 @@ inline std::shared_ptr<Response> Client::Post(const char *path,
   return Post(path, Headers(), params);
 }
 
+inline std::shared_ptr<Response> Client::Post(const char *path,
+                                              size_t content_length,
+                                              ContentProvider content_provider,
+                                              const char *content_type) {
+  return Post(path, Headers(), content_length, content_provider, content_type);
+}
+
+inline std::shared_ptr<Response>
+Client::Post(const char *path, const Headers &headers, size_t content_length,
+             ContentProvider content_provider, const char *content_type) {
+  return send_with_content_provider("POST", path, headers, std::string(),
+                                    content_length, content_provider,
+                                    content_type);
+}
+
 inline std::shared_ptr<Response>
 Client::Post(const char *path, const Headers &headers, const Params &params) {
   std::string query;
@@ -3146,35 +4129,28 @@ Client::Post(const char *path, const MultipartFormDataItems &items) {
 inline std::shared_ptr<Response>
 Client::Post(const char *path, const Headers &headers,
              const MultipartFormDataItems &items) {
-  Request req;
-  req.method = "POST";
-  req.headers = headers;
-  req.path = path;
-
   auto boundary = detail::make_multipart_data_boundary();
 
-  req.headers.emplace("Content-Type",
-                      "multipart/form-data; boundary=" + boundary);
+  std::string body;
 
   for (const auto &item : items) {
-    req.body += "--" + boundary + "\r\n";
-    req.body += "Content-Disposition: form-data; name=\"" + item.name + "\"";
+    body += "--" + boundary + "\r\n";
+    body += "Content-Disposition: form-data; name=\"" + item.name + "\"";
     if (!item.filename.empty()) {
-      req.body += "; filename=\"" + item.filename + "\"";
+      body += "; filename=\"" + item.filename + "\"";
     }
-    req.body += "\r\n";
+    body += "\r\n";
     if (!item.content_type.empty()) {
-      req.body += "Content-Type: " + item.content_type + "\r\n";
+      body += "Content-Type: " + item.content_type + "\r\n";
     }
-    req.body += "\r\n";
-    req.body += item.content + "\r\n";
+    body += "\r\n";
+    body += item.content + "\r\n";
   }
 
-  req.body += "--" + boundary + "--\r\n";
+  body += "--" + boundary + "--\r\n";
 
-  auto res = std::make_shared<Response>();
-
-  return send(req, *res) ? res : nullptr;
+  std::string content_type = "multipart/form-data; boundary=" + boundary;
+  return Post(path, headers, body, content_type.c_str());
 }
 
 inline std::shared_ptr<Response> Client::Put(const char *path,
@@ -3187,17 +4163,41 @@ inline std::shared_ptr<Response> Client::Put(const char *path,
                                              const Headers &headers,
                                              const std::string &body,
                                              const char *content_type) {
-  Request req;
-  req.method = "PUT";
-  req.headers = headers;
-  req.path = path;
+  return send_with_content_provider("PUT", path, headers, body, 0, nullptr,
+                                    content_type);
+}
 
-  req.headers.emplace("Content-Type", content_type);
-  req.body = body;
+inline std::shared_ptr<Response> Client::Put(const char *path,
+                                             size_t content_length,
+                                             ContentProvider content_provider,
+                                             const char *content_type) {
+  return Put(path, Headers(), content_length, content_provider, content_type);
+}
 
-  auto res = std::make_shared<Response>();
+inline std::shared_ptr<Response>
+Client::Put(const char *path, const Headers &headers, size_t content_length,
+            ContentProvider content_provider, const char *content_type) {
+  return send_with_content_provider("PUT", path, headers, std::string(),
+                                    content_length, content_provider,
+                                    content_type);
+}
 
-  return send(req, *res) ? res : nullptr;
+inline std::shared_ptr<Response> Client::Put(const char *path,
+                                             const Params &params) {
+  return Put(path, Headers(), params);
+}
+
+inline std::shared_ptr<Response>
+Client::Put(const char *path, const Headers &headers, const Params &params) {
+  std::string query;
+  for (auto it = params.begin(); it != params.end(); ++it) {
+    if (it != params.begin()) { query += "&"; }
+    query += it->first;
+    query += "=";
+    query += detail::encode_url(it->second);
+  }
+
+  return Put(path, headers, query, "application/x-www-form-urlencoded");
 }
 
 inline std::shared_ptr<Response> Client::Patch(const char *path,
@@ -3210,17 +4210,23 @@ inline std::shared_ptr<Response> Client::Patch(const char *path,
                                                const Headers &headers,
                                                const std::string &body,
                                                const char *content_type) {
-  Request req;
-  req.method = "PATCH";
-  req.headers = headers;
-  req.path = path;
-
-  req.headers.emplace("Content-Type", content_type);
-  req.body = body;
+  return send_with_content_provider("PATCH", path, headers, body, 0, nullptr,
+                                    content_type);
+}
 
-  auto res = std::make_shared<Response>();
+inline std::shared_ptr<Response> Client::Patch(const char *path,
+                                               size_t content_length,
+                                               ContentProvider content_provider,
+                                               const char *content_type) {
+  return Patch(path, Headers(), content_length, content_provider, content_type);
+}
 
-  return send(req, *res) ? res : nullptr;
+inline std::shared_ptr<Response>
+Client::Patch(const char *path, const Headers &headers, size_t content_length,
+              ContentProvider content_provider, const char *content_type) {
+  return send_with_content_provider("PATCH", path, headers, std::string(),
+                                    content_length, content_provider,
+                                    content_type);
 }
 
 inline std::shared_ptr<Response> Client::Delete(const char *path) {
@@ -3271,11 +4277,58 @@ inline std::shared_ptr<Response> Client::Options(const char *path,
   return send(req, *res) ? res : nullptr;
 }
 
+inline void Client::set_timeout_sec(time_t timeout_sec) {
+  timeout_sec_ = timeout_sec;
+}
+
+inline void Client::set_read_timeout(time_t sec, time_t usec) {
+  read_timeout_sec_ = sec;
+  read_timeout_usec_ = usec;
+}
+
 inline void Client::set_keep_alive_max_count(size_t count) {
   keep_alive_max_count_ = count;
 }
 
-inline void Client::follow_location(bool on) { follow_location_ = on; }
+inline void Client::set_basic_auth(const char *username, const char *password) {
+  basic_auth_username_ = username;
+  basic_auth_password_ = password;
+}
+
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+inline void Client::set_digest_auth(const char *username,
+                                    const char *password) {
+  digest_auth_username_ = username;
+  digest_auth_password_ = password;
+}
+#endif
+
+inline void Client::set_follow_location(bool on) { follow_location_ = on; }
+
+inline void Client::set_compress(bool on) { compress_ = on; }
+
+inline void Client::set_interface(const char *intf) { interface_ = intf; }
+
+inline void Client::set_proxy(const char *host, int port) {
+  proxy_host_ = host;
+  proxy_port_ = port;
+}
+
+inline void Client::set_proxy_basic_auth(const char *username,
+                                         const char *password) {
+  proxy_basic_auth_username_ = username;
+  proxy_basic_auth_password_ = password;
+}
+
+#ifdef CPPHTTPLIB_OPENSSL_SUPPORT
+inline void Client::set_proxy_digest_auth(const char *username,
+                                          const char *password) {
+  proxy_digest_auth_username_ = username;
+  proxy_digest_auth_password_ = password;
+}
+#endif
+
+inline void Client::set_logger(Logger logger) { logger_ = std::move(logger); }
 
 /*
  * SSL Implementation
@@ -3284,11 +4337,10 @@ inline void Client::follow_location(bool on) { follow_location_ = on; }
 namespace detail {
 
 template <typename U, typename V, typename T>
-inline bool process_and_close_socket_ssl(bool is_client_request, socket_t sock,
-                                         size_t keep_alive_max_count,
-                                         SSL_CTX *ctx, std::mutex &ctx_mutex,
-                                         U SSL_connect_or_accept, V setup,
-                                         T callback) {
+inline bool process_and_close_socket_ssl(
+    bool is_client_request, socket_t sock, size_t keep_alive_max_count,
+    time_t read_timeout_sec, time_t read_timeout_usec, SSL_CTX *ctx,
+    std::mutex &ctx_mutex, U SSL_connect_or_accept, V setup, T callback) {
   assert(keep_alive_max_count > 0);
 
   SSL *ssl = nullptr;
@@ -3316,7 +4368,7 @@ inline bool process_and_close_socket_ssl(bool is_client_request, socket_t sock,
     return false;
   }
 
-  bool ret = false;
+  auto ret = false;
 
   if (SSL_connect_or_accept(ssl) == 1) {
     if (keep_alive_max_count > 1) {
@@ -3325,7 +4377,7 @@ inline bool process_and_close_socket_ssl(bool is_client_request, socket_t sock,
              (is_client_request ||
               detail::select_read(sock, CPPHTTPLIB_KEEPALIVE_TIMEOUT_SECOND,
                                   CPPHTTPLIB_KEEPALIVE_TIMEOUT_USECOND) > 0)) {
-        SSLSocketStream strm(sock, ssl);
+        SSLSocketStream strm(sock, ssl, read_timeout_sec, read_timeout_usec);
         auto last_connection = count == 1;
         auto connection_close = false;
 
@@ -3335,7 +4387,7 @@ inline bool process_and_close_socket_ssl(bool is_client_request, socket_t sock,
         count--;
       }
     } else {
-      SSLSocketStream strm(sock, ssl);
+      SSLSocketStream strm(sock, ssl, read_timeout_sec, read_timeout_usec);
       auto dummy_connection_close = false;
       ret = callback(ssl, strm, true, dummy_connection_close);
     }
@@ -3382,11 +4434,20 @@ private:
 class SSLInit {
 public:
   SSLInit() {
+#if OPENSSL_VERSION_NUMBER < 0x1010001fL
     SSL_load_error_strings();
     SSL_library_init();
+#else
+    OPENSSL_init_ssl(
+        OPENSSL_INIT_LOAD_SSL_STRINGS | OPENSSL_INIT_LOAD_CRYPTO_STRINGS, NULL);
+#endif
   }
 
-  ~SSLInit() { ERR_free_strings(); }
+  ~SSLInit() {
+#if OPENSSL_VERSION_NUMBER < 0x1010001fL
+    ERR_free_strings();
+#endif
+  }
 
 private:
 #if OPENSSL_VERSION_NUMBER < 0x10100000L
@@ -3394,41 +4455,44 @@ private:
 #endif
 };
 
-static SSLInit sslinit_;
-
-} // namespace detail
-
 // SSL socket stream implementation
-inline SSLSocketStream::SSLSocketStream(socket_t sock, SSL *ssl)
-    : sock_(sock), ssl_(ssl) {}
+inline SSLSocketStream::SSLSocketStream(socket_t sock, SSL *ssl,
+                                        time_t read_timeout_sec,
+                                        time_t read_timeout_usec)
+    : sock_(sock), ssl_(ssl), read_timeout_sec_(read_timeout_sec),
+      read_timeout_usec_(read_timeout_usec) {}
 
 inline SSLSocketStream::~SSLSocketStream() {}
 
+inline bool SSLSocketStream::is_readable() const {
+  return detail::select_read(sock_, read_timeout_sec_, read_timeout_usec_) > 0;
+}
+
+inline bool SSLSocketStream::is_writable() const {
+  return detail::select_write(sock_, 0, 0) > 0;
+}
+
 inline int SSLSocketStream::read(char *ptr, size_t size) {
   if (SSL_pending(ssl_) > 0 ||
-      detail::select_read(sock_, CPPHTTPLIB_READ_TIMEOUT_SECOND,
-                          CPPHTTPLIB_READ_TIMEOUT_USECOND) > 0) {
+      select_read(sock_, read_timeout_sec_, read_timeout_usec_) > 0) {
     return SSL_read(ssl_, ptr, static_cast<int>(size));
   }
   return -1;
 }
 
 inline int SSLSocketStream::write(const char *ptr, size_t size) {
-  return SSL_write(ssl_, ptr, static_cast<int>(size));
-}
-
-inline int SSLSocketStream::write(const char *ptr) {
-  return write(ptr, strlen(ptr));
-}
-
-inline int SSLSocketStream::write(const std::string &s) {
-  return write(s.data(), s.size());
+  if (is_writable()) { return SSL_write(ssl_, ptr, static_cast<int>(size)); }
+  return -1;
 }
 
 inline std::string SSLSocketStream::get_remote_addr() const {
   return detail::get_remote_addr(sock_);
 }
 
+static SSLInit sslinit_;
+
+} // namespace detail
+
 // SSL HTTP server implementation
 inline SSLServer::SSLServer(const char *cert_path, const char *private_key_path,
                             const char *client_ca_cert_file_path,
@@ -3476,8 +4540,8 @@ inline bool SSLServer::is_valid() const { return ctx_; }
 
 inline bool SSLServer::process_and_close_socket(socket_t sock) {
   return detail::process_and_close_socket_ssl(
-      false, sock, keep_alive_max_count_, ctx_, ctx_mutex_, SSL_accept,
-      [](SSL * /*ssl*/) { return true; },
+      false, sock, keep_alive_max_count_, read_timeout_sec_, read_timeout_usec_,
+      ctx_, ctx_mutex_, SSL_accept, [](SSL * /*ssl*/) { return true; },
       [this](SSL *ssl, Stream &strm, bool last_connection,
              bool &connection_close) {
         return process_request(strm, last_connection, connection_close,
@@ -3486,21 +4550,21 @@ inline bool SSLServer::process_and_close_socket(socket_t sock) {
 }
 
 // SSL HTTP client implementation
-inline SSLClient::SSLClient(const char *host, int port, time_t timeout_sec,
-                            const char *client_cert_path,
-                            const char *client_key_path)
-    : Client(host, port, timeout_sec) {
+inline SSLClient::SSLClient(const std::string &host, int port,
+                            const std::string &client_cert_path,
+                            const std::string &client_key_path)
+    : Client(host, port, client_cert_path, client_key_path) {
   ctx_ = SSL_CTX_new(SSLv23_client_method());
 
   detail::split(&host_[0], &host_[host_.size()], '.',
                 [&](const char *b, const char *e) {
                   host_components_.emplace_back(std::string(b, e));
                 });
-  if (client_cert_path && client_key_path) {
-    if (SSL_CTX_use_certificate_file(ctx_, client_cert_path,
+  if (!client_cert_path.empty() && !client_key_path.empty()) {
+    if (SSL_CTX_use_certificate_file(ctx_, client_cert_path.c_str(),
                                      SSL_FILETYPE_PEM) != 1 ||
-        SSL_CTX_use_PrivateKey_file(ctx_, client_key_path, SSL_FILETYPE_PEM) !=
-            1) {
+        SSL_CTX_use_PrivateKey_file(ctx_, client_key_path.c_str(),
+                                    SSL_FILETYPE_PEM) != 1) {
       SSL_CTX_free(ctx_);
       ctx_ = nullptr;
     }
@@ -3527,9 +4591,7 @@ inline long SSLClient::get_openssl_verify_result() const {
   return verify_result_;
 }
 
-inline SSL_CTX* SSLClient::ssl_context() const noexcept {
-        return ctx_;
-}
+inline SSL_CTX *SSLClient::ssl_context() const noexcept { return ctx_; }
 
 inline bool SSLClient::process_and_close_socket(
     socket_t sock, size_t request_count,
@@ -3541,7 +4603,8 @@ inline bool SSLClient::process_and_close_socket(
 
   return is_valid() &&
          detail::process_and_close_socket_ssl(
-             true, sock, request_count, ctx_, ctx_mutex_,
+             true, sock, request_count, read_timeout_sec_, read_timeout_usec_,
+             ctx_, ctx_mutex_,
              [&](SSL *ssl) {
                if (ca_cert_file_path_.empty()) {
                  SSL_CTX_set_verify(ctx_, SSL_VERIFY_NONE, nullptr);
@@ -3712,6 +4775,8 @@ inline bool SSLClient::check_host_name(const char *pattern,
 }
 #endif
 
+// ----------------------------------------------------------------------------
+
 } // namespace httplib
 
 #endif // CPPHTTPLIB_HTTPLIB_H
diff --git a/src/audio_core/algorithm/interpolate.cpp b/src/audio_core/algorithm/interpolate.cpp
index 5005ba519..a58f24169 100644
--- a/src/audio_core/algorithm/interpolate.cpp
+++ b/src/audio_core/algorithm/interpolate.cpp
@@ -5,6 +5,7 @@
 #define _USE_MATH_DEFINES
 
 #include <algorithm>
+#include <climits>
 #include <cmath>
 #include <vector>
 #include "audio_core/algorithm/interpolate.h"
@@ -13,13 +14,131 @@
 
 namespace AudioCore {
 
-/// The Lanczos kernel
-static double Lanczos(std::size_t a, double x) {
-    if (x == 0.0)
-        return 1.0;
-    const double px = M_PI * x;
-    return a * std::sin(px) * std::sin(px / a) / (px * px);
-}
+constexpr std::array<s16, 512> curve_lut0 = {
+    6600,  19426, 6722,  3,     6479,  19424, 6845,  9,     6359,  19419, 6968,  15,    6239,
+    19412, 7093,  22,    6121,  19403, 7219,  28,    6004,  19391, 7345,  34,    5888,  19377,
+    7472,  41,    5773,  19361, 7600,  48,    5659,  19342, 7728,  55,    5546,  19321, 7857,
+    62,    5434,  19298, 7987,  69,    5323,  19273, 8118,  77,    5213,  19245, 8249,  84,
+    5104,  19215, 8381,  92,    4997,  19183, 8513,  101,   4890,  19148, 8646,  109,   4785,
+    19112, 8780,  118,   4681,  19073, 8914,  127,   4579,  19031, 9048,  137,   4477,  18988,
+    9183,  147,   4377,  18942, 9318,  157,   4277,  18895, 9454,  168,   4179,  18845, 9590,
+    179,   4083,  18793, 9726,  190,   3987,  18738, 9863,  202,   3893,  18682, 10000, 215,
+    3800,  18624, 10137, 228,   3709,  18563, 10274, 241,   3618,  18500, 10411, 255,   3529,
+    18436, 10549, 270,   3441,  18369, 10687, 285,   3355,  18300, 10824, 300,   3269,  18230,
+    10962, 317,   3186,  18157, 11100, 334,   3103,  18082, 11238, 351,   3022,  18006, 11375,
+    369,   2942,  17927, 11513, 388,   2863,  17847, 11650, 408,   2785,  17765, 11788, 428,
+    2709,  17681, 11925, 449,   2635,  17595, 12062, 471,   2561,  17507, 12198, 494,   2489,
+    17418, 12334, 517,   2418,  17327, 12470, 541,   2348,  17234, 12606, 566,   2280,  17140,
+    12741, 592,   2213,  17044, 12876, 619,   2147,  16946, 13010, 647,   2083,  16846, 13144,
+    675,   2020,  16745, 13277, 704,   1958,  16643, 13409, 735,   1897,  16539, 13541, 766,
+    1838,  16434, 13673, 798,   1780,  16327, 13803, 832,   1723,  16218, 13933, 866,   1667,
+    16109, 14062, 901,   1613,  15998, 14191, 937,   1560,  15885, 14318, 975,   1508,  15772,
+    14445, 1013,  1457,  15657, 14571, 1052,  1407,  15540, 14695, 1093,  1359,  15423, 14819,
+    1134,  1312,  15304, 14942, 1177,  1266,  15185, 15064, 1221,  1221,  15064, 15185, 1266,
+    1177,  14942, 15304, 1312,  1134,  14819, 15423, 1359,  1093,  14695, 15540, 1407,  1052,
+    14571, 15657, 1457,  1013,  14445, 15772, 1508,  975,   14318, 15885, 1560,  937,   14191,
+    15998, 1613,  901,   14062, 16109, 1667,  866,   13933, 16218, 1723,  832,   13803, 16327,
+    1780,  798,   13673, 16434, 1838,  766,   13541, 16539, 1897,  735,   13409, 16643, 1958,
+    704,   13277, 16745, 2020,  675,   13144, 16846, 2083,  647,   13010, 16946, 2147,  619,
+    12876, 17044, 2213,  592,   12741, 17140, 2280,  566,   12606, 17234, 2348,  541,   12470,
+    17327, 2418,  517,   12334, 17418, 2489,  494,   12198, 17507, 2561,  471,   12062, 17595,
+    2635,  449,   11925, 17681, 2709,  428,   11788, 17765, 2785,  408,   11650, 17847, 2863,
+    388,   11513, 17927, 2942,  369,   11375, 18006, 3022,  351,   11238, 18082, 3103,  334,
+    11100, 18157, 3186,  317,   10962, 18230, 3269,  300,   10824, 18300, 3355,  285,   10687,
+    18369, 3441,  270,   10549, 18436, 3529,  255,   10411, 18500, 3618,  241,   10274, 18563,
+    3709,  228,   10137, 18624, 3800,  215,   10000, 18682, 3893,  202,   9863,  18738, 3987,
+    190,   9726,  18793, 4083,  179,   9590,  18845, 4179,  168,   9454,  18895, 4277,  157,
+    9318,  18942, 4377,  147,   9183,  18988, 4477,  137,   9048,  19031, 4579,  127,   8914,
+    19073, 4681,  118,   8780,  19112, 4785,  109,   8646,  19148, 4890,  101,   8513,  19183,
+    4997,  92,    8381,  19215, 5104,  84,    8249,  19245, 5213,  77,    8118,  19273, 5323,
+    69,    7987,  19298, 5434,  62,    7857,  19321, 5546,  55,    7728,  19342, 5659,  48,
+    7600,  19361, 5773,  41,    7472,  19377, 5888,  34,    7345,  19391, 6004,  28,    7219,
+    19403, 6121,  22,    7093,  19412, 6239,  15,    6968,  19419, 6359,  9,     6845,  19424,
+    6479,  3,     6722,  19426, 6600};
+
+constexpr std::array<s16, 512> curve_lut1 = {
+    -68,   32639, 69,    -5,    -200,  32630, 212,   -15,   -328,  32613, 359,   -26,   -450,
+    32586, 512,   -36,   -568,  32551, 669,   -47,   -680,  32507, 832,   -58,   -788,  32454,
+    1000,  -69,   -891,  32393, 1174,  -80,   -990,  32323, 1352,  -92,   -1084, 32244, 1536,
+    -103,  -1173, 32157, 1724,  -115,  -1258, 32061, 1919,  -128,  -1338, 31956, 2118,  -140,
+    -1414, 31844, 2322,  -153,  -1486, 31723, 2532,  -167,  -1554, 31593, 2747,  -180,  -1617,
+    31456, 2967,  -194,  -1676, 31310, 3192,  -209,  -1732, 31157, 3422,  -224,  -1783, 30995,
+    3657,  -240,  -1830, 30826, 3897,  -256,  -1874, 30649, 4143,  -272,  -1914, 30464, 4393,
+    -289,  -1951, 30272, 4648,  -307,  -1984, 30072, 4908,  -325,  -2014, 29866, 5172,  -343,
+    -2040, 29652, 5442,  -362,  -2063, 29431, 5716,  -382,  -2083, 29203, 5994,  -403,  -2100,
+    28968, 6277,  -424,  -2114, 28727, 6565,  -445,  -2125, 28480, 6857,  -468,  -2133, 28226,
+    7153,  -490,  -2139, 27966, 7453,  -514,  -2142, 27700, 7758,  -538,  -2142, 27428, 8066,
+    -563,  -2141, 27151, 8378,  -588,  -2136, 26867, 8694,  -614,  -2130, 26579, 9013,  -641,
+    -2121, 26285, 9336,  -668,  -2111, 25987, 9663,  -696,  -2098, 25683, 9993,  -724,  -2084,
+    25375, 10326, -753,  -2067, 25063, 10662, -783,  -2049, 24746, 11000, -813,  -2030, 24425,
+    11342, -844,  -2009, 24100, 11686, -875,  -1986, 23771, 12033, -907,  -1962, 23438, 12382,
+    -939,  -1937, 23103, 12733, -972,  -1911, 22764, 13086, -1005, -1883, 22422, 13441, -1039,
+    -1855, 22077, 13798, -1072, -1825, 21729, 14156, -1107, -1795, 21380, 14516, -1141, -1764,
+    21027, 14877, -1176, -1732, 20673, 15239, -1211, -1700, 20317, 15602, -1246, -1667, 19959,
+    15965, -1282, -1633, 19600, 16329, -1317, -1599, 19239, 16694, -1353, -1564, 18878, 17058,
+    -1388, -1530, 18515, 17423, -1424, -1495, 18151, 17787, -1459, -1459, 17787, 18151, -1495,
+    -1424, 17423, 18515, -1530, -1388, 17058, 18878, -1564, -1353, 16694, 19239, -1599, -1317,
+    16329, 19600, -1633, -1282, 15965, 19959, -1667, -1246, 15602, 20317, -1700, -1211, 15239,
+    20673, -1732, -1176, 14877, 21027, -1764, -1141, 14516, 21380, -1795, -1107, 14156, 21729,
+    -1825, -1072, 13798, 22077, -1855, -1039, 13441, 22422, -1883, -1005, 13086, 22764, -1911,
+    -972,  12733, 23103, -1937, -939,  12382, 23438, -1962, -907,  12033, 23771, -1986, -875,
+    11686, 24100, -2009, -844,  11342, 24425, -2030, -813,  11000, 24746, -2049, -783,  10662,
+    25063, -2067, -753,  10326, 25375, -2084, -724,  9993,  25683, -2098, -696,  9663,  25987,
+    -2111, -668,  9336,  26285, -2121, -641,  9013,  26579, -2130, -614,  8694,  26867, -2136,
+    -588,  8378,  27151, -2141, -563,  8066,  27428, -2142, -538,  7758,  27700, -2142, -514,
+    7453,  27966, -2139, -490,  7153,  28226, -2133, -468,  6857,  28480, -2125, -445,  6565,
+    28727, -2114, -424,  6277,  28968, -2100, -403,  5994,  29203, -2083, -382,  5716,  29431,
+    -2063, -362,  5442,  29652, -2040, -343,  5172,  29866, -2014, -325,  4908,  30072, -1984,
+    -307,  4648,  30272, -1951, -289,  4393,  30464, -1914, -272,  4143,  30649, -1874, -256,
+    3897,  30826, -1830, -240,  3657,  30995, -1783, -224,  3422,  31157, -1732, -209,  3192,
+    31310, -1676, -194,  2967,  31456, -1617, -180,  2747,  31593, -1554, -167,  2532,  31723,
+    -1486, -153,  2322,  31844, -1414, -140,  2118,  31956, -1338, -128,  1919,  32061, -1258,
+    -115,  1724,  32157, -1173, -103,  1536,  32244, -1084, -92,   1352,  32323, -990,  -80,
+    1174,  32393, -891,  -69,   1000,  32454, -788,  -58,   832,   32507, -680,  -47,   669,
+    32551, -568,  -36,   512,   32586, -450,  -26,   359,   32613, -328,  -15,   212,   32630,
+    -200,  -5,    69,    32639, -68};
+
+constexpr std::array<s16, 512> curve_lut2 = {
+    3195,  26287, 3329,  -32,   3064,  26281, 3467,  -34,   2936,  26270, 3608,  -38,   2811,
+    26253, 3751,  -42,   2688,  26230, 3897,  -46,   2568,  26202, 4046,  -50,   2451,  26169,
+    4199,  -54,   2338,  26130, 4354,  -58,   2227,  26085, 4512,  -63,   2120,  26035, 4673,
+    -67,   2015,  25980, 4837,  -72,   1912,  25919, 5004,  -76,   1813,  25852, 5174,  -81,
+    1716,  25780, 5347,  -87,   1622,  25704, 5522,  -92,   1531,  25621, 5701,  -98,   1442,
+    25533, 5882,  -103,  1357,  25440, 6066,  -109,  1274,  25342, 6253,  -115,  1193,  25239,
+    6442,  -121,  1115,  25131, 6635,  -127,  1040,  25018, 6830,  -133,  967,   24899, 7027,
+    -140,  897,   24776, 7227,  -146,  829,   24648, 7430,  -153,  764,   24516, 7635,  -159,
+    701,   24379, 7842,  -166,  641,   24237, 8052,  -174,  583,   24091, 8264,  -181,  526,
+    23940, 8478,  -187,  472,   23785, 8695,  -194,  420,   23626, 8914,  -202,  371,   23462,
+    9135,  -209,  324,   23295, 9358,  -215,  279,   23123, 9583,  -222,  236,   22948, 9809,
+    -230,  194,   22769, 10038, -237,  154,   22586, 10269, -243,  117,   22399, 10501, -250,
+    81,    22208, 10735, -258,  47,    22015, 10970, -265,  15,    21818, 11206, -271,  -16,
+    21618, 11444, -277,  -44,   21415, 11684, -283,  -71,   21208, 11924, -290,  -97,   20999,
+    12166, -296,  -121,  20786, 12409, -302,  -143,  20571, 12653, -306,  -163,  20354, 12898,
+    -311,  -183,  20134, 13143, -316,  -201,  19911, 13389, -321,  -218,  19686, 13635, -325,
+    -234,  19459, 13882, -328,  -248,  19230, 14130, -332,  -261,  18998, 14377, -335,  -273,
+    18765, 14625, -337,  -284,  18531, 14873, -339,  -294,  18295, 15121, -341,  -302,  18057,
+    15369, -341,  -310,  17817, 15617, -341,  -317,  17577, 15864, -340,  -323,  17335, 16111,
+    -340,  -328,  17092, 16357, -338,  -332,  16848, 16603, -336,  -336,  16603, 16848, -332,
+    -338,  16357, 17092, -328,  -340,  16111, 17335, -323,  -340,  15864, 17577, -317,  -341,
+    15617, 17817, -310,  -341,  15369, 18057, -302,  -341,  15121, 18295, -294,  -339,  14873,
+    18531, -284,  -337,  14625, 18765, -273,  -335,  14377, 18998, -261,  -332,  14130, 19230,
+    -248,  -328,  13882, 19459, -234,  -325,  13635, 19686, -218,  -321,  13389, 19911, -201,
+    -316,  13143, 20134, -183,  -311,  12898, 20354, -163,  -306,  12653, 20571, -143,  -302,
+    12409, 20786, -121,  -296,  12166, 20999, -97,   -290,  11924, 21208, -71,   -283,  11684,
+    21415, -44,   -277,  11444, 21618, -16,   -271,  11206, 21818, 15,    -265,  10970, 22015,
+    47,    -258,  10735, 22208, 81,    -250,  10501, 22399, 117,   -243,  10269, 22586, 154,
+    -237,  10038, 22769, 194,   -230,  9809,  22948, 236,   -222,  9583,  23123, 279,   -215,
+    9358,  23295, 324,   -209,  9135,  23462, 371,   -202,  8914,  23626, 420,   -194,  8695,
+    23785, 472,   -187,  8478,  23940, 526,   -181,  8264,  24091, 583,   -174,  8052,  24237,
+    641,   -166,  7842,  24379, 701,   -159,  7635,  24516, 764,   -153,  7430,  24648, 829,
+    -146,  7227,  24776, 897,   -140,  7027,  24899, 967,   -133,  6830,  25018, 1040,  -127,
+    6635,  25131, 1115,  -121,  6442,  25239, 1193,  -115,  6253,  25342, 1274,  -109,  6066,
+    25440, 1357,  -103,  5882,  25533, 1442,  -98,   5701,  25621, 1531,  -92,   5522,  25704,
+    1622,  -87,   5347,  25780, 1716,  -81,   5174,  25852, 1813,  -76,   5004,  25919, 1912,
+    -72,   4837,  25980, 2015,  -67,   4673,  26035, 2120,  -63,   4512,  26085, 2227,  -58,
+    4354,  26130, 2338,  -54,   4199,  26169, 2451,  -50,   4046,  26202, 2568,  -46,   3897,
+    26230, 2688,  -42,   3751,  26253, 2811,  -38,   3608,  26270, 2936,  -34,   3467,  26281,
+    3064,  -32,   3329,  26287, 3195};
 
 std::vector<s16> Interpolate(InterpolationState& state, std::vector<s16> input, double ratio) {
     if (input.size() < 2)
@@ -30,40 +149,39 @@ std::vector<s16> Interpolate(InterpolationState& state, std::vector<s16> input,
         ratio = 1.0;
     }
 
-    if (ratio != state.current_ratio) {
-        const double cutoff_frequency = std::min(0.5 / ratio, 0.5 * ratio);
-        state.nyquist = CascadingFilter::LowPass(std::clamp(cutoff_frequency, 0.0, 0.4), 3);
-        state.current_ratio = ratio;
-    }
-    state.nyquist.Process(input);
-
-    constexpr std::size_t taps = InterpolationState::lanczos_taps;
-    const std::size_t num_frames = input.size() / 2;
-
-    std::vector<s16> output;
-    output.reserve(static_cast<std::size_t>(input.size() / ratio + 4));
-
-    double& pos = state.position;
-    auto& h = state.history;
-    for (std::size_t i = 0; i < num_frames; ++i) {
-        std::rotate(h.begin(), h.end() - 1, h.end());
-        h[0][0] = input[i * 2 + 0];
-        h[0][1] = input[i * 2 + 1];
-
-        while (pos <= 1.0) {
-            double l = 0.0;
-            double r = 0.0;
-            for (std::size_t j = 0; j < h.size(); j++) {
-                const double lanczos_calc = Lanczos(taps, pos + j - taps + 1);
-                l += lanczos_calc * h[j][0];
-                r += lanczos_calc * h[j][1];
-            }
-            output.emplace_back(static_cast<s16>(std::clamp(l, -32768.0, 32767.0)));
-            output.emplace_back(static_cast<s16>(std::clamp(r, -32768.0, 32767.0)));
-
-            pos += ratio;
+    const int step = static_cast<int>(ratio * 0x8000);
+    const std::array<s16, 512>& lut = [step] {
+        if (step > 0xaaaa) {
+            return curve_lut0;
+        }
+        if (step <= 0x8000) {
+            return curve_lut1;
         }
-        pos -= 1.0;
+        return curve_lut2;
+    }();
+
+    std::vector<s16> output(static_cast<std::size_t>(input.size() / ratio));
+    int in_offset = 0;
+    for (std::size_t out_offset = 0; out_offset < output.size(); out_offset += 2) {
+        const int lut_index = (state.fraction >> 8) * 4;
+
+        const int l = input[(in_offset + 0) * 2 + 0] * lut[lut_index + 0] +
+                      input[(in_offset + 1) * 2 + 0] * lut[lut_index + 1] +
+                      input[(in_offset + 2) * 2 + 0] * lut[lut_index + 2] +
+                      input[(in_offset + 3) * 2 + 0] * lut[lut_index + 3];
+
+        const int r = input[(in_offset + 0) * 2 + 1] * lut[lut_index + 0] +
+                      input[(in_offset + 1) * 2 + 1] * lut[lut_index + 1] +
+                      input[(in_offset + 2) * 2 + 1] * lut[lut_index + 2] +
+                      input[(in_offset + 3) * 2 + 1] * lut[lut_index + 3];
+
+        const int new_offset = state.fraction + step;
+
+        in_offset += new_offset >> 15;
+        state.fraction = new_offset & 0x7fff;
+
+        output[out_offset + 0] = static_cast<s16>(std::clamp(l >> 15, SHRT_MIN, SHRT_MAX));
+        output[out_offset + 1] = static_cast<s16>(std::clamp(r >> 15, SHRT_MIN, SHRT_MAX));
     }
 
     return output;
diff --git a/src/audio_core/algorithm/interpolate.h b/src/audio_core/algorithm/interpolate.h
index edbd6460f..1b9831a75 100644
--- a/src/audio_core/algorithm/interpolate.h
+++ b/src/audio_core/algorithm/interpolate.h
@@ -6,19 +6,12 @@
 
 #include <array>
 #include <vector>
-#include "audio_core/algorithm/filter.h"
 #include "common/common_types.h"
 
 namespace AudioCore {
 
 struct InterpolationState {
-    static constexpr std::size_t lanczos_taps = 4;
-    static constexpr std::size_t history_size = lanczos_taps * 2 - 1;
-
-    double current_ratio = 0.0;
-    CascadingFilter nyquist;
-    std::array<std::array<s16, 2>, history_size> history = {};
-    double position = 0;
+    int fraction = 0;
 };
 
 /// Interpolates input signal to produce output signal.
diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt
index 26612e692..88c06b2ce 100644
--- a/src/core/CMakeLists.txt
+++ b/src/core/CMakeLists.txt
@@ -187,6 +187,8 @@ add_library(core STATIC
     hle/kernel/synchronization.h
     hle/kernel/thread.cpp
     hle/kernel/thread.h
+    hle/kernel/time_manager.cpp
+    hle/kernel/time_manager.h
     hle/kernel/transfer_memory.cpp
     hle/kernel/transfer_memory.h
     hle/kernel/vm_manager.cpp
diff --git a/src/core/core.cpp b/src/core/core.cpp
index 0eb0c0dca..86e314c94 100644
--- a/src/core/core.cpp
+++ b/src/core/core.cpp
@@ -707,4 +707,12 @@ const Service::SM::ServiceManager& System::ServiceManager() const {
     return *impl->service_manager;
 }
 
+void System::RegisterCoreThread(std::size_t id) {
+    impl->kernel.RegisterCoreThread(id);
+}
+
+void System::RegisterHostThread() {
+    impl->kernel.RegisterHostThread();
+}
+
 } // namespace Core
diff --git a/src/core/core.h b/src/core/core.h
index e69d68fcf..8d862a8e6 100644
--- a/src/core/core.h
+++ b/src/core/core.h
@@ -360,6 +360,12 @@ public:
 
     const CurrentBuildProcessID& GetCurrentProcessBuildID() const;
 
+    /// Register a host thread as an emulated CPU Core.
+    void RegisterCoreThread(std::size_t id);
+
+    /// Register a host thread as an auxiliary thread.
+    void RegisterHostThread();
+
 private:
     System();
 
diff --git a/src/core/frontend/framebuffer_layout.cpp b/src/core/frontend/framebuffer_layout.cpp
index d6d2cf3f0..2dc795d56 100644
--- a/src/core/frontend/framebuffer_layout.cpp
+++ b/src/core/frontend/framebuffer_layout.cpp
@@ -27,9 +27,9 @@ FramebufferLayout DefaultFrameLayout(u32 width, u32 height) {
     // so just calculate them both even if the other isn't showing.
     FramebufferLayout res{width, height};
 
-    const float emulation_aspect_ratio{static_cast<float>(ScreenUndocked::Height) /
-                                       ScreenUndocked::Width};
-    const auto window_aspect_ratio = static_cast<float>(height) / width;
+    const float window_aspect_ratio = static_cast<float>(height) / width;
+    const float emulation_aspect_ratio = EmulationAspectRatio(
+        static_cast<AspectRatio>(Settings::values.aspect_ratio), window_aspect_ratio);
 
     const Common::Rectangle<u32> screen_window_area{0, 0, width, height};
     Common::Rectangle<u32> screen = MaxRectangle(screen_window_area, emulation_aspect_ratio);
@@ -58,4 +58,19 @@ FramebufferLayout FrameLayoutFromResolutionScale(u32 res_scale) {
     return DefaultFrameLayout(width, height);
 }
 
+float EmulationAspectRatio(AspectRatio aspect, float window_aspect_ratio) {
+    switch (aspect) {
+    case AspectRatio::Default:
+        return static_cast<float>(ScreenUndocked::Height) / ScreenUndocked::Width;
+    case AspectRatio::R4_3:
+        return 3.0f / 4.0f;
+    case AspectRatio::R21_9:
+        return 9.0f / 21.0f;
+    case AspectRatio::StretchToWindow:
+        return window_aspect_ratio;
+    default:
+        return static_cast<float>(ScreenUndocked::Height) / ScreenUndocked::Width;
+    }
+}
+
 } // namespace Layout
diff --git a/src/core/frontend/framebuffer_layout.h b/src/core/frontend/framebuffer_layout.h
index d2370adde..1d39c1faf 100644
--- a/src/core/frontend/framebuffer_layout.h
+++ b/src/core/frontend/framebuffer_layout.h
@@ -18,6 +18,13 @@ enum ScreenDocked : u32 {
     HeightDocked = 1080,
 };
 
+enum class AspectRatio {
+    Default,
+    R4_3,
+    R21_9,
+    StretchToWindow,
+};
+
 /// Describes the layout of the window framebuffer
 struct FramebufferLayout {
     u32 width{ScreenUndocked::Width};
@@ -48,4 +55,12 @@ FramebufferLayout DefaultFrameLayout(u32 width, u32 height);
  */
 FramebufferLayout FrameLayoutFromResolutionScale(u32 res_scale);
 
+/**
+ * Convenience method to determine emulation aspect ratio
+ * @param aspect Represents the index of aspect ratio stored in Settings::values.aspect_ratio
+ * @param window_aspect_ratio Current window aspect ratio
+ * @return Emulation render window aspect ratio
+ */
+float EmulationAspectRatio(AspectRatio aspect, float window_aspect_ratio);
+
 } // namespace Layout
diff --git a/src/core/hardware_properties.h b/src/core/hardware_properties.h
index 213461b6a..b04e046ed 100644
--- a/src/core/hardware_properties.h
+++ b/src/core/hardware_properties.h
@@ -20,6 +20,8 @@ constexpr u32 NUM_CPU_CORES = 4;            // Number of CPU Cores
 
 } // namespace Hardware
 
+constexpr u32 INVALID_HOST_THREAD_ID = 0xFFFFFFFF;
+
 struct EmuThreadHandle {
     u32 host_handle;
     u32 guest_handle;
diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp
index 4eb1d8703..9232f4d7e 100644
--- a/src/core/hle/kernel/kernel.cpp
+++ b/src/core/hle/kernel/kernel.cpp
@@ -3,9 +3,12 @@
 // Refer to the license.txt file included.
 
 #include <atomic>
+#include <bitset>
 #include <functional>
 #include <memory>
 #include <mutex>
+#include <thread>
+#include <unordered_map>
 #include <utility>
 
 #include "common/assert.h"
@@ -15,6 +18,7 @@
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
+#include "core/hardware_properties.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
@@ -25,6 +29,7 @@
 #include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/synchronization.h"
 #include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/time_manager.h"
 #include "core/hle/lock.h"
 #include "core/hle/result.h"
 #include "core/memory.h"
@@ -44,7 +49,7 @@ static void ThreadWakeupCallback(u64 thread_handle, [[maybe_unused]] s64 cycles_
     std::lock_guard lock{HLE::g_hle_lock};
 
     std::shared_ptr<Thread> thread =
-        system.Kernel().RetrieveThreadFromWakeupCallbackHandleTable(proper_handle);
+        system.Kernel().RetrieveThreadFromGlobalHandleTable(proper_handle);
     if (thread == nullptr) {
         LOG_CRITICAL(Kernel, "Callback fired for invalid thread {:08X}", proper_handle);
         return;
@@ -97,8 +102,8 @@ static void ThreadWakeupCallback(u64 thread_handle, [[maybe_unused]] s64 cycles_
 }
 
 struct KernelCore::Impl {
-    explicit Impl(Core::System& system)
-        : system{system}, global_scheduler{system}, synchronization{system} {}
+    explicit Impl(Core::System& system, KernelCore& kernel)
+        : system{system}, global_scheduler{kernel}, synchronization{system}, time_manager{system} {}
 
     void Initialize(KernelCore& kernel) {
         Shutdown();
@@ -120,7 +125,7 @@ struct KernelCore::Impl {
 
         system_resource_limit = nullptr;
 
-        thread_wakeup_callback_handle_table.Clear();
+        global_handle_table.Clear();
         thread_wakeup_event_type = nullptr;
         preemption_event = nullptr;
 
@@ -138,8 +143,8 @@ struct KernelCore::Impl {
 
     void InitializePhysicalCores() {
         exclusive_monitor =
-            Core::MakeExclusiveMonitor(system.Memory(), global_scheduler.CpuCoresCount());
-        for (std::size_t i = 0; i < global_scheduler.CpuCoresCount(); i++) {
+            Core::MakeExclusiveMonitor(system.Memory(), Core::Hardware::NUM_CPU_CORES);
+        for (std::size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
             cores.emplace_back(system, i, *exclusive_monitor);
         }
     }
@@ -184,6 +189,50 @@ struct KernelCore::Impl {
         system.Memory().SetCurrentPageTable(*process);
     }
 
+    void RegisterCoreThread(std::size_t core_id) {
+        std::unique_lock lock{register_thread_mutex};
+        const std::thread::id this_id = std::this_thread::get_id();
+        const auto it = host_thread_ids.find(this_id);
+        ASSERT(core_id < Core::Hardware::NUM_CPU_CORES);
+        ASSERT(it == host_thread_ids.end());
+        ASSERT(!registered_core_threads[core_id]);
+        host_thread_ids[this_id] = static_cast<u32>(core_id);
+        registered_core_threads.set(core_id);
+    }
+
+    void RegisterHostThread() {
+        std::unique_lock lock{register_thread_mutex};
+        const std::thread::id this_id = std::this_thread::get_id();
+        const auto it = host_thread_ids.find(this_id);
+        ASSERT(it == host_thread_ids.end());
+        host_thread_ids[this_id] = registered_thread_ids++;
+    }
+
+    u32 GetCurrentHostThreadID() const {
+        const std::thread::id this_id = std::this_thread::get_id();
+        const auto it = host_thread_ids.find(this_id);
+        if (it == host_thread_ids.end()) {
+            return Core::INVALID_HOST_THREAD_ID;
+        }
+        return it->second;
+    }
+
+    Core::EmuThreadHandle GetCurrentEmuThreadID() const {
+        Core::EmuThreadHandle result = Core::EmuThreadHandle::InvalidHandle();
+        result.host_handle = GetCurrentHostThreadID();
+        if (result.host_handle >= Core::Hardware::NUM_CPU_CORES) {
+            return result;
+        }
+        const Kernel::Scheduler& sched = cores[result.host_handle].Scheduler();
+        const Kernel::Thread* current = sched.GetCurrentThread();
+        if (current != nullptr) {
+            result.guest_handle = current->GetGlobalHandle();
+        } else {
+            result.guest_handle = InvalidHandle;
+        }
+        return result;
+    }
+
     std::atomic<u32> next_object_id{0};
     std::atomic<u64> next_kernel_process_id{Process::InitialKIPIDMin};
     std::atomic<u64> next_user_process_id{Process::ProcessIDMin};
@@ -194,15 +243,16 @@ struct KernelCore::Impl {
     Process* current_process = nullptr;
     Kernel::GlobalScheduler global_scheduler;
     Kernel::Synchronization synchronization;
+    Kernel::TimeManager time_manager;
 
     std::shared_ptr<ResourceLimit> system_resource_limit;
 
     std::shared_ptr<Core::Timing::EventType> thread_wakeup_event_type;
     std::shared_ptr<Core::Timing::EventType> preemption_event;
 
-    // TODO(yuriks): This can be removed if Thread objects are explicitly pooled in the future,
-    // allowing us to simply use a pool index or similar.
-    Kernel::HandleTable thread_wakeup_callback_handle_table;
+    // This is the kernel's handle table or supervisor handle table which
+    // stores all the objects in place.
+    Kernel::HandleTable global_handle_table;
 
     /// Map of named ports managed by the kernel, which can be retrieved using
     /// the ConnectToPort SVC.
@@ -211,11 +261,17 @@ struct KernelCore::Impl {
     std::unique_ptr<Core::ExclusiveMonitor> exclusive_monitor;
     std::vector<Kernel::PhysicalCore> cores;
 
+    // 0-3 IDs represent core threads, >3 represent others
+    std::unordered_map<std::thread::id, u32> host_thread_ids;
+    u32 registered_thread_ids{Core::Hardware::NUM_CPU_CORES};
+    std::bitset<Core::Hardware::NUM_CPU_CORES> registered_core_threads;
+    std::mutex register_thread_mutex;
+
     // System context
     Core::System& system;
 };
 
-KernelCore::KernelCore(Core::System& system) : impl{std::make_unique<Impl>(system)} {}
+KernelCore::KernelCore(Core::System& system) : impl{std::make_unique<Impl>(system, *this)} {}
 KernelCore::~KernelCore() {
     Shutdown();
 }
@@ -232,9 +288,8 @@ std::shared_ptr<ResourceLimit> KernelCore::GetSystemResourceLimit() const {
     return impl->system_resource_limit;
 }
 
-std::shared_ptr<Thread> KernelCore::RetrieveThreadFromWakeupCallbackHandleTable(
-    Handle handle) const {
-    return impl->thread_wakeup_callback_handle_table.Get<Thread>(handle);
+std::shared_ptr<Thread> KernelCore::RetrieveThreadFromGlobalHandleTable(Handle handle) const {
+    return impl->global_handle_table.Get<Thread>(handle);
 }
 
 void KernelCore::AppendNewProcess(std::shared_ptr<Process> process) {
@@ -265,6 +320,14 @@ const Kernel::GlobalScheduler& KernelCore::GlobalScheduler() const {
     return impl->global_scheduler;
 }
 
+Kernel::Scheduler& KernelCore::Scheduler(std::size_t id) {
+    return impl->cores[id].Scheduler();
+}
+
+const Kernel::Scheduler& KernelCore::Scheduler(std::size_t id) const {
+    return impl->cores[id].Scheduler();
+}
+
 Kernel::PhysicalCore& KernelCore::PhysicalCore(std::size_t id) {
     return impl->cores[id];
 }
@@ -281,6 +344,14 @@ const Kernel::Synchronization& KernelCore::Synchronization() const {
     return impl->synchronization;
 }
 
+Kernel::TimeManager& KernelCore::TimeManager() {
+    return impl->time_manager;
+}
+
+const Kernel::TimeManager& KernelCore::TimeManager() const {
+    return impl->time_manager;
+}
+
 Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() {
     return *impl->exclusive_monitor;
 }
@@ -338,12 +409,28 @@ const std::shared_ptr<Core::Timing::EventType>& KernelCore::ThreadWakeupCallback
     return impl->thread_wakeup_event_type;
 }
 
-Kernel::HandleTable& KernelCore::ThreadWakeupCallbackHandleTable() {
-    return impl->thread_wakeup_callback_handle_table;
+Kernel::HandleTable& KernelCore::GlobalHandleTable() {
+    return impl->global_handle_table;
+}
+
+const Kernel::HandleTable& KernelCore::GlobalHandleTable() const {
+    return impl->global_handle_table;
+}
+
+void KernelCore::RegisterCoreThread(std::size_t core_id) {
+    impl->RegisterCoreThread(core_id);
+}
+
+void KernelCore::RegisterHostThread() {
+    impl->RegisterHostThread();
+}
+
+u32 KernelCore::GetCurrentHostThreadID() const {
+    return impl->GetCurrentHostThreadID();
 }
 
-const Kernel::HandleTable& KernelCore::ThreadWakeupCallbackHandleTable() const {
-    return impl->thread_wakeup_callback_handle_table;
+Core::EmuThreadHandle KernelCore::GetCurrentEmuThreadID() const {
+    return impl->GetCurrentEmuThreadID();
 }
 
 } // namespace Kernel
diff --git a/src/core/hle/kernel/kernel.h b/src/core/hle/kernel/kernel.h
index 1eede3063..c4f78ab71 100644
--- a/src/core/hle/kernel/kernel.h
+++ b/src/core/hle/kernel/kernel.h
@@ -11,6 +11,7 @@
 #include "core/hle/kernel/object.h"
 
 namespace Core {
+struct EmuThreadHandle;
 class ExclusiveMonitor;
 class System;
 } // namespace Core
@@ -29,8 +30,10 @@ class HandleTable;
 class PhysicalCore;
 class Process;
 class ResourceLimit;
+class Scheduler;
 class Synchronization;
 class Thread;
+class TimeManager;
 
 /// Represents a single instance of the kernel.
 class KernelCore {
@@ -64,7 +67,7 @@ public:
     std::shared_ptr<ResourceLimit> GetSystemResourceLimit() const;
 
     /// Retrieves a shared pointer to a Thread instance within the thread wakeup handle table.
-    std::shared_ptr<Thread> RetrieveThreadFromWakeupCallbackHandleTable(Handle handle) const;
+    std::shared_ptr<Thread> RetrieveThreadFromGlobalHandleTable(Handle handle) const;
 
     /// Adds the given shared pointer to an internal list of active processes.
     void AppendNewProcess(std::shared_ptr<Process> process);
@@ -87,6 +90,12 @@ public:
     /// Gets the sole instance of the global scheduler
     const Kernel::GlobalScheduler& GlobalScheduler() const;
 
+    /// Gets the sole instance of the Scheduler assoviated with cpu core 'id'
+    Kernel::Scheduler& Scheduler(std::size_t id);
+
+    /// Gets the sole instance of the Scheduler assoviated with cpu core 'id'
+    const Kernel::Scheduler& Scheduler(std::size_t id) const;
+
     /// Gets the an instance of the respective physical CPU core.
     Kernel::PhysicalCore& PhysicalCore(std::size_t id);
 
@@ -99,6 +108,12 @@ public:
     /// Gets the an instance of the Synchronization Interface.
     const Kernel::Synchronization& Synchronization() const;
 
+    /// Gets the an instance of the TimeManager Interface.
+    Kernel::TimeManager& TimeManager();
+
+    /// Gets the an instance of the TimeManager Interface.
+    const Kernel::TimeManager& TimeManager() const;
+
     /// Stops execution of 'id' core, in order to reschedule a new thread.
     void PrepareReschedule(std::size_t id);
 
@@ -120,6 +135,18 @@ public:
     /// Determines whether or not the given port is a valid named port.
     bool IsValidNamedPort(NamedPortTable::const_iterator port) const;
 
+    /// Gets the current host_thread/guest_thread handle.
+    Core::EmuThreadHandle GetCurrentEmuThreadID() const;
+
+    /// Gets the current host_thread handle.
+    u32 GetCurrentHostThreadID() const;
+
+    /// Register the current thread as a CPU Core Thread.
+    void RegisterCoreThread(std::size_t core_id);
+
+    /// Register the current thread as a non CPU core thread.
+    void RegisterHostThread();
+
 private:
     friend class Object;
     friend class Process;
@@ -140,11 +167,11 @@ private:
     /// Retrieves the event type used for thread wakeup callbacks.
     const std::shared_ptr<Core::Timing::EventType>& ThreadWakeupCallbackEventType() const;
 
-    /// Provides a reference to the thread wakeup callback handle table.
-    Kernel::HandleTable& ThreadWakeupCallbackHandleTable();
+    /// Provides a reference to the global handle table.
+    Kernel::HandleTable& GlobalHandleTable();
 
-    /// Provides a const reference to the thread wakeup callback handle table.
-    const Kernel::HandleTable& ThreadWakeupCallbackHandleTable() const;
+    /// Provides a const reference to the global handle table.
+    const Kernel::HandleTable& GlobalHandleTable() const;
 
     struct Impl;
     std::unique_ptr<Impl> impl;
diff --git a/src/core/hle/kernel/scheduler.cpp b/src/core/hle/kernel/scheduler.cpp
index 86f1421bf..c65f82fb7 100644
--- a/src/core/hle/kernel/scheduler.cpp
+++ b/src/core/hle/kernel/scheduler.cpp
@@ -18,10 +18,11 @@
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/scheduler.h"
+#include "core/hle/kernel/time_manager.h"
 
 namespace Kernel {
 
-GlobalScheduler::GlobalScheduler(Core::System& system) : system{system} {}
+GlobalScheduler::GlobalScheduler(KernelCore& kernel) : kernel{kernel} {}
 
 GlobalScheduler::~GlobalScheduler() = default;
 
@@ -35,7 +36,7 @@ void GlobalScheduler::RemoveThread(std::shared_ptr<Thread> thread) {
 }
 
 void GlobalScheduler::UnloadThread(std::size_t core) {
-    Scheduler& sched = system.Scheduler(core);
+    Scheduler& sched = kernel.Scheduler(core);
     sched.UnloadThread();
 }
 
@@ -50,7 +51,7 @@ void GlobalScheduler::SelectThread(std::size_t core) {
         sched.is_context_switch_pending = sched.selected_thread != sched.current_thread;
         std::atomic_thread_fence(std::memory_order_seq_cst);
     };
-    Scheduler& sched = system.Scheduler(core);
+    Scheduler& sched = kernel.Scheduler(core);
     Thread* current_thread = nullptr;
     // Step 1: Get top thread in schedule queue.
     current_thread = scheduled_queue[core].empty() ? nullptr : scheduled_queue[core].front();
@@ -356,6 +357,32 @@ void GlobalScheduler::Shutdown() {
     thread_list.clear();
 }
 
+void GlobalScheduler::Lock() {
+    Core::EmuThreadHandle current_thread = kernel.GetCurrentEmuThreadID();
+    if (current_thread == current_owner) {
+        ++scope_lock;
+    } else {
+        inner_lock.lock();
+        current_owner = current_thread;
+        ASSERT(current_owner != Core::EmuThreadHandle::InvalidHandle());
+        scope_lock = 1;
+    }
+}
+
+void GlobalScheduler::Unlock() {
+    if (--scope_lock != 0) {
+        ASSERT(scope_lock > 0);
+        return;
+    }
+    for (std::size_t i = 0; i < Core::Hardware::NUM_CPU_CORES; i++) {
+        SelectThread(i);
+    }
+    current_owner = Core::EmuThreadHandle::InvalidHandle();
+    scope_lock = 1;
+    inner_lock.unlock();
+    // TODO(Blinkhawk): Setup the interrupts and change context on current core.
+}
+
 Scheduler::Scheduler(Core::System& system, Core::ARM_Interface& cpu_core, std::size_t core_id)
     : system(system), cpu_core(cpu_core), core_id(core_id) {}
 
@@ -485,4 +512,27 @@ void Scheduler::Shutdown() {
     selected_thread = nullptr;
 }
 
+SchedulerLock::SchedulerLock(KernelCore& kernel) : kernel{kernel} {
+    kernel.GlobalScheduler().Lock();
+}
+
+SchedulerLock::~SchedulerLock() {
+    kernel.GlobalScheduler().Unlock();
+}
+
+SchedulerLockAndSleep::SchedulerLockAndSleep(KernelCore& kernel, Handle& event_handle,
+                                             Thread* time_task, s64 nanoseconds)
+    : SchedulerLock{kernel}, event_handle{event_handle}, time_task{time_task}, nanoseconds{
+                                                                                   nanoseconds} {
+    event_handle = InvalidHandle;
+}
+
+SchedulerLockAndSleep::~SchedulerLockAndSleep() {
+    if (sleep_cancelled) {
+        return;
+    }
+    auto& time_manager = kernel.TimeManager();
+    time_manager.ScheduleTimeEvent(event_handle, time_task, nanoseconds);
+}
+
 } // namespace Kernel
diff --git a/src/core/hle/kernel/scheduler.h b/src/core/hle/kernel/scheduler.h
index 96db049cb..1c93a838c 100644
--- a/src/core/hle/kernel/scheduler.h
+++ b/src/core/hle/kernel/scheduler.h
@@ -6,6 +6,7 @@
 
 #include <atomic>
 #include <memory>
+#include <mutex>
 #include <vector>
 
 #include "common/common_types.h"
@@ -20,11 +21,13 @@ class System;
 
 namespace Kernel {
 
+class KernelCore;
 class Process;
+class SchedulerLock;
 
 class GlobalScheduler final {
 public:
-    explicit GlobalScheduler(Core::System& system);
+    explicit GlobalScheduler(KernelCore& kernel);
     ~GlobalScheduler();
 
     /// Adds a new thread to the scheduler
@@ -138,6 +141,14 @@ public:
     void Shutdown();
 
 private:
+    friend class SchedulerLock;
+
+    /// Lock the scheduler to the current thread.
+    void Lock();
+
+    /// Unlocks the scheduler, reselects threads, interrupts cores for rescheduling
+    /// and reschedules current core if needed.
+    void Unlock();
     /**
      * Transfers a thread into an specific core. If the destination_core is -1
      * it will be unscheduled from its source code and added into its suggested
@@ -158,9 +169,14 @@ private:
     // ordered from Core 0 to Core 3.
     std::array<u32, Core::Hardware::NUM_CPU_CORES> preemption_priorities = {59, 59, 59, 62};
 
+    /// Scheduler lock mechanisms.
+    std::mutex inner_lock{}; // TODO(Blinkhawk): Replace for a SpinLock
+    std::atomic<s64> scope_lock{};
+    Core::EmuThreadHandle current_owner{Core::EmuThreadHandle::InvalidHandle()};
+
     /// Lists all thread ids that aren't deleted/etc.
     std::vector<std::shared_ptr<Thread>> thread_list;
-    Core::System& system;
+    KernelCore& kernel;
 };
 
 class Scheduler final {
@@ -227,4 +243,30 @@ private:
     bool is_context_switch_pending = false;
 };
 
+class SchedulerLock {
+public:
+    explicit SchedulerLock(KernelCore& kernel);
+    ~SchedulerLock();
+
+protected:
+    KernelCore& kernel;
+};
+
+class SchedulerLockAndSleep : public SchedulerLock {
+public:
+    explicit SchedulerLockAndSleep(KernelCore& kernel, Handle& event_handle, Thread* time_task,
+                                   s64 nanoseconds);
+    ~SchedulerLockAndSleep();
+
+    void CancelSleep() {
+        sleep_cancelled = true;
+    }
+
+private:
+    Handle& event_handle;
+    Thread* time_task;
+    s64 nanoseconds;
+    bool sleep_cancelled{};
+};
+
 } // namespace Kernel
diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp
index ae5f2c8bd..bf850e0b2 100644
--- a/src/core/hle/kernel/thread.cpp
+++ b/src/core/hle/kernel/thread.cpp
@@ -46,9 +46,9 @@ Thread::~Thread() = default;
 void Thread::Stop() {
     // Cancel any outstanding wakeup events for this thread
     Core::System::GetInstance().CoreTiming().UnscheduleEvent(kernel.ThreadWakeupCallbackEventType(),
-                                                             callback_handle);
-    kernel.ThreadWakeupCallbackHandleTable().Close(callback_handle);
-    callback_handle = 0;
+                                                             global_handle);
+    kernel.GlobalHandleTable().Close(global_handle);
+    global_handle = 0;
     SetStatus(ThreadStatus::Dead);
     Signal();
 
@@ -73,12 +73,12 @@ void Thread::WakeAfterDelay(s64 nanoseconds) {
     // thread-safe version of ScheduleEvent.
     const s64 cycles = Core::Timing::nsToCycles(std::chrono::nanoseconds{nanoseconds});
     Core::System::GetInstance().CoreTiming().ScheduleEvent(
-        cycles, kernel.ThreadWakeupCallbackEventType(), callback_handle);
+        cycles, kernel.ThreadWakeupCallbackEventType(), global_handle);
 }
 
 void Thread::CancelWakeupTimer() {
     Core::System::GetInstance().CoreTiming().UnscheduleEvent(kernel.ThreadWakeupCallbackEventType(),
-                                                             callback_handle);
+                                                             global_handle);
 }
 
 void Thread::ResumeFromWait() {
@@ -190,7 +190,7 @@ ResultVal<std::shared_ptr<Thread>> Thread::Create(KernelCore& kernel, std::strin
     thread->condvar_wait_address = 0;
     thread->wait_handle = 0;
     thread->name = std::move(name);
-    thread->callback_handle = kernel.ThreadWakeupCallbackHandleTable().Create(thread).Unwrap();
+    thread->global_handle = kernel.GlobalHandleTable().Create(thread).Unwrap();
     thread->owner_process = &owner_process;
     auto& scheduler = kernel.GlobalScheduler();
     scheduler.AddThread(thread);
diff --git a/src/core/hle/kernel/thread.h b/src/core/hle/kernel/thread.h
index 7a4916318..129e7858a 100644
--- a/src/core/hle/kernel/thread.h
+++ b/src/core/hle/kernel/thread.h
@@ -453,6 +453,10 @@ public:
         is_sync_cancelled = value;
     }
 
+    Handle GetGlobalHandle() const {
+        return global_handle;
+    }
+
 private:
     void SetSchedulingStatus(ThreadSchedStatus new_status);
     void SetCurrentPriority(u32 new_priority);
@@ -514,7 +518,7 @@ private:
     VAddr arb_wait_address{0};
 
     /// Handle used as userdata to reference this object when inserting into the CoreTiming queue.
-    Handle callback_handle = 0;
+    Handle global_handle = 0;
 
     /// Callback that will be invoked when the thread is resumed from a waiting state. If the thread
     /// was waiting via WaitSynchronization then the object will be the last object that became
diff --git a/src/core/hle/kernel/time_manager.cpp b/src/core/hle/kernel/time_manager.cpp
new file mode 100644
index 000000000..21b290468
--- /dev/null
+++ b/src/core/hle/kernel/time_manager.cpp
@@ -0,0 +1,44 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "core/core.h"
+#include "core/core_timing.h"
+#include "core/core_timing_util.h"
+#include "core/hle/kernel/handle_table.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/kernel/time_manager.h"
+
+namespace Kernel {
+
+TimeManager::TimeManager(Core::System& system) : system{system} {
+    time_manager_event_type = Core::Timing::CreateEvent(
+        "Kernel::TimeManagerCallback", [this](u64 thread_handle, [[maybe_unused]] s64 cycles_late) {
+            Handle proper_handle = static_cast<Handle>(thread_handle);
+            std::shared_ptr<Thread> thread =
+                this->system.Kernel().RetrieveThreadFromGlobalHandleTable(proper_handle);
+            thread->ResumeFromWait();
+        });
+}
+
+void TimeManager::ScheduleTimeEvent(Handle& event_handle, Thread* timetask, s64 nanoseconds) {
+    if (nanoseconds > 0) {
+        ASSERT(timetask);
+        event_handle = timetask->GetGlobalHandle();
+        const s64 cycles = Core::Timing::nsToCycles(std::chrono::nanoseconds{nanoseconds});
+        system.CoreTiming().ScheduleEvent(cycles, time_manager_event_type, event_handle);
+    } else {
+        event_handle = InvalidHandle;
+    }
+}
+
+void TimeManager::UnscheduleTimeEvent(Handle event_handle) {
+    if (event_handle == InvalidHandle) {
+        return;
+    }
+    system.CoreTiming().UnscheduleEvent(time_manager_event_type, event_handle);
+}
+
+} // namespace Kernel
diff --git a/src/core/hle/kernel/time_manager.h b/src/core/hle/kernel/time_manager.h
new file mode 100644
index 000000000..eaec486d1
--- /dev/null
+++ b/src/core/hle/kernel/time_manager.h
@@ -0,0 +1,43 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+
+#include "core/hle/kernel/object.h"
+
+namespace Core {
+class System;
+} // namespace Core
+
+namespace Core::Timing {
+struct EventType;
+} // namespace Core::Timing
+
+namespace Kernel {
+
+class Thread;
+
+/**
+ * The `TimeManager` takes care of scheduling time events on threads and executes their TimeUp
+ * method when the event is triggered.
+ */
+class TimeManager {
+public:
+    explicit TimeManager(Core::System& system);
+
+    /// Schedule a time event on `timetask` thread that will expire in 'nanoseconds'
+    /// returns a non-invalid handle in `event_handle` if correctly scheduled
+    void ScheduleTimeEvent(Handle& event_handle, Thread* timetask, s64 nanoseconds);
+
+    /// Unschedule an existing time event
+    void UnscheduleTimeEvent(Handle event_handle);
+
+private:
+    Core::System& system;
+    std::shared_ptr<Core::Timing::EventType> time_manager_event_type;
+};
+
+} // namespace Kernel
diff --git a/src/core/hle/service/bcat/backend/boxcat.cpp b/src/core/hle/service/bcat/backend/boxcat.cpp
index 67e39a5c4..f589864ee 100644
--- a/src/core/hle/service/bcat/backend/boxcat.cpp
+++ b/src/core/hle/service/bcat/backend/boxcat.cpp
@@ -200,7 +200,8 @@ private:
     DownloadResult DownloadInternal(const std::string& resolved_path, u32 timeout_seconds,
                                     const std::string& content_type_name) {
         if (client == nullptr) {
-            client = std::make_unique<httplib::SSLClient>(BOXCAT_HOSTNAME, PORT, timeout_seconds);
+            client = std::make_unique<httplib::SSLClient>(BOXCAT_HOSTNAME, PORT);
+            client->set_timeout_sec(timeout_seconds);
         }
 
         httplib::Headers headers{
@@ -448,8 +449,8 @@ std::optional<std::vector<u8>> Boxcat::GetLaunchParameter(TitleIDVersion title)
 
 Boxcat::StatusResult Boxcat::GetStatus(std::optional<std::string>& global,
                                        std::map<std::string, EventStatus>& games) {
-    httplib::SSLClient client{BOXCAT_HOSTNAME, static_cast<int>(PORT),
-                              static_cast<int>(TIMEOUT_SECONDS)};
+    httplib::SSLClient client{BOXCAT_HOSTNAME, static_cast<int>(PORT)};
+    client.set_timeout_sec(static_cast<int>(TIMEOUT_SECONDS));
 
     httplib::Headers headers{
         {std::string("Game-Assets-API-Version"), std::string(BOXCAT_API_VERSION)},
diff --git a/src/core/hle/service/ldn/ldn.cpp b/src/core/hle/service/ldn/ldn.cpp
index ed5059047..92adde6d4 100644
--- a/src/core/hle/service/ldn/ldn.cpp
+++ b/src/core/hle/service/ldn/ldn.cpp
@@ -129,12 +129,20 @@ public:
             {304, nullptr, "Disconnect"},
             {400, nullptr, "Initialize"},
             {401, nullptr, "Finalize"},
-            {402, nullptr, "SetOperationMode"},
+            {402, &IUserLocalCommunicationService::Initialize2, "Initialize2"}, // 7.0.0+
         };
         // clang-format on
 
         RegisterHandlers(functions);
     }
+
+    void Initialize2(Kernel::HLERequestContext& ctx) {
+        LOG_WARNING(Service_LDN, "(STUBBED) called");
+        // Result success seem make this services start network and continue.
+        // If we just pass result error then it will stop and maybe try again and again.
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_UNKNOWN);
+    }
 };
 
 class LDNS final : public ServiceFramework<LDNS> {
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
index 6d8bca8bb..f1966ac0e 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.cpp
@@ -44,6 +44,8 @@ u32 nvhost_gpu::ioctl(Ioctl command, const std::vector<u8>& input, const std::ve
         return GetWaitbase(input, output);
     case IoctlCommand::IocChannelSetTimeoutCommand:
         return ChannelSetTimeout(input, output);
+    case IoctlCommand::IocChannelSetTimeslice:
+        return ChannelSetTimeslice(input, output);
     default:
         break;
     }
@@ -228,4 +230,14 @@ u32 nvhost_gpu::ChannelSetTimeout(const std::vector<u8>& input, std::vector<u8>&
     return 0;
 }
 
+u32 nvhost_gpu::ChannelSetTimeslice(const std::vector<u8>& input, std::vector<u8>& output) {
+    IoctlSetTimeslice params{};
+    std::memcpy(&params, input.data(), sizeof(IoctlSetTimeslice));
+    LOG_INFO(Service_NVDRV, "called, timeslice=0x{:X}", params.timeslice);
+
+    channel_timeslice = params.timeslice;
+
+    return 0;
+}
+
 } // namespace Service::Nvidia::Devices
diff --git a/src/core/hle/service/nvdrv/devices/nvhost_gpu.h b/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
index d056dd046..2ac74743f 100644
--- a/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
+++ b/src/core/hle/service/nvdrv/devices/nvhost_gpu.h
@@ -48,6 +48,7 @@ private:
         IocAllocObjCtxCommand = 0xC0104809,
         IocChannelGetWaitbaseCommand = 0xC0080003,
         IocChannelSetTimeoutCommand = 0x40044803,
+        IocChannelSetTimeslice = 0xC004481D,
     };
 
     enum class CtxObjects : u32_le {
@@ -101,6 +102,11 @@ private:
     static_assert(sizeof(IoctlChannelSetPriority) == 4,
                   "IoctlChannelSetPriority is incorrect size");
 
+    struct IoctlSetTimeslice {
+        u32_le timeslice;
+    };
+    static_assert(sizeof(IoctlSetTimeslice) == 4, "IoctlSetTimeslice is incorrect size");
+
     struct IoctlEventIdControl {
         u32_le cmd; // 0=disable, 1=enable, 2=clear
         u32_le id;
@@ -174,6 +180,7 @@ private:
     u64_le user_data{};
     IoctlZCullBind zcull_params{};
     u32_le channel_priority{};
+    u32_le channel_timeslice{};
 
     u32 SetNVMAPfd(const std::vector<u8>& input, std::vector<u8>& output);
     u32 SetClientData(const std::vector<u8>& input, std::vector<u8>& output);
@@ -188,6 +195,7 @@ private:
                   const std::vector<u8>& input2, IoctlVersion version);
     u32 GetWaitbase(const std::vector<u8>& input, std::vector<u8>& output);
     u32 ChannelSetTimeout(const std::vector<u8>& input, std::vector<u8>& output);
+    u32 ChannelSetTimeslice(const std::vector<u8>& input, std::vector<u8>& output);
 
     std::shared_ptr<nvmap> nvmap_dev;
     u32 assigned_syncpoints{};
diff --git a/src/core/settings.h b/src/core/settings.h
index e1a9a0ffa..f837d3fbc 100644
--- a/src/core/settings.h
+++ b/src/core/settings.h
@@ -429,6 +429,7 @@ struct Values {
     int vulkan_device;
 
     float resolution_factor;
+    int aspect_ratio;
     bool use_frame_limit;
     u16 frame_limit;
     bool use_disk_shader_cache;
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index db9332d00..4b0c6346f 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -37,6 +37,7 @@ add_library(video_core STATIC
     memory_manager.h
     morton.cpp
     morton.h
+    query_cache.h
     rasterizer_accelerated.cpp
     rasterizer_accelerated.h
     rasterizer_cache.cpp
@@ -74,6 +75,8 @@ add_library(video_core STATIC
     renderer_opengl/gl_stream_buffer.h
     renderer_opengl/gl_texture_cache.cpp
     renderer_opengl/gl_texture_cache.h
+    renderer_opengl/gl_query_cache.cpp
+    renderer_opengl/gl_query_cache.h
     renderer_opengl/maxwell_to_gl.h
     renderer_opengl/renderer_opengl.cpp
     renderer_opengl/renderer_opengl.h
@@ -177,6 +180,8 @@ if (ENABLE_VULKAN)
         renderer_vulkan/vk_memory_manager.h
         renderer_vulkan/vk_pipeline_cache.cpp
         renderer_vulkan/vk_pipeline_cache.h
+        renderer_vulkan/vk_query_cache.cpp
+        renderer_vulkan/vk_query_cache.h
         renderer_vulkan/vk_rasterizer.cpp
         renderer_vulkan/vk_rasterizer.h
         renderer_vulkan/vk_renderpass_cache.cpp
diff --git a/src/video_core/engines/maxwell_3d.cpp b/src/video_core/engines/maxwell_3d.cpp
index 0b3e8749b..b28de1092 100644
--- a/src/video_core/engines/maxwell_3d.cpp
+++ b/src/video_core/engines/maxwell_3d.cpp
@@ -4,6 +4,7 @@
 
 #include <cinttypes>
 #include <cstring>
+#include <optional>
 #include "common/assert.h"
 #include "core/core.h"
 #include "core/core_timing.h"
@@ -16,6 +17,8 @@
 
 namespace Tegra::Engines {
 
+using VideoCore::QueryType;
+
 /// First register id that is actually a Macro call.
 constexpr u32 MacroRegistersStart = 0xE00;
 
@@ -400,6 +403,10 @@ void Maxwell3D::CallMethod(const GPU::MethodCall& method_call) {
         ProcessQueryCondition();
         break;
     }
+    case MAXWELL3D_REG_INDEX(counter_reset): {
+        ProcessCounterReset();
+        break;
+    }
     case MAXWELL3D_REG_INDEX(sync_info): {
         ProcessSyncPoint();
         break;
@@ -482,7 +489,7 @@ void Maxwell3D::FlushMMEInlineDraw() {
 
     const bool is_indexed = mme_draw.current_mode == MMEDrawMode::Indexed;
     if (ShouldExecute()) {
-        rasterizer.DrawMultiBatch(is_indexed);
+        rasterizer.Draw(is_indexed, true);
     }
 
     // TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if
@@ -544,40 +551,28 @@ void Maxwell3D::ProcessQueryGet() {
                "Units other than CROP are unimplemented");
 
     switch (regs.query.query_get.operation) {
-    case Regs::QueryOperation::Release: {
-        const u64 result = regs.query.query_sequence;
-        StampQueryResult(result, regs.query.query_get.short_query == 0);
+    case Regs::QueryOperation::Release:
+        StampQueryResult(regs.query.query_sequence, regs.query.query_get.short_query == 0);
         break;
-    }
-    case Regs::QueryOperation::Acquire: {
-        // Todo(Blinkhawk): Under this operation, the GPU waits for the CPU
-        // to write a value that matches the current payload.
+    case Regs::QueryOperation::Acquire:
+        // TODO(Blinkhawk): Under this operation, the GPU waits for the CPU to write a value that
+        // matches the current payload.
         UNIMPLEMENTED_MSG("Unimplemented query operation ACQUIRE");
         break;
-    }
-    case Regs::QueryOperation::Counter: {
-        u64 result{};
-        switch (regs.query.query_get.select) {
-        case Regs::QuerySelect::Zero:
-            result = 0;
-            break;
-        default:
-            result = 1;
-            UNIMPLEMENTED_MSG("Unimplemented query select type {}",
-                              static_cast<u32>(regs.query.query_get.select.Value()));
+    case Regs::QueryOperation::Counter:
+        if (const std::optional<u64> result = GetQueryResult()) {
+            // If the query returns an empty optional it means it's cached and deferred.
+            // In this case we have a non-empty result, so we stamp it immediately.
+            StampQueryResult(*result, regs.query.query_get.short_query == 0);
         }
-        StampQueryResult(result, regs.query.query_get.short_query == 0);
         break;
-    }
-    case Regs::QueryOperation::Trap: {
+    case Regs::QueryOperation::Trap:
         UNIMPLEMENTED_MSG("Unimplemented query operation TRAP");
         break;
-    }
-    default: {
+    default:
         UNIMPLEMENTED_MSG("Unknown query operation");
         break;
     }
-    }
 }
 
 void Maxwell3D::ProcessQueryCondition() {
@@ -593,20 +588,20 @@ void Maxwell3D::ProcessQueryCondition() {
     }
     case Regs::ConditionMode::ResNonZero: {
         Regs::QueryCompare cmp;
-        memory_manager.ReadBlockUnsafe(condition_address, &cmp, sizeof(cmp));
+        memory_manager.ReadBlock(condition_address, &cmp, sizeof(cmp));
         execute_on = cmp.initial_sequence != 0U && cmp.initial_mode != 0U;
         break;
     }
     case Regs::ConditionMode::Equal: {
         Regs::QueryCompare cmp;
-        memory_manager.ReadBlockUnsafe(condition_address, &cmp, sizeof(cmp));
+        memory_manager.ReadBlock(condition_address, &cmp, sizeof(cmp));
         execute_on =
             cmp.initial_sequence == cmp.current_sequence && cmp.initial_mode == cmp.current_mode;
         break;
     }
     case Regs::ConditionMode::NotEqual: {
         Regs::QueryCompare cmp;
-        memory_manager.ReadBlockUnsafe(condition_address, &cmp, sizeof(cmp));
+        memory_manager.ReadBlock(condition_address, &cmp, sizeof(cmp));
         execute_on =
             cmp.initial_sequence != cmp.current_sequence || cmp.initial_mode != cmp.current_mode;
         break;
@@ -619,6 +614,18 @@ void Maxwell3D::ProcessQueryCondition() {
     }
 }
 
+void Maxwell3D::ProcessCounterReset() {
+    switch (regs.counter_reset) {
+    case Regs::CounterReset::SampleCnt:
+        rasterizer.ResetCounter(QueryType::SamplesPassed);
+        break;
+    default:
+        LOG_WARNING(Render_OpenGL, "Unimplemented counter reset={}",
+                    static_cast<int>(regs.counter_reset));
+        break;
+    }
+}
+
 void Maxwell3D::ProcessSyncPoint() {
     const u32 sync_point = regs.sync_info.sync_point.Value();
     const u32 increment = regs.sync_info.increment.Value();
@@ -647,7 +654,7 @@ void Maxwell3D::DrawArrays() {
 
     const bool is_indexed{regs.index_array.count && !regs.vertex_buffer.count};
     if (ShouldExecute()) {
-        rasterizer.DrawBatch(is_indexed);
+        rasterizer.Draw(is_indexed, false);
     }
 
     // TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if
@@ -661,6 +668,22 @@ void Maxwell3D::DrawArrays() {
     }
 }
 
+std::optional<u64> Maxwell3D::GetQueryResult() {
+    switch (regs.query.query_get.select) {
+    case Regs::QuerySelect::Zero:
+        return 0;
+    case Regs::QuerySelect::SamplesPassed:
+        // Deferred.
+        rasterizer.Query(regs.query.QueryAddress(), VideoCore::QueryType::SamplesPassed,
+                         system.GPU().GetTicks());
+        return {};
+    default:
+        UNIMPLEMENTED_MSG("Unimplemented query select type {}",
+                          static_cast<u32>(regs.query.query_get.select.Value()));
+        return 1;
+    }
+}
+
 void Maxwell3D::ProcessCBBind(std::size_t stage_index) {
     // Bind the buffer currently in CB_ADDRESS to the specified index in the desired shader stage.
     auto& shader = state.shader_stages[stage_index];
diff --git a/src/video_core/engines/maxwell_3d.h b/src/video_core/engines/maxwell_3d.h
index 0a2af54e5..6ea7cc6a5 100644
--- a/src/video_core/engines/maxwell_3d.h
+++ b/src/video_core/engines/maxwell_3d.h
@@ -6,6 +6,7 @@
 
 #include <array>
 #include <bitset>
+#include <optional>
 #include <type_traits>
 #include <unordered_map>
 #include <vector>
@@ -409,6 +410,27 @@ public:
             Linear = 1,
         };
 
+        enum class CounterReset : u32 {
+            SampleCnt = 0x01,
+            Unk02 = 0x02,
+            Unk03 = 0x03,
+            Unk04 = 0x04,
+            EmittedPrimitives = 0x10, // Not tested
+            Unk11 = 0x11,
+            Unk12 = 0x12,
+            Unk13 = 0x13,
+            Unk15 = 0x15,
+            Unk16 = 0x16,
+            Unk17 = 0x17,
+            Unk18 = 0x18,
+            Unk1A = 0x1A,
+            Unk1B = 0x1B,
+            Unk1C = 0x1C,
+            Unk1D = 0x1D,
+            Unk1E = 0x1E,
+            GeneratedPrimitives = 0x1F,
+        };
+
         struct Cull {
             enum class FrontFace : u32 {
                 ClockWise = 0x0900,
@@ -520,7 +542,7 @@ public:
                 BitField<12, 1, InvMemoryLayout> type;
             } memory_layout;
             union {
-                BitField<0, 16, u32> array_mode;
+                BitField<0, 16, u32> layers;
                 BitField<16, 1, u32> volume;
             };
             u32 layer_stride;
@@ -778,8 +800,12 @@ public:
 
                 u32 zeta_width;
                 u32 zeta_height;
+                union {
+                    BitField<0, 16, u32> zeta_layers;
+                    BitField<16, 1, u32> zeta_volume;
+                };
 
-                INSERT_UNION_PADDING_WORDS(0x27);
+                INSERT_UNION_PADDING_WORDS(0x26);
 
                 u32 depth_test_enable;
 
@@ -857,7 +883,7 @@ public:
                     BitField<7, 1, u32> c7;
                 } clip_distance_enabled;
 
-                INSERT_UNION_PADDING_WORDS(0x1);
+                u32 samplecnt_enable;
 
                 float point_size;
 
@@ -865,7 +891,11 @@ public:
 
                 u32 point_sprite_enable;
 
-                INSERT_UNION_PADDING_WORDS(0x5);
+                INSERT_UNION_PADDING_WORDS(0x3);
+
+                CounterReset counter_reset;
+
+                INSERT_UNION_PADDING_WORDS(0x1);
 
                 u32 zeta_enable;
 
@@ -1412,12 +1442,15 @@ private:
     /// Handles a write to the QUERY_GET register.
     void ProcessQueryGet();
 
-    // Writes the query result accordingly
+    /// Writes the query result accordingly.
     void StampQueryResult(u64 payload, bool long_query);
 
-    // Handles Conditional Rendering
+    /// Handles conditional rendering.
     void ProcessQueryCondition();
 
+    /// Handles counter resets.
+    void ProcessCounterReset();
+
     /// Handles writes to syncing register.
     void ProcessSyncPoint();
 
@@ -1434,6 +1467,9 @@ private:
 
     // Handles a instance drawcall from MME
     void StepInstance(MMEDrawMode expected_mode, u32 count);
+
+    /// Returns a query's value or an empty object if the value will be deferred through a cache.
+    std::optional<u64> GetQueryResult();
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
@@ -1475,6 +1511,7 @@ ASSERT_REG_POSITION(vertex_attrib_format, 0x458);
 ASSERT_REG_POSITION(rt_control, 0x487);
 ASSERT_REG_POSITION(zeta_width, 0x48a);
 ASSERT_REG_POSITION(zeta_height, 0x48b);
+ASSERT_REG_POSITION(zeta_layers, 0x48c);
 ASSERT_REG_POSITION(depth_test_enable, 0x4B3);
 ASSERT_REG_POSITION(independent_blend_enable, 0x4B9);
 ASSERT_REG_POSITION(depth_write_enabled, 0x4BA);
@@ -1499,8 +1536,10 @@ ASSERT_REG_POSITION(screen_y_control, 0x4EB);
 ASSERT_REG_POSITION(vb_element_base, 0x50D);
 ASSERT_REG_POSITION(vb_base_instance, 0x50E);
 ASSERT_REG_POSITION(clip_distance_enabled, 0x544);
+ASSERT_REG_POSITION(samplecnt_enable, 0x545);
 ASSERT_REG_POSITION(point_size, 0x546);
 ASSERT_REG_POSITION(point_sprite_enable, 0x548);
+ASSERT_REG_POSITION(counter_reset, 0x54C);
 ASSERT_REG_POSITION(zeta_enable, 0x54E);
 ASSERT_REG_POSITION(multisample_control, 0x54F);
 ASSERT_REG_POSITION(condition, 0x554);
diff --git a/src/video_core/gpu.cpp b/src/video_core/gpu.cpp
index 4419ab735..7d7137109 100644
--- a/src/video_core/gpu.cpp
+++ b/src/video_core/gpu.cpp
@@ -24,7 +24,7 @@ MICROPROFILE_DEFINE(GPU_wait, "GPU", "Wait for the GPU", MP_RGB(128, 128, 192));
 GPU::GPU(Core::System& system, VideoCore::RendererBase& renderer, bool is_async)
     : system{system}, renderer{renderer}, is_async{is_async} {
     auto& rasterizer{renderer.Rasterizer()};
-    memory_manager = std::make_unique<Tegra::MemoryManager>(system);
+    memory_manager = std::make_unique<Tegra::MemoryManager>(system, rasterizer);
     dma_pusher = std::make_unique<Tegra::DmaPusher>(*this);
     maxwell_3d = std::make_unique<Engines::Maxwell3D>(system, rasterizer, *memory_manager);
     fermi_2d = std::make_unique<Engines::Fermi2D>(rasterizer);
diff --git a/src/video_core/memory_manager.cpp b/src/video_core/memory_manager.cpp
index f1d50be3e..f5d33f27a 100644
--- a/src/video_core/memory_manager.cpp
+++ b/src/video_core/memory_manager.cpp
@@ -11,10 +11,12 @@
 #include "core/memory.h"
 #include "video_core/gpu.h"
 #include "video_core/memory_manager.h"
+#include "video_core/rasterizer_interface.h"
 
 namespace Tegra {
 
-MemoryManager::MemoryManager(Core::System& system) : system{system} {
+MemoryManager::MemoryManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer)
+    : rasterizer{rasterizer}, system{system} {
     std::fill(page_table.pointers.begin(), page_table.pointers.end(), nullptr);
     std::fill(page_table.attributes.begin(), page_table.attributes.end(),
               Common::PageType::Unmapped);
@@ -83,6 +85,7 @@ GPUVAddr MemoryManager::UnmapBuffer(GPUVAddr gpu_addr, u64 size) {
     const auto cpu_addr = GpuToCpuAddress(gpu_addr);
     ASSERT(cpu_addr);
 
+    // Flush and invalidate through the GPU interface, to be asynchronous if possible.
     system.GPU().FlushAndInvalidateRegion(cache_addr, aligned_size);
 
     UnmapRange(gpu_addr, aligned_size);
@@ -242,7 +245,9 @@ void MemoryManager::ReadBlock(GPUVAddr src_addr, void* dest_buffer, const std::s
         switch (page_table.attributes[page_index]) {
         case Common::PageType::Memory: {
             const u8* src_ptr{page_table.pointers[page_index] + page_offset};
-            system.GPU().FlushRegion(ToCacheAddr(src_ptr), copy_amount);
+            // Flush must happen on the rasterizer interface, such that memory is always synchronous
+            // when it is read (even when in asynchronous GPU mode). Fixes Dead Cells title menu.
+            rasterizer.FlushRegion(ToCacheAddr(src_ptr), copy_amount);
             std::memcpy(dest_buffer, src_ptr, copy_amount);
             break;
         }
@@ -292,7 +297,9 @@ void MemoryManager::WriteBlock(GPUVAddr dest_addr, const void* src_buffer, const
         switch (page_table.attributes[page_index]) {
         case Common::PageType::Memory: {
             u8* dest_ptr{page_table.pointers[page_index] + page_offset};
-            system.GPU().InvalidateRegion(ToCacheAddr(dest_ptr), copy_amount);
+            // Invalidate must happen on the rasterizer interface, such that memory is always
+            // synchronous when it is written (even when in asynchronous GPU mode).
+            rasterizer.InvalidateRegion(ToCacheAddr(dest_ptr), copy_amount);
             std::memcpy(dest_ptr, src_buffer, copy_amount);
             break;
         }
@@ -339,8 +346,10 @@ void MemoryManager::CopyBlock(GPUVAddr dest_addr, GPUVAddr src_addr, const std::
 
         switch (page_table.attributes[page_index]) {
         case Common::PageType::Memory: {
+            // Flush must happen on the rasterizer interface, such that memory is always synchronous
+            // when it is copied (even when in asynchronous GPU mode).
             const u8* src_ptr{page_table.pointers[page_index] + page_offset};
-            system.GPU().FlushRegion(ToCacheAddr(src_ptr), copy_amount);
+            rasterizer.FlushRegion(ToCacheAddr(src_ptr), copy_amount);
             WriteBlock(dest_addr, src_ptr, copy_amount);
             break;
         }
diff --git a/src/video_core/memory_manager.h b/src/video_core/memory_manager.h
index 393447eb4..aea010087 100644
--- a/src/video_core/memory_manager.h
+++ b/src/video_core/memory_manager.h
@@ -10,6 +10,10 @@
 #include "common/common_types.h"
 #include "common/page_table.h"
 
+namespace VideoCore {
+class RasterizerInterface;
+}
+
 namespace Core {
 class System;
 }
@@ -47,7 +51,7 @@ struct VirtualMemoryArea {
 
 class MemoryManager final {
 public:
-    explicit MemoryManager(Core::System& system);
+    explicit MemoryManager(Core::System& system, VideoCore::RasterizerInterface& rasterizer);
     ~MemoryManager();
 
     GPUVAddr AllocateSpace(u64 size, u64 align);
@@ -172,6 +176,7 @@ private:
 
     Common::PageTable page_table{page_bits};
     VMAMap vma_map;
+    VideoCore::RasterizerInterface& rasterizer;
 
     Core::System& system;
 };
diff --git a/src/video_core/query_cache.h b/src/video_core/query_cache.h
new file mode 100644
index 000000000..e66054ed0
--- /dev/null
+++ b/src/video_core/query_cache.h
@@ -0,0 +1,359 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <algorithm>
+#include <array>
+#include <cstring>
+#include <iterator>
+#include <memory>
+#include <mutex>
+#include <optional>
+#include <unordered_map>
+#include <vector>
+
+#include "common/assert.h"
+#include "core/core.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/gpu.h"
+#include "video_core/memory_manager.h"
+#include "video_core/rasterizer_interface.h"
+
+namespace VideoCommon {
+
+template <class QueryCache, class HostCounter>
+class CounterStreamBase {
+public:
+    explicit CounterStreamBase(QueryCache& cache, VideoCore::QueryType type)
+        : cache{cache}, type{type} {}
+
+    /// Updates the state of the stream, enabling or disabling as needed.
+    void Update(bool enabled) {
+        if (enabled) {
+            Enable();
+        } else {
+            Disable();
+        }
+    }
+
+    /// Resets the stream to zero. It doesn't disable the query after resetting.
+    void Reset() {
+        if (current) {
+            current->EndQuery();
+
+            // Immediately start a new query to avoid disabling its state.
+            current = cache.Counter(nullptr, type);
+        }
+        last = nullptr;
+    }
+
+    /// Returns the current counter slicing as needed.
+    std::shared_ptr<HostCounter> Current() {
+        if (!current) {
+            return nullptr;
+        }
+        current->EndQuery();
+        last = std::move(current);
+        current = cache.Counter(last, type);
+        return last;
+    }
+
+    /// Returns true when the counter stream is enabled.
+    bool IsEnabled() const {
+        return current != nullptr;
+    }
+
+private:
+    /// Enables the stream.
+    void Enable() {
+        if (current) {
+            return;
+        }
+        current = cache.Counter(last, type);
+    }
+
+    // Disables the stream.
+    void Disable() {
+        if (current) {
+            current->EndQuery();
+        }
+        last = std::exchange(current, nullptr);
+    }
+
+    QueryCache& cache;
+    const VideoCore::QueryType type;
+
+    std::shared_ptr<HostCounter> current;
+    std::shared_ptr<HostCounter> last;
+};
+
+template <class QueryCache, class CachedQuery, class CounterStream, class HostCounter,
+          class QueryPool>
+class QueryCacheBase {
+public:
+    explicit QueryCacheBase(Core::System& system, VideoCore::RasterizerInterface& rasterizer)
+        : system{system}, rasterizer{rasterizer}, streams{{CounterStream{
+                                                      static_cast<QueryCache&>(*this),
+                                                      VideoCore::QueryType::SamplesPassed}}} {}
+
+    void InvalidateRegion(CacheAddr addr, std::size_t size) {
+        std::unique_lock lock{mutex};
+        FlushAndRemoveRegion(addr, size);
+    }
+
+    void FlushRegion(CacheAddr addr, std::size_t size) {
+        std::unique_lock lock{mutex};
+        FlushAndRemoveRegion(addr, size);
+    }
+
+    /**
+     * Records a query in GPU mapped memory, potentially marked with a timestamp.
+     * @param gpu_addr  GPU address to flush to when the mapped memory is read.
+     * @param type      Query type, e.g. SamplesPassed.
+     * @param timestamp Timestamp, when empty the flushed query is assumed to be short.
+     */
+    void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) {
+        std::unique_lock lock{mutex};
+        auto& memory_manager = system.GPU().MemoryManager();
+        const auto host_ptr = memory_manager.GetPointer(gpu_addr);
+
+        CachedQuery* query = TryGet(ToCacheAddr(host_ptr));
+        if (!query) {
+            const auto cpu_addr = memory_manager.GpuToCpuAddress(gpu_addr);
+            ASSERT_OR_EXECUTE(cpu_addr, return;);
+
+            query = Register(type, *cpu_addr, host_ptr, timestamp.has_value());
+        }
+
+        query->BindCounter(Stream(type).Current(), timestamp);
+    }
+
+    /// Updates counters from GPU state. Expected to be called once per draw, clear or dispatch.
+    void UpdateCounters() {
+        std::unique_lock lock{mutex};
+        const auto& regs = system.GPU().Maxwell3D().regs;
+        Stream(VideoCore::QueryType::SamplesPassed).Update(regs.samplecnt_enable);
+    }
+
+    /// Resets a counter to zero. It doesn't disable the query after resetting.
+    void ResetCounter(VideoCore::QueryType type) {
+        std::unique_lock lock{mutex};
+        Stream(type).Reset();
+    }
+
+    /// Disable all active streams. Expected to be called at the end of a command buffer.
+    void DisableStreams() {
+        std::unique_lock lock{mutex};
+        for (auto& stream : streams) {
+            stream.Update(false);
+        }
+    }
+
+    /// Returns a new host counter.
+    std::shared_ptr<HostCounter> Counter(std::shared_ptr<HostCounter> dependency,
+                                         VideoCore::QueryType type) {
+        return std::make_shared<HostCounter>(static_cast<QueryCache&>(*this), std::move(dependency),
+                                             type);
+    }
+
+    /// Returns the counter stream of the specified type.
+    CounterStream& Stream(VideoCore::QueryType type) {
+        return streams[static_cast<std::size_t>(type)];
+    }
+
+    /// Returns the counter stream of the specified type.
+    const CounterStream& Stream(VideoCore::QueryType type) const {
+        return streams[static_cast<std::size_t>(type)];
+    }
+
+protected:
+    std::array<QueryPool, VideoCore::NumQueryTypes> query_pools;
+
+private:
+    /// Flushes a memory range to guest memory and removes it from the cache.
+    void FlushAndRemoveRegion(CacheAddr addr, std::size_t size) {
+        const u64 addr_begin = static_cast<u64>(addr);
+        const u64 addr_end = addr_begin + static_cast<u64>(size);
+        const auto in_range = [addr_begin, addr_end](CachedQuery& query) {
+            const u64 cache_begin = query.GetCacheAddr();
+            const u64 cache_end = cache_begin + query.SizeInBytes();
+            return cache_begin < addr_end && addr_begin < cache_end;
+        };
+
+        const u64 page_end = addr_end >> PAGE_SHIFT;
+        for (u64 page = addr_begin >> PAGE_SHIFT; page <= page_end; ++page) {
+            const auto& it = cached_queries.find(page);
+            if (it == std::end(cached_queries)) {
+                continue;
+            }
+            auto& contents = it->second;
+            for (auto& query : contents) {
+                if (!in_range(query)) {
+                    continue;
+                }
+                rasterizer.UpdatePagesCachedCount(query.CpuAddr(), query.SizeInBytes(), -1);
+                query.Flush();
+            }
+            contents.erase(std::remove_if(std::begin(contents), std::end(contents), in_range),
+                           std::end(contents));
+        }
+    }
+
+    /// Registers the passed parameters as cached and returns a pointer to the stored cached query.
+    CachedQuery* Register(VideoCore::QueryType type, VAddr cpu_addr, u8* host_ptr, bool timestamp) {
+        rasterizer.UpdatePagesCachedCount(cpu_addr, CachedQuery::SizeInBytes(timestamp), 1);
+        const u64 page = static_cast<u64>(ToCacheAddr(host_ptr)) >> PAGE_SHIFT;
+        return &cached_queries[page].emplace_back(static_cast<QueryCache&>(*this), type, cpu_addr,
+                                                  host_ptr);
+    }
+
+    /// Tries to a get a cached query. Returns nullptr on failure.
+    CachedQuery* TryGet(CacheAddr addr) {
+        const u64 page = static_cast<u64>(addr) >> PAGE_SHIFT;
+        const auto it = cached_queries.find(page);
+        if (it == std::end(cached_queries)) {
+            return nullptr;
+        }
+        auto& contents = it->second;
+        const auto found =
+            std::find_if(std::begin(contents), std::end(contents),
+                         [addr](auto& query) { return query.GetCacheAddr() == addr; });
+        return found != std::end(contents) ? &*found : nullptr;
+    }
+
+    static constexpr std::uintptr_t PAGE_SIZE = 4096;
+    static constexpr unsigned PAGE_SHIFT = 12;
+
+    Core::System& system;
+    VideoCore::RasterizerInterface& rasterizer;
+
+    std::recursive_mutex mutex;
+
+    std::unordered_map<u64, std::vector<CachedQuery>> cached_queries;
+
+    std::array<CounterStream, VideoCore::NumQueryTypes> streams;
+};
+
+template <class QueryCache, class HostCounter>
+class HostCounterBase {
+public:
+    explicit HostCounterBase(std::shared_ptr<HostCounter> dependency_)
+        : dependency{std::move(dependency_)}, depth{dependency ? (dependency->Depth() + 1) : 0} {
+        // Avoid nesting too many dependencies to avoid a stack overflow when these are deleted.
+        constexpr u64 depth_threshold = 96;
+        if (depth > depth_threshold) {
+            depth = 0;
+            base_result = dependency->Query();
+            dependency = nullptr;
+        }
+    }
+    virtual ~HostCounterBase() = default;
+
+    /// Returns the current value of the query.
+    u64 Query() {
+        if (result) {
+            return *result;
+        }
+
+        u64 value = BlockingQuery() + base_result;
+        if (dependency) {
+            value += dependency->Query();
+            dependency = nullptr;
+        }
+
+        result = value;
+        return *result;
+    }
+
+    /// Returns true when flushing this query will potentially wait.
+    bool WaitPending() const noexcept {
+        return result.has_value();
+    }
+
+    u64 Depth() const noexcept {
+        return depth;
+    }
+
+protected:
+    /// Returns the value of query from the backend API blocking as needed.
+    virtual u64 BlockingQuery() const = 0;
+
+private:
+    std::shared_ptr<HostCounter> dependency; ///< Counter to add to this value.
+    std::optional<u64> result;               ///< Filled with the already returned value.
+    u64 depth;                               ///< Number of nested dependencies.
+    u64 base_result = 0;                     ///< Equivalent to nested dependencies value.
+};
+
+template <class HostCounter>
+class CachedQueryBase {
+public:
+    explicit CachedQueryBase(VAddr cpu_addr, u8* host_ptr)
+        : cpu_addr{cpu_addr}, host_ptr{host_ptr} {}
+    virtual ~CachedQueryBase() = default;
+
+    CachedQueryBase(CachedQueryBase&&) noexcept = default;
+    CachedQueryBase(const CachedQueryBase&) = delete;
+
+    CachedQueryBase& operator=(CachedQueryBase&&) noexcept = default;
+    CachedQueryBase& operator=(const CachedQueryBase&) = delete;
+
+    /// Flushes the query to guest memory.
+    virtual void Flush() {
+        // When counter is nullptr it means that it's just been reseted. We are supposed to write a
+        // zero in these cases.
+        const u64 value = counter ? counter->Query() : 0;
+        std::memcpy(host_ptr, &value, sizeof(u64));
+
+        if (timestamp) {
+            std::memcpy(host_ptr + TIMESTAMP_OFFSET, &*timestamp, sizeof(u64));
+        }
+    }
+
+    /// Binds a counter to this query.
+    void BindCounter(std::shared_ptr<HostCounter> counter_, std::optional<u64> timestamp_) {
+        if (counter) {
+            // If there's an old counter set it means the query is being rewritten by the game.
+            // To avoid losing the data forever, flush here.
+            Flush();
+        }
+        counter = std::move(counter_);
+        timestamp = timestamp_;
+    }
+
+    VAddr CpuAddr() const noexcept {
+        return cpu_addr;
+    }
+
+    CacheAddr GetCacheAddr() const noexcept {
+        return ToCacheAddr(host_ptr);
+    }
+
+    u64 SizeInBytes() const noexcept {
+        return SizeInBytes(timestamp.has_value());
+    }
+
+    static constexpr u64 SizeInBytes(bool with_timestamp) noexcept {
+        return with_timestamp ? LARGE_QUERY_SIZE : SMALL_QUERY_SIZE;
+    }
+
+protected:
+    /// Returns true when querying the counter may potentially block.
+    bool WaitPending() const noexcept {
+        return counter && counter->WaitPending();
+    }
+
+private:
+    static constexpr std::size_t SMALL_QUERY_SIZE = 8;   // Query size without timestamp.
+    static constexpr std::size_t LARGE_QUERY_SIZE = 16;  // Query size with timestamp.
+    static constexpr std::intptr_t TIMESTAMP_OFFSET = 8; // Timestamp offset in a large query.
+
+    VAddr cpu_addr;                       ///< Guest CPU address.
+    u8* host_ptr;                         ///< Writable host pointer.
+    std::shared_ptr<HostCounter> counter; ///< Host counter to query, owns the dependency tree.
+    std::optional<u64> timestamp;         ///< Timestamp to flush to guest memory.
+};
+
+} // namespace VideoCommon
diff --git a/src/video_core/rasterizer_interface.h b/src/video_core/rasterizer_interface.h
index c586cd6fe..f18eaf4bc 100644
--- a/src/video_core/rasterizer_interface.h
+++ b/src/video_core/rasterizer_interface.h
@@ -6,6 +6,7 @@
 
 #include <atomic>
 #include <functional>
+#include <optional>
 #include "common/common_types.h"
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/gpu.h"
@@ -17,6 +18,11 @@ class MemoryManager;
 
 namespace VideoCore {
 
+enum class QueryType {
+    SamplesPassed,
+};
+constexpr std::size_t NumQueryTypes = 1;
+
 enum class LoadCallbackStage {
     Prepare,
     Decompile,
@@ -29,11 +35,8 @@ class RasterizerInterface {
 public:
     virtual ~RasterizerInterface() {}
 
-    /// Draw the current batch of vertex arrays
-    virtual bool DrawBatch(bool is_indexed) = 0;
-
-    /// Draw the current batch of multiple instances of vertex arrays
-    virtual bool DrawMultiBatch(bool is_indexed) = 0;
+    /// Dispatches a draw invocation
+    virtual void Draw(bool is_indexed, bool is_instanced) = 0;
 
     /// Clear the current framebuffer
     virtual void Clear() = 0;
@@ -41,6 +44,12 @@ public:
     /// Dispatches a compute shader invocation
     virtual void DispatchCompute(GPUVAddr code_addr) = 0;
 
+    /// Resets the counter of a query
+    virtual void ResetCounter(QueryType type) = 0;
+
+    /// Records a GPU query and caches it
+    virtual void Query(GPUVAddr gpu_addr, QueryType type, std::optional<u64> timestamp) = 0;
+
     /// Notify rasterizer that all caches should be flushed to Switch memory
     virtual void FlushAll() = 0;
 
diff --git a/src/video_core/renderer_opengl/gl_query_cache.cpp b/src/video_core/renderer_opengl/gl_query_cache.cpp
new file mode 100644
index 000000000..f12e9f55f
--- /dev/null
+++ b/src/video_core/renderer_opengl/gl_query_cache.cpp
@@ -0,0 +1,120 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cstring>
+#include <memory>
+#include <unordered_map>
+#include <utility>
+#include <vector>
+
+#include <glad/glad.h>
+
+#include "common/assert.h"
+#include "core/core.h"
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/memory_manager.h"
+#include "video_core/renderer_opengl/gl_query_cache.h"
+#include "video_core/renderer_opengl/gl_rasterizer.h"
+
+namespace OpenGL {
+
+namespace {
+
+constexpr std::array<GLenum, VideoCore::NumQueryTypes> QueryTargets = {GL_SAMPLES_PASSED};
+
+constexpr GLenum GetTarget(VideoCore::QueryType type) {
+    return QueryTargets[static_cast<std::size_t>(type)];
+}
+
+} // Anonymous namespace
+
+QueryCache::QueryCache(Core::System& system, RasterizerOpenGL& gl_rasterizer)
+    : VideoCommon::QueryCacheBase<
+          QueryCache, CachedQuery, CounterStream, HostCounter,
+          std::vector<OGLQuery>>{system,
+                                 static_cast<VideoCore::RasterizerInterface&>(gl_rasterizer)},
+      gl_rasterizer{gl_rasterizer} {}
+
+QueryCache::~QueryCache() = default;
+
+OGLQuery QueryCache::AllocateQuery(VideoCore::QueryType type) {
+    auto& reserve = query_pools[static_cast<std::size_t>(type)];
+    OGLQuery query;
+    if (reserve.empty()) {
+        query.Create(GetTarget(type));
+        return query;
+    }
+
+    query = std::move(reserve.back());
+    reserve.pop_back();
+    return query;
+}
+
+void QueryCache::Reserve(VideoCore::QueryType type, OGLQuery&& query) {
+    query_pools[static_cast<std::size_t>(type)].push_back(std::move(query));
+}
+
+bool QueryCache::AnyCommandQueued() const noexcept {
+    return gl_rasterizer.AnyCommandQueued();
+}
+
+HostCounter::HostCounter(QueryCache& cache, std::shared_ptr<HostCounter> dependency,
+                         VideoCore::QueryType type)
+    : VideoCommon::HostCounterBase<QueryCache, HostCounter>{std::move(dependency)}, cache{cache},
+      type{type}, query{cache.AllocateQuery(type)} {
+    glBeginQuery(GetTarget(type), query.handle);
+}
+
+HostCounter::~HostCounter() {
+    cache.Reserve(type, std::move(query));
+}
+
+void HostCounter::EndQuery() {
+    if (!cache.AnyCommandQueued()) {
+        // There are chances a query waited on without commands (glDraw, glClear, glDispatch). Not
+        // having any of these causes a lock. glFlush is considered a command, so we can safely wait
+        // for this. Insert to the OpenGL command stream a flush.
+        glFlush();
+    }
+    glEndQuery(GetTarget(type));
+}
+
+u64 HostCounter::BlockingQuery() const {
+    GLint64 value;
+    glGetQueryObjecti64v(query.handle, GL_QUERY_RESULT, &value);
+    return static_cast<u64>(value);
+}
+
+CachedQuery::CachedQuery(QueryCache& cache, VideoCore::QueryType type, VAddr cpu_addr, u8* host_ptr)
+    : VideoCommon::CachedQueryBase<HostCounter>{cpu_addr, host_ptr}, cache{&cache}, type{type} {}
+
+CachedQuery::CachedQuery(CachedQuery&& rhs) noexcept
+    : VideoCommon::CachedQueryBase<HostCounter>(std::move(rhs)), cache{rhs.cache}, type{rhs.type} {}
+
+CachedQuery& CachedQuery::operator=(CachedQuery&& rhs) noexcept {
+    VideoCommon::CachedQueryBase<HostCounter>::operator=(std::move(rhs));
+    cache = rhs.cache;
+    type = rhs.type;
+    return *this;
+}
+
+void CachedQuery::Flush() {
+    // Waiting for a query while another query of the same target is enabled locks Nvidia's driver.
+    // To avoid this disable and re-enable keeping the dependency stream.
+    // But we only have to do this if we have pending waits to be done.
+    auto& stream = cache->Stream(type);
+    const bool slice_counter = WaitPending() && stream.IsEnabled();
+    if (slice_counter) {
+        stream.Update(false);
+    }
+
+    VideoCommon::CachedQueryBase<HostCounter>::Flush();
+
+    if (slice_counter) {
+        stream.Update(true);
+    }
+}
+
+} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_query_cache.h b/src/video_core/renderer_opengl/gl_query_cache.h
new file mode 100644
index 000000000..d8e7052a1
--- /dev/null
+++ b/src/video_core/renderer_opengl/gl_query_cache.h
@@ -0,0 +1,78 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <memory>
+#include <vector>
+
+#include "common/common_types.h"
+#include "video_core/query_cache.h"
+#include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_opengl/gl_resource_manager.h"
+
+namespace Core {
+class System;
+}
+
+namespace OpenGL {
+
+class CachedQuery;
+class HostCounter;
+class QueryCache;
+class RasterizerOpenGL;
+
+using CounterStream = VideoCommon::CounterStreamBase<QueryCache, HostCounter>;
+
+class QueryCache final : public VideoCommon::QueryCacheBase<QueryCache, CachedQuery, CounterStream,
+                                                            HostCounter, std::vector<OGLQuery>> {
+public:
+    explicit QueryCache(Core::System& system, RasterizerOpenGL& rasterizer);
+    ~QueryCache();
+
+    OGLQuery AllocateQuery(VideoCore::QueryType type);
+
+    void Reserve(VideoCore::QueryType type, OGLQuery&& query);
+
+    bool AnyCommandQueued() const noexcept;
+
+private:
+    RasterizerOpenGL& gl_rasterizer;
+};
+
+class HostCounter final : public VideoCommon::HostCounterBase<QueryCache, HostCounter> {
+public:
+    explicit HostCounter(QueryCache& cache, std::shared_ptr<HostCounter> dependency,
+                         VideoCore::QueryType type);
+    ~HostCounter();
+
+    void EndQuery();
+
+private:
+    u64 BlockingQuery() const override;
+
+    QueryCache& cache;
+    const VideoCore::QueryType type;
+    OGLQuery query;
+};
+
+class CachedQuery final : public VideoCommon::CachedQueryBase<HostCounter> {
+public:
+    explicit CachedQuery(QueryCache& cache, VideoCore::QueryType type, VAddr cpu_addr,
+                         u8* host_ptr);
+    CachedQuery(CachedQuery&& rhs) noexcept;
+    CachedQuery(const CachedQuery&) = delete;
+
+    CachedQuery& operator=(CachedQuery&& rhs) noexcept;
+    CachedQuery& operator=(const CachedQuery&) = delete;
+
+    void Flush() override;
+
+private:
+    QueryCache* cache;
+    VideoCore::QueryType type;
+};
+
+} // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp
index b0eb14c8b..e1965fb21 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp
@@ -25,6 +25,7 @@
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/engines/shader_type.h"
 #include "video_core/memory_manager.h"
+#include "video_core/renderer_opengl/gl_query_cache.h"
 #include "video_core/renderer_opengl/gl_rasterizer.h"
 #include "video_core/renderer_opengl/gl_shader_cache.h"
 #include "video_core/renderer_opengl/gl_shader_gen.h"
@@ -92,8 +93,8 @@ std::size_t GetConstBufferSize(const Tegra::Engines::ConstBufferInfo& buffer,
 RasterizerOpenGL::RasterizerOpenGL(Core::System& system, Core::Frontend::EmuWindow& emu_window,
                                    ScreenInfo& info)
     : RasterizerAccelerated{system.Memory()}, texture_cache{system, *this, device},
-      shader_cache{*this, system, emu_window, device}, system{system}, screen_info{info},
-      buffer_cache{*this, system, device, STREAM_BUFFER_SIZE} {
+      shader_cache{*this, system, emu_window, device}, query_cache{system, *this}, system{system},
+      screen_info{info}, buffer_cache{*this, system, device, STREAM_BUFFER_SIZE} {
     shader_program_manager = std::make_unique<GLShader::ProgramManager>();
     state.draw.shader_program = 0;
     state.Apply();
@@ -541,11 +542,16 @@ void RasterizerOpenGL::Clear() {
     } else if (use_stencil) {
         glClearBufferiv(GL_STENCIL, 0, &regs.clear_stencil);
     }
+
+    ++num_queued_commands;
 }
 
 void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     MICROPROFILE_SCOPE(OpenGL_Drawing);
     auto& gpu = system.GPU().Maxwell3D();
+    const auto& regs = gpu.regs;
+
+    query_cache.UpdateCounters();
 
     SyncRasterizeEnable(state);
     SyncColorMask();
@@ -611,7 +617,7 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
 
     // Setup shaders and their used resources.
     texture_cache.GuardSamplers(true);
-    const auto primitive_mode = MaxwellToGL::PrimitiveTopology(gpu.regs.draw.topology);
+    const GLenum primitive_mode = MaxwellToGL::PrimitiveTopology(gpu.regs.draw.topology);
     SetupShaders(primitive_mode);
     texture_cache.GuardSamplers(false);
 
@@ -638,35 +644,47 @@ void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
         glTextureBarrier();
     }
 
+    ++num_queued_commands;
+
     const GLuint base_instance = static_cast<GLuint>(gpu.regs.vb_base_instance);
     const GLsizei num_instances =
         static_cast<GLsizei>(is_instanced ? gpu.mme_draw.instance_count : 1);
     if (is_indexed) {
-        const GLenum index_format = MaxwellToGL::IndexFormat(gpu.regs.index_array.format);
         const GLint base_vertex = static_cast<GLint>(gpu.regs.vb_element_base);
         const GLsizei num_vertices = static_cast<GLsizei>(gpu.regs.index_array.count);
-        glDrawElementsInstancedBaseVertexBaseInstance(
-            primitive_mode, num_vertices, index_format,
-            reinterpret_cast<const void*>(index_buffer_offset), num_instances, base_vertex,
-            base_instance);
+        const GLvoid* offset = reinterpret_cast<const GLvoid*>(index_buffer_offset);
+        const GLenum format = MaxwellToGL::IndexFormat(gpu.regs.index_array.format);
+        if (num_instances == 1 && base_instance == 0 && base_vertex == 0) {
+            glDrawElements(primitive_mode, num_vertices, format, offset);
+        } else if (num_instances == 1 && base_instance == 0) {
+            glDrawElementsBaseVertex(primitive_mode, num_vertices, format, offset, base_vertex);
+        } else if (base_vertex == 0 && base_instance == 0) {
+            glDrawElementsInstanced(primitive_mode, num_vertices, format, offset, num_instances);
+        } else if (base_vertex == 0) {
+            glDrawElementsInstancedBaseInstance(primitive_mode, num_vertices, format, offset,
+                                                num_instances, base_instance);
+        } else if (base_instance == 0) {
+            glDrawElementsInstancedBaseVertex(primitive_mode, num_vertices, format, offset,
+                                              num_instances, base_vertex);
+        } else {
+            glDrawElementsInstancedBaseVertexBaseInstance(primitive_mode, num_vertices, format,
+                                                          offset, num_instances, base_vertex,
+                                                          base_instance);
+        }
     } else {
         const GLint base_vertex = static_cast<GLint>(gpu.regs.vertex_buffer.first);
         const GLsizei num_vertices = static_cast<GLsizei>(gpu.regs.vertex_buffer.count);
-        glDrawArraysInstancedBaseInstance(primitive_mode, base_vertex, num_vertices, num_instances,
-                                          base_instance);
+        if (num_instances == 1 && base_instance == 0) {
+            glDrawArrays(primitive_mode, base_vertex, num_vertices);
+        } else if (base_instance == 0) {
+            glDrawArraysInstanced(primitive_mode, base_vertex, num_vertices, num_instances);
+        } else {
+            glDrawArraysInstancedBaseInstance(primitive_mode, base_vertex, num_vertices,
+                                              num_instances, base_instance);
+        }
     }
 }
 
-bool RasterizerOpenGL::DrawBatch(bool is_indexed) {
-    Draw(is_indexed, false);
-    return true;
-}
-
-bool RasterizerOpenGL::DrawMultiBatch(bool is_indexed) {
-    Draw(is_indexed, true);
-    return true;
-}
-
 void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
     if (device.HasBrokenCompute()) {
         return;
@@ -707,6 +725,16 @@ void RasterizerOpenGL::DispatchCompute(GPUVAddr code_addr) {
     state.ApplyProgramPipeline();
 
     glDispatchCompute(launch_desc.grid_dim_x, launch_desc.grid_dim_y, launch_desc.grid_dim_z);
+    ++num_queued_commands;
+}
+
+void RasterizerOpenGL::ResetCounter(VideoCore::QueryType type) {
+    query_cache.ResetCounter(type);
+}
+
+void RasterizerOpenGL::Query(GPUVAddr gpu_addr, VideoCore::QueryType type,
+                             std::optional<u64> timestamp) {
+    query_cache.Query(gpu_addr, type, timestamp);
 }
 
 void RasterizerOpenGL::FlushAll() {}
@@ -718,6 +746,7 @@ void RasterizerOpenGL::FlushRegion(CacheAddr addr, u64 size) {
     }
     texture_cache.FlushRegion(addr, size);
     buffer_cache.FlushRegion(addr, size);
+    query_cache.FlushRegion(addr, size);
 }
 
 void RasterizerOpenGL::InvalidateRegion(CacheAddr addr, u64 size) {
@@ -728,6 +757,7 @@ void RasterizerOpenGL::InvalidateRegion(CacheAddr addr, u64 size) {
     texture_cache.InvalidateRegion(addr, size);
     shader_cache.InvalidateRegion(addr, size);
     buffer_cache.InvalidateRegion(addr, size);
+    query_cache.InvalidateRegion(addr, size);
 }
 
 void RasterizerOpenGL::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
@@ -738,10 +768,18 @@ void RasterizerOpenGL::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
 }
 
 void RasterizerOpenGL::FlushCommands() {
+    // Only flush when we have commands queued to OpenGL.
+    if (num_queued_commands == 0) {
+        return;
+    }
+    num_queued_commands = 0;
     glFlush();
 }
 
 void RasterizerOpenGL::TickFrame() {
+    // Ticking a frame means that buffers will be swapped, calling glFlush implicitly.
+    num_queued_commands = 0;
+
     buffer_cache.TickFrame();
 }
 
diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h
index 0501f3828..68abe9a21 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer.h
+++ b/src/video_core/renderer_opengl/gl_rasterizer.h
@@ -24,6 +24,7 @@
 #include "video_core/renderer_opengl/gl_buffer_cache.h"
 #include "video_core/renderer_opengl/gl_device.h"
 #include "video_core/renderer_opengl/gl_framebuffer_cache.h"
+#include "video_core/renderer_opengl/gl_query_cache.h"
 #include "video_core/renderer_opengl/gl_resource_manager.h"
 #include "video_core/renderer_opengl/gl_sampler_cache.h"
 #include "video_core/renderer_opengl/gl_shader_cache.h"
@@ -57,10 +58,11 @@ public:
                               ScreenInfo& info);
     ~RasterizerOpenGL() override;
 
-    bool DrawBatch(bool is_indexed) override;
-    bool DrawMultiBatch(bool is_indexed) override;
+    void Draw(bool is_indexed, bool is_instanced) override;
     void Clear() override;
     void DispatchCompute(GPUVAddr code_addr) override;
+    void ResetCounter(VideoCore::QueryType type) override;
+    void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override;
     void FlushAll() override;
     void FlushRegion(CacheAddr addr, u64 size) override;
     void InvalidateRegion(CacheAddr addr, u64 size) override;
@@ -75,6 +77,11 @@ public:
     void LoadDiskResources(const std::atomic_bool& stop_loading,
                            const VideoCore::DiskResourceLoadCallback& callback) override;
 
+    /// Returns true when there are commands queued to the OpenGL server.
+    bool AnyCommandQueued() const {
+        return num_queued_commands > 0;
+    }
+
 private:
     /// Configures the color and depth framebuffer states.
     void ConfigureFramebuffers();
@@ -102,9 +109,6 @@ private:
     void SetupGlobalMemory(u32 binding, const GLShader::GlobalMemoryEntry& entry, GPUVAddr gpu_addr,
                            std::size_t size);
 
-    /// Syncs all the state, shaders, render targets and textures setting before a draw call.
-    void Draw(bool is_indexed, bool is_instanced);
-
     /// Configures the current textures to use for the draw command.
     void SetupDrawTextures(std::size_t stage_index, const Shader& shader);
 
@@ -180,10 +184,23 @@ private:
     /// Syncs the alpha test state to match the guest state
     void SyncAlphaTest();
 
-    /// Check for extension that are not strictly required
-    /// but are needed for correct emulation
+    /// Check for extension that are not strictly required but are needed for correct emulation
     void CheckExtensions();
 
+    std::size_t CalculateVertexArraysSize() const;
+
+    std::size_t CalculateIndexBufferSize() const;
+
+    /// Updates and returns a vertex array object representing current vertex format
+    GLuint SetupVertexFormat();
+
+    void SetupVertexBuffer(GLuint vao);
+    void SetupVertexInstances(GLuint vao);
+
+    GLintptr SetupIndexBuffer();
+
+    void SetupShaders(GLenum primitive_mode);
+
     const Device device;
     OpenGLState state;
 
@@ -191,6 +208,7 @@ private:
     ShaderCacheOpenGL shader_cache;
     SamplerCacheOpenGL sampler_cache;
     FramebufferCacheOpenGL framebuffer_cache;
+    QueryCache query_cache;
 
     Core::System& system;
     ScreenInfo& screen_info;
@@ -208,19 +226,8 @@ private:
     BindBuffersRangePushBuffer bind_ubo_pushbuffer{GL_UNIFORM_BUFFER};
     BindBuffersRangePushBuffer bind_ssbo_pushbuffer{GL_SHADER_STORAGE_BUFFER};
 
-    std::size_t CalculateVertexArraysSize() const;
-
-    std::size_t CalculateIndexBufferSize() const;
-
-    /// Updates and returns a vertex array object representing current vertex format
-    GLuint SetupVertexFormat();
-
-    void SetupVertexBuffer(GLuint vao);
-    void SetupVertexInstances(GLuint vao);
-
-    GLintptr SetupIndexBuffer();
-
-    void SetupShaders(GLenum primitive_mode);
+    /// Number of commands queued to the OpenGL driver. Reseted on flush.
+    std::size_t num_queued_commands = 0;
 };
 
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_resource_manager.cpp b/src/video_core/renderer_opengl/gl_resource_manager.cpp
index 5c96c1d46..f0ddfb276 100644
--- a/src/video_core/renderer_opengl/gl_resource_manager.cpp
+++ b/src/video_core/renderer_opengl/gl_resource_manager.cpp
@@ -207,4 +207,21 @@ void OGLFramebuffer::Release() {
     handle = 0;
 }
 
+void OGLQuery::Create(GLenum target) {
+    if (handle != 0)
+        return;
+
+    MICROPROFILE_SCOPE(OpenGL_ResourceCreation);
+    glCreateQueries(target, 1, &handle);
+}
+
+void OGLQuery::Release() {
+    if (handle == 0)
+        return;
+
+    MICROPROFILE_SCOPE(OpenGL_ResourceDeletion);
+    glDeleteQueries(1, &handle);
+    handle = 0;
+}
+
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_resource_manager.h b/src/video_core/renderer_opengl/gl_resource_manager.h
index 3a85a1d4c..514d1d165 100644
--- a/src/video_core/renderer_opengl/gl_resource_manager.h
+++ b/src/video_core/renderer_opengl/gl_resource_manager.h
@@ -266,4 +266,29 @@ public:
     GLuint handle = 0;
 };
 
+class OGLQuery : private NonCopyable {
+public:
+    OGLQuery() = default;
+
+    OGLQuery(OGLQuery&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
+
+    ~OGLQuery() {
+        Release();
+    }
+
+    OGLQuery& operator=(OGLQuery&& o) noexcept {
+        Release();
+        handle = std::exchange(o.handle, 0);
+        return *this;
+    }
+
+    /// Creates a new internal OpenGL resource and stores the handle
+    void Create(GLenum target);
+
+    /// Deletes the internal OpenGL resource
+    void Release();
+
+    GLuint handle = 0;
+};
+
 } // namespace OpenGL
diff --git a/src/video_core/renderer_opengl/gl_texture_cache.cpp b/src/video_core/renderer_opengl/gl_texture_cache.cpp
index c9d8aeca9..cf934b0d8 100644
--- a/src/video_core/renderer_opengl/gl_texture_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_texture_cache.cpp
@@ -261,6 +261,13 @@ CachedSurface::~CachedSurface() = default;
 void CachedSurface::DownloadTexture(std::vector<u8>& staging_buffer) {
     MICROPROFILE_SCOPE(OpenGL_Texture_Download);
 
+    if (params.IsBuffer()) {
+        glGetNamedBufferSubData(texture_buffer.handle, 0,
+                                static_cast<GLsizeiptr>(params.GetHostSizeInBytes()),
+                                staging_buffer.data());
+        return;
+    }
+
     SCOPE_EXIT({ glPixelStorei(GL_PACK_ROW_LENGTH, 0); });
 
     for (u32 level = 0; level < params.emulated_levels; ++level) {
@@ -399,24 +406,36 @@ CachedSurfaceView::CachedSurfaceView(CachedSurface& surface, const ViewParams& p
 CachedSurfaceView::~CachedSurfaceView() = default;
 
 void CachedSurfaceView::Attach(GLenum attachment, GLenum target) const {
-    ASSERT(params.num_layers == 1 && params.num_levels == 1);
+    ASSERT(params.num_levels == 1);
 
-    const auto& owner_params = surface.GetSurfaceParams();
+    const GLuint texture = surface.GetTexture();
+    if (params.num_layers > 1) {
+        // Layered framebuffer attachments
+        UNIMPLEMENTED_IF(params.base_layer != 0);
+
+        switch (params.target) {
+        case SurfaceTarget::Texture2DArray:
+            glFramebufferTexture(target, attachment, texture, params.base_level);
+            break;
+        default:
+            UNIMPLEMENTED();
+        }
+        return;
+    }
 
-    switch (owner_params.target) {
+    const GLenum view_target = surface.GetTarget();
+    switch (surface.GetSurfaceParams().target) {
     case SurfaceTarget::Texture1D:
-        glFramebufferTexture1D(target, attachment, surface.GetTarget(), surface.GetTexture(),
-                               params.base_level);
+        glFramebufferTexture1D(target, attachment, view_target, texture, params.base_level);
         break;
     case SurfaceTarget::Texture2D:
-        glFramebufferTexture2D(target, attachment, surface.GetTarget(), surface.GetTexture(),
-                               params.base_level);
+        glFramebufferTexture2D(target, attachment, view_target, texture, params.base_level);
         break;
     case SurfaceTarget::Texture1DArray:
     case SurfaceTarget::Texture2DArray:
     case SurfaceTarget::TextureCubemap:
     case SurfaceTarget::TextureCubeArray:
-        glFramebufferTextureLayer(target, attachment, surface.GetTexture(), params.base_level,
+        glFramebufferTextureLayer(target, attachment, texture, params.base_level,
                                   params.base_layer);
         break;
     default:
diff --git a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
index 8c49c66a7..ef66dd141 100644
--- a/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
+++ b/src/video_core/renderer_vulkan/maxwell_to_vk.cpp
@@ -165,7 +165,7 @@ struct FormatTuple {
     {vk::Format::eUndefined, {}},                                // ASTC_2D_5X4
     {vk::Format::eUndefined, {}},                                // BGRA8_SRGB
     {vk::Format::eBc1RgbaSrgbBlock, {}},                         // DXT1_SRGB
-    {vk::Format::eUndefined, {}},                                // DXT23_SRGB
+    {vk::Format::eBc2SrgbBlock, {}},                             // DXT23_SRGB
     {vk::Format::eBc3SrgbBlock, {}},                             // DXT45_SRGB
     {vk::Format::eBc7SrgbBlock, {}},                             // BC7U_SRGB
     {vk::Format::eR4G4B4A4UnormPack16, Attachable},              // R4G4B4A4U
@@ -364,6 +364,8 @@ vk::Format VertexFormat(Maxwell::VertexAttribute::Type type, Maxwell::VertexAttr
             return vk::Format::eR8G8B8A8Uint;
         case Maxwell::VertexAttribute::Size::Size_32:
             return vk::Format::eR32Uint;
+        case Maxwell::VertexAttribute::Size::Size_32_32_32_32:
+            return vk::Format::eR32G32B32A32Uint;
         default:
             break;
         }
diff --git a/src/video_core/renderer_vulkan/vk_device.cpp b/src/video_core/renderer_vulkan/vk_device.cpp
index de712223e..886bde3b9 100644
--- a/src/video_core/renderer_vulkan/vk_device.cpp
+++ b/src/video_core/renderer_vulkan/vk_device.cpp
@@ -104,8 +104,11 @@ bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instan
     features.depthBiasClamp = true;
     features.geometryShader = true;
     features.tessellationShader = true;
+    features.occlusionQueryPrecise = true;
     features.fragmentStoresAndAtomics = true;
     features.shaderImageGatherExtended = true;
+    features.shaderStorageImageReadWithoutFormat =
+        is_shader_storage_img_read_without_format_supported;
     features.shaderStorageImageWriteWithoutFormat = true;
     features.textureCompressionASTC_LDR = is_optimal_astc_supported;
 
@@ -117,6 +120,10 @@ bool VKDevice::Create(const vk::DispatchLoaderDynamic& dldi, vk::Instance instan
     bit8_storage.uniformAndStorageBuffer8BitAccess = true;
     SetNext(next, bit8_storage);
 
+    vk::PhysicalDeviceHostQueryResetFeaturesEXT host_query_reset;
+    host_query_reset.hostQueryReset = true;
+    SetNext(next, host_query_reset);
+
     vk::PhysicalDeviceFloat16Int8FeaturesKHR float16_int8;
     if (is_float16_supported) {
         float16_int8.shaderFloat16 = true;
@@ -273,6 +280,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
         VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME,
         VK_EXT_SHADER_SUBGROUP_BALLOT_EXTENSION_NAME,
         VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME,
+        VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME,
     };
     std::bitset<required_extensions.size()> available_extensions{};
 
@@ -340,6 +348,7 @@ bool VKDevice::IsSuitable(const vk::DispatchLoaderDynamic& dldi, vk::PhysicalDev
         std::make_pair(features.depthBiasClamp, "depthBiasClamp"),
         std::make_pair(features.geometryShader, "geometryShader"),
         std::make_pair(features.tessellationShader, "tessellationShader"),
+        std::make_pair(features.occlusionQueryPrecise, "occlusionQueryPrecise"),
         std::make_pair(features.fragmentStoresAndAtomics, "fragmentStoresAndAtomics"),
         std::make_pair(features.shaderImageGatherExtended, "shaderImageGatherExtended"),
         std::make_pair(features.shaderStorageImageWriteWithoutFormat,
@@ -376,7 +385,7 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
         }
     };
 
-    extensions.reserve(13);
+    extensions.reserve(14);
     extensions.push_back(VK_KHR_SWAPCHAIN_EXTENSION_NAME);
     extensions.push_back(VK_KHR_16BIT_STORAGE_EXTENSION_NAME);
     extensions.push_back(VK_KHR_8BIT_STORAGE_EXTENSION_NAME);
@@ -384,6 +393,7 @@ std::vector<const char*> VKDevice::LoadExtensions(const vk::DispatchLoaderDynami
     extensions.push_back(VK_EXT_VERTEX_ATTRIBUTE_DIVISOR_EXTENSION_NAME);
     extensions.push_back(VK_EXT_SHADER_SUBGROUP_BALLOT_EXTENSION_NAME);
     extensions.push_back(VK_EXT_SHADER_SUBGROUP_VOTE_EXTENSION_NAME);
+    extensions.push_back(VK_EXT_HOST_QUERY_RESET_EXTENSION_NAME);
 
     [[maybe_unused]] const bool nsight =
         std::getenv("NVTX_INJECTION64_PATH") || std::getenv("NSIGHT_LAUNCHED");
@@ -457,6 +467,8 @@ void VKDevice::SetupFamilies(const vk::DispatchLoaderDynamic& dldi, vk::SurfaceK
 
 void VKDevice::SetupFeatures(const vk::DispatchLoaderDynamic& dldi) {
     const auto supported_features{physical.getFeatures(dldi)};
+    is_shader_storage_img_read_without_format_supported =
+        supported_features.shaderStorageImageReadWithoutFormat;
     is_optimal_astc_supported = IsOptimalAstcSupported(supported_features, dldi);
 }
 
@@ -531,6 +543,7 @@ std::unordered_map<vk::Format, vk::FormatProperties> VKDevice::GetFormatProperti
                                         vk::Format::eBc6HUfloatBlock,
                                         vk::Format::eBc6HSfloatBlock,
                                         vk::Format::eBc1RgbaSrgbBlock,
+                                        vk::Format::eBc2SrgbBlock,
                                         vk::Format::eBc3SrgbBlock,
                                         vk::Format::eBc7SrgbBlock,
                                         vk::Format::eAstc4x4SrgbBlock,
diff --git a/src/video_core/renderer_vulkan/vk_device.h b/src/video_core/renderer_vulkan/vk_device.h
index 72603f9f6..2c27ad730 100644
--- a/src/video_core/renderer_vulkan/vk_device.h
+++ b/src/video_core/renderer_vulkan/vk_device.h
@@ -122,6 +122,11 @@ public:
         return properties.limits.maxPushConstantsSize;
     }
 
+    /// Returns true if Shader storage Image Read Without Format supported.
+    bool IsShaderStorageImageReadWithoutFormatSupported() const {
+        return is_shader_storage_img_read_without_format_supported;
+    }
+
     /// Returns true if ASTC is natively supported.
     bool IsOptimalAstcSupported() const {
         return is_optimal_astc_supported;
@@ -227,6 +232,8 @@ private:
     bool ext_depth_range_unrestricted{};       ///< Support for VK_EXT_depth_range_unrestricted.
     bool ext_shader_viewport_index_layer{};    ///< Support for VK_EXT_shader_viewport_index_layer.
     bool nv_device_diagnostic_checkpoints{};   ///< Support for VK_NV_device_diagnostic_checkpoints.
+    bool is_shader_storage_img_read_without_format_supported{}; ///< Support for shader storage
+                                                                ///< image read without format
 
     // Telemetry parameters
     std::string vendor_name;                      ///< Device's driver name.
diff --git a/src/video_core/renderer_vulkan/vk_query_cache.cpp b/src/video_core/renderer_vulkan/vk_query_cache.cpp
new file mode 100644
index 000000000..ffbf60dda
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_query_cache.cpp
@@ -0,0 +1,122 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <cstddef>
+#include <cstdint>
+#include <utility>
+#include <vector>
+
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_query_cache.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
+
+namespace Vulkan {
+
+namespace {
+
+constexpr std::array QUERY_TARGETS = {vk::QueryType::eOcclusion};
+
+constexpr vk::QueryType GetTarget(VideoCore::QueryType type) {
+    return QUERY_TARGETS[static_cast<std::size_t>(type)];
+}
+
+} // Anonymous namespace
+
+QueryPool::QueryPool() : VKFencedPool{GROW_STEP} {}
+
+QueryPool::~QueryPool() = default;
+
+void QueryPool::Initialize(const VKDevice& device_, VideoCore::QueryType type_) {
+    device = &device_;
+    type = type_;
+}
+
+std::pair<vk::QueryPool, std::uint32_t> QueryPool::Commit(VKFence& fence) {
+    std::size_t index;
+    do {
+        index = CommitResource(fence);
+    } while (usage[index]);
+    usage[index] = true;
+
+    return {*pools[index / GROW_STEP], static_cast<std::uint32_t>(index % GROW_STEP)};
+}
+
+void QueryPool::Allocate(std::size_t begin, std::size_t end) {
+    usage.resize(end);
+
+    const auto dev = device->GetLogical();
+    const u32 size = static_cast<u32>(end - begin);
+    const vk::QueryPoolCreateInfo query_pool_ci({}, GetTarget(type), size, {});
+    pools.push_back(dev.createQueryPoolUnique(query_pool_ci, nullptr, device->GetDispatchLoader()));
+}
+
+void QueryPool::Reserve(std::pair<vk::QueryPool, std::uint32_t> query) {
+    const auto it =
+        std::find_if(std::begin(pools), std::end(pools),
+                     [query_pool = query.first](auto& pool) { return query_pool == *pool; });
+    ASSERT(it != std::end(pools));
+
+    const std::ptrdiff_t pool_index = std::distance(std::begin(pools), it);
+    usage[pool_index * GROW_STEP + static_cast<std::ptrdiff_t>(query.second)] = false;
+}
+
+VKQueryCache::VKQueryCache(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                           const VKDevice& device, VKScheduler& scheduler)
+    : VideoCommon::QueryCacheBase<VKQueryCache, CachedQuery, CounterStream, HostCounter,
+                                  QueryPool>{system, rasterizer},
+      device{device}, scheduler{scheduler} {
+    for (std::size_t i = 0; i < static_cast<std::size_t>(VideoCore::NumQueryTypes); ++i) {
+        query_pools[i].Initialize(device, static_cast<VideoCore::QueryType>(i));
+    }
+}
+
+VKQueryCache::~VKQueryCache() = default;
+
+std::pair<vk::QueryPool, std::uint32_t> VKQueryCache::AllocateQuery(VideoCore::QueryType type) {
+    return query_pools[static_cast<std::size_t>(type)].Commit(scheduler.GetFence());
+}
+
+void VKQueryCache::Reserve(VideoCore::QueryType type,
+                           std::pair<vk::QueryPool, std::uint32_t> query) {
+    query_pools[static_cast<std::size_t>(type)].Reserve(query);
+}
+
+HostCounter::HostCounter(VKQueryCache& cache, std::shared_ptr<HostCounter> dependency,
+                         VideoCore::QueryType type)
+    : VideoCommon::HostCounterBase<VKQueryCache, HostCounter>{std::move(dependency)}, cache{cache},
+      type{type}, query{cache.AllocateQuery(type)}, ticks{cache.Scheduler().Ticks()} {
+    const auto dev = cache.Device().GetLogical();
+    cache.Scheduler().Record([dev, query = query](vk::CommandBuffer cmdbuf, auto& dld) {
+        dev.resetQueryPoolEXT(query.first, query.second, 1, dld);
+        cmdbuf.beginQuery(query.first, query.second, vk::QueryControlFlagBits::ePrecise, dld);
+    });
+}
+
+HostCounter::~HostCounter() {
+    cache.Reserve(type, query);
+}
+
+void HostCounter::EndQuery() {
+    cache.Scheduler().Record([query = query](auto cmdbuf, auto& dld) {
+        cmdbuf.endQuery(query.first, query.second, dld);
+    });
+}
+
+u64 HostCounter::BlockingQuery() const {
+    if (ticks >= cache.Scheduler().Ticks()) {
+        cache.Scheduler().Flush();
+    }
+
+    const auto dev = cache.Device().GetLogical();
+    const auto& dld = cache.Device().GetDispatchLoader();
+    u64 value;
+    dev.getQueryPoolResults(query.first, query.second, 1, sizeof(value), &value, sizeof(value),
+                            vk::QueryResultFlagBits::e64 | vk::QueryResultFlagBits::eWait, dld);
+    return value;
+}
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_query_cache.h b/src/video_core/renderer_vulkan/vk_query_cache.h
new file mode 100644
index 000000000..c3092ee96
--- /dev/null
+++ b/src/video_core/renderer_vulkan/vk_query_cache.h
@@ -0,0 +1,104 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <cstddef>
+#include <cstdint>
+#include <memory>
+#include <utility>
+#include <vector>
+
+#include "common/common_types.h"
+#include "video_core/query_cache.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+
+namespace VideoCore {
+class RasterizerInterface;
+}
+
+namespace Vulkan {
+
+class CachedQuery;
+class HostCounter;
+class VKDevice;
+class VKQueryCache;
+class VKScheduler;
+
+using CounterStream = VideoCommon::CounterStreamBase<VKQueryCache, HostCounter>;
+
+class QueryPool final : public VKFencedPool {
+public:
+    explicit QueryPool();
+    ~QueryPool() override;
+
+    void Initialize(const VKDevice& device, VideoCore::QueryType type);
+
+    std::pair<vk::QueryPool, std::uint32_t> Commit(VKFence& fence);
+
+    void Reserve(std::pair<vk::QueryPool, std::uint32_t> query);
+
+protected:
+    void Allocate(std::size_t begin, std::size_t end) override;
+
+private:
+    static constexpr std::size_t GROW_STEP = 512;
+
+    const VKDevice* device = nullptr;
+    VideoCore::QueryType type = {};
+
+    std::vector<UniqueQueryPool> pools;
+    std::vector<bool> usage;
+};
+
+class VKQueryCache final
+    : public VideoCommon::QueryCacheBase<VKQueryCache, CachedQuery, CounterStream, HostCounter,
+                                         QueryPool> {
+public:
+    explicit VKQueryCache(Core::System& system, VideoCore::RasterizerInterface& rasterizer,
+                          const VKDevice& device, VKScheduler& scheduler);
+    ~VKQueryCache();
+
+    std::pair<vk::QueryPool, std::uint32_t> AllocateQuery(VideoCore::QueryType type);
+
+    void Reserve(VideoCore::QueryType type, std::pair<vk::QueryPool, std::uint32_t> query);
+
+    const VKDevice& Device() const noexcept {
+        return device;
+    }
+
+    VKScheduler& Scheduler() const noexcept {
+        return scheduler;
+    }
+
+private:
+    const VKDevice& device;
+    VKScheduler& scheduler;
+};
+
+class HostCounter final : public VideoCommon::HostCounterBase<VKQueryCache, HostCounter> {
+public:
+    explicit HostCounter(VKQueryCache& cache, std::shared_ptr<HostCounter> dependency,
+                         VideoCore::QueryType type);
+    ~HostCounter();
+
+    void EndQuery();
+
+private:
+    u64 BlockingQuery() const override;
+
+    VKQueryCache& cache;
+    const VideoCore::QueryType type;
+    const std::pair<vk::QueryPool, std::uint32_t> query;
+    const u64 ticks;
+};
+
+class CachedQuery : public VideoCommon::CachedQueryBase<HostCounter> {
+public:
+    explicit CachedQuery(VKQueryCache&, VideoCore::QueryType, VAddr cpu_addr, u8* host_ptr)
+        : VideoCommon::CachedQueryBase<HostCounter>{cpu_addr, host_ptr} {}
+};
+
+} // namespace Vulkan
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.cpp b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
index aada38702..3bf86da87 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.cpp
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.cpp
@@ -289,25 +289,19 @@ RasterizerVulkan::RasterizerVulkan(Core::System& system, Core::Frontend::EmuWind
                     staging_pool),
       pipeline_cache(system, *this, device, scheduler, descriptor_pool, update_descriptor_queue),
       buffer_cache(*this, system, device, memory_manager, scheduler, staging_pool),
-      sampler_cache(device) {}
-
-RasterizerVulkan::~RasterizerVulkan() = default;
-
-bool RasterizerVulkan::DrawBatch(bool is_indexed) {
-    Draw(is_indexed, false);
-    return true;
+      sampler_cache(device), query_cache(system, *this, device, scheduler) {
+    scheduler.SetQueryCache(query_cache);
 }
 
-bool RasterizerVulkan::DrawMultiBatch(bool is_indexed) {
-    Draw(is_indexed, true);
-    return true;
-}
+RasterizerVulkan::~RasterizerVulkan() = default;
 
 void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
     MICROPROFILE_SCOPE(Vulkan_Drawing);
 
     FlushWork();
 
+    query_cache.UpdateCounters();
+
     const auto& gpu = system.GPU().Maxwell3D();
     GraphicsPipelineCacheKey key{GetFixedPipelineState(gpu.regs)};
 
@@ -362,6 +356,8 @@ void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
 void RasterizerVulkan::Clear() {
     MICROPROFILE_SCOPE(Vulkan_Clearing);
 
+    query_cache.UpdateCounters();
+
     const auto& gpu = system.GPU().Maxwell3D();
     if (!system.GPU().Maxwell3D().ShouldExecute()) {
         return;
@@ -429,6 +425,8 @@ void RasterizerVulkan::DispatchCompute(GPUVAddr code_addr) {
     sampled_views.clear();
     image_views.clear();
 
+    query_cache.UpdateCounters();
+
     const auto& launch_desc = system.GPU().KeplerCompute().launch_description;
     const ComputePipelineCacheKey key{
         code_addr,
@@ -471,17 +469,28 @@ void RasterizerVulkan::DispatchCompute(GPUVAddr code_addr) {
     });
 }
 
+void RasterizerVulkan::ResetCounter(VideoCore::QueryType type) {
+    query_cache.ResetCounter(type);
+}
+
+void RasterizerVulkan::Query(GPUVAddr gpu_addr, VideoCore::QueryType type,
+                             std::optional<u64> timestamp) {
+    query_cache.Query(gpu_addr, type, timestamp);
+}
+
 void RasterizerVulkan::FlushAll() {}
 
 void RasterizerVulkan::FlushRegion(CacheAddr addr, u64 size) {
     texture_cache.FlushRegion(addr, size);
     buffer_cache.FlushRegion(addr, size);
+    query_cache.FlushRegion(addr, size);
 }
 
 void RasterizerVulkan::InvalidateRegion(CacheAddr addr, u64 size) {
     texture_cache.InvalidateRegion(addr, size);
     pipeline_cache.InvalidateRegion(addr, size);
     buffer_cache.InvalidateRegion(addr, size);
+    query_cache.InvalidateRegion(addr, size);
 }
 
 void RasterizerVulkan::FlushAndInvalidateRegion(CacheAddr addr, u64 size) {
@@ -602,33 +611,34 @@ bool RasterizerVulkan::WalkAttachmentOverlaps(const CachedSurfaceView& attachmen
 std::tuple<vk::Framebuffer, vk::Extent2D> RasterizerVulkan::ConfigureFramebuffers(
     vk::RenderPass renderpass) {
     FramebufferCacheKey key{renderpass, std::numeric_limits<u32>::max(),
-                            std::numeric_limits<u32>::max()};
+                            std::numeric_limits<u32>::max(), std::numeric_limits<u32>::max()};
 
-    const auto MarkAsModifiedAndPush = [&](const View& view) {
-        if (view == nullptr) {
+    const auto try_push = [&](const View& view) {
+        if (!view) {
             return false;
         }
         key.views.push_back(view->GetHandle());
         key.width = std::min(key.width, view->GetWidth());
         key.height = std::min(key.height, view->GetHeight());
+        key.layers = std::min(key.layers, view->GetNumLayers());
         return true;
     };
 
     for (std::size_t index = 0; index < std::size(color_attachments); ++index) {
-        if (MarkAsModifiedAndPush(color_attachments[index])) {
+        if (try_push(color_attachments[index])) {
             texture_cache.MarkColorBufferInUse(index);
         }
     }
-    if (MarkAsModifiedAndPush(zeta_attachment)) {
+    if (try_push(zeta_attachment)) {
         texture_cache.MarkDepthBufferInUse();
     }
 
     const auto [fbentry, is_cache_miss] = framebuffer_cache.try_emplace(key);
     auto& framebuffer = fbentry->second;
     if (is_cache_miss) {
-        const vk::FramebufferCreateInfo framebuffer_ci({}, key.renderpass,
-                                                       static_cast<u32>(key.views.size()),
-                                                       key.views.data(), key.width, key.height, 1);
+        const vk::FramebufferCreateInfo framebuffer_ci(
+            {}, key.renderpass, static_cast<u32>(key.views.size()), key.views.data(), key.width,
+            key.height, key.layers);
         const auto dev = device.GetLogical();
         const auto& dld = device.GetDispatchLoader();
         framebuffer = dev.createFramebufferUnique(framebuffer_ci, nullptr, dld);
diff --git a/src/video_core/renderer_vulkan/vk_rasterizer.h b/src/video_core/renderer_vulkan/vk_rasterizer.h
index 7be71e734..4dc8af6e8 100644
--- a/src/video_core/renderer_vulkan/vk_rasterizer.h
+++ b/src/video_core/renderer_vulkan/vk_rasterizer.h
@@ -24,6 +24,7 @@
 #include "video_core/renderer_vulkan/vk_descriptor_pool.h"
 #include "video_core/renderer_vulkan/vk_memory_manager.h"
 #include "video_core/renderer_vulkan/vk_pipeline_cache.h"
+#include "video_core/renderer_vulkan/vk_query_cache.h"
 #include "video_core/renderer_vulkan/vk_renderpass_cache.h"
 #include "video_core/renderer_vulkan/vk_resource_manager.h"
 #include "video_core/renderer_vulkan/vk_sampler_cache.h"
@@ -55,6 +56,7 @@ struct FramebufferCacheKey {
     vk::RenderPass renderpass{};
     u32 width = 0;
     u32 height = 0;
+    u32 layers = 0;
     ImageViewsPack views;
 
     std::size_t Hash() const noexcept {
@@ -65,12 +67,17 @@ struct FramebufferCacheKey {
         }
         boost::hash_combine(hash, width);
         boost::hash_combine(hash, height);
+        boost::hash_combine(hash, layers);
         return hash;
     }
 
     bool operator==(const FramebufferCacheKey& rhs) const noexcept {
-        return std::tie(renderpass, views, width, height) ==
-               std::tie(rhs.renderpass, rhs.views, rhs.width, rhs.height);
+        return std::tie(renderpass, views, width, height, layers) ==
+               std::tie(rhs.renderpass, rhs.views, rhs.width, rhs.height, rhs.layers);
+    }
+
+    bool operator!=(const FramebufferCacheKey& rhs) const noexcept {
+        return !operator==(rhs);
     }
 };
 
@@ -96,7 +103,7 @@ struct ImageView {
     vk::ImageLayout* layout = nullptr;
 };
 
-class RasterizerVulkan : public VideoCore::RasterizerAccelerated {
+class RasterizerVulkan final : public VideoCore::RasterizerAccelerated {
 public:
     explicit RasterizerVulkan(Core::System& system, Core::Frontend::EmuWindow& render_window,
                               VKScreenInfo& screen_info, const VKDevice& device,
@@ -104,10 +111,11 @@ public:
                               VKScheduler& scheduler);
     ~RasterizerVulkan() override;
 
-    bool DrawBatch(bool is_indexed) override;
-    bool DrawMultiBatch(bool is_indexed) override;
+    void Draw(bool is_indexed, bool is_instanced) override;
     void Clear() override;
     void DispatchCompute(GPUVAddr code_addr) override;
+    void ResetCounter(VideoCore::QueryType type) override;
+    void Query(GPUVAddr gpu_addr, VideoCore::QueryType type, std::optional<u64> timestamp) override;
     void FlushAll() override;
     void FlushRegion(CacheAddr addr, u64 size) override;
     void InvalidateRegion(CacheAddr addr, u64 size) override;
@@ -140,8 +148,6 @@ private:
 
     static constexpr std::size_t ZETA_TEXCEPTION_INDEX = 8;
 
-    void Draw(bool is_indexed, bool is_instanced);
-
     void FlushWork();
 
     Texceptions UpdateAttachments();
@@ -247,6 +253,7 @@ private:
     VKPipelineCache pipeline_cache;
     VKBufferCache buffer_cache;
     VKSamplerCache sampler_cache;
+    VKQueryCache query_cache;
 
     std::array<View, Maxwell::NumRenderTargets> color_attachments;
     View zeta_attachment;
diff --git a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp
index 0a8ec8398..204b7c39c 100644
--- a/src/video_core/renderer_vulkan/vk_sampler_cache.cpp
+++ b/src/video_core/renderer_vulkan/vk_sampler_cache.cpp
@@ -23,7 +23,14 @@ static std::optional<vk::BorderColor> TryConvertBorderColor(std::array<float, 4>
     } else if (color == std::array<float, 4>{1, 1, 1, 1}) {
         return vk::BorderColor::eFloatOpaqueWhite;
     } else {
-        return {};
+        if (color[0] + color[1] + color[2] > 1.35f) {
+            // If color elements are brighter than roughly 0.5 average, use white border
+            return vk::BorderColor::eFloatOpaqueWhite;
+        }
+        if (color[3] > 0.5f) {
+            return vk::BorderColor::eFloatOpaqueBlack;
+        }
+        return vk::BorderColor::eFloatTransparentBlack;
     }
 }
 
@@ -37,8 +44,6 @@ UniqueSampler VKSamplerCache::CreateSampler(const Tegra::Texture::TSCEntry& tsc)
 
     const auto border_color{tsc.GetBorderColor()};
     const auto vk_border_color{TryConvertBorderColor(border_color)};
-    UNIMPLEMENTED_IF_MSG(!vk_border_color, "Unimplemented border color {} {} {} {}",
-                         border_color[0], border_color[1], border_color[2], border_color[3]);
 
     constexpr bool unnormalized_coords{false};
 
diff --git a/src/video_core/renderer_vulkan/vk_scheduler.cpp b/src/video_core/renderer_vulkan/vk_scheduler.cpp
index d66133ad1..92bd6c344 100644
--- a/src/video_core/renderer_vulkan/vk_scheduler.cpp
+++ b/src/video_core/renderer_vulkan/vk_scheduler.cpp
@@ -6,6 +6,7 @@
 #include "common/microprofile.h"
 #include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_query_cache.h"
 #include "video_core/renderer_vulkan/vk_resource_manager.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 
@@ -139,6 +140,8 @@ void VKScheduler::SubmitExecution(vk::Semaphore semaphore) {
 }
 
 void VKScheduler::AllocateNewContext() {
+    ++ticks;
+
     std::unique_lock lock{mutex};
     current_fence = next_fence;
     next_fence = &resource_manager.CommitFence();
@@ -146,6 +149,10 @@ void VKScheduler::AllocateNewContext() {
     current_cmdbuf = resource_manager.CommitCommandBuffer(*current_fence);
     current_cmdbuf.begin({vk::CommandBufferUsageFlagBits::eOneTimeSubmit},
                          device.GetDispatchLoader());
+    // Enable counters once again. These are disabled when a command buffer is finished.
+    if (query_cache) {
+        query_cache->UpdateCounters();
+    }
 }
 
 void VKScheduler::InvalidateState() {
@@ -159,6 +166,7 @@ void VKScheduler::InvalidateState() {
 }
 
 void VKScheduler::EndPendingOperations() {
+    query_cache->DisableStreams();
     EndRenderPass();
 }
 
diff --git a/src/video_core/renderer_vulkan/vk_scheduler.h b/src/video_core/renderer_vulkan/vk_scheduler.h
index bcdffbba0..62fd7858b 100644
--- a/src/video_core/renderer_vulkan/vk_scheduler.h
+++ b/src/video_core/renderer_vulkan/vk_scheduler.h
@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <atomic>
 #include <condition_variable>
 #include <memory>
 #include <optional>
@@ -18,6 +19,7 @@ namespace Vulkan {
 
 class VKDevice;
 class VKFence;
+class VKQueryCache;
 class VKResourceManager;
 
 class VKFenceView {
@@ -67,6 +69,11 @@ public:
     /// Binds a pipeline to the current execution context.
     void BindGraphicsPipeline(vk::Pipeline pipeline);
 
+    /// Assigns the query cache.
+    void SetQueryCache(VKQueryCache& query_cache_) {
+        query_cache = &query_cache_;
+    }
+
     /// Returns true when viewports have been set in the current command buffer.
     bool TouchViewports() {
         return std::exchange(state.viewports, true);
@@ -112,6 +119,11 @@ public:
         return current_fence;
     }
 
+    /// Returns the current command buffer tick.
+    u64 Ticks() const {
+        return ticks;
+    }
+
 private:
     class Command {
     public:
@@ -205,6 +217,8 @@ private:
 
     const VKDevice& device;
     VKResourceManager& resource_manager;
+    VKQueryCache* query_cache = nullptr;
+
     vk::CommandBuffer current_cmdbuf;
     VKFence* current_fence = nullptr;
     VKFence* next_fence = nullptr;
@@ -227,6 +241,7 @@ private:
     Common::SPSCQueue<std::unique_ptr<CommandChunk>> chunk_reserve;
     std::mutex mutex;
     std::condition_variable cv;
+    std::atomic<u64> ticks = 0;
     bool quit = false;
 };
 
diff --git a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
index 24a658dce..2da622d15 100644
--- a/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
+++ b/src/video_core/renderer_vulkan/vk_shader_decompiler.cpp
@@ -86,6 +86,7 @@ struct AttributeType {
 
 struct VertexIndices {
     std::optional<u32> position;
+    std::optional<u32> layer;
     std::optional<u32> viewport;
     std::optional<u32> point_size;
     std::optional<u32> clip_distances;
@@ -275,21 +276,29 @@ public:
         AddCapability(spv::Capability::ImageGatherExtended);
         AddCapability(spv::Capability::SampledBuffer);
         AddCapability(spv::Capability::StorageImageWriteWithoutFormat);
+        AddCapability(spv::Capability::DrawParameters);
         AddCapability(spv::Capability::SubgroupBallotKHR);
         AddCapability(spv::Capability::SubgroupVoteKHR);
         AddExtension("SPV_KHR_shader_ballot");
         AddExtension("SPV_KHR_subgroup_vote");
         AddExtension("SPV_KHR_storage_buffer_storage_class");
         AddExtension("SPV_KHR_variable_pointers");
+        AddExtension("SPV_KHR_shader_draw_parameters");
 
-        if (ir.UsesViewportIndex()) {
-            AddCapability(spv::Capability::MultiViewport);
-            if (device.IsExtShaderViewportIndexLayerSupported()) {
+        if (ir.UsesLayer() || ir.UsesViewportIndex()) {
+            if (ir.UsesViewportIndex()) {
+                AddCapability(spv::Capability::MultiViewport);
+            }
+            if (stage != ShaderType::Geometry && device.IsExtShaderViewportIndexLayerSupported()) {
                 AddExtension("SPV_EXT_shader_viewport_index_layer");
                 AddCapability(spv::Capability::ShaderViewportIndexLayerEXT);
             }
         }
 
+        if (device.IsShaderStorageImageReadWithoutFormatSupported()) {
+            AddCapability(spv::Capability::StorageImageReadWithoutFormat);
+        }
+
         if (device.IsFloat16Supported()) {
             AddCapability(spv::Capability::Float16);
         }
@@ -492,9 +501,11 @@ private:
         interfaces.push_back(AddGlobalVariable(Name(out_vertex, "out_vertex")));
 
         // Declare input attributes
-        vertex_index = DeclareInputBuiltIn(spv::BuiltIn::VertexIndex, t_in_uint, "vertex_index");
+        vertex_index = DeclareInputBuiltIn(spv::BuiltIn::VertexIndex, t_in_int, "vertex_index");
         instance_index =
-            DeclareInputBuiltIn(spv::BuiltIn::InstanceIndex, t_in_uint, "instance_index");
+            DeclareInputBuiltIn(spv::BuiltIn::InstanceIndex, t_in_int, "instance_index");
+        base_vertex = DeclareInputBuiltIn(spv::BuiltIn::BaseVertex, t_in_int, "base_vertex");
+        base_instance = DeclareInputBuiltIn(spv::BuiltIn::BaseInstance, t_in_int, "base_instance");
     }
 
     void DeclareTessControl() {
@@ -920,13 +931,22 @@ private:
         VertexIndices indices;
         indices.position = AddBuiltIn(t_float4, spv::BuiltIn::Position, "position");
 
+        if (ir.UsesLayer()) {
+            if (stage != ShaderType::Vertex || device.IsExtShaderViewportIndexLayerSupported()) {
+                indices.layer = AddBuiltIn(t_int, spv::BuiltIn::Layer, "layer");
+            } else {
+                LOG_ERROR(
+                    Render_Vulkan,
+                    "Shader requires Layer but it's not supported on this stage with this device.");
+            }
+        }
+
         if (ir.UsesViewportIndex()) {
             if (stage != ShaderType::Vertex || device.IsExtShaderViewportIndexLayerSupported()) {
                 indices.viewport = AddBuiltIn(t_int, spv::BuiltIn::ViewportIndex, "viewport_index");
             } else {
-                LOG_ERROR(Render_Vulkan,
-                          "Shader requires ViewportIndex but it's not supported on this "
-                          "stage with this device.");
+                LOG_ERROR(Render_Vulkan, "Shader requires ViewportIndex but it's not supported on "
+                                         "this stage with this device.");
             }
         }
 
@@ -1068,9 +1088,12 @@ private:
                     return {OpLoad(t_float, AccessElement(t_in_float, tess_coord, element)),
                             Type::Float};
                 case 2:
-                    return {OpLoad(t_uint, instance_index), Type::Uint};
+                    return {
+                        OpISub(t_int, OpLoad(t_int, instance_index), OpLoad(t_int, base_instance)),
+                        Type::Int};
                 case 3:
-                    return {OpLoad(t_uint, vertex_index), Type::Uint};
+                    return {OpISub(t_int, OpLoad(t_int, vertex_index), OpLoad(t_int, base_vertex)),
+                            Type::Int};
                 }
                 UNIMPLEMENTED_MSG("Unmanaged TessCoordInstanceIDVertexID element={}", element);
                 return {Constant(t_uint, 0U), Type::Uint};
@@ -1285,6 +1308,13 @@ private:
                 }
                 case Attribute::Index::LayerViewportPointSize:
                     switch (element) {
+                    case 1: {
+                        if (!out_indices.layer) {
+                            return {};
+                        }
+                        const u32 index = out_indices.layer.value();
+                        return {AccessElement(t_out_int, out_vertex, index), Type::Int};
+                    }
                     case 2: {
                         if (!out_indices.viewport) {
                             return {};
@@ -1355,6 +1385,11 @@ private:
             UNIMPLEMENTED();
         }
 
+        if (!target.id) {
+            // On failure we return a nullptr target.id, skip these stores.
+            return {};
+        }
+
         OpStore(target.id, As(Visit(src), target.type));
         return {};
     }
@@ -1748,8 +1783,16 @@ private:
     }
 
     Expression ImageLoad(Operation operation) {
-        UNIMPLEMENTED();
-        return {};
+        if (!device.IsShaderStorageImageReadWithoutFormatSupported()) {
+            return {v_float_zero, Type::Float};
+        }
+
+        const auto& meta{std::get<MetaImage>(operation.GetMeta())};
+
+        const Id coords = GetCoordinates(operation, Type::Int);
+        const Id texel = OpImageRead(t_uint4, GetImage(operation), coords);
+
+        return {OpCompositeExtract(t_uint, texel, meta.element), Type::Uint};
     }
 
     Expression ImageStore(Operation operation) {
@@ -2542,6 +2585,8 @@ private:
 
     Id instance_index{};
     Id vertex_index{};
+    Id base_instance{};
+    Id base_vertex{};
     std::array<Id, Maxwell::NumRenderTargets> frag_colors{};
     Id frag_depth{};
     Id frag_coord{};
diff --git a/src/video_core/renderer_vulkan/vk_texture_cache.h b/src/video_core/renderer_vulkan/vk_texture_cache.h
index d3edbe80c..22e3d34de 100644
--- a/src/video_core/renderer_vulkan/vk_texture_cache.h
+++ b/src/video_core/renderer_vulkan/vk_texture_cache.h
@@ -151,6 +151,10 @@ public:
         return params.GetMipHeight(base_level);
     }
 
+    u32 GetNumLayers() const {
+        return num_layers;
+    }
+
     bool IsBufferView() const {
         return buffer_view;
     }
diff --git a/src/video_core/shader/decode/conversion.cpp b/src/video_core/shader/decode/conversion.cpp
index 0eeb75559..6ead42070 100644
--- a/src/video_core/shader/decode/conversion.cpp
+++ b/src/video_core/shader/decode/conversion.cpp
@@ -83,14 +83,14 @@ u32 ShaderIR::DecodeConversion(NodeBlock& bb, u32 pc) {
 
         const bool input_signed = instr.conversion.is_input_signed;
 
-        if (instr.conversion.src_size == Register::Size::Byte) {
-            const u32 offset = static_cast<u32>(instr.conversion.int_src.selector) * 8;
-            if (offset > 0) {
-                value = SignedOperation(OperationCode::ILogicalShiftRight, input_signed,
-                                        std::move(value), Immediate(offset));
+        if (const u32 offset = static_cast<u32>(instr.conversion.int_src.selector); offset > 0) {
+            ASSERT(instr.conversion.src_size == Register::Size::Byte ||
+                   instr.conversion.src_size == Register::Size::Short);
+            if (instr.conversion.src_size == Register::Size::Short) {
+                ASSERT(offset == 0 || offset == 2);
             }
-        } else {
-            UNIMPLEMENTED_IF(instr.conversion.int_src.selector != 0);
+            value = SignedOperation(OperationCode::ILogicalShiftRight, input_signed,
+                                    std::move(value), Immediate(offset * 8));
         }
 
         value = ConvertIntegerSize(value, instr.conversion.src_size, input_signed);
diff --git a/src/video_core/shader/decode/texture.cpp b/src/video_core/shader/decode/texture.cpp
index 351c8c2f1..bee7d8cad 100644
--- a/src/video_core/shader/decode/texture.cpp
+++ b/src/video_core/shader/decode/texture.cpp
@@ -522,68 +522,53 @@ Node4 ShaderIR::GetTextureCode(Instruction instr, TextureType texture_type,
                                Node array, Node depth_compare, u32 bias_offset,
                                std::vector<Node> aoffi,
                                std::optional<Tegra::Shader::Register> bindless_reg) {
-    const auto is_array = static_cast<bool>(array);
-    const auto is_shadow = static_cast<bool>(depth_compare);
+    const bool is_array = array != nullptr;
+    const bool is_shadow = depth_compare != nullptr;
     const bool is_bindless = bindless_reg.has_value();
 
-    UNIMPLEMENTED_IF_MSG((texture_type == TextureType::Texture3D && (is_array || is_shadow)) ||
-                             (texture_type == TextureType::TextureCube && is_array && is_shadow),
-                         "This method is not supported.");
+    UNIMPLEMENTED_IF(texture_type == TextureType::TextureCube && is_array && is_shadow);
+    ASSERT_MSG(texture_type != TextureType::Texture3D || !is_array || !is_shadow,
+               "Illegal texture type");
 
     const SamplerInfo info{texture_type, is_array, is_shadow, false};
-    Node index_var{};
+    Node index_var;
     const Sampler* sampler = is_bindless ? GetBindlessSampler(*bindless_reg, index_var, info)
                                          : GetSampler(instr.sampler, info);
-    Node4 values;
-    if (sampler == nullptr) {
-        for (u32 element = 0; element < values.size(); ++element) {
-            values[element] = Immediate(0);
-        }
-        return values;
+    if (!sampler) {
+        return {Immediate(0), Immediate(0), Immediate(0), Immediate(0)};
     }
 
     const bool lod_needed = process_mode == TextureProcessMode::LZ ||
                             process_mode == TextureProcessMode::LL ||
                             process_mode == TextureProcessMode::LLA;
-
-    // LOD selection (either via bias or explicit textureLod) not supported in GL for
-    // sampler2DArrayShadow and samplerCubeArrayShadow.
-    const bool gl_lod_supported =
-        !((texture_type == Tegra::Shader::TextureType::Texture2D && is_array && is_shadow) ||
-          (texture_type == Tegra::Shader::TextureType::TextureCube && is_array && is_shadow));
-
-    const OperationCode read_method =
-        (lod_needed && gl_lod_supported) ? OperationCode::TextureLod : OperationCode::Texture;
-
-    UNIMPLEMENTED_IF(process_mode != TextureProcessMode::None && !gl_lod_supported);
+    const OperationCode opcode = lod_needed ? OperationCode::TextureLod : OperationCode::Texture;
 
     Node bias;
     Node lod;
-    if (process_mode != TextureProcessMode::None && gl_lod_supported) {
-        switch (process_mode) {
-        case TextureProcessMode::LZ:
-            lod = Immediate(0.0f);
-            break;
-        case TextureProcessMode::LB:
-            // If present, lod or bias are always stored in the register
-            // indexed by the gpr20 field with an offset depending on the
-            // usage of the other registers
-            bias = GetRegister(instr.gpr20.Value() + bias_offset);
-            break;
-        case TextureProcessMode::LL:
-            lod = GetRegister(instr.gpr20.Value() + bias_offset);
-            break;
-        default:
-            UNIMPLEMENTED_MSG("Unimplemented process mode={}", static_cast<u32>(process_mode));
-            break;
-        }
+    switch (process_mode) {
+    case TextureProcessMode::None:
+        break;
+    case TextureProcessMode::LZ:
+        lod = Immediate(0.0f);
+        break;
+    case TextureProcessMode::LB:
+        // If present, lod or bias are always stored in the register indexed by the gpr20 field with
+        // an offset depending on the usage of the other registers.
+        bias = GetRegister(instr.gpr20.Value() + bias_offset);
+        break;
+    case TextureProcessMode::LL:
+        lod = GetRegister(instr.gpr20.Value() + bias_offset);
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Unimplemented process mode={}", static_cast<u32>(process_mode));
+        break;
     }
 
+    Node4 values;
     for (u32 element = 0; element < values.size(); ++element) {
-        auto copy_coords = coords;
         MetaTexture meta{*sampler, array, depth_compare, aoffi,    {}, {}, bias,
                          lod,      {},    element,       index_var};
-        values[element] = Operation(read_method, meta, std::move(copy_coords));
+        values[element] = Operation(opcode, meta, coords);
     }
 
     return values;
diff --git a/src/video_core/texture_cache/surface_base.cpp b/src/video_core/texture_cache/surface_base.cpp
index 84469b7ba..002df414f 100644
--- a/src/video_core/texture_cache/surface_base.cpp
+++ b/src/video_core/texture_cache/surface_base.cpp
@@ -277,6 +277,10 @@ void SurfaceBaseImpl::FlushBuffer(Tegra::MemoryManager& memory_manager,
             SwizzleFunc(MortonSwizzleMode::LinearToMorton, host_ptr, params,
                         staging_buffer.data() + host_offset, level);
         }
+    } else if (params.IsBuffer()) {
+        // Buffers don't have pitch or any fancy layout property. We can just memcpy them to guest
+        // memory.
+        std::memcpy(host_ptr, staging_buffer.data(), guest_memory_size);
     } else {
         ASSERT(params.target == SurfaceTarget::Texture2D);
         ASSERT(params.num_levels == 1);
diff --git a/src/video_core/texture_cache/surface_params.cpp b/src/video_core/texture_cache/surface_params.cpp
index 38b3a4ba8..f00839313 100644
--- a/src/video_core/texture_cache/surface_params.cpp
+++ b/src/video_core/texture_cache/surface_params.cpp
@@ -84,19 +84,16 @@ SurfaceParams SurfaceParams::CreateForTexture(const FormatLookupTable& lookup_ta
     if (entry.IsShadow() && params.type == SurfaceType::ColorTexture) {
         switch (params.pixel_format) {
         case PixelFormat::R16U:
-        case PixelFormat::R16F: {
+        case PixelFormat::R16F:
             params.pixel_format = PixelFormat::Z16;
             break;
-        }
-        case PixelFormat::R32F: {
+        case PixelFormat::R32F:
             params.pixel_format = PixelFormat::Z32F;
             break;
-        }
-        default: {
+        default:
             UNIMPLEMENTED_MSG("Unimplemented shadow convert format: {}",
                               static_cast<u32>(params.pixel_format));
         }
-        }
         params.type = GetFormatType(params.pixel_format);
     }
     params.type = GetFormatType(params.pixel_format);
@@ -168,27 +165,29 @@ SurfaceParams SurfaceParams::CreateForImage(const FormatLookupTable& lookup_tabl
     return params;
 }
 
-SurfaceParams SurfaceParams::CreateForDepthBuffer(
-    Core::System& system, u32 zeta_width, u32 zeta_height, Tegra::DepthFormat format,
-    u32 block_width, u32 block_height, u32 block_depth,
-    Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type) {
+SurfaceParams SurfaceParams::CreateForDepthBuffer(Core::System& system) {
+    const auto& regs = system.GPU().Maxwell3D().regs;
+    regs.zeta_width, regs.zeta_height, regs.zeta.format, regs.zeta.memory_layout.type;
     SurfaceParams params;
-    params.is_tiled = type == Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear;
+    params.is_tiled = regs.zeta.memory_layout.type ==
+                      Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout::BlockLinear;
     params.srgb_conversion = false;
-    params.block_width = std::min(block_width, 5U);
-    params.block_height = std::min(block_height, 5U);
-    params.block_depth = std::min(block_depth, 5U);
+    params.block_width = std::min(regs.zeta.memory_layout.block_width.Value(), 5U);
+    params.block_height = std::min(regs.zeta.memory_layout.block_height.Value(), 5U);
+    params.block_depth = std::min(regs.zeta.memory_layout.block_depth.Value(), 5U);
     params.tile_width_spacing = 1;
-    params.pixel_format = PixelFormatFromDepthFormat(format);
+    params.pixel_format = PixelFormatFromDepthFormat(regs.zeta.format);
     params.type = GetFormatType(params.pixel_format);
-    params.width = zeta_width;
-    params.height = zeta_height;
-    params.target = SurfaceTarget::Texture2D;
-    params.depth = 1;
+    params.width = regs.zeta_width;
+    params.height = regs.zeta_height;
     params.pitch = 0;
     params.num_levels = 1;
     params.emulated_levels = 1;
-    params.is_layered = false;
+
+    const bool is_layered = regs.zeta_layers > 1 && params.block_depth == 0;
+    params.is_layered = is_layered;
+    params.target = is_layered ? SurfaceTarget::Texture2DArray : SurfaceTarget::Texture2D;
+    params.depth = is_layered ? regs.zeta_layers.Value() : 1U;
     return params;
 }
 
@@ -214,11 +213,13 @@ SurfaceParams SurfaceParams::CreateForFramebuffer(Core::System& system, std::siz
         params.width = params.pitch / bpp;
     }
     params.height = config.height;
-    params.depth = 1;
-    params.target = SurfaceTarget::Texture2D;
     params.num_levels = 1;
     params.emulated_levels = 1;
-    params.is_layered = false;
+
+    const bool is_layered = config.layers > 1 && params.block_depth == 0;
+    params.is_layered = is_layered;
+    params.depth = is_layered ? config.layers.Value() : 1;
+    params.target = is_layered ? SurfaceTarget::Texture2DArray : SurfaceTarget::Texture2D;
     return params;
 }
 
diff --git a/src/video_core/texture_cache/surface_params.h b/src/video_core/texture_cache/surface_params.h
index 9256fd6d9..995cc3818 100644
--- a/src/video_core/texture_cache/surface_params.h
+++ b/src/video_core/texture_cache/surface_params.h
@@ -35,10 +35,7 @@ public:
                                         const VideoCommon::Shader::Image& entry);
 
     /// Creates SurfaceCachedParams for a depth buffer configuration.
-    static SurfaceParams CreateForDepthBuffer(
-        Core::System& system, u32 zeta_width, u32 zeta_height, Tegra::DepthFormat format,
-        u32 block_width, u32 block_height, u32 block_depth,
-        Tegra::Engines::Maxwell3D::Regs::InvMemoryLayout type);
+    static SurfaceParams CreateForDepthBuffer(Core::System& system);
 
     /// Creates SurfaceCachedParams from a framebuffer configuration.
     static SurfaceParams CreateForFramebuffer(Core::System& system, std::size_t index);
diff --git a/src/video_core/texture_cache/texture_cache.h b/src/video_core/texture_cache/texture_cache.h
index f4c015635..c70e4aec2 100644
--- a/src/video_core/texture_cache/texture_cache.h
+++ b/src/video_core/texture_cache/texture_cache.h
@@ -160,10 +160,7 @@ public:
             SetEmptyDepthBuffer();
             return {};
         }
-        const auto depth_params{SurfaceParams::CreateForDepthBuffer(
-            system, regs.zeta_width, regs.zeta_height, regs.zeta.format,
-            regs.zeta.memory_layout.block_width, regs.zeta.memory_layout.block_height,
-            regs.zeta.memory_layout.block_depth, regs.zeta.memory_layout.type)};
+        const auto depth_params{SurfaceParams::CreateForDepthBuffer(system)};
         auto surface_view = GetSurface(gpu_addr, cache_addr, depth_params, preserve_contents, true);
         if (depth_buffer.target)
             depth_buffer.target->MarkAsRenderTarget(false, NO_RT);
@@ -721,7 +718,6 @@ private:
     std::pair<TSurface, TView> GetSurface(const GPUVAddr gpu_addr, const CacheAddr cache_addr,
                                           const SurfaceParams& params, bool preserve_contents,
                                           bool is_render) {
-
         // Step 1
         // Check Level 1 Cache for a fast structural match. If candidate surface
         // matches at certain level we are pretty much done.
@@ -733,14 +729,18 @@ private:
                 return RecycleSurface(overlaps, params, gpu_addr, preserve_contents,
                                       topological_result);
             }
+
             const auto struct_result = current_surface->MatchesStructure(params);
-            if (struct_result != MatchStructureResult::None &&
-                (params.target != SurfaceTarget::Texture3D ||
-                 current_surface->MatchTarget(params.target))) {
-                if (struct_result == MatchStructureResult::FullMatch) {
-                    return ManageStructuralMatch(current_surface, params, is_render);
-                } else {
-                    return RebuildSurface(current_surface, params, is_render);
+            if (struct_result != MatchStructureResult::None) {
+                const auto& old_params = current_surface->GetSurfaceParams();
+                const bool not_3d = params.target != SurfaceTarget::Texture3D &&
+                                    old_params.target != SurfaceTarget::Texture3D;
+                if (not_3d || current_surface->MatchTarget(params.target)) {
+                    if (struct_result == MatchStructureResult::FullMatch) {
+                        return ManageStructuralMatch(current_surface, params, is_render);
+                    } else {
+                        return RebuildSurface(current_surface, params, is_render);
+                    }
                 }
             }
         }
diff --git a/src/web_service/web_backend.cpp b/src/web_service/web_backend.cpp
index 6683f459f..737ffe409 100644
--- a/src/web_service/web_backend.cpp
+++ b/src/web_service/web_backend.cpp
@@ -73,14 +73,12 @@ struct Client::Impl {
                 if (!parsedUrl.GetPort(&port)) {
                     port = HTTP_PORT;
                 }
-                cli = std::make_unique<httplib::Client>(parsedUrl.m_Host.c_str(), port,
-                                                        TIMEOUT_SECONDS);
+                cli = std::make_unique<httplib::Client>(parsedUrl.m_Host.c_str(), port);
             } else if (parsedUrl.m_Scheme == "https") {
                 if (!parsedUrl.GetPort(&port)) {
                     port = HTTPS_PORT;
                 }
-                cli = std::make_unique<httplib::SSLClient>(parsedUrl.m_Host.c_str(), port,
-                                                           TIMEOUT_SECONDS);
+                cli = std::make_unique<httplib::SSLClient>(parsedUrl.m_Host.c_str(), port);
             } else {
                 LOG_ERROR(WebService, "Bad URL scheme {}", parsedUrl.m_Scheme);
                 return Common::WebResult{Common::WebResult::Code::InvalidURL, "Bad URL scheme"};
@@ -90,6 +88,7 @@ struct Client::Impl {
             LOG_ERROR(WebService, "Invalid URL {}", host + path);
             return Common::WebResult{Common::WebResult::Code::InvalidURL, "Invalid URL"};
         }
+        cli->set_timeout_sec(TIMEOUT_SECONDS);
 
         httplib::Headers params;
         if (!jwt.empty()) {
diff --git a/src/yuzu/configuration/config.cpp b/src/yuzu/configuration/config.cpp
index cd94693c1..6209fff75 100644
--- a/src/yuzu/configuration/config.cpp
+++ b/src/yuzu/configuration/config.cpp
@@ -630,6 +630,7 @@ void Config::ReadRendererValues() {
     Settings::values.vulkan_device = ReadSetting(QStringLiteral("vulkan_device"), 0).toInt();
     Settings::values.resolution_factor =
         ReadSetting(QStringLiteral("resolution_factor"), 1.0).toFloat();
+    Settings::values.aspect_ratio = ReadSetting(QStringLiteral("aspect_ratio"), 0).toInt();
     Settings::values.use_frame_limit =
         ReadSetting(QStringLiteral("use_frame_limit"), true).toBool();
     Settings::values.frame_limit = ReadSetting(QStringLiteral("frame_limit"), 100).toInt();
@@ -1064,6 +1065,7 @@ void Config::SaveRendererValues() {
     WriteSetting(QStringLiteral("vulkan_device"), Settings::values.vulkan_device, 0);
     WriteSetting(QStringLiteral("resolution_factor"),
                  static_cast<double>(Settings::values.resolution_factor), 1.0);
+    WriteSetting(QStringLiteral("aspect_ratio"), Settings::values.aspect_ratio, 0);
     WriteSetting(QStringLiteral("use_frame_limit"), Settings::values.use_frame_limit, true);
     WriteSetting(QStringLiteral("frame_limit"), Settings::values.frame_limit, 100);
     WriteSetting(QStringLiteral("use_disk_shader_cache"), Settings::values.use_disk_shader_cache,
diff --git a/src/yuzu/configuration/configure_graphics.cpp b/src/yuzu/configuration/configure_graphics.cpp
index f57a24e36..ea899c080 100644
--- a/src/yuzu/configuration/configure_graphics.cpp
+++ b/src/yuzu/configuration/configure_graphics.cpp
@@ -97,6 +97,7 @@ void ConfigureGraphics::SetConfiguration() {
     ui->api->setCurrentIndex(static_cast<int>(Settings::values.renderer_backend));
     ui->resolution_factor_combobox->setCurrentIndex(
         static_cast<int>(FromResolutionFactor(Settings::values.resolution_factor)));
+    ui->aspect_ratio_combobox->setCurrentIndex(Settings::values.aspect_ratio);
     ui->use_disk_shader_cache->setEnabled(runtime_lock);
     ui->use_disk_shader_cache->setChecked(Settings::values.use_disk_shader_cache);
     ui->use_accurate_gpu_emulation->setChecked(Settings::values.use_accurate_gpu_emulation);
@@ -114,6 +115,7 @@ void ConfigureGraphics::ApplyConfiguration() {
     Settings::values.vulkan_device = vulkan_device;
     Settings::values.resolution_factor =
         ToResolutionFactor(static_cast<Resolution>(ui->resolution_factor_combobox->currentIndex()));
+    Settings::values.aspect_ratio = ui->aspect_ratio_combobox->currentIndex();
     Settings::values.use_disk_shader_cache = ui->use_disk_shader_cache->isChecked();
     Settings::values.use_accurate_gpu_emulation = ui->use_accurate_gpu_emulation->isChecked();
     Settings::values.use_asynchronous_gpu_emulation =
diff --git a/src/yuzu/configuration/configure_graphics.ui b/src/yuzu/configuration/configure_graphics.ui
index e24372204..db60426ab 100644
--- a/src/yuzu/configuration/configure_graphics.ui
+++ b/src/yuzu/configuration/configure_graphics.ui
@@ -139,6 +139,41 @@
          </layout>
         </item>
         <item>
+         <layout class="QHBoxLayout" name="horizontalLayout_6">
+          <item>
+           <widget class="QLabel" name="ar_label">
+            <property name="text">
+             <string>Aspect Ratio:</string>
+            </property>
+           </widget>
+          </item>
+          <item>
+           <widget class="QComboBox" name="aspect_ratio_combobox">
+            <item>
+             <property name="text">
+              <string>Default (16:9)</string>
+             </property>
+            </item>
+            <item>
+             <property name="text">
+              <string>Force 4:3</string>
+             </property>
+            </item>
+            <item>
+             <property name="text">
+              <string>Force 21:9</string>
+             </property>
+            </item>
+            <item>
+             <property name="text">
+              <string>Stretch to Window</string>
+             </property>
+            </item>
+           </widget>
+          </item>
+         </layout>
+        </item>
+        <item>
          <layout class="QHBoxLayout" name="horizontalLayout_3">
           <item>
            <widget class="QLabel" name="bg_label">
diff --git a/src/yuzu_cmd/config.cpp b/src/yuzu_cmd/config.cpp
index b01a36023..96f1ce3af 100644
--- a/src/yuzu_cmd/config.cpp
+++ b/src/yuzu_cmd/config.cpp
@@ -379,6 +379,8 @@ void Config::ReadValues() {
 
     Settings::values.resolution_factor =
         static_cast<float>(sdl2_config->GetReal("Renderer", "resolution_factor", 1.0));
+    Settings::values.aspect_ratio =
+        static_cast<int>(sdl2_config->GetInteger("Renderer", "aspect_ratio", 0));
     Settings::values.use_frame_limit = sdl2_config->GetBoolean("Renderer", "use_frame_limit", true);
     Settings::values.frame_limit =
         static_cast<u16>(sdl2_config->GetInteger("Renderer", "frame_limit", 100));
diff --git a/src/yuzu_cmd/default_ini.h b/src/yuzu_cmd/default_ini.h
index 00fd88279..8a2b658cd 100644
--- a/src/yuzu_cmd/default_ini.h
+++ b/src/yuzu_cmd/default_ini.h
@@ -122,6 +122,10 @@ use_shader_jit =
 # factor for the Switch resolution
 resolution_factor =
 
+# Aspect ratio
+# 0: Default (16:9), 1: Force 4:3, 2: Force 21:9, 3: Stretch to Window
+aspect_ratio =
+
 # Whether to enable V-Sync (caps the framerate at 60FPS) or not.
 # 0 (default): Off, 1: On
 use_vsync =
diff --git a/src/yuzu_tester/config.cpp b/src/yuzu_tester/config.cpp
index 84ab4d687..0ac93b62a 100644
--- a/src/yuzu_tester/config.cpp
+++ b/src/yuzu_tester/config.cpp
@@ -118,6 +118,8 @@ void Config::ReadValues() {
     // Renderer
     Settings::values.resolution_factor =
         static_cast<float>(sdl2_config->GetReal("Renderer", "resolution_factor", 1.0));
+    Settings::values.aspect_ratio =
+        static_cast<int>(sdl2_config->GetInteger("Renderer", "aspect_ratio", 0));
     Settings::values.use_frame_limit = false;
     Settings::values.frame_limit = 100;
     Settings::values.use_disk_shader_cache =
diff --git a/src/yuzu_tester/default_ini.h b/src/yuzu_tester/default_ini.h
index 9a3e86d68..8d93f7b88 100644
--- a/src/yuzu_tester/default_ini.h
+++ b/src/yuzu_tester/default_ini.h
@@ -26,6 +26,10 @@ use_shader_jit =
 # factor for the Switch resolution
 resolution_factor =
 
+# Aspect ratio
+# 0: Default (16:9), 1: Force 4:3, 2: Force 21:9, 3: Stretch to Window
+aspect_ratio =
+
 # Whether to enable V-Sync (caps the framerate at 60FPS) or not.
 # 0 (default): Off, 1: On
 use_vsync =