cfg: asio driver selector must list both 32-bit and 64-bit drivers (#734)

## Link to GitHub Issue or related Pull Request, if one exists
#730 

## Description of change
spicecfg is a 32-bit application. When it uses the ASIO SDK it only saw
the 32-bit drivers.

There are some ASIO drivers that have different names under 32-bit and
64-bit (Xonar AE is one of them)... so we have to manually scan the
registry and surface both the WOW32 and WOW64 nodes.

## Testing
<img width="329" height="309" alt="image"
src="https://github.com/user-attachments/assets/7abecf76-6835-4df5-8c2e-e7b425130f4c"
/>

Checked both 32-bit and 64-bit configurator, identical results.

src/spice2x/CMakeLists.txt

@@ -235,6 +235,13 @@ add_compile_definitions(
         _WIN32_IE=0x0400
 )
 
+# SPICE_XP: set by build_all.sh for the WinXP-compat toolchains. Gates out
+# the DX11 overlay backend (WinXP lacks D3DCOMPILER_47.dll).
+option(SPICE_XP "Build for the WinXP-compat toolchain" OFF)
+if(SPICE_XP)
+    add_compile_definitions(SPICE_XP=1)
+endif()
+
 # acioemu log
 #set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -DACIOEMU_LOG")
 
@@ -409,6 +416,7 @@ set(SOURCE_FILES ${SOURCE_FILES}
         games/nost/io.cpp
         games/nost/poke.cpp
         games/gitadora/gitadora.cpp
+        games/gitadora/asio.cpp
         games/gitadora/io.cpp
         games/gitadora/handle.cpp
         games/gitadora/j32d.cpp
@@ -486,6 +494,8 @@ set(SOURCE_FILES ${SOURCE_FILES}
         # hooks
         hooks/audio/acm.cpp
         hooks/audio/audio.cpp
+        hooks/audio/asio_driver_scan.cpp
+        hooks/audio/asio_proxy.cpp
         hooks/audio/buffer.cpp
         hooks/audio/mme.cpp
         hooks/audio/util.cpp
@@ -494,8 +504,12 @@ set(SOURCE_FILES ${SOURCE_FILES}
         hooks/audio/backends/mmdevice/device.cpp
         hooks/audio/backends/mmdevice/device_collection.cpp
         hooks/audio/backends/mmdevice/device_enumerator.cpp
+        hooks/audio/backends/mmdevice/null_device.cpp
+        hooks/audio/backends/mmdevice/null_discard_backend.cpp
         hooks/audio/backends/wasapi/audio_client.cpp
         hooks/audio/backends/wasapi/audio_render_client.cpp
+        hooks/audio/backends/wasapi/downmix.cpp
+        hooks/audio/backends/wasapi/resample.cpp
         hooks/audio/backends/wasapi/dummy_audio_client.cpp
         hooks/audio/backends/wasapi/dummy_audio_clock.cpp
         hooks/audio/backends/wasapi/dummy_audio_render_client.cpp
@@ -517,12 +531,19 @@ set(SOURCE_FILES ${SOURCE_FILES}
         hooks/graphics/backends/d3d9/d3d9_fake_swapchain.cpp
         hooks/graphics/backends/d3d9/d3d9_swapchain.cpp
         hooks/graphics/backends/d3d9/d3d9_texture.cpp
+        hooks/graphics/backends/d3d11/d3d11_backend.cpp
+        hooks/graphics/backends/d3d11/d3d11_swapchain.cpp
+        hooks/graphics/backends/d3d11/d3d11_factory.cpp
+        hooks/graphics/backends/d3d11/d3d11_vtable_capture.cpp
+        hooks/graphics/backends/d3d11/d3d11_screenshot.cpp
         hooks/input/dinput8/fake_backend.cpp
         hooks/input/dinput8/fake_device.cpp
         hooks/input/dinput8/hook.cpp
         hooks/lang.cpp
         hooks/libraryhook.cpp
         hooks/networkhook.cpp
+        hooks/icmphook_net.cpp
+        hooks/icmphook_iphlpapi.cpp
         hooks/powrprof.cpp
         #hooks/rom.cpp
         hooks/setupapihook.cpp
@@ -556,6 +577,7 @@ set(SOURCE_FILES ${SOURCE_FILES}
 
         # overlay
         overlay/overlay.cpp
+        overlay/notifications.cpp
         overlay/window.cpp
         overlay/imgui/extensions.cpp
         overlay/imgui/impl_spice.cpp
@@ -607,6 +629,9 @@ set(SOURCE_FILES ${SOURCE_FILES}
         reader/structuredmessage.cpp
         reader/crypt.cpp
 
+        # sdk
+        sdk/sdk.cpp
+
         # stubs
         stubs/stubs.cpp
 
@@ -633,6 +658,7 @@ set(SOURCE_FILES ${SOURCE_FILES}
         util/time.cpp
         util/cpuutils.cpp
         util/netutils.cpp
+        util/precise_timer.cpp
         util/sysutils.cpp
         util/lz77.cpp
         util/tapeled.cpp
@@ -707,6 +733,10 @@ target_link_libraries(spicetools_spice64 PUBLIC winscard)
 set_target_properties(spicetools_spice64 PROPERTIES PREFIX "")
 set_target_properties(spicetools_spice64 PROPERTIES OUTPUT_NAME "spice64")
 target_compile_definitions(spicetools_spice64 PRIVATE SPICE64=1)
+if(NOT SPICE_XP)
+    target_link_libraries(spicetools_spice64 PUBLIC d3d11 dxgi d3dcompiler)
+    target_compile_definitions(spicetools_spice64 PRIVATE SPICE_D3D11=1)
+endif()
 
 IF(NOT MSVC)
     set_target_properties(spicetools_spice64 PROPERTIES COMPILE_FLAGS "-m64" LINK_FLAGS "-m64")
@@ -726,6 +756,9 @@ set_target_properties(spicetools_spice64_linux PROPERTIES PREFIX "")
 set_target_properties(spicetools_spice64_linux PROPERTIES OUTPUT_NAME "spice64_linux")
 target_compile_definitions(spicetools_spice64_linux PRIVATE SPICE64=1)
 target_compile_definitions(spicetools_spice64_linux PRIVATE NO_SCARD=1 PRIVATE SPICE_LINUX=1)
+# spice64_linux is never built under the WinXP toolchain, so dx11 is always on.
+target_link_libraries(spicetools_spice64_linux PUBLIC d3d11 dxgi d3dcompiler)
+target_compile_definitions(spicetools_spice64_linux PRIVATE SPICE_D3D11=1)
 
 IF(NOT MSVC)
     set_target_properties(spicetools_spice64_linux PROPERTIES COMPILE_FLAGS "-m64" LINK_FLAGS "-m64")
@@ -738,7 +771,7 @@ set(SOURCE_FILES ${SOURCE_FILES} launcher/options.h launcher/options.cpp)
 set(RESOURCE_FILES cfg/manifest.manifest cfg/manifest.rc cfg/icon.rc cfg/Win32D.rc)
 add_executable(spicetools_cfg WIN32 ${SOURCE_FILES} ${RESOURCE_FILES})
 target_link_libraries(spicetools_cfg
-        PUBLIC ws2_32 version comctl32 shlwapi iphlpapi hid secur32 setupapi psapi winmm winhttp strmiids
+        PUBLIC d3d9 ws2_32 version comctl32 shlwapi iphlpapi hid secur32 setupapi psapi winmm winhttp strmiids
         PRIVATE fmt::fmt-header-only discord-rpc imgui hash-library minhook imm32 dwmapi CpuFeatures::cpu_features)
 target_link_libraries(spicetools_cfg PUBLIC winscard)
 set_target_properties(spicetools_cfg PROPERTIES PREFIX "")
@@ -756,7 +789,7 @@ set(SOURCE_FILES ${SOURCE_FILES} launcher/options.h launcher/options.cpp)
 set(RESOURCE_FILES cfg/manifest.manifest cfg/manifest.rc cfg/icon.rc cfg/Win32D.rc)
 add_executable(spicetools_cfg_linux WIN32 ${SOURCE_FILES} ${RESOURCE_FILES})
 target_link_libraries(spicetools_cfg_linux
-        PUBLIC ws2_32 version comctl32 shlwapi iphlpapi hid secur32 setupapi psapi winmm winhttp strmiids
+        PUBLIC d3d9 ws2_32 version comctl32 shlwapi iphlpapi hid secur32 setupapi psapi winmm winhttp strmiids
         PRIVATE fmt::fmt-header-only discord-rpc imgui hash-library minhook imm32 dwmapi CpuFeatures::cpu_features)
 set_target_properties(spicetools_cfg_linux PROPERTIES PREFIX "")
 set_target_properties(spicetools_cfg_linux PROPERTIES OUTPUT_NAME "spicecfg_linux")
@@ -860,6 +893,36 @@ if(NOT MSVC)
     set_target_properties(spicetools_stubs_cpusbxpkm PROPERTIES COMPILE_FLAGS "-m32" LINK_FLAGS "-m32")
 endif()
 
+# sdk_sample_v0_flat_c.dll (32 bit)
+set(SOURCE_FILES sdk/sample/v0/flat_c/v0_flat_c.c)
+add_library(spicetools_sdk_sample_v0_flat_c_32 SHARED ${SOURCE_FILES} ${RESOURCE_FILES} sdk/sample/v0/flat_c/v0_flat_c.def)
+set_target_properties(spicetools_sdk_sample_v0_flat_c_32 PROPERTIES PREFIX "")
+set_target_properties(spicetools_sdk_sample_v0_flat_c_32 PROPERTIES OUTPUT_NAME "sdk_sample_v0_flat_c")
+
+if(NOT MSVC)
+    set_target_properties(spicetools_sdk_sample_v0_flat_c_32 PROPERTIES COMPILE_FLAGS "-m32" LINK_FLAGS "-m32")
+endif()
+
+# sdk_sample_v0_flat_c.dll (64 bit)
+set(SOURCE_FILES sdk/sample/v0/flat_c/v0_flat_c.c)
+add_library(spicetools_sdk_sample_v0_flat_c_64 SHARED ${SOURCE_FILES} ${RESOURCE_FILES} sdk/sample/v0/flat_c/v0_flat_c.def)
+set_target_properties(spicetools_sdk_sample_v0_flat_c_64 PROPERTIES PREFIX "")
+set_target_properties(spicetools_sdk_sample_v0_flat_c_64 PROPERTIES OUTPUT_NAME "sdk_sample_v0_flat_c")
+
+if(NOT MSVC)
+    set_target_properties(spicetools_sdk_sample_v0_flat_c_64 PROPERTIES COMPILE_FLAGS "-m64" LINK_FLAGS "-m64")
+endif()
+
+# sdk_sample_v0_cpp.dll (64 bit)
+set(SOURCE_FILES sdk/sample/v0/cpp/v0_cpp.cpp)
+add_library(spicetools_sdk_sample_v0_cpp_64 SHARED ${SOURCE_FILES} ${RESOURCE_FILES} sdk/sample/v0/cpp/v0_cpp.def)
+set_target_properties(spicetools_sdk_sample_v0_cpp_64 PROPERTIES PREFIX "")
+set_target_properties(spicetools_sdk_sample_v0_cpp_64 PROPERTIES OUTPUT_NAME "sdk_sample_v0_cpp")
+
+if(NOT MSVC)
+    set_target_properties(spicetools_sdk_sample_v0_cpp_64 PROPERTIES COMPILE_FLAGS "-m64" LINK_FLAGS "-m64")
+endif()
+
 # output directories
 ####################
 
@@ -869,14 +932,14 @@ set_target_properties(spicetools_cfg spicetools_cfg_linux
         RUNTIME_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/spicetools")
 
 # output 32bit
-set_target_properties(spicetools_spice spicetools_spice_laa spicetools_spice_linux spicetools_stubs_kbt spicetools_stubs_kld spicetools_stubs_cpusbxpkm
+set_target_properties(spicetools_spice spicetools_spice_laa spicetools_spice_linux spicetools_stubs_kbt spicetools_stubs_kld spicetools_stubs_cpusbxpkm spicetools_sdk_sample_v0_flat_c_32
         PROPERTIES
         ARCHIVE_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/archive32"
         LIBRARY_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/spicetools/32"
         RUNTIME_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/spicetools/32")
 
 # output 64bit
-set_target_properties(spicetools_spice64 spicetools_spice64_linux spicetools_stubs_kbt64 spicetools_stubs_kld64 spicetools_stubs_nvcuda spicetools_stubs_nvcuvid spicetools_stubs_nvEncodeAPI64
+set_target_properties(spicetools_spice64 spicetools_spice64_linux spicetools_stubs_kbt64 spicetools_stubs_kld64 spicetools_stubs_nvcuda spicetools_stubs_nvcuvid spicetools_stubs_nvEncodeAPI64 spicetools_sdk_sample_v0_flat_c_64 spicetools_sdk_sample_v0_cpp_64
         PROPERTIES
         ARCHIVE_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/archive64"
         LIBRARY_OUTPUT_DIRECTORY "${CMAKE_BINARY_DIR}/spicetools/64"

src/spice2x/README.md

@@ -283,6 +283,16 @@ Python is the only one that is fully spec compliant.
 Other libraries may be missing features and contain bugs; please feel free to
 contribute code to fill the gaps if you work on a project using these libraries.
 
+## Spice SDK (for DLL hooks / plugins)
+
+As an alternative to Spice API, Spice SDK is available.
+
+This is a flat C, header-only library that can be included in your DLL. Once
+initialized, you can call directly into helper routines provided by Spicetools
+executable, making low-latency and high-throughput interactions possible.
+
+Check the documentation on the wiki for more details.
+
 ## License
 Unless otherwise noted, all files are licensed under the GPLv3.
 See the LICENSE file for the full license text.

src/spice2x/acio/mdxf/mdxf.cpp

@@ -1,13 +1,16 @@
 #include "mdxf.h"
 #include "mdxf_poll.h"
 
+#include <mutex>
+
 #include "avs/game.h"
+#include "games/ddr/ddr.h"
 #include "games/ddr/io.h"
 #include "launcher/launcher.h"
 #include "rawinput/rawinput.h"
 #include "util/logging.h"
+#include "util/precise_timer.h"
 #include "util/utils.h"
-#include <mutex>
 
 #define MDFX_DEBUG_VERBOSE 0
 
@@ -108,9 +111,10 @@ static void mdxf_thread_start() {
     MDXF_THREAD = std::thread([] {
         SetThreadPriority(GetCurrentThread(), THREAD_PRIORITY_BELOW_NORMAL);
 
+        timeutils::PreciseSleepTimer timer;
         while (MDXF_THREAD_RUNNING.load(std::memory_order_acquire)) {
             mdxf_poll(false);
-            std::this_thread::sleep_for(THREAD_PERIOD);
+            timer.sleep(THREAD_PERIOD);
         }
     });
 }
@@ -357,14 +361,17 @@ static bool __cdecl ac_io_mdxf_update_control_status_buffer_impl(int node, MDXFP
 
         // decide on button map
         const size_t *button_map = nullptr;
+        int player = 0;
         switch (node) {
             case 17:
             case 25:
                 button_map = &buttons_p1[0];
+                player = 1;
                 break;
             case 18:
             case 26:
                 button_map = &buttons_p2[0];
+                player = 2;
                 break;
         }
         
@@ -374,19 +381,27 @@ static bool __cdecl ac_io_mdxf_update_control_status_buffer_impl(int node, MDXFP
         if (source == EXTERNAL_POLL) {
             // get buttons
             auto &buttons = games::ddr::get_buttons();
+            
+            // get analogs
+            bool analog_left = false;
+            bool analog_right = false;
+            games::ddr::get_analog_x_axis(player, analog_left, analog_right);
+            bool analog_up = false;
+            bool analog_down = false;
+            games::ddr::get_analog_y_axis(player, analog_up, analog_down);
+            
             uint8_t up_down = 0;
             uint8_t left_right = 0;
-            
-            if (GameAPI::Buttons::getState(RI_MGR, buttons.at(button_map[0]))) {
+            if (GameAPI::Buttons::getState(RI_MGR, buttons.at(button_map[0])) || analog_up) {
                 up_down |= 0xF0;
             }
-            if (GameAPI::Buttons::getState(RI_MGR, buttons.at(button_map[1]))) {
+            if (GameAPI::Buttons::getState(RI_MGR, buttons.at(button_map[1])) || analog_down) {
                 up_down |= 0x0F;
             }
-            if (GameAPI::Buttons::getState(RI_MGR, buttons.at(button_map[2]))) {
+            if (GameAPI::Buttons::getState(RI_MGR, buttons.at(button_map[2])) || analog_left) {
                 left_right |= 0xF0;
             }
-            if (GameAPI::Buttons::getState(RI_MGR, buttons.at(button_map[3]))) {
+            if (GameAPI::Buttons::getState(RI_MGR, buttons.at(button_map[3])) || analog_right) {
                 left_right |= 0x0F;
             }
             current_state = (uint16_t(up_down) << 8) | left_right;
@@ -500,4 +515,4 @@ acio::MDXFModule::~MDXFModule() {
     if (IS_THREAD_NEEDED) {
         mdxf_thread_stop();
     }
-}
+}

src/spice2x/acioemu/device.h

@@ -1,5 +1,6 @@
 #pragma once
 
+#include <cstdint>
 #include <string>
 
 #include "util/circular_buffer.h"

src/spice2x/acioemu/icca.cpp

@@ -5,6 +5,7 @@
 #include "games/sdvx/sdvx.h"
 #include "misc/eamuse.h"
 #include "util/logging.h"
+#include "util/precise_timer.h"
 #include "util/utils.h"
 
 using namespace acioemu;
@@ -46,11 +47,12 @@ ICCADevice::ICCADevice(bool flip_order, bool keypad_thread, uint8_t node_count)
     // keypad thread for faster polling
     if (keypad_thread) {
         this->keypad_thread = new std::thread([this]() {
+            timeutils::PreciseSleepTimer timer;
             while (this->cards) {
                 for (int unit = 0; unit < this->node_count; unit++) {
                     this->update_keypad(unit, false);
                 }
-                Sleep(7);
+                timer.sleep(7);
             }
         });
     }

src/spice2x/api/controller.cpp

@@ -11,6 +11,8 @@
 #include "util/logging.h"
 #include "util/utils.h"
 
+#include "overlay/notifications.h"
+
 #include "module.h"
 #include "modules/analogs.h"
 #include "modules/buttons.h"
@@ -197,6 +199,9 @@ void Controller::connection_handler(api::ClientState client_state) {
 
     // log connection
     log_info("api", "client connected: {}", client_address_str);
+    overlay::notifications::add(
+            overlay::notifications::Severity::Success,
+            fmt::format("API client connected ({})", client_address_str));
     client_states_m.lock();
     client_states.emplace_back(&client_state);
     client_states_m.unlock();
@@ -277,6 +282,9 @@ void Controller::connection_handler(api::ClientState client_state) {
 
     // log disconnect
     log_info("api", "client disconnected: {}", client_address_str);
+    overlay::notifications::add(
+            overlay::notifications::Severity::Info,
+            fmt::format("API client disconnected ({})", client_address_str));
     client_states_m.lock();
     client_states.erase(std::remove(client_states.begin(), client_states.end(), &client_state));
     client_states_m.unlock();

src/spice2x/api/modules/analogs.cpp

@@ -6,6 +6,7 @@
 #include "launcher/launcher.h"
 #include "games/io.h"
 #include "util/utils.h"
+#include "acio/mdxf/mdxf_poll.h"
 
 using namespace std::placeholders;
 using namespace rapidjson;
@@ -104,6 +105,7 @@ namespace api::modules {
                 for (auto &analog : *this->analogs) {
                     analog.override_enabled = false;
                 }
+                mdxf_poll(true);
             }
             return;
         }
@@ -148,6 +150,7 @@ namespace api::modules {
             if (analog.getName() == name) {
                 analog.override_state = CLAMP(state, 0.f, 1.f);
                 analog.override_enabled = true;
+                mdxf_poll(true);
                 return true;
             }
         }
@@ -167,6 +170,7 @@ namespace api::modules {
         for (auto &analog : *this->analogs) {
             if (analog.getName() == name) {
                 analog.override_enabled = false;
+                mdxf_poll(true);
                 return true;
             }
         }

src/spice2x/api/modules/buttons.cpp

@@ -6,6 +6,7 @@
 #include "launcher/launcher.h"
 #include "games/io.h"
 #include "util/utils.h"
+#include "acio/mdxf/mdxf_poll.h"
 
 using namespace std::placeholders;
 using namespace rapidjson;
@@ -107,6 +108,7 @@ namespace api::modules {
                 for (auto &button : *this->buttons) {
                     button.override_enabled = false;
                 }
+                mdxf_poll(true);
             }
             return;
         }
@@ -153,6 +155,7 @@ namespace api::modules {
                         GameAPI::Buttons::BUTTON_PRESSED : GameAPI::Buttons::BUTTON_NOT_PRESSED;
                 button.override_velocity = CLAMP(state, 0.f, 1.f);
                 button.override_enabled = true;
+                mdxf_poll(true);
                 return true;
             }
         }
@@ -172,6 +175,7 @@ namespace api::modules {
         for (auto &button : *this->buttons) {
             if (button.getName() == name) {
                 button.override_enabled = false;
+                mdxf_poll(true);
                 return true;
             }
         }

src/spice2x/api/modules/keypads.cpp

@@ -7,6 +7,7 @@
 #include "avs/game.h"
 #include "external/rapidjson/document.h"
 #include "misc/eamuse.h"
+#include "util/precise_timer.h"
 
 using namespace std::placeholders;
 using namespace rapidjson;
@@ -58,6 +59,7 @@ namespace api::modules {
         // get params
         auto keypad = req.params[0].GetUint();
         auto input = std::string(req.params[1].GetString());
+        timeutils::PreciseSleepTimer timer;
 
         // process all chars
         for (auto c : input) {
@@ -91,11 +93,11 @@ namespace api::modules {
 
             // set
             eamuse_set_keypad_overrides(keypad, state);
-            Sleep(sleep_time);
+            timer.sleep(sleep_time);
 
             // unset
             eamuse_set_keypad_overrides(keypad, 0);
-            Sleep(sleep_time);
+            timer.sleep(sleep_time);
         }
     }
 

src/spice2x/api/modules/memory.cpp

@@ -58,7 +58,7 @@ namespace api::modules {
 
         // check if file exists in modules
         if (!fileutils::file_exists(dll_path)) {
-            return error(res, "Couldn't find " + dll_path.string());
+            return error(res, fmt::format("Couldn't find {}", dll_path));
         }
 
         // get module
@@ -118,7 +118,7 @@ namespace api::modules {
 
         // check if file exists in modules
         if (!fileutils::file_exists(dll_path)) {
-            return error(res, "Couldn't find " + dll_path.string());
+            return error(res, fmt::format("Couldn't find {}", dll_path));
         }
 
         // get module
@@ -196,7 +196,7 @@ namespace api::modules {
 
         // check if file exists in modules
         if (!fileutils::file_exists(dll_path)) {
-            return error(res, "Couldn't find " + dll_path.string());
+            return error(res, fmt::format("Couldn't find {}", dll_path));
         }
 
         // get module

src/spice2x/api/serial.cpp

@@ -5,6 +5,7 @@
 #include <utility>
 
 #include "util/logging.h"
+#include "util/precise_timer.h"
 #include "util/utils.h"
 
 
@@ -16,6 +17,7 @@ namespace api {
         controller->init_state(this->state);
         this->thread = new std::thread([this] () {
             log_warning("api::serial", "listening on {} (baud: {})", this->port, this->baud);
+            timeutils::PreciseSleepTimer timer;
 
             // read buffer
             uint8_t read_buffer[16*1024];
@@ -160,7 +162,7 @@ namespace api {
 
                 // slow down on reconnect
                 if (this->running) {
-                    Sleep(retry_time);
+                    timer.sleep(retry_time);
                 }
             }
         });

src/spice2x/api/websocket.cpp

@@ -5,6 +5,7 @@
 #include "util/utils.h"
 #include "util/rc4.h"
 #include "util/logging.h"
+#include "overlay/notifications.h"
 #include "controller.h"
 
 using namespace headsocket;
@@ -91,12 +92,18 @@ namespace api {
 
         // log connection
         log_info("api::websocket", "client connected");
+        overlay::notifications::add(
+                overlay::notifications::Severity::Success,
+                "API websocket client connected");
     }
 
     void WebSocketClient::on_disconnect() {
 
         // log disconnection
         log_info("api::websocket", "client disconnected");
+        overlay::notifications::add(
+                overlay::notifications::Severity::Info,
+                "API websocket client disconnected");
 
         // get pointer to server
         auto srv = reinterpret_cast<WebSocketServer *>(server().get());

src/spice2x/avs/core.cpp

@@ -1168,7 +1168,7 @@ namespace avs {
             // Ongaku Paradise
             {"arkjc9.dll", 0xA00000},
 
-            // KAMUNITY 
+            // KAMUNITY
             {"kamunity.dll", 33554432},
         };
 
@@ -1516,7 +1516,7 @@ namespace avs {
                         "    * It's also possible that you have incomplete game data\n"
                         "    * Do NOT copy over random DLLs from another game installation; DLL must match game version\n"
                         "\n"
-                    , DLL_NAME, MODULE_PATH.string()) };
+                    , DLL_NAME, MODULE_PATH) };
                     log_warning("avs-ea3", "{}", info_str);
                     log_fatal("avs-ea3", "Failed to find critical avs DLL on disk (avs2-core.dll OR {})", DLL_NAME);
                 }
@@ -1757,9 +1757,9 @@ namespace avs {
         {
             deferredlogs::defer_error_messages({
                 "AVS filesystem initialization failure was previously detected during boot!",
-                fmt::format("    ERROR: directory could not be created: {}", src_path.string().c_str()),
-                fmt::format("    if you see a crash, it may have been caused by bad <mounttable> contents in {}", avs::core::CFG_PATH.c_str()),
-                "    fix the XML file and try again"  
+                fmt::format("    ERROR: directory could not be created: {}", src_path),
+                fmt::format("    if you see a crash, it may have been caused by bad <mounttable> contents in {}", avs::core::CFG_PATH),
+                "    fix the XML file and try again"
             });
         }
 
@@ -1780,14 +1780,14 @@ namespace avs {
             auto created = std::filesystem::create_directories(real_path, err);
 
             if (created) {
-                log_info("avs-core", "created '{}' at '{}'", avs_path, real_path.string());
+                log_info("avs-core", "created '{}' at '{}'", avs_path, real_path);
             }
 
             if (err) {
                 avs_dir_err(real_path);
                 log_warning("avs-core", "failed to create '{}' folder at '{}': {}",
                         avs_path,
-                        real_path.string(),
+                        real_path,
                         err.message());
             }
         }
@@ -1921,6 +1921,15 @@ namespace avs {
             // create nvram and raw directories if possible for mounttable configurations
             create_avs_config_fs_table(config, config_node);
 
+#if !SPICE64
+            // sdvx4 bad log config fix
+            if (avs::game::DLL_NAME == "soundvoltex.dll" &&  // it's too early for avs::game::is_model
+                property_search_safe(config, config_node, "/log/enable_console")) {
+                log_info("avs-core", "applying SDVX4 avs-config.xml fix for <log><enable_console>");
+                property_search_remove_safe(config, config_node, "/log/enable_console");
+            }
+#endif
+
             // set log level
             if (!LOG_LEVEL_CUSTOM.empty()) {
                 property_search_remove_safe(config, config_node, "/log/level");

src/spice2x/avs/ea3.cpp

@@ -151,8 +151,8 @@ namespace avs {
                         "    * It's also possible that you have incomplete game data\n"
                         "    * Do NOT copy over random DLLs from another game installation; DLL must match game version\n"
                         "\n"
-                    , DLL_NAME, MODULE_PATH.string()) };
-                    
+                    , DLL_NAME, MODULE_PATH) };
+
                     log_warning("avs-ea3", "{}", info_str);
                     log_fatal("avs-ea3", "Failed to find critical ea3 DLL on disk (avs2-ea3.dll OR {})", DLL_NAME);
                 }
@@ -217,8 +217,10 @@ namespace avs {
                 ea3_config_name = "prop/eamuse-config.xml";
             }
 
-            if (avs::core::file_exists(ea3_config_name)) {
-                log_info("avs-ea3", "booting (using {})", ea3_config_name);
+            if (fileutils::file_exists(ea3_config_name)) {
+                log_info("avs-ea3", "found {} on disk", ea3_config_name);
+            } else if (CFG_PATH.size()) {
+                log_fatal("avs-ea3", "user-specified ea3 config file is missing: {}", ea3_config_name);
             } else {
                 log_warning("avs-ea3", "looked for the following files in order:");
                 log_warning("avs-ea3", "  * prop/ea3-config.xml");

src/spice2x/avs/game.cpp

@@ -73,11 +73,10 @@ namespace avs {
 
             // load game instance
             const auto dll_path = MODULE_PATH / DLL_NAME;
-            const auto dll_path_s = dll_path.string();
-            log_info("avs-game", "DLL path: {}", dll_path_s.c_str());
+            log_info("avs-game", "DLL path: {}", dll_path);
 
             // MAX_PATH is 260
-            if (130 <= dll_path_s.length()) {
+            if (const auto dll_path_len = dll_path.wstring().length(); 130 <= dll_path_len) {
                 log_warning(
                     "avs-game",
                     "PATH TOO LONG WARNING\n\n"
@@ -93,7 +92,7 @@ namespace avs {
                     "long, often resulting in random crashes. Move the game contents to\n"
                     "a directory with shorter path.\n"
                     "-------------------------------------------------------------------\n\n",
-                    dll_path_s, dll_path_s.length());
+                    dll_path, dll_path_len);
             }
 
             // ddr gamemdx.dll user error
@@ -112,7 +111,7 @@ namespace avs {
 
             // file not found on disk
             if (!fileutils::file_exists(dll_path)) {
-                log_warning("avs-game", "game DLL could not be found on disk: {}", dll_path.string().c_str());
+                log_warning("avs-game", "game DLL could not be found on disk: {}", dll_path);
                 log_warning("avs-game", "double check -exec and -modules parameters; unless you know what you're doing, leave them blank");
             }
 

src/spice2x/build/controller_presets.xml

@@ -39,8 +39,8 @@
             <button name="VEFX" vkey="10" analogtype="0" devid="Player 1" invert="false" debounce_up="0" debounce_down="0" velocity_threshold="0"/>
         </buttons>
         <analogs>
-            <analog name="Turntable P1" devid="Player 1" index="4" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Turntable P2" devid="Player 2" index="4" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
+            <analog name="Turntable P1" devid="Player 1" index="4" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Turntable P2" devid="Player 2" index="4" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
         </analogs>
         <lights>
             <light name="P1 1" devid="Player 1" index="0"/>
@@ -117,8 +117,8 @@
         <keypad_buttons/>
         <!-- Assumes analog mode. Need to invert direction since this FW is designed for INF. -->
         <analogs>
-            <analog name="Turntable P1" devid="arcin (1p)" index="0" sensivity="1" deadzone="0" deadzone_mirror="false" invert="true" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Turntable P2" devid="arcin (2p)" index="0" sensivity="1" deadzone="0" deadzone_mirror="false" invert="true" smoothing="false" multiplier="1" relative="false" delay="0"/>
+            <analog name="Turntable P1" devid="arcin (1p)" index="0" sensivity="1" deadzone="0" deadzone_mirror="false" invert="true" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Turntable P2" devid="arcin (2p)" index="0" sensivity="1" deadzone="0" deadzone_mirror="false" invert="true" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
         </analogs>
         <lights>
             <light name="P1 1" devid="arcin (1p)" index="0"/>
@@ -144,34 +144,34 @@
         <buttons/>
         <keypad_buttons/>
         <analogs>
-            <analog name="Key 1" devid="Analogs key 01-28" index="4" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 2" devid="Analogs key 01-28" index="5" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 3" devid="Analogs key 01-28" index="6" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 4" devid="Analogs key 01-28" index="7" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 5" devid="Analogs key 01-28" index="8" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 6" devid="Analogs key 01-28" index="9" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 7" devid="Analogs key 01-28" index="10" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 8" devid="Analogs key 01-28" index="11" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 9" devid="Analogs key 01-28" index="12" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 10" devid="Analogs key 01-28" index="13" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 11" devid="Analogs key 01-28" index="14" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 12" devid="Analogs key 01-28" index="15" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 13" devid="Analogs key 01-28" index="16" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 14" devid="Analogs key 01-28" index="17" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 15" devid="Analogs key 01-28" index="18" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 16" devid="Analogs key 01-28" index="19" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 17" devid="Analogs key 01-28" index="20" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 18" devid="Analogs key 01-28" index="21" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 19" devid="Analogs key 01-28" index="22" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 20" devid="Analogs key 01-28" index="23" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 21" devid="Analogs key 01-28" index="24" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 22" devid="Analogs key 01-28" index="25" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 23" devid="Analogs key 01-28" index="26" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 24" devid="Analogs key 01-28" index="27" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 25" devid="Analogs key 01-28" index="28" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 26" devid="Analogs key 01-28" index="29" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 27" devid="Analogs key 01-28" index="30" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Key 28" devid="Analogs key 01-28" index="31" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
+            <analog name="Key 1" devid="Analogs key 01-28" index="4" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 2" devid="Analogs key 01-28" index="5" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 3" devid="Analogs key 01-28" index="6" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 4" devid="Analogs key 01-28" index="7" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 5" devid="Analogs key 01-28" index="8" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 6" devid="Analogs key 01-28" index="9" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 7" devid="Analogs key 01-28" index="10" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 8" devid="Analogs key 01-28" index="11" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 9" devid="Analogs key 01-28" index="12" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 10" devid="Analogs key 01-28" index="13" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 11" devid="Analogs key 01-28" index="14" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 12" devid="Analogs key 01-28" index="15" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 13" devid="Analogs key 01-28" index="16" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 14" devid="Analogs key 01-28" index="17" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 15" devid="Analogs key 01-28" index="18" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 16" devid="Analogs key 01-28" index="19" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 17" devid="Analogs key 01-28" index="20" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 18" devid="Analogs key 01-28" index="21" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 19" devid="Analogs key 01-28" index="22" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 20" devid="Analogs key 01-28" index="23" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 21" devid="Analogs key 01-28" index="24" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 22" devid="Analogs key 01-28" index="25" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 23" devid="Analogs key 01-28" index="26" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 24" devid="Analogs key 01-28" index="27" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 25" devid="Analogs key 01-28" index="28" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 26" devid="Analogs key 01-28" index="29" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 27" devid="Analogs key 01-28" index="30" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Key 28" devid="Analogs key 01-28" index="31" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
         </analogs>
         <lights>
             <light name="Key 1 R" devid="Lights 01-14" index="0"/>
@@ -277,9 +277,9 @@
         </buttons>
         <keypad_buttons/>
         <analogs>
-            <analog name="Guitar P1 Wail X" devid="Gitaller P1" index="1" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Guitar P1 Wail Y" devid="Gitaller P1" index="0" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="Guitar P1 Wail Z" devid="Gitaller P1" index="5" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
+            <analog name="Guitar P1 Wail X" devid="Gitaller P1" index="1" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Guitar P1 Wail Y" devid="Gitaller P1" index="0" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="Guitar P1 Wail Z" devid="Gitaller P1" index="5" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
         </analogs>
         <lights/>
     <!-- Gen 1 FAUCETWO in HID mode; should also work for later gens -->
@@ -296,8 +296,8 @@
         </buttons>
         <keypad_buttons/>
         <analogs>
-            <analog name="VOL-L" devid="Buttons (MI_01)" index="1" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
-            <analog name="VOL-R" devid="Buttons (MI_01)" index="0" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay="0"/>
+            <analog name="VOL-L" devid="Buttons (MI_01)" index="1" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
+            <analog name="VOL-R" devid="Buttons (MI_01)" index="0" sensivity="1" deadzone="0" deadzone_mirror="false" invert="false" smoothing="false" multiplier="1" relative="false" delay_ms="0"/>
         </analogs>
         <lights>
             <light name="BT-A" devid="Lights (MI_02)" index="24"/>

src/spice2x/build/manifest.manifest

@@ -23,7 +23,7 @@
     <trustInfo xmlns="urn:schemas-microsoft-com:asm.v2">
         <security>
             <requestedPrivileges>
-                <requestedExecutionLevel level="requireAdministrator" uiAccess="false"/>
+                <requestedExecutionLevel level="asInvoker" uiAccess="false"/>
             </requestedPrivileges>
         </security>
     </trustInfo>

src/spice2x/build_all.sh

@@ -68,8 +68,8 @@ DIST_FOLDER="./dist"
 DIST_NAME="spice2x-$(date +%y)-$(date +%m)-$(date +%d).zip"
 DIST_NAME_EXTRAS="spice2x-$(date +%y)-$(date +%m)-$(date +%d)-full.zip"
 DIST_COMMENT=${DIST_NAME}$'\n'"$GIT_BRANCH - $GIT_HEAD"$'\nThank you for playing.'
-TARGETS_32="spicetools_stubs_kbt spicetools_stubs_kld spicetools_cfg spicetools_cfg_linux spicetools_spice spicetools_spice_laa spicetools_spice_linux spicetools_stubs_cpusbxpkm"
-TARGETS_64="spicetools_stubs_kbt64 spicetools_stubs_kld64 spicetools_stubs_nvEncodeAPI64 spicetools_stubs_nvcuvid spicetools_stubs_nvcuda spicetools_spice64 spicetools_spice64_linux"
+TARGETS_32="spicetools_stubs_kbt spicetools_stubs_kld spicetools_cfg spicetools_cfg_linux spicetools_spice spicetools_spice_laa spicetools_spice_linux spicetools_stubs_cpusbxpkm spicetools_sdk_sample_v0_flat_c_32"
+TARGETS_64="spicetools_stubs_kbt64 spicetools_stubs_kld64 spicetools_stubs_nvEncodeAPI64 spicetools_stubs_nvcuvid spicetools_stubs_nvcuda spicetools_spice64 spicetools_spice64_linux spicetools_sdk_sample_v0_flat_c_64 spicetools_sdk_sample_v0_cpp_64"
 TARGETS_XP32="spicetools_cfg spicetools_spice"
 TARGETS_XP64="spicetools_spice64"
 
@@ -163,6 +163,7 @@ time (
 	if ((BUILD_XP > 0))
 	then
 		# 32 bit Windows XP
+		echo ""
 		echo "Building 32bit targets (WinXP toolchain)..."
 		echo "========================="
 		if ((CLEAN_BUILD > 0))
@@ -171,7 +172,7 @@ time (
 		fi
 		mkdir -p ${BUILDDIR_WINXP_32}
 		pushd ${BUILDDIR_WINXP_32} > /dev/null
-		CXXFLAGS="$CXXFLAGS -DSPICE_XP=1" cmake -G "Ninja" -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_WINXP_32} -DCMAKE_BUILD_TYPE=${BUILD_TYPE} "$OLDPWD" && ninja ${TARGETS_XP32}
+		cmake -G "Ninja" -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_WINXP_32} -DCMAKE_BUILD_TYPE=${BUILD_TYPE} -DSPICE_XP=ON "$OLDPWD" && ninja ${TARGETS_XP32}
 		popd > /dev/null
 
 		# 64 bit Windows XP
@@ -184,9 +185,10 @@ time (
 		fi
 		mkdir -p ${BUILDDIR_WINXP_64}
 		pushd ${BUILDDIR_WINXP_64} > /dev/null
-		CXXFLAGS="$CXXFLAGS -DSPICE_XP=1" cmake -G "Ninja" -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_WINXP_64} -DCMAKE_BUILD_TYPE=${BUILD_TYPE} "$OLDPWD" && ninja ${TARGETS_XP64}
+		cmake -G "Ninja" -DCMAKE_TOOLCHAIN_FILE=${TOOLCHAIN_WINXP_64} -DCMAKE_BUILD_TYPE=${BUILD_TYPE} -DSPICE_XP=ON "$OLDPWD" && ninja ${TARGETS_XP64}
 		popd > /dev/null
 	else
+		echo ""
 		echo "Skipping WinXP builds, toolchain not specified"
 	fi
 
@@ -267,6 +269,8 @@ rm -rf ${OUTDIR_EXTRAS}
 mkdir -p ${OUTDIR_EXTRAS}
 mkdir -p ${OUTDIR_EXTRAS}/largeaddressaware
 mkdir -p ${OUTDIR_EXTRAS}/linux
+mkdir -p ${OUTDIR_EXTRAS}/sdk/samples/32
+mkdir -p ${OUTDIR_EXTRAS}/sdk/samples/64
 if ((BUILD_XP > 0))
 then
 mkdir -p ${OUTDIR_EXTRAS}/winxp
@@ -279,16 +283,8 @@ then
     cp ${BUILDDIR_32}/spicetools/spicecfg-pdb.pdb ${OUTDIR} 2>/dev/null
     cp ${BUILDDIR_32}/spicetools/32/spice-pdb.exe ${OUTDIR} 2>/dev/null
     cp ${BUILDDIR_32}/spicetools/32/spice-pdb.pdb ${OUTDIR} 2>/dev/null
-    #cp ${BUILDDIR_32}/spicetools/32/kbt.dll ${OUTDIR}/stubs/32 2>/dev/null
-    #cp ${BUILDDIR_32}/spicetools/32/kld.dll ${OUTDIR}/stubs/32 2>/dev/null
     cp ${BUILDDIR_64}/spicetools/64/spice64-pdb.exe ${OUTDIR} 2>/dev/null
     cp ${BUILDDIR_64}/spicetools/64/spice64-pdb.pdb ${OUTDIR} 2>/dev/null
-    #cp ${BUILDDIR_64}/spicetools/64/kbt.dll ${OUTDIR}/stubs/64 2>/dev/null
-    #cp ${BUILDDIR_64}/spicetools/64/kld.dll ${OUTDIR}/stubs/64 2>/dev/null
-    #cp ${BUILDDIR_64}/spicetools/64/nvEncodeAPI64.dll ${OUTDIR}/stubs/64 2>/dev/null
-    #cp ${BUILDDIR_64}/spicetools/64/nvcuda.dll ${OUTDIR}/stubs/64 2>/dev/null
-    #cp ${BUILDDIR_64}/spicetools/64/nvcuvid.dll ${OUTDIR}/stubs/64 2>/dev/null
-	#cp ${BUILDDIR_32}/spicetools/32/cpusbxpkm.dll ${OUTDIR}/stubs/32 2>/dev/null
 else
     # release files
     cp ${BUILDDIR_32}/spicetools/spicecfg.exe ${OUTDIR} 2>/dev/null
@@ -296,16 +292,15 @@ else
     cp ${BUILDDIR_32}/spicetools/32/spice.exe ${OUTDIR} 2>/dev/null
     cp ${BUILDDIR_32}/spicetools/32/spice_laa.exe ${OUTDIR_EXTRAS}/largeaddressaware/spice.exe 2>/dev/null
     cp ${BUILDDIR_32}/spicetools/32/spice_linux.exe ${OUTDIR_EXTRAS}/linux/spice.exe 2>/dev/null
-    #cp ${BUILDDIR_32}/spicetools/32/kbt.dll ${OUTDIR}/stubs/32 2>/dev/null
-    #cp ${BUILDDIR_32}/spicetools/32/kld.dll ${OUTDIR}/stubs/32 2>/dev/null
     cp ${BUILDDIR_64}/spicetools/64/spice64.exe ${OUTDIR} 2>/dev/null
     cp ${BUILDDIR_64}/spicetools/64/spice64_linux.exe ${OUTDIR_EXTRAS}/linux/spice64.exe 2>/dev/null
-    #cp ${BUILDDIR_64}/spicetools/64/kbt.dll ${OUTDIR}/stubs/64 2>/dev/null
-    #cp ${BUILDDIR_64}/spicetools/64/kld.dll ${OUTDIR}/stubs/64 2>/dev/null
     cp ${BUILDDIR_64}/spicetools/64/nvEncodeAPI64.dll ${OUTDIR}/stubs/64 2>/dev/null
     cp ${BUILDDIR_64}/spicetools/64/nvcuda.dll ${OUTDIR}/stubs/64 2>/dev/null
     cp ${BUILDDIR_64}/spicetools/64/nvcuvid.dll ${OUTDIR}/stubs/64 2>/dev/null
     cp ${BUILDDIR_32}/spicetools/32/cpusbxpkm.dll ${OUTDIR}/stubs/32 2>/dev/null
+	cp ${BUILDDIR_32}/spicetools/32/sdk_sample_v0_flat_c.dll ${OUTDIR_EXTRAS}/sdk/samples/32/v0_flat_c.dll 2>/dev/null
+	cp ${BUILDDIR_64}/spicetools/64/sdk_sample_v0_flat_c.dll ${OUTDIR_EXTRAS}/sdk/samples/64/v0_flat_c.dll 2>/dev/null
+	cp ${BUILDDIR_64}/spicetools/64/sdk_sample_v0_cpp.dll ${OUTDIR_EXTRAS}/sdk/samples/64/v0_cpp.dll 2>/dev/null
 	if ((BUILD_XP > 0))
 	then
 		cp ${BUILDDIR_WINXP_32}/spicetools/spicecfg.exe ${OUTDIR_EXTRAS}/winxp 2>/dev/null
@@ -324,6 +319,9 @@ then
 		echo "WARNING: Couldn't get git to create the archive. Is this a git repository?"
 fi
 
+# sdk headers
+cp -r ./sdk/include ${OUTDIR_EXTRAS}/sdk
+
 # copy resources
 rm -rf ${OUTDIR_EXTRAS}/api
 mkdir ${OUTDIR_EXTRAS}/api

src/spice2x/build_docker.bat

@@ -1,7 +1,7 @@
 docker build --pull external/docker -t spicetools/deps
 docker build --build-context gitroot=%cd%/../../.git . -t spicetools/spice
-docker run --rm -it -v %cd%/dist:/src/src/spice2x/dist -v %cd%/bin:/src/src/spice2x/bin -v %cd%/.ccache:/src/src/spice2x/.ccache spicetools/spice %*
+docker run --rm -i -v %cd%/dist:/src/src/spice2x/dist -v %cd%/bin:/src/src/spice2x/bin -v %cd%/.ccache:/src/src/spice2x/.ccache spicetools/spice %*
 @REM to generate PDBs, set DEBUG to 1 in build_all.sh, place cv2pdb in external\cv2pdb, and run below
 @REM external\cv2pdb\cv2pdb.exe bin\spice2x\spicecfg.exe bin\spice2x\spicecfg-pdb.exe bin\spice2x\spicecfg-pdb.pdb
 @REM external\cv2pdb\cv2pdb.exe bin\spice2x\spice.exe bin\spice2x\spice-pdb.exe bin\spice2x\spice-pdb.pdb
-@REM external\cv2pdb\cv2pdb.exe bin\spice2x\spice64.exe bin\spice2x\spice64-pdb.exe bin\spice2x\spice64-pdb.pdb
+@REM external\cv2pdb\cv2pdb.exe bin\spice2x\spice64.exe bin\spice2x\spice64-pdb.exe bin\spice2x\spice64-pdb.pdb

src/spice2x/cfg/analog.cpp

@@ -207,22 +207,38 @@ float Analog::applyMultiplier(float value) {
 }
 
 float Analog::normalizeAnalogValue(float value) {
-    // effectively the same as fmodf(value, 1.f)
-    // for small values, this is MUCH faster than fmodf.
-    float new_value = value;
-    while (new_value > 1.f) {
-        new_value -= 1.f;
+    if (getType() == GameAPI::Analogs::AnalogType::Circular) {
+        // effectively the same as fmodf(value, 1.f)
+        // for small values, this is MUCH faster than fmodf.
+        float new_value = value;
+        while (new_value > 1.f) {
+            new_value -= 1.f;
+        }
+        while (new_value < 0.f) {
+            new_value += 1.f;
+        }
+        return new_value;
+
+    } else {
+        // clamp to [0, 1] range
+        return std::clamp(value, 0.f, 1.f);
     }
-    while (new_value < 0.f) {
-        new_value += 1.f;
-    }
-    return new_value;
 }
 
 float Analog::applyDeadzone(float raw_value) {
     float value = raw_value;
-    const auto deadzone = this->getDeadzone();
-    if (deadzone > 0) {
+    auto deadzone = this->getDeadzone();
+
+    // in the past, positive deadzone applied in the center, negative deadzone applied to 0
+    // after each analog value received a type (circular/linear) this has been simpliifed to
+    // positive values only since we can figure out where the rest value is
+    // for back compat, treat negative value as positive
+    if (deadzone < 0.f) {
+        deadzone = -deadzone;
+    }
+
+    // relative mode assumes that user is using a stick, so center is neutral regardless of analog type
+    if (getType() != GameAPI::Analogs::AnalogType::LinearPositive || isRelativeMode()) {
 
         // calculate values
         const auto delta = value - 0.5f;
@@ -255,7 +271,7 @@ float Analog::applyDeadzone(float raw_value) {
             }
         }
 
-    } else if (deadzone < 0) {
+    } else {
 
         // invert for mirror
         if (this->getDeadzoneMirror()) {
@@ -263,8 +279,8 @@ float Analog::applyDeadzone(float raw_value) {
         }
 
         // deadzone from minimum value
-        if (deadzone > -1 && value > -deadzone) {
-            value = std::min(1.f, (value + deadzone) / (1.f + deadzone));
+        if (deadzone < 1.f && deadzone < value) {
+            value = std::max(0.f, (value - deadzone) / (1.f - deadzone));
         } else {
             value = 0.f;
         }
@@ -275,4 +291,75 @@ float Analog::applyDeadzone(float raw_value) {
         }
     }
     return value;
-}
+}
+
+float Analog::getRelativeModeValue(float raw_value) {
+    const auto now = get_performance_seconds();
+    auto delta_time = now - this->rel_mode_last_read_time_s;
+
+    if (this->rel_mode_last_read_time_s != 0.f) {
+        // some heuristics to prevent huge jumps:
+        //  * if we went for more than 250ms without polls, discard it (e.g., during loading screens)
+        //  * cap the delta at 100ms to prevent huge jumps in case of very infrequent polling
+        if (delta_time < 0.f || 0.25f < delta_time) {
+            delta_time = 0.f;
+        } else if (delta_time > 0.1f) {
+            delta_time = 0.1f;
+        }
+
+        // scale [0, 1] to [-1, 1] to simplify calculations
+        const auto delta_raw_value = (raw_value - 0.5f) * 2.f;
+
+        // target is one revolution per second at max speed at 1.0 sensitivity
+        auto adjusted_delta_value = delta_raw_value * delta_time;
+
+        // multiplier / divisor
+        if (this->getMultiplier() > 1) {
+            adjusted_delta_value *= this->getMultiplier();
+        } else if (this->getMultiplier() < -1) {
+            adjusted_delta_value /= -this->getMultiplier();
+        }
+
+        // sensitivity
+        if (this->isSensitivitySet()) {
+            adjusted_delta_value *= this->getSensitivity();
+        }
+
+        // calculate the new absolute value
+        this->rel_mode_absolute_value += adjusted_delta_value;
+    }
+
+    // update for next poll
+    this->rel_mode_last_read_time_s = now;
+    this->rel_mode_absolute_value = normalizeAnalogValue(this->rel_mode_absolute_value);
+    return this->rel_mode_absolute_value;
+}
+
+float Analog::getDelayedValue(float raw_value) {
+    const double delay_ms = static_cast<double>(this->getDelayMs());
+    if (delay_ms == 0.0) {
+        return raw_value;
+    }
+
+    // always push a new value
+    const auto now = get_performance_milliseconds();
+    this->delayed_inputs.emplace(now, raw_value);
+    
+    // drain the queue down to reasonable length to prevent unconstrained growth
+    // this would accommodate 1 second at ~1000Hz which is overkill
+    // (UI only allows for 250ms of delay)
+    while (this->delayed_inputs.size() > 1024) {
+        this->delayed_inputs.pop();
+    }
+
+    // pop until we find the oldest value still inside the delay window
+    while (this->delayed_inputs.size() > 1) {
+        const auto delta_t = now - this->delayed_inputs.front().time_in_ms;
+        if (delta_t <= delay_ms) {
+            break;
+        }
+        this->delayed_inputs.pop();
+    }
+
+    return this->delayed_inputs.front().value;
+}

src/spice2x/cfg/analog.h

@@ -1,5 +1,6 @@
 #pragma once
 
+#include <cstdint>
 #include <array>
 #include <string>
 #include <cmath>
@@ -13,12 +14,32 @@ namespace rawinput {
     class RawInputManager;
 }
 
+namespace GameAPI::Analogs {
+    enum class AnalogType {
+        // default; values wrap around (below 0 turns into 1, over 1 is 0)
+        // knobs, turntables
+        Circular = 0,
+
+        // typical joystick that rests at the center and caps at [0, 1]
+        LinearCentered = 1,
+
+        // one-directional value (sliders and instruments like piano/drum velocity)
+        // starts at 0 and goes up to 1
+        LinearPositive = 2,
+    };
+}
+
 struct AnalogMovingAverage {
     double time_in_ms;
     float sine;
     float cosine;
 };
 
+struct AnalogDelayEntry {
+    double time_in_ms;
+    float value;
+};
+
 class Analog {
 private:
     std::string name;
@@ -31,10 +52,11 @@ private:
     float last_state = 0.5f;
     bool sensitivity_set = false;
     bool deadzone_set = false;
+    GameAPI::Analogs::AnalogType type = GameAPI::Analogs::AnalogType::Circular;
     
     // smoothing function
     bool smoothing = false;
-    std::array<AnalogMovingAverage, ANALOG_HISTORY_CNT> vector_history;
+    std::array<AnalogMovingAverage, ANALOG_HISTORY_CNT> vector_history = {};
     int vector_history_index = 0;
     float smoothed_last_state = 0.f;
 
@@ -48,12 +70,13 @@ private:
     unsigned short divisor_region = 0;
 
     // relative input mode
-    float absolute_value_for_rel_mode = 0.5f;
+    float rel_mode_absolute_value = 0.5f;
+    float rel_mode_last_read_time_s = 0.f;
     bool relative_mode = false;
 
-    // circular buffer (delayed input)
-    int delay_buffer_depth = 0;
-    std::queue<float> delay_buffer;
+    // delay
+    uint32_t delay_ms = 0;
+    std::queue<AnalogDelayEntry> delayed_inputs;
 
     float calculateAngularDifference(float old_rads, float new_rads);
     float normalizeAngle(float rads);
@@ -66,7 +89,8 @@ public:
     float override_state = 0.5f;
 
     explicit Analog(std::string name) : name(std::move(name)) {
-        vector_history.fill({0.0, 0.f, 0.f});
+    };
+    explicit Analog(std::string name, GameAPI::Analogs::AnalogType type) : name(std::move(name)), type(type) {
     };
 
     std::string getDisplayString(rawinput::RawInputManager* manager);
@@ -74,6 +98,8 @@ public:
     float applyAngularSensitivity(float raw_rads);
     float applyMultiplier(float raw_value);
     float applyDeadzone(float raw_value);
+    float getRelativeModeValue(float raw_value);
+    float getDelayedValue(float raw_value);
 
     inline bool isSet() {
         if (this->override_enabled) {
@@ -90,8 +116,8 @@ public:
         deadzone_mirror = false;
         setMultiplier(1);
         setRelativeMode(false);
-        setDelayBufferDepth(0);
         setLastState(0.5f);
+        setDelayMs(0);
     }
 
     inline void clearBindings() {
@@ -193,25 +219,27 @@ public:
     }
 
     inline void setRelativeMode(bool relative_mode) {
+        if (relative_mode) {
+            this->smoothing = false;
+        }
         this->relative_mode = relative_mode;
-        this->absolute_value_for_rel_mode = 0.5f;
+        this->rel_mode_absolute_value = 0.5f;
+        this->rel_mode_last_read_time_s = 0.f;
     }
 
-    inline float getAbsoluteValue(float relative_delta) {
-        this->absolute_value_for_rel_mode =
-            normalizeAnalogValue(this->absolute_value_for_rel_mode + relative_delta);
-        return this->absolute_value_for_rel_mode;
+    inline GameAPI::Analogs::AnalogType getType() const {
+        return this->type;
     }
 
-    inline int getDelayBufferDepth() const {
-        return this->delay_buffer_depth;
+    inline void setType(GameAPI::Analogs::AnalogType type) {
+        this->type = type;
     }
 
-    inline void setDelayBufferDepth(int depth) {
-        this->delay_buffer_depth = depth;
+    inline uint32_t getDelayMs() const {
+        return this->delay_ms;
     }
 
-    inline std::queue<float> &getDelayBuffer() {
-        return this->delay_buffer;
+    inline void setDelayMs(uint32_t delay_ms) {
+        this->delay_ms = delay_ms;
     }
 };

src/spice2x/cfg/api.cpp

@@ -199,9 +199,19 @@ GameAPI::Buttons::State GameAPI::Buttons::getState(rawinput::RawInputManager *ma
                             if (vKey < value_states->size()) {
                                 auto value = value_states->at(vKey);
                                 if (current_button->getAnalogType() == BAT_POSITIVE) {
-                                    state = value > 0.6f ? BUTTON_PRESSED : BUTTON_NOT_PRESSED;
+                                    float threshold = 0.6f;
+                                    if (current_button->getBatThreshold() > 0) {
+                                        threshold = std::clamp(current_button->getBatThreshold() / 100.f, 0.01f, 0.99f);
+                                    }
+                                    state = value > threshold ? BUTTON_PRESSED : BUTTON_NOT_PRESSED;
+
                                 } else if (current_button->getAnalogType() == BAT_NEGATIVE) {
-                                    state = value < 0.4f ? BUTTON_PRESSED : BUTTON_NOT_PRESSED;
+                                    float threshold = 0.4f;
+                                    if (current_button->getBatThreshold() > 0) {
+                                        threshold = std::clamp((100 - current_button->getBatThreshold()) / 100.f, 0.01f, 0.99f);
+                                    }
+                                    state = value < threshold ? BUTTON_PRESSED : BUTTON_NOT_PRESSED;
+
                                 } else {
                                     state = value > 0.01f ? BUTTON_PRESSED : BUTTON_NOT_PRESSED;
                                 }
@@ -669,81 +679,98 @@ float GameAPI::Analogs::getState(rawinput::RawInputManager *manager, rawinput::D
             }
 
             // deadzone
-            if (analog.isDeadzoneSet()) {
+            // do not apply deadzone to circular analogs since it doesn't make sense (except in relative mode)
+            if (analog.isDeadzoneSet() &&
+                (analog.getType() != AnalogType::Circular || analog.isRelativeMode())) {
                 value = analog.applyDeadzone(value);
             }
 
-            if (analog.isRelativeMode()) {
-                float relative_delta = value - 0.5f;
-                // built-in scaling to make values reasonable
-                relative_delta /= 80.f;
+            if (analog.getType() == AnalogType::Circular) {
 
-                // integer multiplier/divisor
-                const auto mult = analog.getMultiplier();
-                if (mult < -1) {
-                    relative_delta /= -mult;
-                } else if (1 < mult) {
-                    relative_delta *= mult;
-                }
+                if (analog.isRelativeMode()) {
+                    value = analog.getRelativeModeValue(value);
 
-                // sensitivity (ranges from 0.0 to 4.0)
-                if (analog.isSensitivitySet()) {
-                    relative_delta *= analog.getSensitivity();
-                }
+                } else {
+                    // integer multiplier
+                    value = analog.applyMultiplier(value);
 
-                // translate relative movement to absolute value
-                value = analog.getAbsoluteValue(relative_delta);
+                    // smoothing/sensitivity
+                    if (analog.getSmoothing() || analog.isSensitivitySet()) {
+                        float rads = value * (float) M_TAU;
 
-            } else {
-                // integer multiplier
-                value = analog.applyMultiplier(value);
+                        // smoothing
+                        if (analog.getSmoothing()) {
 
-                // smoothing/sensitivity
-                if (analog.getSmoothing() || analog.isSensitivitySet()) {
-                    float rads = value * (float) M_TAU;
+                            // preserve direction
+                            if (rads >= M_TAU) {
+                                rads -= 0.0001f;
+                            }
 
-                    // smoothing
-                    if (analog.getSmoothing()) {
-
-                        // preserve direction
-                        if (rads >= M_TAU) {
-                            rads -= 0.0001f;
+                            // calculate angle
+                            rads = analog.getSmoothedValue(rads);
                         }
 
-                        // calculate angle
-                        rads = analog.getSmoothedValue(rads);
+                        // sensitivity
+                        if (analog.isSensitivitySet()) {
+                            rads = analog.applyAngularSensitivity(rads);
+                        }
+
+                        // apply to value
+                        value = rads * (float) M_1_TAU;
+                    }
+                }
+            } else {
+                // sensitivity
+                if (analog.isSensitivitySet()) {
+                    // adjust curve
+                    // values < 1.f : less sensitive around neutral
+                    // values > 1.f : more sensitive around neutral
+                    float curve = analog.getSensitivity();
+                    if (curve <= 0.f) {
+                        curve = 0.01f;
+                    }
+                    curve = 1.f / curve;
+
+                    if (analog.getType() == AnalogType::LinearCentered) {
+                        // convert 0.0..1.0 to -1.0..+1.0
+                        float signed_raw = (value - 0.5f) * 2.0f;
+                        // apply curve
+                        float sign = signed_raw < 0.0f ? -1.0f : 1.0f;
+                        float magnitude = fabsf(signed_raw);
+                        float curved = sign * powf(magnitude, curve);
+                        // convert back to 0.0..1.0
+                        value = curved * 0.5f + 0.5f;
+                    } else {
+                        value = powf(value, curve);
                     }
 
-                    // sensitivity
-                    if (analog.isSensitivitySet()) {
-                        rads = analog.applyAngularSensitivity(rads);
+                    value = std::clamp(value, 0.f, 1.f);
+                }
+
+                // multiplier / divisor
+                if (analog.getMultiplier() < -1) {
+                    if (analog.getType() == AnalogType::LinearCentered) {
+                        value = (value - 0.5f) / (-analog.getMultiplier()) + 0.5f;
+                    } else {
+                        value /= -analog.getMultiplier();
                     }
 
-                    // apply to value
-                    value = rads * (float) M_1_TAU;
+                    value = std::clamp(value, 0.f, 1.f);
+
+                } else if (analog.getMultiplier() > 1) {
+                    if (analog.getType() == AnalogType::LinearCentered) {
+                        value = (value - 0.5f) * analog.getMultiplier() + 0.5f;
+                    } else {
+                        value *= analog.getMultiplier();
+                    }
+
+                    value = std::clamp(value, 0.f, 1.f);
                 }
             }
 
             // delay
-            if (0 < analog.getDelayBufferDepth()) {
-                auto& queue = analog.getDelayBuffer();
-
-                // ensure the queue isn't too long; drop old values
-                while (analog.getDelayBufferDepth() <= (int)queue.size()) {
-                    queue.pop();
-                }
-
-                // always push new value
-                queue.push(value);
-
-                // get a new value to return
-                if ((int)queue.size() < analog.getDelayBufferDepth()) {
-                    // not enough in the queue, stall for now, shouldn't happen often
-                    value = analog.getLastState();
-                } else {
-                    value = queue.front();
-                    queue.pop();
-                }
+            if (analog.getDelayMs() > 0) {
+                value = analog.getDelayedValue(value);
             }
 
             break;
@@ -829,6 +856,42 @@ std::vector<Analog> GameAPI::Analogs::sortAnalogs(
     return sorted;
 }
 
+static std::vector<Analog> sortAnalogsWithTypeInternal(
+    std::vector<Analog> &analogs,
+    const std::initializer_list<GameAPI::Analogs::AnalogWithType> list) {
+
+    std::vector<Analog> sorted;
+
+    bool analog_found;
+    for (auto &a : list) {
+        analog_found = false;
+
+        for (auto &analog : analogs) {
+            if (a.name == analog.getName()) {
+                analog_found = true;
+                analog.setType(a.type);
+                sorted.push_back(analog);
+                break;
+            }
+        }
+
+        if (!analog_found) {
+            sorted.emplace_back(a.name, a.type);
+        }
+    }
+
+    return sorted;
+}
+
+void GameAPI::Analogs::sortAnalogsWithType(
+    std::vector<Analog> *analogs,
+    const std::initializer_list<AnalogWithType> list) {
+
+    if (analogs) {
+        *analogs = sortAnalogsWithTypeInternal(*analogs, list);
+    }
+}
+
 float GameAPI::Analogs::getState(rawinput::RawInputManager *manager, Analog &analog) {
 
     // check override

src/spice2x/cfg/api.h

@@ -85,14 +85,15 @@ namespace GameAPI {
                 const std::vector<Analog> &analogs,
                 const std::vector<std::string> &analog_names);
 
-        template<typename T>
-        void sortAnalogs(std::vector<Analog> *analogs, T t) {
-            const std::vector<std::string> analog_names { t };
+        struct AnalogWithType {
+            const std::string name;
+            const AnalogType type;
 
-            if (analogs) {
-                *analogs = GameAPI::Analogs::sortAnalogs(*analogs, analog_names);
-            }
-        }
+            AnalogWithType(std::string name, AnalogType type) :
+                name(std::move(name)), type(type) {}
+        };
+
+        void sortAnalogsWithType(std::vector<Analog> *analogs, const std::initializer_list<AnalogWithType> list);
 
         template<typename T, typename... Rest>
         void sortAnalogs(std::vector<Analog> *analogs, T t, Rest... rest) {

src/spice2x/cfg/button.cpp

@@ -306,7 +306,10 @@ std::string Button::getDisplayString(rawinput::RawInputManager* manager) {
     std::string vKeyString = fmt::format("{:#x}", vKey);
 
     // device must be existing
-    if (this->device_identifier.empty() && vKey == INVALID_VKEY) {
+    if (this->isNaive() && vKey == INVALID_VKEY) {
+        if (this->getInvert()) {
+            return "(always on)";
+        }
         return "";
     }
 

src/spice2x/cfg/button.h

@@ -56,6 +56,7 @@ private:
     ButtonAnalogType analog_type = BAT_NONE;
     double debounce_up = 0.0;
     double debounce_down = 0.0;
+    int bat_threshold = 0; // for positive/negative only, value in percentage, zero is default
     bool invert = false;
     bool is_temporary = false;
 
@@ -109,6 +110,7 @@ public:
         alternatives.clear();
         device_identifier = "";
         analog_type = BAT_NONE;
+        bat_threshold = 0;
     }
 
     std::string getDisplayString(rawinput::RawInputManager* manager);
@@ -169,6 +171,14 @@ public:
         this->debounce_down = debounce_time_down;
     }
 
+    inline void setBatThreshold(int bat_threshold) {
+        this->bat_threshold = bat_threshold;
+    }
+
+    inline int getBatThreshold() const {
+        return this->bat_threshold;
+    }
+
     inline bool getInvert() const {
         return this->invert;
     }

src/spice2x/cfg/config.cpp

@@ -28,7 +28,7 @@ Config::Config() {
     this->status = false;
     if (CONFIG_PATH_OVERRIDE.length() > 0) {
         this->configLocation = CONFIG_PATH_OVERRIDE;
-        log_info("cfg", "using custom config file: {}", this->configLocation.string());
+        log_info("cfg", "using custom config file: {}", this->configLocation);
     } else {
         this->configLocation = std::filesystem::path(_wgetenv(L"APPDATA")) / L"spicetools.xml";
         // avoids logging the expanded appdata path as it contains user name
@@ -57,7 +57,7 @@ Config::Config() {
                 this->firstFillConfigFile();
                 break;
             case tinyxml2::XMLError::XML_ERROR_FILE_COULD_NOT_BE_OPENED:
-                log_fatal("cfg", "could not open config file: {}", this->configLocation.string());
+                log_fatal("cfg", "could not open config file: {}", this->configLocation);
                 break;
             case tinyxml2::XMLError::XML_ERROR_FILE_NOT_FOUND:
                 this->createConfigFile();
@@ -72,7 +72,7 @@ Config::Config() {
             case tinyxml2::XMLError::XML_ERROR_PARSING_UNKNOWN:
             case tinyxml2::XMLError::XML_ERROR_MISMATCHED_ELEMENT:
             case tinyxml2::XMLError::XML_ERROR_PARSING:
-                log_warning("cfg", "Couldn't read config file: {}", this->configLocation.string());
+                log_warning("cfg", "Couldn't read config file: {}", this->configLocation);
                 this->createConfigFile();
                 this->firstFillConfigFile();
                 break;
@@ -98,7 +98,7 @@ bool Config::getStatus() {
     return this->status;
 }
 
-bool Config::addGame(Game &game) {
+bool Config::addGame(Game &game, bool save) {
     tinyxml2::XMLNode *rootNode = this->configFile.LastChild();
     tinyxml2::XMLElement *gameNodes = rootNode->FirstChildElement("game");
 
@@ -162,6 +162,7 @@ bool Config::addGame(Game &game) {
                     auto analogType = (int) BAT_NONE;
                     double debounce_up = 0.0;
                     double debounce_down = 0.0;
+                    int bat_threshold = 0;
                     int velocity_threshold = 0;
                     bool invert = false;
                     tinyxml2::XMLError attrError = gameButtonNode->QueryIntAttribute("vkey", &vKey);
@@ -169,6 +170,7 @@ bool Config::addGame(Game &game) {
                     gameButtonNode->QueryIntAttribute("analogtype", &analogType);
                     gameButtonNode->QueryDoubleAttribute("debounce_up", &debounce_up);
                     gameButtonNode->QueryDoubleAttribute("debounce_down", &debounce_down);
+                    gameButtonNode->QueryIntAttribute("bat_threshold", &bat_threshold);
                     gameButtonNode->QueryIntAttribute("velocity_threshold", &velocity_threshold);
                     gameButtonNode->QueryBoolAttribute("invert", &invert);
                     if (attrError != tinyxml2::XMLError::XML_SUCCESS) {
@@ -180,6 +182,7 @@ bool Config::addGame(Game &game) {
                         gameButtonNode->SetAttribute("devid", button->getDeviceIdentifier().c_str());
                         gameButtonNode->SetAttribute("debounce_up", debounce_up);
                         gameButtonNode->SetAttribute("debounce_down", debounce_down);
+                        gameButtonNode->SetAttribute("bat_threshold", bat_threshold);
                         gameButtonNode->SetAttribute("velocity_threshold", velocity_threshold);
                         gameButtonNode->SetAttribute("invert", invert);
                         gameButtonsNode->InsertEndChild(gameButtonNode);
@@ -188,6 +191,7 @@ bool Config::addGame(Game &game) {
                         button->setAnalogType((ButtonAnalogType) analogType);
                         button->setDebounceUp(debounce_up);
                         button->setDebounceDown(debounce_down);
+                        button->setBatThreshold(bat_threshold);
                         button->setVelocityThreshold(velocity_threshold);
                         button->setInvert(invert);
                         if (devid) {
@@ -207,6 +211,7 @@ bool Config::addGame(Game &game) {
                 gameButtonNode->SetAttribute("analogtype", (int) it.getAnalogType());
                 gameButtonNode->SetAttribute("debounce_up", it.getDebounceUp());
                 gameButtonNode->SetAttribute("debounce_down", it.getDebounceDown());
+                gameButtonNode->SetAttribute("bat_threshold", it.getBatThreshold());
                 gameButtonNode->SetAttribute("velocity_threshold", it.getVelocityThreshold());
                 gameButtonNode->SetAttribute("invert", it.getInvert());
                 gameButtonNode->SetAttribute("devid", it.getDeviceIdentifier().c_str());
@@ -248,7 +253,7 @@ bool Config::addGame(Game &game) {
                 bool smoothing = false;
                 int multiplier = 1;
                 bool relative_mode = false;
-                int delay_buffer_depth = 0;
+                uint32_t delay = 0;
                 tinyxml2::XMLError err1 = gameAnalogNode->QueryIntAttribute("index", &index);
                 gameAnalogNode->QueryFloatAttribute("sensivity", &sensitivity);
                 gameAnalogNode->QueryFloatAttribute("deadzone", &deadzone);
@@ -257,7 +262,7 @@ bool Config::addGame(Game &game) {
                 gameAnalogNode->QueryBoolAttribute("smoothing", &smoothing);
                 gameAnalogNode->QueryIntAttribute("multiplier", &multiplier);
                 gameAnalogNode->QueryBoolAttribute("relative", &relative_mode);
-                gameAnalogNode->QueryIntAttribute("delay", &delay_buffer_depth);
+                gameAnalogNode->QueryUnsignedAttribute("delay_ms", &delay);
                 const char *devid = gameAnalogNode->Attribute("devid");
 
                 if (err1 != tinyxml2::XMLError::XML_SUCCESS || !devid) {
@@ -273,7 +278,7 @@ bool Config::addGame(Game &game) {
                     gameAnalogNode->SetAttribute("smoothing", it.getSmoothing());
                     gameAnalogNode->SetAttribute("multiplier", it.getMultiplier());
                     gameAnalogNode->SetAttribute("relative", it.isRelativeMode());
-                    gameAnalogNode->SetAttribute("delay", it.getDelayBufferDepth());
+                    gameAnalogNode->SetAttribute("delay_ms", it.getDelayMs());
                     gameAnalogsNode->InsertEndChild(gameAnalogNode);
                 } else {
                     it.setIndex(static_cast<unsigned short int>(index));
@@ -285,7 +290,7 @@ bool Config::addGame(Game &game) {
                     it.setSmoothing(smoothing);
                     it.setMultiplier(multiplier);
                     it.setRelativeMode(relative_mode);
-                    it.setDelayBufferDepth(delay_buffer_depth);
+                    it.setDelayMs(delay);
                 }
             } else {
                 gameAnalogNode = this->configFile.NewElement("analog");
@@ -298,7 +303,7 @@ bool Config::addGame(Game &game) {
                 gameAnalogNode->SetAttribute("smoothing", it.getSmoothing());
                 gameAnalogNode->SetAttribute("multiplier", it.getMultiplier());
                 gameAnalogNode->SetAttribute("relative", it.isRelativeMode());
-                gameAnalogNode->SetAttribute("delay", it.getDelayBufferDepth());
+                gameAnalogNode->SetAttribute("delay_ms", it.getDelayMs());
                 gameAnalogNode->SetAttribute("devid", it.getDeviceIdentifier().c_str());
                 gameAnalogsNode->InsertEndChild(gameAnalogNode);
             }
@@ -432,6 +437,7 @@ bool Config::addGame(Game &game) {
             gameButtonNode->SetAttribute("analogtype", it.getAnalogType());
             gameButtonNode->SetAttribute("debounce_up", it.getDebounceUp());
             gameButtonNode->SetAttribute("debounce_down", it.getDebounceDown());
+            gameButtonNode->SetAttribute("bat_threshold", it.getBatThreshold());
             gameButtonNode->SetAttribute("velocity_threshold", it.getVelocityThreshold());
             gameButtonNode->SetAttribute("invert", it.getInvert());
             gameButtonNode->SetAttribute("devid", it.getDeviceIdentifier().c_str());
@@ -452,7 +458,7 @@ bool Config::addGame(Game &game) {
             gameAnalogNode->SetAttribute("smoothing", it.getSmoothing());
             gameAnalogNode->SetAttribute("multiplier", it.getMultiplier());
             gameAnalogNode->SetAttribute("relative", it.isRelativeMode());
-            gameAnalogNode->SetAttribute("delay", it.getDelayBufferDepth());
+            gameAnalogNode->SetAttribute("delay_ms", it.getDelayMs());
             gameAnalogsNode->InsertEndChild(gameAnalogNode);
         }
 
@@ -480,13 +486,20 @@ bool Config::addGame(Game &game) {
         rootNode->InsertEndChild(gameNode);
     }
 
-    // save config
-    this->saveConfigFile();
+    // save config (skipped when caller batches multiple addGame calls
+    // and flushes once at the end via Config::save())
+    if (save) {
+        this->saveConfigFile();
+    }
 
     // return success
     return true;
 }
 
+void Config::save() {
+    this->saveConfigFile();
+}
+
 bool Config::updateBinding(const Game &game, const Button &button, int alternative) {
 
     // get root node
@@ -535,6 +548,7 @@ bool Config::updateBinding(const Game &game, const Button &button, int alternati
                 gameButtonNode->SetAttribute("analogtype", (int) button.getAnalogType());
                 gameButtonNode->SetAttribute("debounce_up", button.getDebounceUp());
                 gameButtonNode->SetAttribute("debounce_down", button.getDebounceDown());
+                gameButtonNode->SetAttribute("bat_threshold", button.getBatThreshold());
                 gameButtonNode->SetAttribute("velocity_threshold", button.getVelocityThreshold());
                 gameButtonNode->SetAttribute("invert", button.getInvert());
                 gameButtonNode->SetAttribute("devid", button.getDeviceIdentifier().c_str());
@@ -551,6 +565,7 @@ bool Config::updateBinding(const Game &game, const Button &button, int alternati
             gameButtonNode->SetAttribute("analogtype", 0);
             gameButtonNode->SetAttribute("debounce_up", 0.0);
             gameButtonNode->SetAttribute("debounce_down", 0.0);
+            gameButtonNode->SetAttribute("bat_threshold", 0);
             gameButtonNode->SetAttribute("velocity_threshold", 0);
             gameButtonNode->SetAttribute("invert", false);
             gameButtonNode->SetAttribute("devid", "");
@@ -684,7 +699,7 @@ bool Config::updateBinding(const Game &game, const Analog &analog) {
         gameAnalogNode->SetAttribute("smoothing", analog.getSmoothing());
         gameAnalogNode->SetAttribute("multiplier", analog.getMultiplier());
         gameAnalogNode->SetAttribute("relative", analog.isRelativeMode());
-        gameAnalogNode->SetAttribute("delay", analog.getDelayBufferDepth());
+        gameAnalogNode->SetAttribute("delay_ms", analog.getDelayMs());
         gameAnalogNode->SetAttribute("devid", analog.getDeviceIdentifier().c_str());
     } else {
         gameAnalogNode = this->configFile.NewElement("analog");
@@ -696,7 +711,7 @@ bool Config::updateBinding(const Game &game, const Analog &analog) {
         gameAnalogNode->SetAttribute("smoothing", analog.getSmoothing());
         gameAnalogNode->SetAttribute("multiplier", analog.getMultiplier());
         gameAnalogNode->SetAttribute("relative", analog.isRelativeMode());
-        gameAnalogNode->SetAttribute("delay", analog.getDelayBufferDepth());
+        gameAnalogNode->SetAttribute("delay_ms", analog.getDelayMs());
         gameAnalogNode->SetAttribute("devid", analog.getDeviceIdentifier().c_str());
         gameAnalogsNode->InsertEndChild(gameAnalogNode);
     }
@@ -946,12 +961,14 @@ std::vector<Button> Config::getButtons(const std::string &gameName) {
                 auto analogType = (int) BAT_NONE;
                 double debounce_up = 0.0;
                 double debounce_down = 0.0;
+                int bat_threshold = 0;
                 int velocity_threshold = 0;
                 bool invert = false;
                 gameButtonNode->QueryIntAttribute("vkey", &vKey);
                 gameButtonNode->QueryIntAttribute("analogtype", &analogType);
                 gameButtonNode->QueryDoubleAttribute("debounce_up", &debounce_up);
                 gameButtonNode->QueryDoubleAttribute("debounce_down", &debounce_down);
+                gameButtonNode->QueryIntAttribute("bat_threshold", &bat_threshold);
                 gameButtonNode->QueryIntAttribute("velocity_threshold", &velocity_threshold);
                 gameButtonNode->QueryBoolAttribute("invert", &invert);
                 const char *devid = gameButtonNode->Attribute("devid");
@@ -966,6 +983,7 @@ std::vector<Button> Config::getButtons(const std::string &gameName) {
                         alt.setAnalogType((ButtonAnalogType) analogType);
                         alt.setDebounceUp(debounce_up);
                         alt.setDebounceDown(debounce_down);
+                        alt.setBatThreshold(bat_threshold);
                         alt.setVelocityThreshold(velocity_threshold);
                         alt.setInvert(invert);
                         if (devid) {
@@ -984,6 +1002,7 @@ std::vector<Button> Config::getButtons(const std::string &gameName) {
                     button.setAnalogType((ButtonAnalogType) analogType);
                     button.setDebounceUp(debounce_up);
                     button.setDebounceDown(debounce_down);
+                    button.setBatThreshold(bat_threshold);
                     button.setVelocityThreshold(velocity_threshold);
                     button.setInvert(invert);
                     if (devid) {
@@ -1130,7 +1149,7 @@ std::vector<Analog> Config::getAnalogs(const std::string &gameName) {
                 bool smoothing = false;
                 int multiplier = 1;
                 bool relative_mode = false;
-                int delay_buffer_depth = 0;
+                uint32_t delay = 0;
                 gameAnalogNode->QueryIntAttribute("index", &index);
                 gameAnalogNode->QueryFloatAttribute("sensivity", &sensitivity);
                 gameAnalogNode->QueryFloatAttribute("deadzone", &deadzone);
@@ -1139,7 +1158,7 @@ std::vector<Analog> Config::getAnalogs(const std::string &gameName) {
                 gameAnalogNode->QueryBoolAttribute("smoothing", &smoothing);
                 gameAnalogNode->QueryIntAttribute("multiplier", &multiplier);
                 gameAnalogNode->QueryBoolAttribute("relative", &relative_mode);
-                gameAnalogNode->QueryIntAttribute("delay", &delay_buffer_depth);
+                gameAnalogNode->QueryUnsignedAttribute("delay_ms", &delay);
                 const char *devid = gameAnalogNode->Attribute("devid");
 
                 // create analog and add to list
@@ -1152,7 +1171,7 @@ std::vector<Analog> Config::getAnalogs(const std::string &gameName) {
                 analog.setSmoothing(smoothing);
                 analog.setMultiplier(multiplier);
                 analog.setRelativeMode(relative_mode);
-                analog.setDelayBufferDepth(delay_buffer_depth);
+                analog.setDelayMs(delay);
                 if (devid) {
                     analog.setDeviceIdentifier(devid);
                 }
@@ -1315,7 +1334,7 @@ void Config::saveConfigFile() {
     // create a .tmp file and write to it...
     const auto xml_result = this->configFile.SaveFile(this->configLocationTemp.c_str(), false);
     if (xml_result != tinyxml2::XMLError::XML_SUCCESS) {
-        log_info("cfg", "failed to write file: {}", this->configLocationTemp.string());
+        log_info("cfg", "failed to write file: {}", this->configLocationTemp);
         return;
     }
     // copy the .tmp file to the main file...
@@ -1325,4 +1344,4 @@ void Config::saveConfigFile() {
     }
     // delete the .tmp file (not critical if this fails)
     DeleteFileW(this->configLocationTemp.c_str());
-}
+}

src/spice2x/cfg/config.h

@@ -22,7 +22,12 @@ public:
     bool getStatus();
     bool createConfigFile();
 
-    bool addGame(Game &game);
+    bool addGame(Game &game, bool save = true);
+
+    // flush pending in-memory changes to disk; intended to be paired with
+    // addGame(game, false) in bulk-init paths so we write the XML once instead
+    // of once per game
+    void save();
 
     bool updateBinding(const Game &game, const Button &button, int alternative);
     bool updateBinding(const Game &game, const Analog &analog);

src/spice2x/cfg/configurator.cpp

@@ -1,11 +1,15 @@
-#include "configurator.h"
+#include "configurator.h"
+
+#include <d3d9.h>
 
 #include "overlay/overlay.h"
+#include "util/logging.h"
 
 namespace cfg {
 
     // globals
     bool CONFIGURATOR_STANDALONE = false;
+    bool CONFIGURATOR_FORCE_SOFTWARE_RENDER = false;
     ConfigType CONFIGURATOR_TYPE = ConfigType::Config;
 
     Configurator::Configurator() {
@@ -16,11 +20,78 @@ namespace cfg {
         CONFIGURATOR_STANDALONE = false;
     }
 
+    // Attempt to bring up a D3D9 device backing the configurator's window so the
+    // overlay can use the hardware-accelerated imgui_impl_dx9 path instead of
+    // the CPU rasterizer. Returns true if both Direct3DCreate9 and CreateDevice
+    // succeed; the caller falls back to overlay::create_software() otherwise so
+    // that environments without D3D9 (Wine without dxvk, headless test boxes,
+    // GPUs whose drivers reject HAL) still get a working configurator.
+    static bool try_init_d3d9(ConfiguratorWindow &wnd) {
+        if (cfg::CONFIGURATOR_FORCE_SOFTWARE_RENDER) {
+            return false;
+        }
+
+        if (wnd.hWnd == nullptr) {
+            return false;
+        }
+
+        IDirect3D9 *d3d = Direct3DCreate9(D3D_SDK_VERSION);
+        if (d3d == nullptr) {
+            log_warning("configurator", "Direct3DCreate9 returned NULL, falling back to software renderer");
+            return false;
+        }
+
+        D3DPRESENT_PARAMETERS pp {};
+        pp.Windowed = TRUE;
+        pp.SwapEffect = D3DSWAPEFFECT_DISCARD;
+        // D3DFMT_UNKNOWN -> driver picks the current desktop format
+        pp.BackBufferFormat = D3DFMT_UNKNOWN;
+        pp.BackBufferWidth = static_cast<UINT>(wnd.client_width);
+        pp.BackBufferHeight = static_cast<UINT>(wnd.client_height);
+        pp.hDeviceWindow = wnd.hWnd;
+        pp.EnableAutoDepthStencil = FALSE;
+        pp.PresentationInterval = D3DPRESENT_INTERVAL_ONE;
+
+        IDirect3DDevice9 *device = nullptr;
+        // SOFTWARE_VERTEXPROCESSING is the most compatible behavior flag; the UI
+        // is tiny so we don't need hardware T&L. FPU_PRESERVE keeps our floating
+        // point environment intact in case other spice code relies on it.
+        const DWORD behavior_flags = D3DCREATE_SOFTWARE_VERTEXPROCESSING | D3DCREATE_FPU_PRESERVE;
+        HRESULT hr = d3d->CreateDevice(
+                D3DADAPTER_DEFAULT,
+                D3DDEVTYPE_HAL,
+                wnd.hWnd,
+                behavior_flags,
+                &pp,
+                &device);
+        if (FAILED(hr) || device == nullptr) {
+            log_warning("configurator",
+                "D3D9 CreateDevice failed (hr={:#x}), falling back to software renderer",
+                static_cast<unsigned int>(hr));
+            d3d->Release();
+            return false;
+        }
+
+        wnd.d3d = d3d;
+        wnd.device = device;
+        wnd.pp = pp;
+        wnd.use_d3d9 = true;
+        return true;
+    }
+
     void Configurator::run() {
 
-        // create instance
+        // bring up the overlay against either a real D3D9 device or the software
+        // rasterizer. The choice is one-shot - the in-game overlay always uses
+        // the same renderer the configurator picked here.
         overlay::ENABLED = true;
-        overlay::create_software(this->wnd.hWnd);
+        if (try_init_d3d9(this->wnd)) {
+            log_info("configurator", "using D3D9 hardware-accelerated renderer");
+            overlay::create_d3d9(this->wnd.hWnd, this->wnd.d3d, this->wnd.device);
+        } else {
+            log_info("configurator", "using software renderer");
+            overlay::create_software(this->wnd.hWnd);
+        }
         overlay::OVERLAY->set_active(true);
         overlay::OVERLAY->hotkeys_enable = false;
         ImGui::GetIO().MouseDrawCursor = false;

src/spice2x/cfg/configurator.h

@@ -11,6 +11,7 @@ namespace cfg {
     // globals
     extern bool CONFIGURATOR_STANDALONE;
     extern ConfigType CONFIGURATOR_TYPE;
+    extern bool CONFIGURATOR_FORCE_SOFTWARE_RENDER;
 
     class Configurator {
     private:

src/spice2x/cfg/configurator_wnd.cpp

@@ -1,11 +1,17 @@
-#include "configurator_wnd.h"
+#include "configurator_wnd.h"
+
+#include <algorithm>
+#include <cstring>
 
 #include <windows.h>
+#include <windowsx.h>
 
 #include "build/defs.h"
+#include "external/imgui/imgui.h"
 #include "launcher/shutdown.h"
 #include "overlay/overlay.h"
 #include "util/logging.h"
+#include "util/precise_timer.h"
 #include "cfg/configurator.h"
 
 #include "icon.h"
@@ -16,6 +22,156 @@ static int WINDOW_SIZE_X = 800;
 static int WINDOW_SIZE_Y = 600;
 static HICON WINDOW_ICON = LoadIcon(GetModuleHandle(nullptr), MAKEINTRESOURCE(MAINICON));
 
+static const UINT_PTR RENDER_TIMER_ID = 1;
+
+// Map a virtual key code to the matching ImGuiKey. Mirrors the table used by
+// the in-game ImGui spice backend so navigation keys behave identically when
+// the configurator runs standalone and drives ImGui from Win32 messages.
+static ImGuiKey vk_to_imgui_key(WPARAM vkey) {
+    switch (vkey) {
+        case VK_TAB: return ImGuiKey_Tab;
+        case VK_LEFT: return ImGuiKey_LeftArrow;
+        case VK_RIGHT: return ImGuiKey_RightArrow;
+        case VK_UP: return ImGuiKey_UpArrow;
+        case VK_DOWN: return ImGuiKey_DownArrow;
+        case VK_PRIOR: return ImGuiKey_PageUp;
+        case VK_NEXT: return ImGuiKey_PageDown;
+        case VK_HOME: return ImGuiKey_Home;
+        case VK_END: return ImGuiKey_End;
+        case VK_INSERT: return ImGuiKey_Insert;
+        case VK_DELETE: return ImGuiKey_Delete;
+        case VK_BACK: return ImGuiKey_Backspace;
+        case VK_SPACE: return ImGuiKey_Space;
+        case VK_RETURN: return ImGuiKey_Enter;
+        case VK_ESCAPE: return ImGuiKey_Escape;
+        case VK_LSHIFT: return ImGuiKey_LeftShift;
+        case VK_RSHIFT: return ImGuiKey_RightShift;
+        case VK_SHIFT: return ImGuiKey_LeftShift;
+        case VK_LCONTROL: return ImGuiKey_LeftCtrl;
+        case VK_RCONTROL: return ImGuiKey_RightCtrl;
+        case VK_CONTROL: return ImGuiKey_LeftCtrl;
+        case 'A': return ImGuiKey_A;
+        case 'C': return ImGuiKey_C;
+        case 'V': return ImGuiKey_V;
+        case 'X': return ImGuiKey_X;
+        case 'Y': return ImGuiKey_Y;
+        case 'Z': return ImGuiKey_Z;
+        default: return ImGuiKey_None;
+    }
+}
+
+static ImGuiKey vk_to_imgui_mod_key(WPARAM vkey) {
+    switch (vkey) {
+        case VK_SHIFT:
+        case VK_LSHIFT:
+        case VK_RSHIFT:
+            return ImGuiMod_Shift;
+        case VK_CONTROL:
+        case VK_LCONTROL:
+        case VK_RCONTROL:
+            return ImGuiMod_Ctrl;
+        case VK_MENU:
+        case VK_LMENU:
+        case VK_RMENU:
+            return ImGuiMod_Alt;
+        default:
+            return ImGuiMod_None;
+    }
+}
+
+static cfg::ConfiguratorWindow *get_state(HWND hWnd) {
+    return reinterpret_cast<cfg::ConfiguratorWindow *>(
+        GetWindowLongPtrW(hWnd, GWLP_USERDATA));
+}
+
+// Returns the refresh rate (Hz) of the monitor the window currently lives on,
+// clamped to a sane range. Falls back to 60 if detection fails or the value
+// looks invalid (DEVMODE may report 0 or 1 to mean "use hardware default").
+static UINT detect_monitor_refresh_hz(HWND hWnd) {
+    UINT hz = 0;
+    HMONITOR mon = MonitorFromWindow(hWnd, MONITOR_DEFAULTTONEAREST);
+    if (mon) {
+        MONITORINFOEXW mi {};
+        mi.cbSize = sizeof(mi);
+        if (GetMonitorInfoW(mon, reinterpret_cast<LPMONITORINFO>(&mi))) {
+            DEVMODEW dm {};
+            dm.dmSize = sizeof(dm);
+            if (EnumDisplaySettingsW(mi.szDevice, ENUM_CURRENT_SETTINGS, &dm)) {
+                hz = dm.dmDisplayFrequency;
+            }
+        }
+    }
+    if (hz <= 1) {
+        HDC hdc = GetDC(hWnd);
+        if (hdc) {
+            int v = GetDeviceCaps(hdc, VREFRESH);
+            if (v > 1) {
+                hz = static_cast<UINT>(v);
+            }
+            ReleaseDC(hWnd, hdc);
+        }
+    }
+    if (hz < 30 || hz > 240) {
+        hz = 60;
+    }
+    return hz;
+}
+
+// Software-path WM_PAINT: blit the overlay's pixel buffer to the window using
+// SetDIBitsToDevice. Avoids creating/destroying a GDI HBITMAP every frame.
+static void paint_software(HWND hWnd) {
+    if (!overlay::OVERLAY) {
+        PAINTSTRUCT ps {};
+        BeginPaint(hWnd, &ps);
+        EndPaint(hWnd, &ps);
+        return;
+    }
+
+    int width = 0;
+    int height = 0;
+    uint32_t *pixel_data = overlay::OVERLAY->sw_get_pixel_data(&width, &height);
+
+    PAINTSTRUCT paint {};
+    HDC hdc = BeginPaint(hWnd, &paint);
+
+    if (pixel_data && width > 0 && height > 0) {
+        BITMAPINFO bmi {};
+        bmi.bmiHeader.biSize = sizeof(BITMAPINFOHEADER);
+        bmi.bmiHeader.biWidth = width;
+        // negative height -> top-down DIB, matching how ImGui packs pixels
+        bmi.bmiHeader.biHeight = -height;
+        bmi.bmiHeader.biPlanes = 1;
+        bmi.bmiHeader.biBitCount = 32;
+        bmi.bmiHeader.biCompression = BI_RGB;
+
+        SetDIBitsToDevice(
+            hdc,
+            0, 0,
+            width, height,
+            0, 0,
+            0, height,
+            pixel_data,
+            &bmi,
+            DIB_RGB_COLORS);
+    }
+
+    EndPaint(hWnd, &paint);
+}
+
+void cfg::ConfiguratorWindow::start_timer() {
+    if (!this->timer_running && this->hWnd) {
+        SetTimer(this->hWnd, RENDER_TIMER_ID, this->timer_interval_ms, nullptr);
+        this->timer_running = true;
+    }
+}
+
+void cfg::ConfiguratorWindow::stop_timer() {
+    if (this->timer_running && this->hWnd) {
+        KillTimer(this->hWnd, RENDER_TIMER_ID);
+        this->timer_running = false;
+    }
+}
+
 cfg::ConfiguratorWindow::ConfiguratorWindow() {
 
     // register the window class
@@ -27,12 +183,22 @@ cfg::ConfiguratorWindow::ConfiguratorWindow() {
     wc.hIcon = WINDOW_ICON;
     RegisterClass(&wc);
 
+    // raise SetTimer resolution so high refresh rates (120/144Hz) are actually
+    // achievable. Without this, USER timers quantize to ~15.6ms and cap us
+    // near ~64 FPS. Uses the shared helper so it respects -notimerhacks
+    // (Use Legacy Timers) and the Win11 timer-throttling opt-out. The helper
+    // intentionally never calls timeEndPeriod; the kernel releases the request
+    // when spicecfg exits.
+    timeutils::set_timer_resolution();
+
     // determine window title
     if (cfg::CONFIGURATOR_TYPE == cfg::ConfigType::Config) {
         WINDOW_TITLE = "spice2x config (" + to_string(VERSION_STRING_CFG) + ")";
         WINDOW_SIZE_X = 800;
         WINDOW_SIZE_Y = 600;
     }
+    this->client_width = WINDOW_SIZE_X;
+    this->client_height = WINDOW_SIZE_Y;
 
     // open window
     this->hWnd = CreateWindowEx(
@@ -53,11 +219,23 @@ cfg::ConfiguratorWindow::ConfiguratorWindow() {
 
 cfg::ConfiguratorWindow::~ConfiguratorWindow() {
 
+    this->stop_timer();
+
     // close window
     DestroyWindow(this->hWnd);
 
     // unregister class
     UnregisterClass(CLASS_NAME, GetModuleHandle(NULL));
+
+    // release D3D9 resources if any
+    if (this->device) {
+        this->device->Release();
+        this->device = nullptr;
+    }
+    if (this->d3d) {
+        this->d3d->Release();
+        this->d3d = nullptr;
+    }
 }
 
 void cfg::ConfiguratorWindow::run() {
@@ -67,8 +245,23 @@ void cfg::ConfiguratorWindow::run() {
     ShowWindow(this->hWnd, SW_SHOWNORMAL);
     UpdateWindow(this->hWnd);
 
-    // draw overlay in 60 FPS
-    SetTimer(this->hWnd, 1, 1000 / 60, nullptr);
+    // SW_SHOWNORMAL usually activates a top-level window, but not always (e.g.,
+    // when the launching process doesn't have foreground rights to transfer).
+    // Force foreground+focus explicitly so WM_MOUSEWHEEL is routed to us from
+    // the very first frame; the Win32 default routes wheel events to the
+    // keyboard-focused window.
+    SetForegroundWindow(this->hWnd);
+    SetFocus(this->hWnd);
+
+    // match the render timer to the monitor refresh rate so scrolling stays smooth
+    // on 60/120/144 Hz panels. Idle CPU is bounded by overlay::sw_pixels_dirty
+    // (software path) and overlay::d3d9_frame_dirty (D3D9 path); the timer is also
+    // paused entirely when the window is minimized (see WM_SIZE handler).
+    const UINT hz = detect_monitor_refresh_hz(this->hWnd);
+    this->timer_interval_ms = std::max<UINT>(1, 1000 / hz);
+    log_info("configurator", "render timer {} ms ({} Hz)",
+        this->timer_interval_ms, hz);
+    this->start_timer();
 
     // window loop
     BOOL ret;
@@ -102,6 +295,72 @@ LRESULT CALLBACK cfg::ConfiguratorWindow::window_proc(HWND hWnd, UINT uMsg, WPAR
 
             break;
         }
+        case WM_SIZE: {
+
+            // pause/resume the render timer based on visibility
+            auto *state = get_state(hWnd);
+            if (state) {
+                if (wParam == SIZE_MINIMIZED) {
+                    state->window_minimized = true;
+                    state->stop_timer();
+                } else {
+                    if (state->window_minimized) {
+                        state->window_minimized = false;
+                        // force a full repaint on the next frame
+                        state->has_valid_draw_hash = false;
+                        state->start_timer();
+                    }
+                    const int new_w = LOWORD(lParam);
+                    const int new_h = HIWORD(lParam);
+                    if (new_w > 0 && new_h > 0
+                            && (new_w != state->client_width || new_h != state->client_height)) {
+                        state->client_width = new_w;
+                        state->client_height = new_h;
+
+                        // reset the D3D9 device with the new back-buffer size so the
+                        // hardware-accelerated path matches the window dimensions.
+                        if (state->use_d3d9 && state->device) {
+                            if (overlay::OVERLAY) {
+                                overlay::OVERLAY->reset_invalidate();
+                            }
+                            state->pp.BackBufferWidth = static_cast<UINT>(new_w);
+                            state->pp.BackBufferHeight = static_cast<UINT>(new_h);
+                            HRESULT hr = state->device->Reset(&state->pp);
+                            if (FAILED(hr)) {
+                                log_warning("configurator", "D3D9 device Reset failed, hr={:#x}",
+                                    static_cast<unsigned int>(hr));
+                            }
+                            if (overlay::OVERLAY) {
+                                overlay::OVERLAY->reset_recreate();
+                            }
+                        }
+
+                        // force a full repaint after resize
+                        state->has_valid_draw_hash = false;
+                    }
+                }
+            }
+            break;
+        }
+        case WM_DISPLAYCHANGE: {
+
+            // monitor refresh rate may have changed (or the window moved to a
+            // different monitor); re-detect and re-arm the render timer.
+            auto *state = get_state(hWnd);
+            if (state && state->timer_running) {
+                const UINT hz = detect_monitor_refresh_hz(hWnd);
+                const UINT new_interval = std::max<UINT>(1, 1000 / hz);
+                if (new_interval != state->timer_interval_ms) {
+                    state->stop_timer();
+                    state->timer_interval_ms = new_interval;
+                    state->start_timer();
+                    log_info("configurator",
+                        "WM_DISPLAYCHANGE: render timer {} ms ({} Hz)",
+                        new_interval, hz);
+                }
+            }
+            break;
+        }
         case WM_CLOSE:
         case WM_DESTROY: {
 
@@ -110,62 +369,192 @@ LRESULT CALLBACK cfg::ConfiguratorWindow::window_proc(HWND hWnd, UINT uMsg, WPAR
             break;
         }
         case WM_TIMER: {
+            if (wParam != RENDER_TIMER_ID) {
+                break;
+            }
 
-            // update overlay
+            auto *state = get_state(hWnd);
+
+            // skip rendering when the window is minimized or hidden - the timer is
+            // already paused on minimize, but defensively skip here as well.
+            if (state && (state->window_minimized || !IsWindowVisible(hWnd))) {
+                break;
+            }
+
+            const bool use_d3d9 = state && state->use_d3d9 && state->device;
+
+            // build the imgui frame (input is already buffered via WM_* messages)
             if (overlay::OVERLAY) {
                 overlay::OVERLAY->update();
                 overlay::OVERLAY->set_active(true);
                 overlay::OVERLAY->new_frame();
-                overlay::OVERLAY->render();
             }
 
-            // repaint window
-            InvalidateRect(hWnd, nullptr, TRUE);
+            if (use_d3d9) {
+                const HRESULT cl = state->device->TestCooperativeLevel();
+                if (cl == D3DERR_DEVICELOST) {
+                    break;
+                }
+                if (cl == D3DERR_DEVICENOTRESET) {
+                    if (overlay::OVERLAY) {
+                        overlay::OVERLAY->reset_invalidate();
+                    }
+                    const HRESULT reset_hr = state->device->Reset(&state->pp);
+                    if (SUCCEEDED(reset_hr) && overlay::OVERLAY) {
+                        overlay::OVERLAY->reset_recreate();
+                    }
+                    break;
+                }
+                if (FAILED(cl)) {
+                    break;
+                }
+
+                if (overlay::OVERLAY) {
+                    overlay::OVERLAY->render();
+                }
+
+                // skip Clear/Present when ImGui draw data is unchanged; the last
+                // presented frame stays visible (same fast path as sw_pixels_dirty).
+                const bool dirty = overlay::OVERLAY && overlay::OVERLAY->d3d9_frame_dirty;
+                const bool force_present = state && !state->has_valid_draw_hash;
+                if (dirty || force_present) {
+                    if (state) {
+                        state->has_valid_draw_hash = true;
+                    }
+                    state->device->Clear(0, nullptr, D3DCLEAR_TARGET,
+                        D3DCOLOR_RGBA(20, 18, 18, 255), 1.0f, 0);
+                    if (SUCCEEDED(state->device->BeginScene())) {
+                        if (overlay::OVERLAY) {
+                            overlay::OVERLAY->d3d9_render_draw(force_present);
+                        }
+                        state->device->EndScene();
+                    }
+                    const HRESULT hr = state->device->Present(nullptr, nullptr, nullptr, nullptr);
+                    if (hr == D3DERR_DEVICELOST) {
+                        break;
+                    }
+                }
+            } else {
+                if (overlay::OVERLAY) {
+                    overlay::OVERLAY->render();
+                }
+                // software path: the overlay's renderer already skipped the costly
+                // paint_imgui call when the draw data was unchanged. Only invalidate
+                // the window when those pixels actually changed, or on the first
+                // frame / after a resize when we lost the previously cached state.
+                const bool dirty = overlay::OVERLAY && overlay::OVERLAY->sw_pixels_dirty;
+                const bool force_repaint = state && !state->has_valid_draw_hash;
+                if (dirty || force_repaint) {
+                    if (state) {
+                        state->has_valid_draw_hash = true;
+                    }
+                    // pass FALSE for bErase - WM_ERASEBKGND already short-circuits, so
+                    // skipping the erase region avoids one extra Win32 round trip.
+                    InvalidateRect(hWnd, nullptr, FALSE);
+                }
+            }
+
             break;
         }
         case WM_ERASEBKGND: {
             return 1;
         }
         case WM_PAINT: {
-
-            // render overlay
-            if (overlay::OVERLAY) {
-
-                // get pixel data
-                int width, height;
-                uint32_t *pixel_data = overlay::OVERLAY->sw_get_pixel_data(&width, &height);
-                if (width > 0 && height > 0) {
-
-                    // create bitmap
-                    HBITMAP bitmap = CreateBitmap(width, height, 1, 8 * sizeof(uint32_t), pixel_data);
-
-                    // prepare paint
-                    PAINTSTRUCT paint{};
-                    HDC hdc = BeginPaint(hWnd, &paint);
-                    HDC hdcMem = CreateCompatibleDC(hdc);
-                    SetBkMode(hdc, TRANSPARENT);
-
-                    // draw bitmap
-                    SelectObject(hdcMem, bitmap);
-                    BitBlt(hdc, paint.rcPaint.left, paint.rcPaint.top,
-                            paint.rcPaint.right - paint.rcPaint.left,
-                            paint.rcPaint.bottom - paint.rcPaint.top,
-                            hdcMem, paint.rcPaint.left, paint.rcPaint.top, SRCCOPY);
-
-                    // delete bitmap
-                    DeleteObject(bitmap);
-
-                    // clean up
-                    DeleteDC(hdcMem);
-                    EndPaint(hWnd, &paint);
-                } else {
-                    return DefWindowProc(hWnd, uMsg, wParam, lParam);
-                }
+            paint_software(hWnd);
+            break;
+        }
+        case WM_ACTIVATE: {
+            const WORD activation = LOWORD(wParam);
+            if (activation == WA_INACTIVE) {
+                // dropping any held mouse buttons so we don't get stuck in a
+                // "button still pressed" state when we come back.
+                auto &io = ImGui::GetIO();
+                io.AddMouseButtonEvent(ImGuiMouseButton_Left, false);
+                io.AddMouseButtonEvent(ImGuiMouseButton_Right, false);
+                io.AddMouseButtonEvent(ImGuiMouseButton_Middle, false);
             } else {
-                return DefWindowProc(hWnd, uMsg, wParam, lParam);
+                // WA_ACTIVE or WA_CLICKACTIVE: make sure the keyboard focus is
+                // actually on this window so WM_MOUSEWHEEL gets routed to us
+                // again. Without this, scrolling silently stops working after
+                // the user alt-tabs back to spicecfg.
+                SetFocus(hWnd);
             }
             break;
         }
+        case WM_KILLFOCUS: {
+            // mirror the WA_INACTIVE mouse-button drop for the keyboard -
+            // without this, a key held during alt-tab stays "down" in ImGui
+            // state until the user presses and releases it again.
+            ImGui::GetIO().ClearInputKeys();
+            break;
+        }
+        // input messages routed straight into ImGui IO. This replaces the previous
+        // per-frame 256-VK rawinput scan in ImGui_ImplSpice_NewFrame for the
+        // standalone configurator (see CONFIGURATOR_STANDALONE branches there).
+        case WM_KEYDOWN:
+        case WM_SYSKEYDOWN:
+        case WM_KEYUP:
+        case WM_SYSKEYUP: {
+            if (wParam == VK_F4) {
+                return DefWindowProc(hWnd, uMsg, wParam, lParam);
+            }
+            const bool down = (uMsg == WM_KEYDOWN) || (uMsg == WM_SYSKEYDOWN);
+            auto &io = ImGui::GetIO();
+            const ImGuiKey key = vk_to_imgui_key(wParam);
+            if (key != ImGuiKey_None) {
+                io.AddKeyEvent(key, down);
+            }
+            const ImGuiKey mod = vk_to_imgui_mod_key(wParam);
+            if (mod != ImGuiMod_None) {
+                io.AddKeyEvent(mod, down);
+            }
+            break;
+        }
+        case WM_MOUSEMOVE: {
+            auto &io = ImGui::GetIO();
+            io.AddMousePosEvent(static_cast<float>(GET_X_LPARAM(lParam)),
+                                static_cast<float>(GET_Y_LPARAM(lParam)));
+            break;
+        }
+        case WM_LBUTTONDOWN:
+        case WM_LBUTTONUP:
+        case WM_RBUTTONDOWN:
+        case WM_RBUTTONUP:
+        case WM_MBUTTONDOWN:
+        case WM_MBUTTONUP: {
+            int button = 0;
+            bool down = false;
+            switch (uMsg) {
+                case WM_LBUTTONDOWN: button = ImGuiMouseButton_Left; down = true; break;
+                case WM_LBUTTONUP: button = ImGuiMouseButton_Left; down = false; break;
+                case WM_RBUTTONDOWN: button = ImGuiMouseButton_Right; down = true; break;
+                case WM_RBUTTONUP: button = ImGuiMouseButton_Right; down = false; break;
+                case WM_MBUTTONDOWN: button = ImGuiMouseButton_Middle; down = true; break;
+                case WM_MBUTTONUP: button = ImGuiMouseButton_Middle; down = false; break;
+            }
+            auto &io = ImGui::GetIO();
+            io.AddMouseButtonEvent(button, down);
+            if (down) {
+                SetCapture(hWnd);
+            } else {
+                ReleaseCapture();
+            }
+            break;
+        }
+        case WM_MOUSEWHEEL: {
+            const float delta = static_cast<float>(GET_WHEEL_DELTA_WPARAM(wParam))
+                / static_cast<float>(WHEEL_DELTA);
+            ImGui::GetIO().AddMouseWheelEvent(0.0f, delta);
+            break;
+        }
+#if !SPICE_XP
+        case WM_MOUSEHWHEEL: {
+            const float delta = static_cast<float>(GET_WHEEL_DELTA_WPARAM(wParam))
+                / static_cast<float>(WHEEL_DELTA);
+            ImGui::GetIO().AddMouseWheelEvent(delta, 0.0f);
+            break;
+        }
+#endif
         default:
             return DefWindowProc(hWnd, uMsg, wParam, lParam);
     }

src/spice2x/cfg/configurator_wnd.h

@@ -1,6 +1,9 @@
-#pragma once
+#pragma once
+
+#include <cstdint>
 
 #include <windows.h>
+#include <d3d9.h>
 
 namespace cfg {
 
@@ -9,11 +12,38 @@ namespace cfg {
 
         HWND hWnd;
 
+        // optional D3D9 device backing the configurator window; nullptr when running
+        // the software-rendered path. Owned by ConfiguratorWindow when set.
+        IDirect3D9 *d3d = nullptr;
+        IDirect3DDevice9 *device = nullptr;
+        D3DPRESENT_PARAMETERS pp {};
+        bool use_d3d9 = false;
+
+        // throttling / pause state. timer_interval_ms is the default until run()
+        // refines it to the actual monitor refresh rate (see detect_monitor_refresh_hz).
+        UINT timer_interval_ms = 1000 / 60;
+        bool timer_running = false;
+        bool window_minimized = false;
+
+        // tracks whether we've issued at least one InvalidateRect since window
+        // creation/resize. The overlay's per-frame "pixels changed" flag suppresses
+        // idle blits; this flag forces the very first blit on startup or after
+        // a resize so the window doesn't show garbage until the user moves the mouse.
+        bool has_valid_draw_hash = false;
+
+        // dimensions of the configurator client area (kept in sync with WM_SIZE)
+        int client_width = 0;
+        int client_height = 0;
+
         ConfiguratorWindow();
         ~ConfiguratorWindow();
 
         void run();
 
+        // start/stop the render timer based on visibility state
+        void start_timer();
+        void stop_timer();
+
         static LRESULT CALLBACK window_proc(HWND hwnd, UINT uMsg, WPARAM wParam, LPARAM lParam);
     };
 }

src/spice2x/cfg/controller_presets.cpp

@@ -24,6 +24,7 @@ namespace overlay::windows {
         el->SetAttribute("invert", entry.invert);
         el->SetAttribute("debounce_up", entry.debounce_up);
         el->SetAttribute("debounce_down", entry.debounce_down);
+        el->SetAttribute("bat_threshold", entry.bat_threshold);
         el->SetAttribute("velocity_threshold", entry.velocity_threshold);
         parent->InsertEndChild(el);
     }
@@ -49,6 +50,10 @@ namespace overlay::windows {
         el->QueryIntAttribute("velocity_threshold", &vel);
         entry.velocity_threshold = (unsigned short)vel;
 
+        int bat = 0;
+        el->QueryIntAttribute("bat_threshold", &bat);
+        entry.bat_threshold = bat;
+
         return entry;
     }
 
@@ -65,7 +70,7 @@ namespace overlay::windows {
         el->SetAttribute("smoothing", analog.smoothing);
         el->SetAttribute("multiplier", analog.multiplier);
         el->SetAttribute("relative", analog.relative_mode);
-        el->SetAttribute("delay", analog.delay_buffer_depth);
+        el->SetAttribute("delay_ms", analog.delay_ms);
         parent->InsertEndChild(el);
     }
 
@@ -88,7 +93,7 @@ namespace overlay::windows {
         el->QueryBoolAttribute("smoothing", &a.smoothing);
         el->QueryIntAttribute("multiplier", &a.multiplier);
         el->QueryBoolAttribute("relative", &a.relative_mode);
-        el->QueryIntAttribute("delay", &a.delay_buffer_depth);
+        el->QueryUnsignedAttribute("delay_ms", &a.delay_ms);
 
         return a;
     }

src/spice2x/cfg/manifest.manifest

@@ -20,10 +20,10 @@
         </dependentAssembly>
     </dependency>
     <trustInfo xmlns="urn:schemas-microsoft-com:asm.v2">
-            <security>
-                <requestedPrivileges>
-                    <requestedExecutionLevel level="requireAdministrator" uiAccess="false"/>
-                </requestedPrivileges>
-            </security>
+        <security>
+            <requestedPrivileges>
+                <requestedExecutionLevel level="asInvoker" uiAccess="false"/>
+            </requestedPrivileges>
+        </security>
     </trustInfo>
 </assembly>

src/spice2x/cfg/screen_resize.cpp

@@ -21,7 +21,7 @@ namespace cfg {
         if (SCREEN_RESIZE_CFG_PATH_OVERRIDE.has_value()) {
             this->config_path = SCREEN_RESIZE_CFG_PATH_OVERRIDE.value();
             if (fileutils::file_exists(this->config_path)) {
-                log_info("ScreenResize", "loading config from: {}", this->config_path.string());
+                log_info("ScreenResize", "loading config from: {}", this->config_path);
                 file_exists = true;
             }
         } else {
@@ -35,7 +35,7 @@ namespace cfg {
 
     ScreenResize::~ScreenResize() {
     }
-    
+
     void ScreenResize::config_load() {
         std::string config = fileutils::text_read(this->config_path);
         if (config.empty()) {
@@ -78,7 +78,7 @@ namespace cfg {
             eamuse_get_game(),
             use_game_setting,
             root);
-        
+
         load_bool_value(doc, root + "enable_screen_resize", this->enable_screen_resize);
         if (this->enable_screen_resize) {
             log_misc("ScreenResize", "enabled by config file");

src/spice2x/external/headsocket.h

@@ -20,6 +20,7 @@ Usage:
 #include <memory>
 #include <string>
 #include <map>
+#include <cstdint>
 
 ///////////////////////////////////////////////////////////////////////////////////////////////////////////////////////
 

src/spice2x/external/imgui/CMakeLists.txt

@@ -21,5 +21,12 @@ set(IMGUI_SOURCES
         misc/cpp/imgui_stdlib.cpp
 )
 
+# spice2x: DX11 backend is only built for non-XP toolchains. WinXP lacks
+# D3DCOMPILER_47.dll, and including this TU would pull in D3DCompile imports.
+if(NOT SPICE_XP)
+    list(APPEND IMGUI_HEADERS backends/imgui_impl_dx11.h)
+    list(APPEND IMGUI_SOURCES backends/imgui_impl_dx11.cpp)
+endif()
+
 add_library(imgui STATIC ${IMGUI_HEADERS} ${IMGUI_SOURCES})
 target_include_directories(imgui PRIVATE ${PROJECT_SOURCE_DIR})

src/spice2x/external/imgui/backends/imgui_impl_dx11.cpp

@@ -0,0 +1,697 @@
+// dear imgui: Renderer Backend for DirectX11
+// This needs to be used along with a Platform Backend (e.g. Win32)
+
+// Implemented features:
+//  [X] Renderer: User texture binding. Use 'ID3D11ShaderResourceView*' as texture identifier. Read the FAQ about ImTextureID/ImTextureRef!
+//  [X] Renderer: Large meshes support (64k+ vertices) even with 16-bit indices (ImGuiBackendFlags_RendererHasVtxOffset).
+//  [X] Renderer: Texture updates support for dynamic font atlas (ImGuiBackendFlags_RendererHasTextures).
+//  [X] Renderer: Expose selected render state for draw callbacks to use. Access in '(ImGui_ImplXXXX_RenderState*)GetPlatformIO().Renderer_RenderState'.
+
+// You can use unmodified imgui_impl_* files in your project. See examples/ folder for examples of using this.
+// Prefer including the entire imgui/ repository into your project (either as a copy or as a submodule), and only build the backends you need.
+// Learn about Dear ImGui:
+// - FAQ                  https://dearimgui.com/faq
+// - Getting Started      https://dearimgui.com/getting-started
+// - Documentation        https://dearimgui.com/docs (same as your local docs/ folder).
+// - Introduction, links and more at the top of imgui.cpp
+
+// CHANGELOG
+// (minor and older changes stripped away, please see git history for details)
+//  2026-01-19: DirectX11: Added 'SamplerNearest' in ImGui_ImplDX11_RenderState. Renamed 'SamplerDefault' to 'SamplerLinear'.
+//  2025-09-18: Call platform_io.ClearRendererHandlers() on shutdown.
+//  2025-06-11: DirectX11: Added support for ImGuiBackendFlags_RendererHasTextures, for dynamic font atlas.
+//  2025-05-07: DirectX11: Honor draw_data->FramebufferScale to allow for custom backends and experiment using it (consistently with other renderer backends, even though in normal condition it is not set under Windows).
+//  2025-01-06: DirectX11: Expose VertexConstantBuffer in ImGui_ImplDX11_RenderState. Reset projection matrix in ImDrawCallback_ResetRenderState handler.
+//  2024-10-07: DirectX11: Changed default texture sampler to Clamp instead of Repeat/Wrap.
+//  2024-10-07: DirectX11: Expose selected render state in ImGui_ImplDX11_RenderState, which you can access in 'void* platform_io.Renderer_RenderState' during draw callbacks.
+//  2022-10-11: Using 'nullptr' instead of 'NULL' as per our switch to C++11.
+//  2021-06-29: Reorganized backend to pull data from a single structure to facilitate usage with multiple-contexts (all g_XXXX access changed to bd->XXXX).
+//  2021-05-19: DirectX11: Replaced direct access to ImDrawCmd::TextureId with a call to ImDrawCmd::GetTexID(). (will become a requirement)
+//  2021-02-18: DirectX11: Change blending equation to preserve alpha in output buffer.
+//  2019-08-01: DirectX11: Fixed code querying the Geometry Shader state (would generally error with Debug layer enabled).
+//  2019-07-21: DirectX11: Backup, clear and restore Geometry Shader is any is bound when calling ImGui_ImplDX11_RenderDrawData. Clearing Hull/Domain/Compute shaders without backup/restore.
+//  2019-05-29: DirectX11: Added support for large mesh (64K+ vertices), enable ImGuiBackendFlags_RendererHasVtxOffset flag.
+//  2019-04-30: DirectX11: Added support for special ImDrawCallback_ResetRenderState callback to reset render state.
+//  2018-12-03: Misc: Added #pragma comment statement to automatically link with d3dcompiler.lib when using D3DCompile().
+//  2018-11-30: Misc: Setting up io.BackendRendererName so it can be displayed in the About Window.
+//  2018-08-01: DirectX11: Querying for IDXGIFactory instead of IDXGIFactory1 to increase compatibility.
+//  2018-07-13: DirectX11: Fixed unreleased resources in Init and Shutdown functions.
+//  2018-06-08: Misc: Extracted imgui_impl_dx11.cpp/.h away from the old combined DX11+Win32 example.
+//  2018-06-08: DirectX11: Use draw_data->DisplayPos and draw_data->DisplaySize to setup projection matrix and clipping rectangle.
+//  2018-02-16: Misc: Obsoleted the io.RenderDrawListsFn callback and exposed ImGui_ImplDX11_RenderDrawData() in the .h file so you can call it yourself.
+//  2018-02-06: Misc: Removed call to ImGui::Shutdown() which is not available from 1.60 WIP, user needs to call CreateContext/DestroyContext themselves.
+//  2016-05-07: DirectX11: Disabling depth-write.
+
+#include "imgui.h"
+#ifndef IMGUI_DISABLE
+#include "imgui_impl_dx11.h"
+
+// DirectX
+#include <stdio.h>
+#include <d3d11.h>
+#include <d3dcompiler.h>
+#ifdef _MSC_VER
+#pragma comment(lib, "d3dcompiler") // Automatically link with d3dcompiler.lib as we are using D3DCompile() below.
+#endif
+
+// Clang/GCC warnings with -Weverything
+#if defined(__clang__)
+#pragma clang diagnostic ignored "-Wold-style-cast"         // warning: use of old-style cast                            // yes, they are more terse.
+#pragma clang diagnostic ignored "-Wsign-conversion"        // warning: implicit conversion changes signedness
+#endif
+
+// DirectX11 data
+struct ImGui_ImplDX11_Texture
+{
+    ID3D11Texture2D*            pTexture;
+    ID3D11ShaderResourceView*   pTextureView;
+};
+
+struct ImGui_ImplDX11_Data
+{
+    ID3D11Device*               pd3dDevice;
+    ID3D11DeviceContext*        pd3dDeviceContext;
+    IDXGIFactory*               pFactory;
+    ID3D11Buffer*               pVB;
+    ID3D11Buffer*               pIB;
+    ID3D11VertexShader*         pVertexShader;
+    ID3D11InputLayout*          pInputLayout;
+    ID3D11Buffer*               pVertexConstantBuffer;
+    ID3D11PixelShader*          pPixelShader;
+    ID3D11SamplerState*         pTexSamplerLinear;
+    ID3D11SamplerState*         pTexSamplerNearest;
+    ID3D11RasterizerState*      pRasterizerState;
+    ID3D11BlendState*           pBlendState;
+    ID3D11DepthStencilState*    pDepthStencilState;
+    int                         VertexBufferSize;
+    int                         IndexBufferSize;
+
+    ImGui_ImplDX11_Data()       { memset((void*)this, 0, sizeof(*this)); VertexBufferSize = 5000; IndexBufferSize = 10000; }
+};
+
+struct VERTEX_CONSTANT_BUFFER_DX11
+{
+    float   mvp[4][4];
+};
+
+// Backend data stored in io.BackendRendererUserData to allow support for multiple Dear ImGui contexts
+// It is STRONGLY preferred that you use docking branch with multi-viewports (== single Dear ImGui context + multiple windows) instead of multiple Dear ImGui contexts.
+static ImGui_ImplDX11_Data* ImGui_ImplDX11_GetBackendData()
+{
+    return ImGui::GetCurrentContext() ? (ImGui_ImplDX11_Data*)ImGui::GetIO().BackendRendererUserData : nullptr;
+}
+
+// Functions
+static void ImGui_ImplDX11_SetupRenderState(const ImDrawData* draw_data, ID3D11DeviceContext* device_ctx)
+{
+    ImGui_ImplDX11_Data* bd = ImGui_ImplDX11_GetBackendData();
+
+    // Setup viewport
+    D3D11_VIEWPORT vp = {};
+    vp.Width = draw_data->DisplaySize.x * draw_data->FramebufferScale.x;
+    vp.Height = draw_data->DisplaySize.y * draw_data->FramebufferScale.y;
+    vp.MinDepth = 0.0f;
+    vp.MaxDepth = 1.0f;
+    vp.TopLeftX = vp.TopLeftY = 0;
+    device_ctx->RSSetViewports(1, &vp);
+
+    // Setup orthographic projection matrix into our constant buffer
+    // Our visible imgui space lies from draw_data->DisplayPos (top left) to draw_data->DisplayPos+data_data->DisplaySize (bottom right). DisplayPos is (0,0) for single viewport apps.
+    D3D11_MAPPED_SUBRESOURCE mapped_resource;
+    if (device_ctx->Map(bd->pVertexConstantBuffer, 0, D3D11_MAP_WRITE_DISCARD, 0, &mapped_resource) == S_OK)
+    {
+        VERTEX_CONSTANT_BUFFER_DX11* constant_buffer = (VERTEX_CONSTANT_BUFFER_DX11*)mapped_resource.pData;
+        float L = draw_data->DisplayPos.x;
+        float R = draw_data->DisplayPos.x + draw_data->DisplaySize.x;
+        float T = draw_data->DisplayPos.y;
+        float B = draw_data->DisplayPos.y + draw_data->DisplaySize.y;
+        float mvp[4][4] =
+        {
+            { 2.0f/(R-L),   0.0f,           0.0f,       0.0f },
+            { 0.0f,         2.0f/(T-B),     0.0f,       0.0f },
+            { 0.0f,         0.0f,           0.5f,       0.0f },
+            { (R+L)/(L-R),  (T+B)/(B-T),    0.5f,       1.0f },
+        };
+        memcpy(&constant_buffer->mvp, mvp, sizeof(mvp));
+        device_ctx->Unmap(bd->pVertexConstantBuffer, 0);
+    }
+
+    // Setup shader and vertex buffers
+    unsigned int stride = sizeof(ImDrawVert);
+    unsigned int offset = 0;
+    device_ctx->IASetInputLayout(bd->pInputLayout);
+    device_ctx->IASetVertexBuffers(0, 1, &bd->pVB, &stride, &offset);
+    device_ctx->IASetIndexBuffer(bd->pIB, sizeof(ImDrawIdx) == 2 ? DXGI_FORMAT_R16_UINT : DXGI_FORMAT_R32_UINT, 0);
+    device_ctx->IASetPrimitiveTopology(D3D11_PRIMITIVE_TOPOLOGY_TRIANGLELIST);
+    device_ctx->VSSetShader(bd->pVertexShader, nullptr, 0);
+    device_ctx->VSSetConstantBuffers(0, 1, &bd->pVertexConstantBuffer);
+    device_ctx->PSSetShader(bd->pPixelShader, nullptr, 0);
+    device_ctx->PSSetSamplers(0, 1, &bd->pTexSamplerLinear);
+    device_ctx->GSSetShader(nullptr, nullptr, 0);
+    device_ctx->HSSetShader(nullptr, nullptr, 0); // In theory we should backup and restore this as well.. very infrequently used..
+    device_ctx->DSSetShader(nullptr, nullptr, 0); // In theory we should backup and restore this as well.. very infrequently used..
+    device_ctx->CSSetShader(nullptr, nullptr, 0); // In theory we should backup and restore this as well.. very infrequently used..
+
+    // Setup render state
+    const float blend_factor[4] = { 0.f, 0.f, 0.f, 0.f };
+    device_ctx->OMSetBlendState(bd->pBlendState, blend_factor, 0xffffffff);
+    device_ctx->OMSetDepthStencilState(bd->pDepthStencilState, 0);
+    device_ctx->RSSetState(bd->pRasterizerState);
+}
+
+// Render function
+void ImGui_ImplDX11_RenderDrawData(ImDrawData* draw_data)
+{
+    // Avoid rendering when minimized
+    if (draw_data->DisplaySize.x <= 0.0f || draw_data->DisplaySize.y <= 0.0f)
+        return;
+
+    ImGui_ImplDX11_Data* bd = ImGui_ImplDX11_GetBackendData();
+    ID3D11DeviceContext* device = bd->pd3dDeviceContext;
+
+    // Catch up with texture updates. Most of the times, the list will have 1 element with an OK status, aka nothing to do.
+    // (This almost always points to ImGui::GetPlatformIO().Textures[] but is part of ImDrawData to allow overriding or disabling texture updates).
+    if (draw_data->Textures != nullptr)
+        for (ImTextureData* tex : *draw_data->Textures)
+            if (tex->Status != ImTextureStatus_OK)
+                ImGui_ImplDX11_UpdateTexture(tex);
+
+    // Create and grow vertex/index buffers if needed
+    if (!bd->pVB || bd->VertexBufferSize < draw_data->TotalVtxCount)
+    {
+        if (bd->pVB) { bd->pVB->Release(); bd->pVB = nullptr; }
+        bd->VertexBufferSize = draw_data->TotalVtxCount + 5000;
+        D3D11_BUFFER_DESC desc = {};
+        desc.Usage = D3D11_USAGE_DYNAMIC;
+        desc.ByteWidth = bd->VertexBufferSize * sizeof(ImDrawVert);
+        desc.BindFlags = D3D11_BIND_VERTEX_BUFFER;
+        desc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE;
+        desc.MiscFlags = 0;
+        if (bd->pd3dDevice->CreateBuffer(&desc, nullptr, &bd->pVB) < 0)
+            return;
+    }
+    if (!bd->pIB || bd->IndexBufferSize < draw_data->TotalIdxCount)
+    {
+        if (bd->pIB) { bd->pIB->Release(); bd->pIB = nullptr; }
+        bd->IndexBufferSize = draw_data->TotalIdxCount + 10000;
+        D3D11_BUFFER_DESC desc = {};
+        desc.Usage = D3D11_USAGE_DYNAMIC;
+        desc.ByteWidth = bd->IndexBufferSize * sizeof(ImDrawIdx);
+        desc.BindFlags = D3D11_BIND_INDEX_BUFFER;
+        desc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE;
+        if (bd->pd3dDevice->CreateBuffer(&desc, nullptr, &bd->pIB) < 0)
+            return;
+    }
+
+    // Upload vertex/index data into a single contiguous GPU buffer
+    D3D11_MAPPED_SUBRESOURCE vtx_resource, idx_resource;
+    if (device->Map(bd->pVB, 0, D3D11_MAP_WRITE_DISCARD, 0, &vtx_resource) != S_OK)
+        return;
+    if (device->Map(bd->pIB, 0, D3D11_MAP_WRITE_DISCARD, 0, &idx_resource) != S_OK)
+        return;
+    ImDrawVert* vtx_dst = (ImDrawVert*)vtx_resource.pData;
+    ImDrawIdx* idx_dst = (ImDrawIdx*)idx_resource.pData;
+    for (const ImDrawList* draw_list : draw_data->CmdLists)
+    {
+        memcpy(vtx_dst, draw_list->VtxBuffer.Data, draw_list->VtxBuffer.Size * sizeof(ImDrawVert));
+        memcpy(idx_dst, draw_list->IdxBuffer.Data, draw_list->IdxBuffer.Size * sizeof(ImDrawIdx));
+        vtx_dst += draw_list->VtxBuffer.Size;
+        idx_dst += draw_list->IdxBuffer.Size;
+    }
+    device->Unmap(bd->pVB, 0);
+    device->Unmap(bd->pIB, 0);
+
+    // Backup DX state that will be modified to restore it afterwards (unfortunately this is very ugly looking and verbose. Close your eyes!)
+    struct BACKUP_DX11_STATE
+    {
+        UINT                        ScissorRectsCount, ViewportsCount;
+        D3D11_RECT                  ScissorRects[D3D11_VIEWPORT_AND_SCISSORRECT_OBJECT_COUNT_PER_PIPELINE];
+        D3D11_VIEWPORT              Viewports[D3D11_VIEWPORT_AND_SCISSORRECT_OBJECT_COUNT_PER_PIPELINE];
+        ID3D11RasterizerState*      RS;
+        ID3D11BlendState*           BlendState;
+        FLOAT                       BlendFactor[4];
+        UINT                        SampleMask;
+        UINT                        StencilRef;
+        ID3D11DepthStencilState*    DepthStencilState;
+        ID3D11ShaderResourceView*   PSShaderResource;
+        ID3D11SamplerState*         PSSampler;
+        ID3D11PixelShader*          PS;
+        ID3D11VertexShader*         VS;
+        ID3D11GeometryShader*       GS;
+        UINT                        PSInstancesCount, VSInstancesCount, GSInstancesCount;
+        ID3D11ClassInstance         *PSInstances[256], *VSInstances[256], *GSInstances[256];   // 256 is max according to PSSetShader documentation
+        D3D11_PRIMITIVE_TOPOLOGY    PrimitiveTopology;
+        ID3D11Buffer*               IndexBuffer, *VertexBuffer, *VSConstantBuffer;
+        UINT                        IndexBufferOffset, VertexBufferStride, VertexBufferOffset;
+        DXGI_FORMAT                 IndexBufferFormat;
+        ID3D11InputLayout*          InputLayout;
+    };
+    BACKUP_DX11_STATE old = {};
+    old.ScissorRectsCount = old.ViewportsCount = D3D11_VIEWPORT_AND_SCISSORRECT_OBJECT_COUNT_PER_PIPELINE;
+    device->RSGetScissorRects(&old.ScissorRectsCount, old.ScissorRects);
+    device->RSGetViewports(&old.ViewportsCount, old.Viewports);
+    device->RSGetState(&old.RS);
+    device->OMGetBlendState(&old.BlendState, old.BlendFactor, &old.SampleMask);
+    device->OMGetDepthStencilState(&old.DepthStencilState, &old.StencilRef);
+    device->PSGetShaderResources(0, 1, &old.PSShaderResource);
+    device->PSGetSamplers(0, 1, &old.PSSampler);
+    old.PSInstancesCount = old.VSInstancesCount = old.GSInstancesCount = 256;
+    device->PSGetShader(&old.PS, old.PSInstances, &old.PSInstancesCount);
+    device->VSGetShader(&old.VS, old.VSInstances, &old.VSInstancesCount);
+    device->VSGetConstantBuffers(0, 1, &old.VSConstantBuffer);
+    device->GSGetShader(&old.GS, old.GSInstances, &old.GSInstancesCount);
+
+    device->IAGetPrimitiveTopology(&old.PrimitiveTopology);
+    device->IAGetIndexBuffer(&old.IndexBuffer, &old.IndexBufferFormat, &old.IndexBufferOffset);
+    device->IAGetVertexBuffers(0, 1, &old.VertexBuffer, &old.VertexBufferStride, &old.VertexBufferOffset);
+    device->IAGetInputLayout(&old.InputLayout);
+
+    // Setup desired DX state
+    ImGui_ImplDX11_SetupRenderState(draw_data, device);
+
+    // Setup render state structure (for callbacks and custom texture bindings)
+    ImGuiPlatformIO& platform_io = ImGui::GetPlatformIO();
+    ImGui_ImplDX11_RenderState render_state;
+    render_state.Device = bd->pd3dDevice;
+    render_state.DeviceContext = bd->pd3dDeviceContext;
+    render_state.SamplerLinear = bd->pTexSamplerLinear;
+    render_state.SamplerNearest = bd->pTexSamplerNearest;
+    render_state.VertexConstantBuffer = bd->pVertexConstantBuffer;
+    platform_io.Renderer_RenderState = &render_state;
+
+    // Render command lists
+    // (Because we merged all buffers into a single one, we maintain our own offset into them)
+    int global_idx_offset = 0;
+    int global_vtx_offset = 0;
+    ImVec2 clip_off = draw_data->DisplayPos;
+    ImVec2 clip_scale = draw_data->FramebufferScale;
+    for (const ImDrawList* draw_list : draw_data->CmdLists)
+    {
+        for (int cmd_i = 0; cmd_i < draw_list->CmdBuffer.Size; cmd_i++)
+        {
+            const ImDrawCmd* pcmd = &draw_list->CmdBuffer[cmd_i];
+            if (pcmd->UserCallback != nullptr)
+            {
+                // User callback, registered via ImDrawList::AddCallback()
+                // (ImDrawCallback_ResetRenderState is a special callback value used by the user to request the renderer to reset render state.)
+                if (pcmd->UserCallback == ImDrawCallback_ResetRenderState)
+                    ImGui_ImplDX11_SetupRenderState(draw_data, device);
+                else
+                    pcmd->UserCallback(draw_list, pcmd);
+            }
+            else
+            {
+                // Project scissor/clipping rectangles into framebuffer space
+                ImVec2 clip_min((pcmd->ClipRect.x - clip_off.x) * clip_scale.x, (pcmd->ClipRect.y - clip_off.y) * clip_scale.y);
+                ImVec2 clip_max((pcmd->ClipRect.z - clip_off.x) * clip_scale.x, (pcmd->ClipRect.w - clip_off.y) * clip_scale.y);
+                if (clip_max.x <= clip_min.x || clip_max.y <= clip_min.y)
+                    continue;
+
+                // Apply scissor/clipping rectangle
+                const D3D11_RECT r = { (LONG)clip_min.x, (LONG)clip_min.y, (LONG)clip_max.x, (LONG)clip_max.y };
+                device->RSSetScissorRects(1, &r);
+
+                // Bind texture, Draw
+                ID3D11ShaderResourceView* texture_srv = (ID3D11ShaderResourceView*)pcmd->GetTexID();
+                device->PSSetShaderResources(0, 1, &texture_srv);
+                device->DrawIndexed(pcmd->ElemCount, pcmd->IdxOffset + global_idx_offset, pcmd->VtxOffset + global_vtx_offset);
+            }
+        }
+        global_idx_offset += draw_list->IdxBuffer.Size;
+        global_vtx_offset += draw_list->VtxBuffer.Size;
+    }
+    platform_io.Renderer_RenderState = nullptr;
+
+    // Restore modified DX state
+    device->RSSetScissorRects(old.ScissorRectsCount, old.ScissorRects);
+    device->RSSetViewports(old.ViewportsCount, old.Viewports);
+    device->RSSetState(old.RS); if (old.RS) old.RS->Release();
+    device->OMSetBlendState(old.BlendState, old.BlendFactor, old.SampleMask); if (old.BlendState) old.BlendState->Release();
+    device->OMSetDepthStencilState(old.DepthStencilState, old.StencilRef); if (old.DepthStencilState) old.DepthStencilState->Release();
+    device->PSSetShaderResources(0, 1, &old.PSShaderResource); if (old.PSShaderResource) old.PSShaderResource->Release();
+    device->PSSetSamplers(0, 1, &old.PSSampler); if (old.PSSampler) old.PSSampler->Release();
+    device->PSSetShader(old.PS, old.PSInstances, old.PSInstancesCount); if (old.PS) old.PS->Release();
+    for (UINT i = 0; i < old.PSInstancesCount; i++) if (old.PSInstances[i]) old.PSInstances[i]->Release();
+    device->VSSetShader(old.VS, old.VSInstances, old.VSInstancesCount); if (old.VS) old.VS->Release();
+    device->VSSetConstantBuffers(0, 1, &old.VSConstantBuffer); if (old.VSConstantBuffer) old.VSConstantBuffer->Release();
+    device->GSSetShader(old.GS, old.GSInstances, old.GSInstancesCount); if (old.GS) old.GS->Release();
+    for (UINT i = 0; i < old.VSInstancesCount; i++) if (old.VSInstances[i]) old.VSInstances[i]->Release();
+    device->IASetPrimitiveTopology(old.PrimitiveTopology);
+    device->IASetIndexBuffer(old.IndexBuffer, old.IndexBufferFormat, old.IndexBufferOffset); if (old.IndexBuffer) old.IndexBuffer->Release();
+    device->IASetVertexBuffers(0, 1, &old.VertexBuffer, &old.VertexBufferStride, &old.VertexBufferOffset); if (old.VertexBuffer) old.VertexBuffer->Release();
+    device->IASetInputLayout(old.InputLayout); if (old.InputLayout) old.InputLayout->Release();
+}
+
+static void ImGui_ImplDX11_DestroyTexture(ImTextureData* tex)
+{
+    if (ImGui_ImplDX11_Texture* backend_tex = (ImGui_ImplDX11_Texture*)tex->BackendUserData)
+    {
+        IM_ASSERT(backend_tex->pTextureView == (ID3D11ShaderResourceView*)(intptr_t)tex->TexID);
+        backend_tex->pTextureView->Release();
+        backend_tex->pTexture->Release();
+        IM_DELETE(backend_tex);
+
+        // Clear identifiers and mark as destroyed (in order to allow e.g. calling InvalidateDeviceObjects while running)
+        tex->SetTexID(ImTextureID_Invalid);
+        tex->BackendUserData = nullptr;
+    }
+    tex->SetStatus(ImTextureStatus_Destroyed);
+}
+
+void ImGui_ImplDX11_UpdateTexture(ImTextureData* tex)
+{
+    ImGui_ImplDX11_Data* bd = ImGui_ImplDX11_GetBackendData();
+    if (tex->Status == ImTextureStatus_WantCreate)
+    {
+        // Create and upload new texture to graphics system
+        //IMGUI_DEBUG_LOG("UpdateTexture #%03d: WantCreate %dx%d\n", tex->UniqueID, tex->Width, tex->Height);
+        IM_ASSERT(tex->TexID == ImTextureID_Invalid && tex->BackendUserData == nullptr);
+        IM_ASSERT(tex->Format == ImTextureFormat_RGBA32);
+        unsigned int* pixels = (unsigned int*)tex->GetPixels();
+        ImGui_ImplDX11_Texture* backend_tex = IM_NEW(ImGui_ImplDX11_Texture)();
+
+        // Create texture
+        D3D11_TEXTURE2D_DESC desc;
+        ZeroMemory(&desc, sizeof(desc));
+        desc.Width = (UINT)tex->Width;
+        desc.Height = (UINT)tex->Height;
+        desc.MipLevels = 1;
+        desc.ArraySize = 1;
+#ifdef IMGUI_USE_BGRA_PACKED_COLOR
+        desc.Format = DXGI_FORMAT_B8G8R8A8_UNORM;
+#else
+        desc.Format = DXGI_FORMAT_R8G8B8A8_UNORM;
+#endif
+        desc.SampleDesc.Count = 1;
+        desc.Usage = D3D11_USAGE_DEFAULT;
+        desc.BindFlags = D3D11_BIND_SHADER_RESOURCE;
+        desc.CPUAccessFlags = 0;
+        D3D11_SUBRESOURCE_DATA subResource;
+        subResource.pSysMem = pixels;
+        subResource.SysMemPitch = desc.Width * 4;
+        subResource.SysMemSlicePitch = 0;
+        bd->pd3dDevice->CreateTexture2D(&desc, &subResource, &backend_tex->pTexture);
+        IM_ASSERT(backend_tex->pTexture != nullptr && "Backend failed to create texture!");
+
+        // Create texture view
+        D3D11_SHADER_RESOURCE_VIEW_DESC srvDesc;
+        ZeroMemory(&srvDesc, sizeof(srvDesc));
+#ifdef IMGUI_USE_BGRA_PACKED_COLOR
+        srvDesc.Format = DXGI_FORMAT_B8G8R8A8_UNORM;
+#else
+        srvDesc.Format = DXGI_FORMAT_R8G8B8A8_UNORM;
+#endif
+        srvDesc.ViewDimension = D3D11_SRV_DIMENSION_TEXTURE2D;
+        srvDesc.Texture2D.MipLevels = desc.MipLevels;
+        srvDesc.Texture2D.MostDetailedMip = 0;
+        bd->pd3dDevice->CreateShaderResourceView(backend_tex->pTexture, &srvDesc, &backend_tex->pTextureView);
+        IM_ASSERT(backend_tex->pTextureView != nullptr && "Backend failed to create texture!");
+
+        // Store identifiers
+        tex->SetTexID((ImTextureID)(intptr_t)backend_tex->pTextureView);
+        tex->SetStatus(ImTextureStatus_OK);
+        tex->BackendUserData = backend_tex;
+    }
+    else if (tex->Status == ImTextureStatus_WantUpdates)
+    {
+        // Update selected blocks. We only ever write to textures regions which have never been used before!
+        // This backend choose to use tex->Updates[] but you can use tex->UpdateRect to upload a single region.
+        ImGui_ImplDX11_Texture* backend_tex = (ImGui_ImplDX11_Texture*)tex->BackendUserData;
+        IM_ASSERT(backend_tex->pTextureView == (ID3D11ShaderResourceView*)(intptr_t)tex->TexID);
+        for (ImTextureRect& r : tex->Updates)
+        {
+            D3D11_BOX box = { (UINT)r.x, (UINT)r.y, (UINT)0, (UINT)(r.x + r.w), (UINT)(r.y + r .h), (UINT)1 };
+            bd->pd3dDeviceContext->UpdateSubresource(backend_tex->pTexture, 0, &box, tex->GetPixelsAt(r.x, r.y), (UINT)tex->GetPitch(), 0);
+        }
+        tex->SetStatus(ImTextureStatus_OK);
+    }
+    if (tex->Status == ImTextureStatus_WantDestroy && tex->UnusedFrames > 0)
+        ImGui_ImplDX11_DestroyTexture(tex);
+}
+
+bool    ImGui_ImplDX11_CreateDeviceObjects()
+{
+    ImGui_ImplDX11_Data* bd = ImGui_ImplDX11_GetBackendData();
+    if (!bd->pd3dDevice)
+        return false;
+    ImGui_ImplDX11_InvalidateDeviceObjects();
+
+    // By using D3DCompile() from <d3dcompiler.h> / d3dcompiler.lib, we introduce a dependency to a given version of d3dcompiler_XX.dll (see D3DCOMPILER_DLL_A)
+    // If you would like to use this DX11 sample code but remove this dependency you can:
+    //  1) compile once, save the compiled shader blobs into a file or source code and pass them to CreateVertexShader()/CreatePixelShader() [preferred solution]
+    //  2) use code to detect any version of the DLL and grab a pointer to D3DCompile from the DLL.
+    // See https://github.com/ocornut/imgui/pull/638 for sources and details.
+
+    // Create the vertex shader
+    {
+        static const char* vertexShader =
+            "cbuffer vertexBuffer : register(b0) \
+            {\
+              float4x4 ProjectionMatrix; \
+            };\
+            struct VS_INPUT\
+            {\
+              float2 pos : POSITION;\
+              float4 col : COLOR0;\
+              float2 uv  : TEXCOORD0;\
+            };\
+            \
+            struct PS_INPUT\
+            {\
+              float4 pos : SV_POSITION;\
+              float4 col : COLOR0;\
+              float2 uv  : TEXCOORD0;\
+            };\
+            \
+            PS_INPUT main(VS_INPUT input)\
+            {\
+              PS_INPUT output;\
+              output.pos = mul( ProjectionMatrix, float4(input.pos.xy, 0.f, 1.f));\
+              output.col = input.col;\
+              output.uv  = input.uv;\
+              return output;\
+            }";
+
+        ID3DBlob* vertexShaderBlob;
+        if (FAILED(D3DCompile(vertexShader, strlen(vertexShader), nullptr, nullptr, nullptr, "main", "vs_4_0", 0, 0, &vertexShaderBlob, nullptr)))
+            return false; // NB: Pass ID3DBlob* pErrorBlob to D3DCompile() to get error showing in (const char*)pErrorBlob->GetBufferPointer(). Make sure to Release() the blob!
+        if (bd->pd3dDevice->CreateVertexShader(vertexShaderBlob->GetBufferPointer(), vertexShaderBlob->GetBufferSize(), nullptr, &bd->pVertexShader) != S_OK)
+        {
+            vertexShaderBlob->Release();
+            return false;
+        }
+
+        // Create the input layout
+        D3D11_INPUT_ELEMENT_DESC local_layout[] =
+        {
+            { "POSITION", 0, DXGI_FORMAT_R32G32_FLOAT,   0, (UINT)offsetof(ImDrawVert, pos), D3D11_INPUT_PER_VERTEX_DATA, 0 },
+            { "TEXCOORD", 0, DXGI_FORMAT_R32G32_FLOAT,   0, (UINT)offsetof(ImDrawVert, uv),  D3D11_INPUT_PER_VERTEX_DATA, 0 },
+#ifdef IMGUI_USE_BGRA_PACKED_COLOR
+            { "COLOR",    0, DXGI_FORMAT_B8G8R8A8_UNORM, 0, (UINT)offsetof(ImDrawVert, col), D3D11_INPUT_PER_VERTEX_DATA, 0 },
+#else
+            { "COLOR",    0, DXGI_FORMAT_R8G8B8A8_UNORM, 0, (UINT)offsetof(ImDrawVert, col), D3D11_INPUT_PER_VERTEX_DATA, 0 },
+#endif
+        };
+        if (bd->pd3dDevice->CreateInputLayout(local_layout, 3, vertexShaderBlob->GetBufferPointer(), vertexShaderBlob->GetBufferSize(), &bd->pInputLayout) != S_OK)
+        {
+            vertexShaderBlob->Release();
+            return false;
+        }
+        vertexShaderBlob->Release();
+
+        // Create the constant buffer
+        {
+            D3D11_BUFFER_DESC desc = {};
+            desc.ByteWidth = sizeof(VERTEX_CONSTANT_BUFFER_DX11);
+            desc.Usage = D3D11_USAGE_DYNAMIC;
+            desc.BindFlags = D3D11_BIND_CONSTANT_BUFFER;
+            desc.CPUAccessFlags = D3D11_CPU_ACCESS_WRITE;
+            desc.MiscFlags = 0;
+            bd->pd3dDevice->CreateBuffer(&desc, nullptr, &bd->pVertexConstantBuffer);
+        }
+    }
+
+    // Create the pixel shader
+    {
+        static const char* pixelShader =
+            "struct PS_INPUT\
+            {\
+            float4 pos : SV_POSITION;\
+            float4 col : COLOR0;\
+            float2 uv  : TEXCOORD0;\
+            };\
+            sampler sampler0;\
+            Texture2D texture0;\
+            \
+            float4 main(PS_INPUT input) : SV_Target\
+            {\
+            float4 out_col = input.col * texture0.Sample(sampler0, input.uv); \
+            return out_col; \
+            }";
+
+        ID3DBlob* pixelShaderBlob;
+        if (FAILED(D3DCompile(pixelShader, strlen(pixelShader), nullptr, nullptr, nullptr, "main", "ps_4_0", 0, 0, &pixelShaderBlob, nullptr)))
+            return false; // NB: Pass ID3DBlob* pErrorBlob to D3DCompile() to get error showing in (const char*)pErrorBlob->GetBufferPointer(). Make sure to Release() the blob!
+        if (bd->pd3dDevice->CreatePixelShader(pixelShaderBlob->GetBufferPointer(), pixelShaderBlob->GetBufferSize(), nullptr, &bd->pPixelShader) != S_OK)
+        {
+            pixelShaderBlob->Release();
+            return false;
+        }
+        pixelShaderBlob->Release();
+    }
+
+    // Create the blending setup
+    {
+        D3D11_BLEND_DESC desc;
+        ZeroMemory(&desc, sizeof(desc));
+        desc.AlphaToCoverageEnable = false;
+        desc.RenderTarget[0].BlendEnable = true;
+        desc.RenderTarget[0].SrcBlend = D3D11_BLEND_SRC_ALPHA;
+        desc.RenderTarget[0].DestBlend = D3D11_BLEND_INV_SRC_ALPHA;
+        desc.RenderTarget[0].BlendOp = D3D11_BLEND_OP_ADD;
+        desc.RenderTarget[0].SrcBlendAlpha = D3D11_BLEND_ONE;
+        desc.RenderTarget[0].DestBlendAlpha = D3D11_BLEND_INV_SRC_ALPHA;
+        desc.RenderTarget[0].BlendOpAlpha = D3D11_BLEND_OP_ADD;
+        desc.RenderTarget[0].RenderTargetWriteMask = D3D11_COLOR_WRITE_ENABLE_ALL;
+        bd->pd3dDevice->CreateBlendState(&desc, &bd->pBlendState);
+    }
+
+    // Create the rasterizer state
+    {
+        D3D11_RASTERIZER_DESC desc;
+        ZeroMemory(&desc, sizeof(desc));
+        desc.FillMode = D3D11_FILL_SOLID;
+        desc.CullMode = D3D11_CULL_NONE;
+        desc.ScissorEnable = true;
+        desc.DepthClipEnable = true;
+        bd->pd3dDevice->CreateRasterizerState(&desc, &bd->pRasterizerState);
+    }
+
+    // Create depth-stencil State
+    {
+        D3D11_DEPTH_STENCIL_DESC desc;
+        ZeroMemory(&desc, sizeof(desc));
+        desc.DepthEnable = false;
+        desc.DepthWriteMask = D3D11_DEPTH_WRITE_MASK_ALL;
+        desc.DepthFunc = D3D11_COMPARISON_ALWAYS;
+        desc.StencilEnable = false;
+        desc.FrontFace.StencilFailOp = desc.FrontFace.StencilDepthFailOp = desc.FrontFace.StencilPassOp = D3D11_STENCIL_OP_KEEP;
+        desc.FrontFace.StencilFunc = D3D11_COMPARISON_ALWAYS;
+        desc.BackFace = desc.FrontFace;
+        bd->pd3dDevice->CreateDepthStencilState(&desc, &bd->pDepthStencilState);
+    }
+
+    // Create texture sampler
+    // (Bilinear sampling is required by default. Set 'io.Fonts->Flags |= ImFontAtlasFlags_NoBakedLines' or 'style.AntiAliasedLinesUseTex = false' to allow point/nearest sampling)
+    {
+        D3D11_SAMPLER_DESC desc;
+        ZeroMemory(&desc, sizeof(desc));
+        desc.Filter = D3D11_FILTER_MIN_MAG_MIP_LINEAR;
+        desc.AddressU = D3D11_TEXTURE_ADDRESS_CLAMP;
+        desc.AddressV = D3D11_TEXTURE_ADDRESS_CLAMP;
+        desc.AddressW = D3D11_TEXTURE_ADDRESS_CLAMP;
+        desc.MipLODBias = 0.f;
+        desc.ComparisonFunc = D3D11_COMPARISON_ALWAYS;
+        desc.MinLOD = 0.f;
+        desc.MaxLOD = 0.f;
+        bd->pd3dDevice->CreateSamplerState(&desc, &bd->pTexSamplerLinear);
+        desc.Filter = D3D11_FILTER_MIN_MAG_MIP_POINT;
+        bd->pd3dDevice->CreateSamplerState(&desc, &bd->pTexSamplerNearest);
+    }
+
+    return true;
+}
+
+void    ImGui_ImplDX11_InvalidateDeviceObjects()
+{
+    ImGui_ImplDX11_Data* bd = ImGui_ImplDX11_GetBackendData();
+    if (!bd->pd3dDevice)
+        return;
+
+    // Destroy all textures
+    for (ImTextureData* tex : ImGui::GetPlatformIO().Textures)
+        if (tex->RefCount == 1)
+            ImGui_ImplDX11_DestroyTexture(tex);
+
+    if (bd->pTexSamplerLinear)      { bd->pTexSamplerLinear->Release(); bd->pTexSamplerLinear = nullptr; }
+    if (bd->pTexSamplerNearest)     { bd->pTexSamplerNearest->Release(); bd->pTexSamplerNearest = nullptr; }
+    if (bd->pIB)                    { bd->pIB->Release(); bd->pIB = nullptr; }
+    if (bd->pVB)                    { bd->pVB->Release(); bd->pVB = nullptr; }
+    if (bd->pBlendState)            { bd->pBlendState->Release(); bd->pBlendState = nullptr; }
+    if (bd->pDepthStencilState)     { bd->pDepthStencilState->Release(); bd->pDepthStencilState = nullptr; }
+    if (bd->pRasterizerState)       { bd->pRasterizerState->Release(); bd->pRasterizerState = nullptr; }
+    if (bd->pPixelShader)           { bd->pPixelShader->Release(); bd->pPixelShader = nullptr; }
+    if (bd->pVertexConstantBuffer)  { bd->pVertexConstantBuffer->Release(); bd->pVertexConstantBuffer = nullptr; }
+    if (bd->pInputLayout)           { bd->pInputLayout->Release(); bd->pInputLayout = nullptr; }
+    if (bd->pVertexShader)          { bd->pVertexShader->Release(); bd->pVertexShader = nullptr; }
+}
+
+bool    ImGui_ImplDX11_Init(ID3D11Device* device, ID3D11DeviceContext* device_context)
+{
+    ImGuiIO& io = ImGui::GetIO();
+    IMGUI_CHECKVERSION();
+    IM_ASSERT(io.BackendRendererUserData == nullptr && "Already initialized a renderer backend!");
+
+    // Setup backend capabilities flags
+    ImGui_ImplDX11_Data* bd = IM_NEW(ImGui_ImplDX11_Data)();
+    io.BackendRendererUserData = (void*)bd;
+    io.BackendRendererName = "imgui_impl_dx11";
+    io.BackendFlags |= ImGuiBackendFlags_RendererHasVtxOffset;  // We can honor the ImDrawCmd::VtxOffset field, allowing for large meshes.
+    io.BackendFlags |= ImGuiBackendFlags_RendererHasTextures;   // We can honor ImGuiPlatformIO::Textures[] requests during render.
+
+    ImGuiPlatformIO& platform_io = ImGui::GetPlatformIO();
+    platform_io.Renderer_TextureMaxWidth = platform_io.Renderer_TextureMaxHeight = D3D11_REQ_TEXTURE2D_U_OR_V_DIMENSION;
+
+    // Get factory from device
+    IDXGIDevice* pDXGIDevice = nullptr;
+    IDXGIAdapter* pDXGIAdapter = nullptr;
+    IDXGIFactory* pFactory = nullptr;
+
+    if (device->QueryInterface(IID_PPV_ARGS(&pDXGIDevice)) == S_OK)
+        if (pDXGIDevice->GetParent(IID_PPV_ARGS(&pDXGIAdapter)) == S_OK)
+            if (pDXGIAdapter->GetParent(IID_PPV_ARGS(&pFactory)) == S_OK)
+            {
+                bd->pd3dDevice = device;
+                bd->pd3dDeviceContext = device_context;
+                bd->pFactory = pFactory;
+            }
+    if (pDXGIDevice) pDXGIDevice->Release();
+    if (pDXGIAdapter) pDXGIAdapter->Release();
+    bd->pd3dDevice->AddRef();
+    bd->pd3dDeviceContext->AddRef();
+
+    return true;
+}
+
+void ImGui_ImplDX11_Shutdown()
+{
+    ImGui_ImplDX11_Data* bd = ImGui_ImplDX11_GetBackendData();
+    IM_ASSERT(bd != nullptr && "No renderer backend to shutdown, or already shutdown?");
+    ImGuiIO& io = ImGui::GetIO();
+    ImGuiPlatformIO& platform_io = ImGui::GetPlatformIO();
+
+    ImGui_ImplDX11_InvalidateDeviceObjects();
+    if (bd->pFactory)             { bd->pFactory->Release(); }
+    if (bd->pd3dDevice)           { bd->pd3dDevice->Release(); }
+    if (bd->pd3dDeviceContext)    { bd->pd3dDeviceContext->Release(); }
+
+    io.BackendRendererName = nullptr;
+    io.BackendRendererUserData = nullptr;
+    io.BackendFlags &= ~(ImGuiBackendFlags_RendererHasVtxOffset | ImGuiBackendFlags_RendererHasTextures);
+    platform_io.ClearRendererHandlers();
+    IM_DELETE(bd);
+}
+
+void ImGui_ImplDX11_NewFrame()
+{
+    ImGui_ImplDX11_Data* bd = ImGui_ImplDX11_GetBackendData();
+    IM_ASSERT(bd != nullptr && "Context or backend not initialized! Did you call ImGui_ImplDX11_Init()?");
+
+    if (!bd->pVertexShader)
+        if (!ImGui_ImplDX11_CreateDeviceObjects())
+            IM_ASSERT(0 && "ImGui_ImplDX11_CreateDeviceObjects() failed!");
+}
+
+//-----------------------------------------------------------------------------
+
+#endif // #ifndef IMGUI_DISABLE

src/spice2x/external/imgui/backends/imgui_impl_dx11.h

@@ -0,0 +1,52 @@
+// dear imgui: Renderer Backend for DirectX11
+// This needs to be used along with a Platform Backend (e.g. Win32)
+
+// Implemented features:
+//  [X] Renderer: User texture binding. Use 'ID3D11ShaderResourceView*' as texture identifier. Read the FAQ about ImTextureID/ImTextureRef!
+//  [X] Renderer: Large meshes support (64k+ vertices) even with 16-bit indices (ImGuiBackendFlags_RendererHasVtxOffset).
+//  [X] Renderer: Texture updates support for dynamic font atlas (ImGuiBackendFlags_RendererHasTextures).
+//  [X] Renderer: Expose selected render state for draw callbacks to use. Access in '(ImGui_ImplXXXX_RenderState*)GetPlatformIO().Renderer_RenderState'.
+
+// You can use unmodified imgui_impl_* files in your project. See examples/ folder for examples of using this.
+// Prefer including the entire imgui/ repository into your project (either as a copy or as a submodule), and only build the backends you need.
+// Learn about Dear ImGui:
+// - FAQ                  https://dearimgui.com/faq
+// - Getting Started      https://dearimgui.com/getting-started
+// - Documentation        https://dearimgui.com/docs (same as your local docs/ folder).
+// - Introduction, links and more at the top of imgui.cpp
+
+#pragma once
+#include "imgui.h"      // IMGUI_IMPL_API
+#ifndef IMGUI_DISABLE
+
+struct ID3D11Device;
+struct ID3D11DeviceContext;
+struct ID3D11SamplerState;
+struct ID3D11Buffer;
+
+// Follow "Getting Started" link and check examples/ folder to learn about using backends!
+IMGUI_IMPL_API bool     ImGui_ImplDX11_Init(ID3D11Device* device, ID3D11DeviceContext* device_context);
+IMGUI_IMPL_API void     ImGui_ImplDX11_Shutdown();
+IMGUI_IMPL_API void     ImGui_ImplDX11_NewFrame();
+IMGUI_IMPL_API void     ImGui_ImplDX11_RenderDrawData(ImDrawData* draw_data);
+
+// Use if you want to reset your rendering device without losing Dear ImGui state.
+IMGUI_IMPL_API bool     ImGui_ImplDX11_CreateDeviceObjects();
+IMGUI_IMPL_API void     ImGui_ImplDX11_InvalidateDeviceObjects();
+
+// (Advanced) Use e.g. if you need to precisely control the timing of texture updates (e.g. for staged rendering), by setting ImDrawData::Textures = nullptr to handle this manually.
+IMGUI_IMPL_API void     ImGui_ImplDX11_UpdateTexture(ImTextureData* tex);
+
+// [BETA] Selected render state data shared with callbacks.
+// This is temporarily stored in GetPlatformIO().Renderer_RenderState during the ImGui_ImplDX11_RenderDrawData() call.
+// (Please open an issue if you feel you need access to more data)
+struct ImGui_ImplDX11_RenderState
+{
+    ID3D11Device*           Device;
+    ID3D11DeviceContext*    DeviceContext;
+    ID3D11SamplerState*     SamplerLinear;
+    ID3D11SamplerState*     SamplerNearest;
+    ID3D11Buffer*           VertexConstantBuffer;
+};
+
+#endif // #ifndef IMGUI_DISABLE

src/spice2x/external/robin_hood.h

@@ -39,6 +39,7 @@
 #define ROBIN_HOOD_VERSION_PATCH 2  // for backwards-compatible bug fixes
 
 #include <algorithm>
+#include <cstdint>
 #include <cstdlib>
 #include <cstring>
 #include <functional>

src/spice2x/games/bc/io.cpp

@@ -38,11 +38,12 @@ std::vector<Analog> &games::bc::get_analogs() {
     if (analogs.empty()) {
         analogs = GameAPI::Analogs::getAnalogs("Busou Shinki: Armored Princess Battle Conductor");
 
-        GameAPI::Analogs::sortAnalogs(
-                &analogs,
-                "Stick X",
-                "Stick Y"
-        );
+        using GameAPI::Analogs::AnalogType;
+
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            { "Stick X", AnalogType::LinearCentered },
+            { "Stick Y", AnalogType::LinearCentered }
+        });
     }
 
     return analogs;

src/spice2x/games/ccj/io.cpp

@@ -45,13 +45,14 @@ std::vector<Analog> &games::ccj::get_analogs() {
     if (analogs.empty()) {
         analogs = GameAPI::Analogs::getAnalogs("Chase Chase Jokers");
 
-        GameAPI::Analogs::sortAnalogs(
-                &analogs,
-                "Joystick X",
-                "Joystick Y",
-                "Trackball DX",
-                "Trackball DY"
-        );
+        using GameAPI::Analogs::AnalogType;
+
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            { "Joystick X", AnalogType::LinearCentered },
+            { "Joystick Y", AnalogType::LinearCentered },
+            { "Trackball DX", AnalogType::LinearCentered },
+            { "Trackball DY", AnalogType::LinearCentered }
+        });
     }
 
     return analogs;

src/spice2x/games/ccj/trackball.cpp

@@ -6,6 +6,7 @@
 #include "util/logging.h"
 #include "rawinput/rawinput.h"
 #include "games/io.h"
+#include "util/precise_timer.h"
 #include "util/utils.h"
 #include "io.h"
 
@@ -112,7 +113,7 @@ namespace games::ccj {
                 static std::chrono::time_point<std::chrono::steady_clock> lastModified = std::chrono::steady_clock::now();
                 static std::chrono::milliseconds debounceDuration(100);
                 auto currentTime = std::chrono::steady_clock::now();
-                bool pressed = get_async_secondary_mouse() & 0x8000;
+                bool pressed = get_async_secondary_mouse();
                 bool focused = GetForegroundWindow() == hWnd;
 
                 if (focused && MOUSE_TRACKBALL_USE_TOGGLE && pressed && (currentTime - lastModified > debounceDuration)) {
@@ -243,6 +244,7 @@ namespace games::ccj {
         log_info("trackball", "thread start, use mouse: {}, toggle: {}", MOUSE_TRACKBALL, MOUSE_TRACKBALL_USE_TOGGLE);
 
         tbThread = new std::thread([&] {
+            timeutils::PreciseSleepTimer timer;
             while (tbThreadRunning) {
                 if (hWnd && wndProc) {
                     wndProc(hWnd, WM_INPUT, RIM_INPUT, (LPARAM)fakeHandle);
@@ -251,7 +253,7 @@ namespace games::ccj {
                 if (!tbThreadRunning)
                     break;
 
-                std::this_thread::sleep_for(10ms);
+                timer.sleep(10);
             }
         });
     }

src/spice2x/games/ddr/ddr.cpp

@@ -23,6 +23,7 @@
 #include "p3io/usbmem.h"
 
 #include "p4io/p4io.h"
+#include <cstring>
 
 using namespace acioemu;
 
@@ -53,9 +54,9 @@ namespace games::ddr {
         return SendMessage_real(hWnd, Msg, wParam, lParam);
     }
 
-    bool contains_only_ascii(const std::string& str) {
+    bool contains_only_ascii(const std::wstring& str) {
         for (auto c: str) {
-            if (static_cast<unsigned char>(c) > 127) {
+            if (c > 127) {
                 return false;
             }
         }
@@ -80,24 +81,24 @@ namespace games::ddr {
         }
 
         if (fileutils::dir_exists(dir)) {
-            log_info("ddr", "looking for codecs in this directory: {}", dir.string());
+            log_info("ddr", "looking for codecs in this directory: {}", dir);
         } else {
-            log_info("ddr", "codecs directory not found: {}", dir.string());
+            log_info("ddr", "codecs directory not found: {}", dir);
             return;
         }
 
         for (const auto &file : std::filesystem::directory_iterator(dir)) {
             const auto &filename = file.path().filename();
-            const auto extension = strtolower(filename.extension().string());
+            const auto extension = filename.extension().wstring();
 
-            if (extension != ".dll") {
+            if (wcsicmp(extension.c_str(), L".dll") != 0) {
                 continue;
             }
 
-            log_info("ddr", "found DLL: {}, size: {} bytes", filename.string(), file.file_size());
-            if (filename == "k-clvsd.dll" || filename.string().find("xactengine") == 0) {
+            log_info("ddr", "found DLL: {}, size: {} bytes", filename, file.file_size());
+            if (filename == "k-clvsd.dll" || filename.wstring().find(L"xactengine") == 0) {
                 const std::wstring wcmd = L"regsvr32.exe /s \"" + file.path().wstring() + L"\"";
-                const std::string cmd = "regsvr32.exe /s \"" + file.path().string() + "\"";
+                const std::string cmd = fmt::format("regsvr32.exe /s \"{}\"", file.path());
 
                 int result = 0;
                 std::thread t([wcmd, &result]() {
@@ -111,12 +112,12 @@ namespace games::ddr {
                     log_warning("ddr", "`{}` failed, returned {}", cmd, result);
                     deferredlogs::defer_error_messages({
                         "DDR codec registration failure",
-                        fmt::format("    {} failed to register with error {}", filename.string(), result)});
+                        fmt::format("    {} failed to register with error {}", filename, result)});
                 }
 
                 static std::once_flag printed;
                 std::call_once(printed, [file]() {
-                    if (!contains_only_ascii(file.path().string())) {
+                    if (!contains_only_ascii(file.path().wstring())) {
                         log_warning(
                             "ddr",
                             "BAD PATH ERROR\n\n"
@@ -290,4 +291,50 @@ namespace games::ddr {
         // dispose device hook
         devicehook_dispose();
     }
+
+    static void get_analog_xy_axis(Analog &analog, bool &less, bool &more) {
+        if (!analog.isSet()) {
+            return;
+        }
+        const auto value = GameAPI::Analogs::getState(RI_MGR, analog);
+
+        // stepmania linux source:
+        // https://github.com/stepmania/stepmania/blob/d55acb1ba26f1c5b5e3048d6d6c0bd116625216f/src/arch/InputHandler/InputHandler_Linux_Event.cpp#L410
+
+        // for example, value cap with range [0, 255]:
+        // 127 = 0.498, neutral
+        // 128 = 0.502, both
+        // apparently some adapters report values like above; this obviously makes a lot of
+        // assumptions about bit width of the HID value cap and how the adapter reports
+        // neutral/both values, but this is good enough for parity with stepmania and its many forks
+        if (0.5001f < value && value < 0.75f) {
+            less |= true;
+            more |= true;
+
+        } else if (value <= 0.25f) {
+            less |= true;
+        } else if (value >= 0.75f) {
+            more |= true;
+        }
+    }
+
+    void get_analog_x_axis(int player, bool &left, bool &right) {
+        if (player != 1 && player != 2) {
+            log_fatal("ddr", "invalid player number: {}, expected 1 or 2", player);
+            return;
+        }
+
+        auto &analog = get_analogs().at(player == 1 ? Analogs::P1_LEFT_RIGHT : Analogs::P2_LEFT_RIGHT);
+        get_analog_xy_axis(analog, left, right);
+    }
+
+    void get_analog_y_axis(int player, bool &up, bool &down) {
+        if (player != 1 && player != 2) {
+            log_fatal("ddr", "invalid player number: {}, expected 1 or 2", player);
+            return;
+        }
+
+        auto &analog = get_analogs().at(player == 1 ? Analogs::P1_UP_DOWN : Analogs::P2_UP_DOWN);
+        get_analog_xy_axis(analog, up, down);
+    }
 }

src/spice2x/games/ddr/ddr.h

@@ -23,4 +23,7 @@ namespace games::ddr {
     private:
         void register_codecs();
     };
+
+    void get_analog_x_axis(int player, bool &left, bool &right);
+    void get_analog_y_axis(int player, bool &up, bool &down);
 }

src/spice2x/games/ddr/io.cpp

@@ -35,6 +35,50 @@ std::vector<Button> &games::ddr::get_buttons() {
     return analogs;
 }
 
+
+std::string games::ddr::get_buttons_help() {
+    // keep to max 100 characters wide
+    return
+        "For DDR pad arrows, double check that simultaneous Left+Right and Up+Down can be detected.\n"
+        "You should boot the game, enter test menu, and use Foot Panel Check.\n\n"
+        "When mapping arrows, try the following in order:\n\n"
+        "  1. If your controller uses face buttons (A/B/X/Y), bind them here.\n"
+        "  2. If your controller supports XInput, use that.\n"
+        "  3. If your controller uses analog axis for arrows, try Analogs tab.\n"
+        "  4. Otherwise, you will need to use remapping software or get a better adapter."
+        ;
+}
+
+std::vector<Analog> &games::ddr::get_analogs() {
+    static std::vector<Analog> analogs;
+
+    if (analogs.empty()) {
+        analogs = GameAPI::Analogs::getAnalogs("Dance Dance Revolution");
+
+        using namespace GameAPI::Analogs;
+
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            { "P1 Left-Right (Axis Fix)", AnalogType::LinearCentered },
+            { "P1 Up-Down (Axis Fix)", AnalogType::LinearCentered },
+            { "P2 Left-Right (Axis Fix)", AnalogType::LinearCentered },
+            { "P2 Up-Down (Axis Fix)", AnalogType::LinearCentered }
+        });
+    }
+    return analogs;
+}
+
+std::string games::ddr::get_analogs_help() {
+    // keep to max 100 characters wide
+    return
+        "Only use this if your DDR pad outputs analog axis for arrows.\n\n"
+        "If the pad uses face buttons (A/B/X/Y), use Buttons tab instead.\n\n"
+        "Spice will treat values <=25% as Left, >=75% as Right, ~=50% as neutral,\n"
+        "and value between 50% and 75% as both arrows.\n\n"
+        "This is the classic Stepmania \"Axis Fix\" which may or may not work with\n"
+        "your dance pad or your adapter."
+        ;
+}
+
 std::vector<Light> &games::ddr::get_lights() {
     static std::vector<Light> lights;
 

src/spice2x/games/ddr/io.h

@@ -33,6 +33,16 @@ namespace games::ddr {
         };
     }
 
+    // all analogs in correct order
+    namespace Analogs {
+        enum {
+            P1_LEFT_RIGHT,
+            P1_UP_DOWN,
+            P2_LEFT_RIGHT,
+            P2_UP_DOWN
+        };
+    }
+
     // all lights in correct order
     namespace Lights {
         enum {
@@ -177,5 +187,8 @@ namespace games::ddr {
 
     // getters
     std::vector<Button> &get_buttons();
+    std::string get_buttons_help();
+    std::string get_analogs_help();
+    std::vector<Analog> &get_analogs();
     std::vector<Light> &get_lights();
 }

src/spice2x/games/ddr/p3io/p3io.cpp

@@ -3,6 +3,7 @@
 #include "cfg/api.h"
 #include "rawinput/rawinput.h"
 #include "util/logging.h"
+#include "util/precise_timer.h"
 #include "util/utils.h"
 
 #include "../ddr.h"
@@ -459,7 +460,8 @@ int games::ddr::DDRP3IOHandle::device_io(
     if (nOutBufferSize >= 4) {
 
         // cool down
-        Sleep(1);
+        static thread_local timeutils::PreciseSleepTimer timer;
+        timer.sleep(1);
 
         // get controls as single variable (4 bytes)
         auto &controls = *(uint32_t*) lpOutBuffer;
@@ -498,32 +500,80 @@ int games::ddr::DDRP3IOHandle::device_io(
 
         // shift table
         static size_t shift_table[] = {
-                30, 28, 29, 8, 9, 10, 11, 12, 24, 25, 14, 15, 16, 17, 18, 19, 20, 26, 27, 22, 23
+                30, 28, 29,     // service, test, coin
+                8,              // p1 start
+                9, 10, 11, 12,  // p1 panel
+                24, 25, 14, 15, // p1 menu
+                16,             // p2 start
+                17, 18, 19, 20, // p2 panel
+                26, 27, 22, 23  // p2 menu 
         };
         static size_t button_table[] = {
                 Buttons::SERVICE,
                 Buttons::TEST,
                 Buttons::COIN_MECH,
+
                 Buttons::P1_START,
+
                 Buttons::P1_PANEL_UP,
                 Buttons::P1_PANEL_DOWN,
                 Buttons::P1_PANEL_LEFT,
                 Buttons::P1_PANEL_RIGHT,
+
                 Buttons::P1_MENU_UP,
                 Buttons::P1_MENU_DOWN,
                 Buttons::P1_MENU_LEFT,
                 Buttons::P1_MENU_RIGHT,
+
                 Buttons::P2_START,
+
                 Buttons::P2_PANEL_UP,
                 Buttons::P2_PANEL_DOWN,
                 Buttons::P2_PANEL_LEFT,
                 Buttons::P2_PANEL_RIGHT,
+
                 Buttons::P2_MENU_UP,
                 Buttons::P2_MENU_DOWN,
                 Buttons::P2_MENU_LEFT,
                 Buttons::P2_MENU_RIGHT,
         };
 
+        // get analogs
+        struct {
+            bool up;
+            bool down;
+            bool left;
+            bool right;
+        } analog_values[2] = {};
+        games::ddr::get_analog_x_axis(1, analog_values[0].left, analog_values[0].right);
+        games::ddr::get_analog_y_axis(1, analog_values[0].up, analog_values[0].down);
+        games::ddr::get_analog_x_axis(2, analog_values[1].left, analog_values[1].right);
+        games::ddr::get_analog_y_axis(2, analog_values[1].up, analog_values[1].down);
+        if (analog_values[0].up) {
+            controls |= 1 << 9;
+        }
+        if (analog_values[0].down) {
+            controls |= 1 << 10;
+        }
+        if (analog_values[0].left) {
+            controls |= 1 << 11;
+        }
+        if (analog_values[0].right) {
+            controls |= 1 << 12;
+        }
+        if (analog_values[1].up) {
+            controls |= 1 << 17;
+        }
+        if (analog_values[1].down) {
+            controls |= 1 << 18;
+        }
+        if (analog_values[1].left) {
+            controls |= 1 << 19;
+        }
+        if (analog_values[1].right) {
+            controls |= 1 << 20;
+        }
+
         // update states
         auto &buttons = get_buttons();
         size_t count = 0;

src/spice2x/games/ddr/p4io/p4io.cpp

@@ -2,6 +2,7 @@
 
 #include "rawinput/rawinput.h"
 #include "util/logging.h"
+#include "util/precise_timer.h"
 #include "util/utils.h"
 
 #include "../io.h"
@@ -255,7 +256,8 @@ namespace games::ddr {
                 case P4IO_IOCTL_GET_INPUTS: {
                     
                     // Prevents this function from being called at its normal 2000-2500 kHz cadence and overloading the CPU, instead reduces it to 250 Hz
-                    Sleep(4);
+                    static thread_local timeutils::PreciseSleepTimer timer;
+                    timer.sleep(4);
                     
                     memset(lpOutBuffer, 0, 16);
                     auto controls = (uint32_t*) lpOutBuffer;

src/spice2x/games/drs/drs.cpp

@@ -7,6 +7,7 @@
 #include "games/game.h"
 #include "util/detour.h"
 #include "util/logging.h"
+#include "util/precise_timer.h"
 #include "util/memutils.h"
 #include "rgb_cam.h"
 
@@ -221,10 +222,11 @@ namespace games::drs {
     void start_touch() {
         std::thread t([] {
             log_info("drs", "starting touch input thread");
+            timeutils::PreciseSleepTimer timer;
 
             // main loop
             while (TRUE) {
-                std::this_thread::sleep_for(std::chrono::milliseconds(1));
+                timer.sleep(1);
 
                 TOUCH_EVENTS.clear();
                 touch_get_events(TOUCH_EVENTS);

src/spice2x/games/ftt/io.cpp

@@ -35,13 +35,14 @@ std::vector<Analog> &games::ftt::get_analogs() {
     if (analogs.empty()) {
         analogs = GameAPI::Analogs::getAnalogs("FutureTomTom");
 
-        GameAPI::Analogs::sortAnalogs(
-                &analogs,
-                "Pad 1",
-                "Pad 2",
-                "Pad 3",
-                "Pad 4"
-        );
+        using GameAPI::Analogs::AnalogType;
+
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            { "Pad 1", AnalogType::LinearPositive },
+            { "Pad 2", AnalogType::LinearPositive },
+            { "Pad 3", AnalogType::LinearPositive },
+            { "Pad 4", AnalogType::LinearPositive }
+        });
     }
 
     return analogs;

src/spice2x/games/gitadora/asio.cpp

@@ -0,0 +1,171 @@
+#include "asio.h"
+
+#include <windows.h>
+#include <cstring>
+
+#include "avs/game.h"
+#include "gitadora.h"
+#include "util/detour.h"
+#include "util/logging.h"
+
+namespace games::gitadora {
+
+    // Redirects the game's hard-coded "XONAR" ASIO driver lookup to the
+    // driver name in ASIO_DRIVER by intercepting registry calls to
+    // HKLM\SOFTWARE\ASIO. Sentinel HKEY values mark the redirected handles
+    // so we can recognise them on subsequent reg* calls.
+
+    static const HKEY PARENT_ASIO_REG_HANDLE = reinterpret_cast<HKEY>(0x4001);
+    static const HKEY DEVICE_ASIO_REG_HANDLE = reinterpret_cast<HKEY>(0x4002);
+    static const char *FAKE_ASIO_DEVICE_NAME = "XONAR";
+
+    static decltype(RegCloseKey) *RegCloseKey_orig = nullptr;
+    static decltype(RegEnumKeyA) *RegEnumKeyA_orig = nullptr;
+    static decltype(RegOpenKeyA) *RegOpenKeyA_orig = nullptr;
+    static decltype(RegOpenKeyExA) *RegOpenKeyExA_orig = nullptr;
+    static decltype(RegQueryValueExA) *RegQueryValueExA_orig = nullptr;
+
+    static HKEY real_asio_reg_handle = nullptr;
+    static HKEY real_asio_device_reg_handle = nullptr;
+
+    static LONG WINAPI RegOpenKeyExA_hook(HKEY hKey, LPCSTR lpSubKey, DWORD ulOptions, REGSAM samDesired,
+            PHKEY phkResult)
+    {
+        if (ASIO_DRIVER.has_value() &&
+            lpSubKey != nullptr &&
+            phkResult != nullptr &&
+            hKey == PARENT_ASIO_REG_HANDLE &&
+            _stricmp(lpSubKey, FAKE_ASIO_DEVICE_NAME) == 0) {
+
+            *phkResult = DEVICE_ASIO_REG_HANDLE;
+
+            log_info("gitadora::asio", "replacing '{}' with '{}'", lpSubKey, ASIO_DRIVER.value());
+            const auto result = RegOpenKeyExA_orig(
+                    real_asio_reg_handle,
+                    ASIO_DRIVER.value().c_str(),
+                    ulOptions,
+                    samDesired,
+                    &real_asio_device_reg_handle);
+
+            if (result != ERROR_SUCCESS) {
+                log_warning(
+                    "gitadora::asio",
+                    "failed to open registry subkey '{}', error=0x{:x}",
+                    ASIO_DRIVER.value(), result);
+                log_warning(
+                    "gitadora::asio",
+                    "due to improper ASIO setting, audio init will fail");
+            }
+
+            return result;
+        }
+
+        return RegOpenKeyExA_orig(hKey, lpSubKey, ulOptions, samDesired, phkResult);
+    }
+
+    static LONG WINAPI RegOpenKeyA_hook(HKEY hKey, LPCSTR lpSubKey, PHKEY phkResult) {
+        if (ASIO_DRIVER.has_value() &&
+            lpSubKey != nullptr &&
+            phkResult != nullptr &&
+            hKey == HKEY_LOCAL_MACHINE &&
+            _stricmp(lpSubKey, "software\\asio") == 0)
+        {
+            *phkResult = PARENT_ASIO_REG_HANDLE;
+            return RegOpenKeyA_orig(hKey, lpSubKey, &real_asio_reg_handle);
+        }
+
+        return RegOpenKeyA_orig(hKey, lpSubKey, phkResult);
+    }
+
+    static LONG WINAPI RegEnumKeyA_hook(HKEY hKey, DWORD dwIndex, LPSTR lpName, DWORD cchName) {
+        if (hKey == PARENT_ASIO_REG_HANDLE && ASIO_DRIVER.has_value()) {
+            if (dwIndex == 0) {
+                // forward to real handle just to verify the key exists; we
+                // overwrite the name with our fake driver string regardless
+                auto ret = RegEnumKeyA_orig(real_asio_reg_handle, dwIndex, lpName, cchName);
+                if (ret == ERROR_SUCCESS && lpName != nullptr && cchName > 0) {
+                    log_info("gitadora::asio", "stubbing '{}' with '{}'", lpName, FAKE_ASIO_DEVICE_NAME);
+                    strncpy(lpName, FAKE_ASIO_DEVICE_NAME, cchName);
+                    lpName[cchName - 1] = '\0';
+                }
+                return ret;
+            } else {
+                return ERROR_NO_MORE_ITEMS;
+            }
+        }
+
+        return RegEnumKeyA_orig(hKey, dwIndex, lpName, cchName);
+    }
+
+    static LONG WINAPI RegQueryValueExA_hook(HKEY hKey, LPCSTR lpValueName, LPDWORD lpReserved, LPDWORD lpType,
+            LPBYTE lpData, LPDWORD lpcbData)
+    {
+        HKEY target = hKey;
+
+        if (ASIO_DRIVER.has_value() &&
+            lpValueName != nullptr &&
+            lpData != nullptr &&
+            lpcbData != nullptr &&
+            hKey == DEVICE_ASIO_REG_HANDLE) {
+            if (_stricmp(lpValueName, "Description") == 0) {
+                // engine may verify the driver name after open; ensure it still
+                // sees something containing "XONAR" so the substring check passes
+                const size_t len = strlen(FAKE_ASIO_DEVICE_NAME) + 1;
+                if (*lpcbData < len) {
+                    *lpcbData = static_cast<DWORD>(len);
+                    return ERROR_MORE_DATA;
+                }
+                memcpy(lpData, FAKE_ASIO_DEVICE_NAME, len);
+                *lpcbData = static_cast<DWORD>(len);
+                if (lpType != nullptr) {
+                    *lpType = REG_SZ;
+                }
+                return ERROR_SUCCESS;
+            }
+
+            // for everything else (CLSID etc.) defer to the real driver subkey
+            target = real_asio_device_reg_handle;
+        }
+
+        return RegQueryValueExA_orig(target, lpValueName, lpReserved, lpType, lpData, lpcbData);
+    }
+
+    static LONG WINAPI RegCloseKey_hook(HKEY hKey) {
+        if (hKey == PARENT_ASIO_REG_HANDLE) {
+            if (real_asio_reg_handle != nullptr) {
+                RegCloseKey_orig(real_asio_reg_handle);
+                real_asio_reg_handle = nullptr;
+            }
+            return ERROR_SUCCESS;
+        }
+
+        if (hKey == DEVICE_ASIO_REG_HANDLE) {
+            if (real_asio_device_reg_handle != nullptr) {
+                RegCloseKey_orig(real_asio_device_reg_handle);
+                real_asio_device_reg_handle = nullptr;
+            }
+            return ERROR_SUCCESS;
+        }
+
+        return RegCloseKey_orig(hKey);
+    }
+
+    void asio_hook_init() {
+        if (!ASIO_DRIVER.has_value()) {
+            return;
+        }
+
+        log_info("gitadora::asio", "installing ASIO driver redirect: XONAR -> {}", ASIO_DRIVER.value());
+
+        RegCloseKey_orig = detour::iat_try(
+                "RegCloseKey", RegCloseKey_hook, avs::game::DLL_INSTANCE);
+        RegEnumKeyA_orig = detour::iat_try(
+                "RegEnumKeyA", RegEnumKeyA_hook, avs::game::DLL_INSTANCE);
+        RegOpenKeyA_orig = detour::iat_try(
+                "RegOpenKeyA", RegOpenKeyA_hook, avs::game::DLL_INSTANCE);
+        RegOpenKeyExA_orig = detour::iat_try(
+                "RegOpenKeyExA", RegOpenKeyExA_hook, avs::game::DLL_INSTANCE);
+        RegQueryValueExA_orig = detour::iat_try(
+                "RegQueryValueExA", RegQueryValueExA_hook, avs::game::DLL_INSTANCE);
+    }
+}

src/spice2x/games/gitadora/asio.h

@@ -0,0 +1,12 @@
+#pragma once
+
+namespace games::gitadora {
+
+    // installs IAT registry hooks in gfdm.dll that redirect the game's
+    // ASIO driver lookup (hard-coded "XONAR" substring) to a user-chosen
+    // driver name read from games::gitadora::ASIO_DRIVER.
+    //
+    // safe to call unconditionally; if ASIO_DRIVER is unset the hooks
+    // forward every call straight through to advapi32.
+    void asio_hook_init();
+}

src/spice2x/games/gitadora/gitadora.cpp

@@ -1,18 +1,27 @@
 #include "gitadora.h"
+#include "asio.h"
 #include "handle.h"
 #include "bi2x_hook.h"
 #include <unordered_map>
+
+#include <ks.h>
+#include <ksmedia.h>
+
 #include "cfg/configurator.h"
+#include "hooks/audio/audio.h"
 #include "hooks/audio/mme.h"
 #include "hooks/graphics/graphics.h"
 #include "misc/wintouchemu.h"
 #include "overlay/overlay.h"
 #include "util/cpuutils.h"
+#include "util/deferlog.h"
 #include "util/detour.h"
 #include "util/libutils.h"
 #include "util/logging.h"
 #include "util/sigscan.h"
 #include "util/socd_cleaner.h"
+#include "util/sysutils.h"
+#include "util/utils.h"
 #include "hooks/setupapihook.h"
 
 namespace games::gitadora {
@@ -20,11 +29,13 @@ namespace games::gitadora {
     // settings
     bool TWOCHANNEL = false;
     std::optional<unsigned int> CAB_TYPE = std::nullopt;
-    bool ARENA_SINGLE_WINDOW = false;
     bool P1_LEFTY = false;
     bool P2_LEFTY = false;
     std::optional<std::string> SUBSCREEN_OVERLAY_SIZE;
     std::optional<socd::SocdAlgorithm> PICK_ALGO = socd::SocdAlgorithm::PreferRecent;
+    std::optional<uint8_t> ARENA_WINDOW_COUNT = std::nullopt;
+    std::optional<std::string> ASIO_DRIVER = std::nullopt;
+    bool ALLOW_REALTEK_AUDIO = false;
 
     /*
      * Prevent GitaDora from creating folders on F drive
@@ -221,13 +232,6 @@ namespace games::gitadora {
             }
 #endif
 
-            // arena model launches a tiny window yet backbuffer at 4k, resulting in unusable overlay
-            // force scaling to make things usable
-            if (!overlay::UI_SCALE_PERCENT.has_value() && is_arena_model() &&
-                GRAPHICS_WINDOWED && !cfg::CONFIGURATOR_STANDALONE) {
-                overlay::UI_SCALE_PERCENT = 250;
-            }
-
             // for guitar wail SOCD cleaning
             socd::ALGORITHM = socd::SocdAlgorithm::PreferRecent;
 
@@ -237,12 +241,334 @@ namespace games::gitadora {
             } else {
                 log_info("gitadora", "guitar pick SOCD algorithm: legacy");
             }
+
+#if SPICE64 && !SPICE_XP
+
+            if (is_arena_model()) {
+                // in full screen, if single-adapter option is checked, it's functionally
+                // the same as forcing a single monitor
+                if (!GRAPHICS_WINDOWED && GRAPHICS_FORCE_SINGLE_ADAPTER) {
+                    ARENA_WINDOW_COUNT = 1;
+                }
+
+                // figure out default settings if user didn't provide one
+                if (!ARENA_WINDOW_COUNT.has_value()) {
+                    if (!GRAPHICS_WINDOWED && sysutils::enumerate_monitors().size() < 4) {
+                        log_info("gitadora", "arena model: <4 monitors, defaulting to single window mode");
+                        ARENA_WINDOW_COUNT = 1;
+                    } else {
+                        ARENA_WINDOW_COUNT = 4;
+                    }
+                }
+
+                const int count = ARENA_WINDOW_COUNT.value();
+                switch (count) {
+                    case 1:
+                        log_info("gitadora", "arena model: single-window mode");
+                        GRAPHICS_FORCE_SINGLE_ADAPTER = true;
+                        GRAPHICS_PREVENT_SECONDARY_WINDOWS = true;
+                        break;
+                    case 2:
+                        if (!GRAPHICS_WINDOWED) {
+                            log_fatal(
+                                "gitadora",
+                                "arena model: 2-window mode is not supported in fullscreen, choose 1 or 4");
+                        }
+                        log_info("gitadora", "arena model: two-window mode");
+                        GRAPHICS_GITADORA_HIDE_SIDE_WINDOWS = true;
+                        break;
+                    case 4:
+                        log_info("gitadora", "arena model: four-window mode");
+                        break;
+                    default:
+                        log_fatal(
+                            "gitadora",
+                            "arena model: unsupported window count: {}", count);
+                }
+            }
+
+#endif
+
         }
     }
 
+#if SPICE64 && !SPICE_XP
+
+    static decltype(GetDisplayConfigBufferSizes) *GetDisplayConfigBufferSizes_orig = nullptr;
+    static decltype(QueryDisplayConfig) *QueryDisplayConfig_orig = nullptr;
+    static decltype(DisplayConfigGetDeviceInfo) *DisplayConfigGetDeviceInfo_orig = nullptr;
+
+    // cached primary real monitor: its path + source/target mode entries.
+    // modeInfoIdx values on the path are renumbered to 0 / 1 so the cache is self-contained.
+    static DISPLAYCONFIG_PATH_INFO real_primary_path = {};
+    static DISPLAYCONFIG_MODE_INFO real_primary_modes[2] = {}; // [0]=source, [1]=target
+
+    // fake monitors appended after the real ones. the game classifies monitors
+    // by outputTechnology + connectorInstance:
+    //   HDMI               -> main 4k monitor (real primary)
+    //   DP connInstance 0  -> left
+    //   DP connInstance 1  -> right
+    //   DP connInstance 2  -> small (sub/touch)
+    // ids are negated on the fake monitor headers so they can be distinguished
+    // from real ones in DisplayConfigGetDeviceInfo.
+    struct FakeMonitor {
+        LONG id;
+        int width;
+        int height;
+        int offset_x;
+        int offset_y;
+        UINT32 connector_instance;
+    };
+
+    // ORDERING MATTERS: the d3d9 wrapper (FAKE_SUBSCREEN_ADAPTER) hands out
+    // adapters as "\\.\DISPLAY_SPICE_FAKE_{N}" for N=1,2,3. The game maps each
+    // adapter to a swap chain role via its DisplayConfig connector instance,
+    // so the entries here must be listed in the same order the wrapper enumerates
+    // them: id=1 -> adapter 1 -> left, id=2 -> adapter 2 -> right, id=3 -> adapter 3 -> small.
+    static constexpr FakeMonitor FAKE_MONITORS[] = {
+        { 1, 1080, 1920, -100000, -100000, 0 }, // left  (DP connector instance 0)
+        { 2, 1080, 1920, -200000, -200000, 1 }, // right (DP connector instance 1)
+        { 3,  800, 1280, -300000, -300000, 2 }, // small (DP connector instance 2, touch)
+    };
+    static constexpr UINT32 FAKE_MONITOR_COUNT = static_cast<UINT32>(std::size(FAKE_MONITORS));
+
+    // call QueryDisplayConfig once, keep only the primary monitor's path and its
+    // two referenced modes (source + target). modeInfoIdx values are rewritten to
+    // 0 and 1 so the cache is self-consistent.
+    static void cache_primary_monitor_info() {
+        UINT32 path_count = 0;
+        UINT32 mode_count = 0;
+        if (GetDisplayConfigBufferSizes_orig(
+                QDC_ONLY_ACTIVE_PATHS, &path_count, &mode_count) != ERROR_SUCCESS) {
+            log_fatal("gitadora", "cache_primary_monitor_info: GetDisplayConfigBufferSizes failed");
+        }
+
+        std::vector<DISPLAYCONFIG_PATH_INFO> all_paths(path_count);
+        std::vector<DISPLAYCONFIG_MODE_INFO> all_modes(mode_count);
+        if (QueryDisplayConfig_orig(
+                QDC_ONLY_ACTIVE_PATHS,
+                &path_count, all_paths.data(),
+                &mode_count, all_modes.data(),
+                nullptr) != ERROR_SUCCESS) {
+            log_fatal("gitadora", "cache_primary_monitor_info: QueryDisplayConfig failed");
+        }
+        all_paths.resize(path_count);
+        all_modes.resize(mode_count);
+
+        // pick the primary monitor: source mode at (0, 0)
+        auto primary = std::find_if(all_paths.begin(), all_paths.end(),
+            [&](const auto &p) {
+                const auto idx = p.sourceInfo.modeInfoIdx;
+                return idx < all_modes.size() &&
+                       all_modes[idx].infoType == DISPLAYCONFIG_MODE_INFO_TYPE_SOURCE &&
+                       all_modes[idx].sourceMode.position.x == 0 &&
+                       all_modes[idx].sourceMode.position.y == 0;
+            });
+        if (primary == all_paths.end()) {
+            log_fatal("gitadora", "cache_primary_monitor_info: no primary monitor found");
+        }
+
+        real_primary_modes[0] = all_modes[primary->sourceInfo.modeInfoIdx];
+        real_primary_modes[1] = all_modes[primary->targetInfo.modeInfoIdx];
+        real_primary_path = *primary;
+        real_primary_path.sourceInfo.modeInfoIdx = 0;
+        real_primary_path.targetInfo.modeInfoIdx = 1;
+
+        log_info("gitadora", "cache_primary_monitor_info: cached primary monitor");
+    }
+
+    static
+    LONG
+    WINAPI
+    GetDisplayConfigBufferSizes_hook(
+        UINT32 Flags,
+        UINT32 *pNumPathArrayElements,
+        UINT32 *pNumModeInfoArrayElements)
+    {
+        // populate cached primary real monitor on the first call
+        static std::once_flag populate_once;
+        std::call_once(populate_once, cache_primary_monitor_info);
+
+        // always report exactly 1 real + N fake monitors
+        *pNumPathArrayElements = 1 + FAKE_MONITOR_COUNT;
+        *pNumModeInfoArrayElements = 2 + FAKE_MONITOR_COUNT * 2;
+
+        log_info(
+            "gitadora",
+            "GetDisplayConfigBufferSizes: 1 real path + {} fake monitor(s)",
+            FAKE_MONITOR_COUNT);
+
+        return ERROR_SUCCESS;
+    }
+
+    // write fake monitor i into the caller's path/mode arrays. layout (single-monitor
+    // assumption): index 0 in both arrays holds the cached primary real monitor, so
+    // fake i occupies path slot (1 + i) and mode slots (2 + i*2) / (2 + i*2 + 1).
+    static void insert_fake_monitor(
+        DISPLAYCONFIG_PATH_INFO *paths,
+        DISPLAYCONFIG_MODE_INFO *modes,
+        UINT32 i)
+    {
+        const FakeMonitor &m = FAKE_MONITORS[i];
+        const UINT32 src_idx = 2 + i * 2;
+        const UINT32 tgt_idx = src_idx + 1;
+        const LUID adapter_id { .LowPart = static_cast<DWORD>(-m.id), .HighPart = -m.id };
+        const UINT32 uid = static_cast<UINT32>(-m.id);
+
+        paths[1 + i] = {
+            .sourceInfo = {
+                .adapterId = adapter_id,
+                .id = uid,
+                .modeInfoIdx = src_idx,
+                .statusFlags = DISPLAYCONFIG_SOURCE_IN_USE,
+            },
+            .targetInfo = {
+                .adapterId = adapter_id,
+                .id = uid,
+                .modeInfoIdx = tgt_idx,
+                .outputTechnology = DISPLAYCONFIG_OUTPUT_TECHNOLOGY_DISPLAYPORT_EXTERNAL,
+                .rotation = DISPLAYCONFIG_ROTATION_IDENTITY,
+                .scaling = DISPLAYCONFIG_SCALING_IDENTITY,
+                .refreshRate = { .Numerator = 60000, .Denominator = 1000 },
+                .scanLineOrdering = DISPLAYCONFIG_SCANLINE_ORDERING_PROGRESSIVE,
+                .targetAvailable = TRUE,
+                .statusFlags = DISPLAYCONFIG_TARGET_IN_USE,
+            },
+            .flags = DISPLAYCONFIG_PATH_ACTIVE,
+        };
+
+        modes[src_idx] = {
+            .infoType = DISPLAYCONFIG_MODE_INFO_TYPE_SOURCE,
+            .id = uid,
+            .adapterId = adapter_id,
+            .sourceMode = {
+                .width = static_cast<UINT32>(m.width),
+                .height = static_cast<UINT32>(m.height),
+                .pixelFormat = DISPLAYCONFIG_PIXELFORMAT_32BPP,
+                .position = { .x = m.offset_x, .y = m.offset_y },
+            },
+        };
+
+        modes[tgt_idx] = {
+            .infoType = DISPLAYCONFIG_MODE_INFO_TYPE_TARGET,
+            .id = uid,
+            .adapterId = adapter_id,
+            .targetMode = {},
+        };
+
+        log_misc(
+            "gitadora",
+            "inserted fake monitor: id={}, width={}, height={}, offset_x={}, offset_y={}",
+            m.id, m.width, m.height, m.offset_x, m.offset_y);
+    }
+
+    static
+    LONG
+    WINAPI
+    QueryDisplayConfig_hook(
+        UINT32 flags,
+        UINT32* numPathArrayElements,
+        DISPLAYCONFIG_PATH_INFO* pathArray,
+        UINT32* numModeInfoArrayElements,
+        DISPLAYCONFIG_MODE_INFO* modeInfoArray,
+        DISPLAYCONFIG_TOPOLOGY_ID* currentTopologyId)
+    {
+        // copy cached primary real monitor into caller buffers at index 0
+        pathArray[0] = real_primary_path;
+        modeInfoArray[0] = real_primary_modes[0];
+        modeInfoArray[1] = real_primary_modes[1];
+        *numPathArrayElements = 1 + FAKE_MONITOR_COUNT;
+        *numModeInfoArrayElements = 2 + FAKE_MONITOR_COUNT * 2;
+        if (currentTopologyId != nullptr) {
+            *currentTopologyId = DISPLAYCONFIG_TOPOLOGY_EXTEND;
+        }
+
+        log_misc("gitadora", "QueryDisplayConfig returning fake monitor paths and modes");
+
+        // append fake monitors after the real one
+        for (UINT32 i = 0; i < FAKE_MONITOR_COUNT; i++) {
+            insert_fake_monitor(pathArray, modeInfoArray, i);
+        }
+
+        return ERROR_SUCCESS;
+    }
+
+    static
+    LONG
+    WINAPI
+    DisplayConfigGetDeviceInfo_hook(DISPLAYCONFIG_DEVICE_INFO_HEADER* requestPacket)
+    {
+        if (requestPacket == nullptr) {
+            return DisplayConfigGetDeviceInfo_orig(requestPacket);
+        }
+
+        // handle fake monitors (negative id) directly without calling orig
+        const auto id = static_cast<int>(requestPacket->id);
+        if (id < 0) {
+            if (requestPacket->type == DISPLAYCONFIG_DEVICE_INFO_GET_SOURCE_NAME) {
+                const auto sourceName = reinterpret_cast<DISPLAYCONFIG_SOURCE_DEVICE_NAME*>(requestPacket);
+                // name must match WrappedIDirect3D9::GetAdapterIdentifier
+                const std::string adapter_name = fmt::format("\\\\.\\DISPLAY_SPICE_FAKE_{}", -id);
+                wcscpy(sourceName->viewGdiDeviceName, s2ws(adapter_name).c_str());
+                log_misc("gitadora",
+                    "DisplayConfigGetDeviceInfo: fake source id={} name={}", id, adapter_name);
+                return ERROR_SUCCESS;
+            }
+            if (requestPacket->type == DISPLAYCONFIG_DEVICE_INFO_GET_TARGET_NAME) {
+                const auto targetName = reinterpret_cast<DISPLAYCONFIG_TARGET_DEVICE_NAME*>(requestPacket);
+                const LONG fake_id = -id;
+                UINT32 conn_inst = 0xff;
+                for (const auto& f : FAKE_MONITORS) {
+                    if (f.id == fake_id) {
+                        conn_inst = f.connector_instance;
+                        break;
+                    }
+                }
+                targetName->outputTechnology = DISPLAYCONFIG_OUTPUT_TECHNOLOGY_DISPLAYPORT_EXTERNAL;
+                targetName->connectorInstance = conn_inst;
+                wcscpy(targetName->monitorFriendlyDeviceName, L"Spice Fake Monitor");
+                wcscpy(targetName->monitorDevicePath, L"\\\\?\\SpiceFakeMonitor");
+                log_misc("gitadora",
+                    "DisplayConfigGetDeviceInfo: fake target id={} -> DP connInst {}",
+                    id, targetName->connectorInstance);
+                return ERROR_SUCCESS;
+            }
+        }
+
+        const auto ret = DisplayConfigGetDeviceInfo_orig(requestPacket);
+        if (ret != ERROR_SUCCESS ||
+            requestPacket->type != DISPLAYCONFIG_DEVICE_INFO_GET_TARGET_NAME) {
+            return ret;
+        }
+
+        // override the cached primary real monitor target info to look like HDMI/0
+        const auto targetName = reinterpret_cast<DISPLAYCONFIG_TARGET_DEVICE_NAME*>(requestPacket);
+        const auto& target = real_primary_path.targetInfo;
+        if (target.id == targetName->header.id &&
+            target.adapterId.HighPart == targetName->header.adapterId.HighPart &&
+            target.adapterId.LowPart == targetName->header.adapterId.LowPart)
+        {
+            targetName->outputTechnology = DISPLAYCONFIG_OUTPUT_TECHNOLOGY_HDMI;
+            targetName->connectorInstance = 0;
+            log_info("gitadora",
+                "overriding primary monitor (id={}) to pretend to be HDMI",
+                targetName->header.id);
+        }
+        return ret;
+    }
+
+#endif
+
     void GitaDoraGame::attach() {
         Game::attach();
 
+        // arena model launches a tiny window yet backbuffer at 4k, resulting in unusable overlay
+        // force scaling to make things usable
+        if (!overlay::UI_SCALE_PERCENT.has_value() && is_arena_model()) {
+            log_info("gitadora", "forcing UI scale to 250% for arena model");
+            overlay::UI_SCALE_PERCENT = 250;
+        }
+
         // modules
         HMODULE sharepj_module = libutils::try_module("libshare-pj.dll");
         HMODULE bmsd2_module = libutils::try_module("libbmsd2.dll");
@@ -269,6 +595,12 @@ namespace games::gitadora {
 #ifdef SPICE64
         // gitadora arena model 
         auto aio = libutils::try_library("libaio.dll");
+
+        // before we start patching and hooking things, detect invalid configuration
+        if (aio != nullptr && !is_arena_model()) {
+            log_fatal("gitadora", "arena model i/o (libaio.dll) detected, but <spec> is not an arena model - bad prop XML files?");
+        }
+
         if (aio != nullptr) {
             SETUPAPI_SETTINGS settings{};
             settings.class_guid[0] = 0x86E0D1E0;
@@ -294,39 +626,76 @@ namespace games::gitadora {
             detour::iat_try("GetDriveTypeA", GetDriveTypeA_hook, avs::game::DLL_INSTANCE);
             detour::iat_try("CreateDirectoryA", CreateDirectoryA_hook, avs::game::DLL_INSTANCE);
 
+            // ASIO driver redirect (XONAR -> user-configured driver)
+            asio_hook_init();
+
             // volume change prevention
             hooks::audio::mme::init(avs::game::DLL_INSTANCE);
 
-            // touch hook
-            if (ARENA_SINGLE_WINDOW) {
+            // fake Realtek audio injection
+            // if ASIO init succeeds, game tries to look for audio device with `Realtek` in friendly name
+            // if ASIO init fails, game opens default audio device
+            // therefore, it's safe to enable this hook by default regardless of ASIO preference
+            // (unless the user explicitly disables it, of course)
+            if (ALLOW_REALTEK_AUDIO) {
+                log_info(
+                    "gitadora",
+                    "fake Realtek audio injection disabled "
+                    "(user's real Realtek audio may be used after successful ASIO init)");
+            } else {
+                log_info(
+                    "gitadora",
+                    "fake Realtek audio injection enabled "
+                    "(create a fake Realtek audio device to prevent crashes after successful ASIO init)");
+                hooks::audio::INJECT_FAKE_REALTEK_AUDIO = true;
+            }
+
+            // monitor/touch hooks (windowed or full screen)
+            if (GRAPHICS_PREVENT_SECONDARY_WINDOWS) {
+                // enable touch hook for subscreen overlay
                 wintouchemu::FORCE = true;
                 wintouchemu::INJECT_MOUSE_AS_WM_TOUCH = true;
                 wintouchemu::hook("GITADORA", avs::game::DLL_INSTANCE);
-            }
-            return;
-        }
 
-        HMODULE gdme_module = libutils::try_module("libgdme.dll");
+#if !SPICE_XP
+
+                if (!GRAPHICS_WINDOWED) {
+                    // monitor hook: always pretend to have 1 primary real monitor + 3 fake monitors
+                    // (LEFT / RIGHT / SMALL) so the game accepts the arena cab topology
+                    GetDisplayConfigBufferSizes_orig =
+                        detour::iat_try("GetDisplayConfigBufferSizes",
+                            GetDisplayConfigBufferSizes_hook, avs::game::DLL_INSTANCE);
+                    QueryDisplayConfig_orig =
+                        detour::iat_try("QueryDisplayConfig",
+                            QueryDisplayConfig_hook, avs::game::DLL_INSTANCE);
+                    DisplayConfigGetDeviceInfo_orig =
+                        detour::iat_try("DisplayConfigGetDeviceInfo",
+                            DisplayConfigGetDeviceInfo_hook, avs::game::DLL_INSTANCE);
+                }
+#endif
+
+            }
+        }
 
         // window patch
-        if (GRAPHICS_WINDOWED && !replace_pattern(
-                gdme_module,
-                "754185ED753D8B4118BF0000CB02",
-                "9090????9090??????????????12", 0, 0))
-        {
-            log_warning("gitadora", "windowed mode failed");
+        if (!is_arena_model()) {
+            HMODULE gdme_module = libutils::try_module("libgdme.dll");
+            if (GRAPHICS_WINDOWED && !replace_pattern(
+                    gdme_module,
+                    "754185ED753D8B4118BF0000CB02",
+                    "9090????9090??????????????12", 0, 0)) {
+                log_warning("gitadora", "windowed mode failed");
+            }
         }
 
-        HMODULE bmsd_engine_module = libutils::try_module("libbmsd-engine.dll");
-        HMODULE bmsd_module = libutils::try_module("libbmsd.dll");
-
         // two channel mod
-        if (TWOCHANNEL) {
-            bmsd2_boot_orig = detour::iat_try("bmsd2_boot", bmsd2_boot_hook, bmsd_module);
+        if (TWOCHANNEL && !is_arena_model()) {
+            HMODULE bmsd_engine_module = libutils::try_module("libbmsd-engine.dll");
+            HMODULE bmsd_module = libutils::try_module("libbmsd.dll");
 
+            bmsd2_boot_orig = detour::iat_try("bmsd2_boot", bmsd2_boot_hook, bmsd_module);
             if (!(replace_pattern(bmsd_engine_module, "33000000488D", "03??????????", 0, 0) ||
-                    replace_pattern(bmsd_engine_module, "330000000F10", "03??????????", 0, 0)))
-            {
+                    replace_pattern(bmsd_engine_module, "330000000F10", "03??????????", 0, 0))) {
                 log_warning("gitadora", "two channel mode failed");
             }
         }
@@ -334,4 +703,18 @@ namespace games::gitadora {
 #endif
 
     }
+
+    void fix_audio_channel_mask(WAVEFORMATEX *format) {
+        if (!format || format->wFormatTag != WAVE_FORMAT_EXTENSIBLE) {
+            return;
+        }
+
+        auto ext = reinterpret_cast<WAVEFORMATEXTENSIBLE *>(format);
+
+        // fix the legacy 7.1 channel mask to the modern surround layout
+        // makes it more compatible with modern audio cards
+        if (ext->dwChannelMask == KSAUDIO_SPEAKER_7POINT1) {
+            ext->dwChannelMask = KSAUDIO_SPEAKER_7POINT1_SURROUND;
+        }
+    }
 }

src/spice2x/games/gitadora/gitadora.h

@@ -2,6 +2,9 @@
 
 #include <optional>
 
+#include <windows.h>
+#include <mmreg.h>
+
 #include "avs/game.h"
 #include "games/game.h"
 #include "util/socd_cleaner.h"
@@ -11,11 +14,13 @@ namespace games::gitadora {
     // settings
     extern bool TWOCHANNEL;
     extern std::optional<unsigned int> CAB_TYPE;
-    extern bool ARENA_SINGLE_WINDOW;
     extern bool P1_LEFTY;
     extern bool P2_LEFTY;
     extern std::optional<std::string> SUBSCREEN_OVERLAY_SIZE;
     extern std::optional<socd::SocdAlgorithm> PICK_ALGO;
+    extern std::optional<uint8_t> ARENA_WINDOW_COUNT;
+    extern std::optional<std::string> ASIO_DRIVER;
+    extern bool ALLOW_REALTEK_AUDIO;
 
     class GitaDoraGame : public games::Game {
     public:
@@ -24,6 +29,8 @@ namespace games::gitadora {
         virtual void attach() override;
     };
 
+    void fix_audio_channel_mask(WAVEFORMATEX *format);
+
     static inline bool is_drum() {
         return (
             avs::game::is_model({ "J32", "K32", "L32" }) ||

src/spice2x/games/gitadora/io.cpp

@@ -86,19 +86,21 @@ std::vector<Button> &games::gitadora::get_buttons() {
 std::vector<Analog> &games::gitadora::get_analogs() {
     static std::vector<Analog> analogs;
 
+    using namespace GameAPI::Analogs;
+
     if (analogs.empty()) {
         analogs = GameAPI::Analogs::getAnalogs("GitaDora");
 
-        GameAPI::Analogs::sortAnalogs(&analogs,
-                "Guitar P1 Wail X",
-                "Guitar P1 Wail Y",
-                "Guitar P1 Wail Z",
-                "Guitar P1 Knob",
-                "Guitar P2 Wail X",
-                "Guitar P2 Wail Y",
-                "Guitar P2 Wail Z",
-                "Guitar P2 Knob"
-        );
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            {"Guitar P1 Wail X", AnalogType::LinearCentered},
+            {"Guitar P1 Wail Y", AnalogType::LinearCentered},
+            {"Guitar P1 Wail Z", AnalogType::LinearCentered},
+            {"Guitar P1 Knob", AnalogType::Circular},
+            {"Guitar P2 Wail X", AnalogType::LinearCentered},
+            {"Guitar P2 Wail Y", AnalogType::LinearCentered},
+            {"Guitar P2 Wail Z", AnalogType::LinearCentered},
+            {"Guitar P2 Knob", AnalogType::Circular}
+        });
     }
 
     return analogs;

src/spice2x/games/iidx/bi2a.cpp

@@ -2,6 +2,7 @@
 
 #include "misc/eamuse.h"
 #include "rawinput/rawinput.h"
+#include "util/precise_timer.h"
 #include "util/utils.h"
 
 #include "iidx.h"
@@ -82,7 +83,8 @@ bool games::iidx::IIDXFMSerialHandle::FMSerialDevice::parse_msg(
             }
 
             // sleep - otherwise the IO thread will go too hard on the CPU
-            Sleep(1);
+            static thread_local timeutils::PreciseSleepTimer timer;
+            timer.sleep(1);
 
             // generate message
             auto msg = this->create_msg(msg_in, 0x2E);

src/spice2x/games/iidx/bi2x.cpp

@@ -2,6 +2,7 @@
 
 #include "misc/eamuse.h"
 #include "rawinput/rawinput.h"
+#include "util/precise_timer.h"
 #include "util/utils.h"
 
 #include "iidx.h"
@@ -82,7 +83,8 @@ bool games::iidx::BI2XSerialHandle::BI2XDevice::parse_msg(
             }
 
             // sleep - otherwise the IO thread will go too hard on the CPU
-            Sleep(1);
+            static thread_local timeutils::PreciseSleepTimer timer;
+            timer.sleep(1);
 
             // generate message
             auto msg = this->create_msg(msg_in, 0x2E);

src/spice2x/games/iidx/iidx.cpp

@@ -67,6 +67,8 @@ namespace games::iidx {
     std::optional<std::string> SOUND_OUTPUT_DEVICE = std::nullopt;
     std::optional<std::string> SOUND_OUTPUT_DEVICE_IN_EFFECT = std::nullopt;
     std::optional<std::string> ASIO_DRIVER = std::nullopt;
+    uint32_t TT_DELAY_P1 = 0;
+    uint32_t TT_DELAY_P2 = 0;
     uint8_t DIGITAL_TT_SENS = 4;
     std::optional<std::string> SUBSCREEN_OVERLAY_SIZE = std::nullopt;
     std::optional<std::string> SCREEN_MODE = std::nullopt;
@@ -781,7 +783,7 @@ namespace games::iidx {
         }
 
         // return higher 8 bit
-        return (uint8_t) (ret_value >> 2);
+        return (uint8_t)(ret_value >> 2);
     }
 
     unsigned char get_slider(uint8_t slider) {
@@ -891,38 +893,10 @@ namespace games::iidx {
             }
         }
 
-        // patch iidx32+ for asio compatibility
-        // only do this if NOT wasapi (as opposed to checking if it's asio)
-        // the patch is only really needed for (some) non-XONAR devices but since people sometimes disguise
-        // other devices as a XONAR, don't check for the exact string (common ASIO workaround for INF)
-        if (avs::game::is_ext(2024090100, INT_MAX) &&
-            !(SOUND_OUTPUT_DEVICE_IN_EFFECT.has_value() &&
-              SOUND_OUTPUT_DEVICE_IN_EFFECT.value() == "wasapi")) {
-
-            // in iidx32 final:
-            // ff 50 08      call   QWORD PTR [rax+0x8]     ; ASIO instance AddRef
-            // 48 8b 4b 08   mov    rcx,QWORD PTR [rbx+0x8]
-            // 48 8b 01      mov    rax,QWORD PTR [rcx]
-            // ff 50 08      call   QWORD PTR [rax+0x8]     ; ASIO instance AddRef
-
-            intptr_t result = replace_pattern(
-                avs::game::DLL_INSTANCE,
-                "FF50????????????????FF50??4533C94533C0418D51",
-                "????????????????????909090??????????????????",
-                0, 0);
-
-            if (result == 0) {
-                log_warning(
-                    "iidx",
-                    "Failed to apply ASIO compatibility fix for iidx32+. "
-                    "Unless patches are applied, your ASIO device may hang or fail to work");
-            } else {
-                log_info(
-                    "iidx",
-                    "Successfully applied ASIO compatibility fix for iidx32+ using signature matching @ 0x{:x}.",
-                    result);
-            }
-        }
+        // note: the iidx32+ ASIO refcount bug (a duplicate AddRef on the ASIO instance with
+        // no matching Release, which leaks the driver and can hang non-XONAR devices) is now
+        // handled transparently by the WrappedAsio proxy (see hooks/audio/asio_proxy.cpp),
+        // so no game-DLL signature patch is needed here anymore
 
 #endif
 

src/spice2x/games/iidx/io.cpp

@@ -63,16 +63,17 @@ std::vector<Analog> &games::iidx::get_analogs() {
     if (analogs.empty()) {
         analogs = GameAPI::Analogs::getAnalogs("Beatmania IIDX");
 
-        GameAPI::Analogs::sortAnalogs(
-                &analogs,
-                "Turntable P1",
-                "Turntable P2",
-                "VEFX",
-                "Low-EQ",
-                "Hi-EQ",
-                "Filter",
-                "Play Volume"
-        );
+        using namespace GameAPI::Analogs;
+
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            { "Turntable P1", AnalogType::Circular },
+            { "Turntable P2", AnalogType::Circular },
+            { "VEFX", AnalogType::LinearPositive },
+            { "Low-EQ", AnalogType::LinearPositive },
+            { "Hi-EQ", AnalogType::LinearPositive },
+            { "Filter", AnalogType::LinearPositive },
+            { "Play Volume", AnalogType::LinearPositive }
+        });
     }
     return analogs;
 }

src/spice2x/games/iidx/local_camera.cpp

@@ -3,6 +3,7 @@
 #if SPICE64 && !SPICE_XP
 
 #include "util/logging.h"
+#include "util/precise_timer.h"
 #include "util/utils.h"
 #include "mf_wrappers.h"
 
@@ -399,6 +400,7 @@ namespace games::iidx {
     void IIDXLocalCamera::CreateThread() {
         // Create thread
         m_drawThread = new std::thread([this]() {
+            timeutils::PreciseSleepTimer timer;
             SetThreadPriority(GetCurrentThread(), THREAD_PRIORITY_BELOW_NORMAL);
 
             double accumulator = 0.0;
@@ -412,7 +414,7 @@ namespace games::iidx {
                     accumulator -= 1.0;
                     floorFrameTimeMicroSec += 1;
                 }
-                std::this_thread::sleep_for(std::chrono::microseconds(floorFrameTimeMicroSec));
+                timer.sleep(std::chrono::microseconds(floorFrameTimeMicroSec));
             }
         });
     }

src/spice2x/games/iidx/poke.cpp

@@ -15,6 +15,7 @@
 #include "touch/touch.h"
 #include "util/libutils.h"
 #include "util/logging.h"
+#include "util/precise_timer.h"
 
 #define POKE_NATIVE_TOUCH 0
 
@@ -200,6 +201,7 @@ namespace games::iidx::poke {
         // create new thread
         THREAD_RUNNING = true;
         THREAD = new std::thread([] {
+            timeutils::PreciseSleepTimer timer;
 
             // log
             log_info("poke", "enabled");
@@ -319,7 +321,7 @@ namespace games::iidx::poke {
                 }
 
                 // slow down
-                Sleep(50);
+                timer.sleep(50);
             }
 
             return nullptr;

src/spice2x/games/io.cpp

@@ -92,6 +92,9 @@ namespace games {
         games.push_back(ddr);
         buttons.insert({ ddr, ddr::get_buttons() });
         lights.insert({ ddr, ddr::get_lights() });
+        buttons_help.insert({ ddr, ddr::get_buttons_help() });
+        analogs_help.insert({ ddr, ddr::get_analogs_help() });
+        analogs.insert({ ddr, ddr::get_analogs() });
         file_hints[ddr].push_back({"ddr.dll"});
         file_hints[ddr].push_back({"mdxja_945.dll"});
         file_hints[ddr].push_back({"arkmdxp3.dll"});

src/spice2x/games/jb/jb.cpp

@@ -19,7 +19,7 @@
 namespace games::jb {
 
     // touch stuff
-    bool TOUCH_LEGACY_BOX = false;
+    JubeatTouchAlgorithm TOUCH_ALGORITHM = Improved;
     static bool TOUCH_ENABLE = false;
     static bool TOUCH_ATTACHED = false;
     static bool IS_PORTRAIT = true;
@@ -78,7 +78,7 @@ namespace games::jb {
         // note that the IO code in device.cpp will correctly compensate for orientation, depending on the model.
         TOUCH_POINTS.clear();
         touch_get_points(TOUCH_POINTS);
-        if (TOUCH_LEGACY_BOX) {
+        if (TOUCH_ALGORITHM == Legacy) {
             auto offset = IS_PORTRAIT ? 580 : 0;
             for (auto &tp : TOUCH_POINTS) {
 
@@ -94,6 +94,17 @@ namespace games::jb {
             }
         } else {
             for (auto &tp : TOUCH_POINTS) {
+                // check window out of bounds
+                if (IS_PORTRAIT) {
+                    if (tp.x > 768 || tp.y > 1360) {
+                        continue;
+                    }
+                } else {
+                    if (tp.x > 1360 || tp.y > 768) {
+                        continue;
+                    }
+                }
+
                 int x_relative = tp.x;
                 int y_relative = tp.y;
 
@@ -110,19 +121,66 @@ namespace games::jb {
                     y_relative -= 8;
                 }
 
-                if (x_relative < 0 || y_relative < 0) {
+                int x_index = -1;
+                int x_hitbox = 0;
+                int y_index = -1;
+                int y_hitbox = 0;
+
+                // x_hitbox and y_hitbox is relative to top-left pixel of each button
+                if (x_relative >= 0) {
+                    x_index = x_relative / JB_BUTTON_HITBOX;
+                    x_hitbox = x_relative % JB_BUTTON_HITBOX;
+                }
+                if (y_relative >= 0) {
+                    y_index = y_relative / JB_BUTTON_HITBOX;
+                    y_hitbox = y_relative % JB_BUTTON_HITBOX;
+                }
+
+                // log_info("jb", "raw={}, idx={}, hitbox={}", tp.x, x_index, x_hitbox);
+
+                if (TOUCH_ALGORITHM == Improved) {
+                    if (IS_PORTRAIT) {
+                        // extend to left border
+                        if (x_relative < 0) {
+                            x_index = 0;
+                        }
+                        // right and bottom borders are covered by the hit box
+                    } else {
+                        // extend to top border
+                        if (y_relative < 0) {
+                            y_index = 0;
+                        }
+                        // extend to left border
+                        if (x_relative < 0) {
+                            x_index = 0;
+                        }
+                        // bottom border is covered by the hit box
+                        // rightmost edge - ignore them entirely
+                        if (x_index == 3 && JB_BUTTON_SIZE < x_hitbox) {
+                            continue;
+                        }
+                    }
+                }
+
+                if (x_index < 0 || y_index < 0 || x_index > 3 || y_index > 3) {
                     continue;
                 }
 
-                // x_hitbox and y_hitbox is relative to top-left pixel of each button
-                int x_index = x_relative / JB_BUTTON_HITBOX;
-                int x_hitbox = x_relative % JB_BUTTON_HITBOX;
-                int y_index = y_relative / JB_BUTTON_HITBOX;
-                int y_hitbox = y_relative % JB_BUTTON_HITBOX;
-
                 // check if the gap was touched
-                if (x_hitbox > JB_BUTTON_SIZE || y_hitbox > JB_BUTTON_SIZE) {
-                    continue;
+                if (TOUCH_ALGORITHM == AcAccurate) {
+                    // in ac-accurate mode, touching the gap is ignored
+                    if (x_hitbox > JB_BUTTON_SIZE || y_hitbox > JB_BUTTON_SIZE) {
+                        continue;
+                    }
+
+                } else if (TOUCH_ALGORITHM == Improved) {
+                    // in improved mode, touching the gap triggers the closest button
+                    if (x_index <= 2 && (JB_BUTTON_SIZE + JB_BUTTON_GAP / 2) < x_hitbox) {
+                        x_index++;
+                    }
+                    if (y_index <= 2 && (JB_BUTTON_SIZE + JB_BUTTON_GAP / 2) < y_hitbox) {
+                        y_index++;
+                    }
                 }
 
                 // set the corresponding state
@@ -167,7 +225,7 @@ namespace games::jb {
     void JBGame::pre_attach() {
         if (!cfg::CONFIGURATOR_STANDALONE) {
             const auto current_path = std::filesystem::current_path();
-            log_misc("jubeat", "current working directory: {}", current_path.string());
+            log_misc("jubeat", "current working directory: {}", current_path);
             if (current_path.parent_path() == current_path.root_path()) {
                 log_warning(
                     "jubeat",
@@ -178,7 +236,7 @@ namespace games::jb {
                     "    c:\\jubeat\\contents\\spice.exe  <- OK\n\n"
                     "To fix this, create a new directory and move ALL game files there.\n\n"
                     "Your current working directory: {}\n",
-                    current_path.string());
+                    current_path);
 
                 log_fatal(
                     "jubeat",
@@ -193,6 +251,21 @@ namespace games::jb {
         // enable touch
         TOUCH_ENABLE = true;
 
+        switch (TOUCH_ALGORITHM) {
+            case Legacy:
+                log_info("jubeat", "using 'legacy' touch targets");
+                break;
+            case Improved:
+                log_info("jubeat", "using 'improved' touch targets");
+                break;
+            case AcAccurate:
+                log_info("jubeat", "using 'ac accurate' touch targets");
+                break;
+            default:
+                log_fatal("jubeat", "unknown touch algo detected in touch_update, this is a bug");
+                break;
+        }
+
         // enable debug logging of gftools
         HMODULE gftools = libutils::try_module("gftools.dll");
         detour::inline_hook((void *) GFDbgSetReportFunc, libutils::try_proc(

src/spice2x/games/jb/jb.h

@@ -4,8 +4,14 @@
 
 namespace games::jb {
 
+    enum JubeatTouchAlgorithm {
+        Legacy,
+        Improved,
+        AcAccurate
+    };
+
     // touch stuff
-    extern bool TOUCH_LEGACY_BOX;
+    extern JubeatTouchAlgorithm TOUCH_ALGORITHM;
     extern bool TOUCH_STATE[16];
     void touch_update();
 

src/spice2x/games/mga/io.cpp

@@ -36,11 +36,11 @@ std::vector<Analog> &games::mga::get_analogs() {
     if (analogs.empty()) {
         analogs = GameAPI::Analogs::getAnalogs("Metal Gear");
 
-        GameAPI::Analogs::sortAnalogs(
-                &analogs,
-                "Joy X",
-                "Joy Y"
-        );
+        using namespace GameAPI::Analogs;
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            { "Joy X", AnalogType::LinearCentered },
+            { "Joy Y", AnalogType::LinearCentered }
+        });
     }
 
     return analogs;

src/spice2x/games/nost/io.cpp

@@ -165,37 +165,38 @@ std::vector<Analog> &games::nost::get_analogs() {
     if (analogs.empty()) {
         analogs = GameAPI::Analogs::getAnalogs("Nostalgia");
 
-        GameAPI::Analogs::sortAnalogs(
-                &analogs,
-                "Key 1",
-                "Key 2",
-                "Key 3",
-                "Key 4",
-                "Key 5",
-                "Key 6",
-                "Key 7",
-                "Key 8",
-                "Key 9",
-                "Key 10",
-                "Key 11",
-                "Key 12",
-                "Key 13",
-                "Key 14",
-                "Key 15",
-                "Key 16",
-                "Key 17",
-                "Key 18",
-                "Key 19",
-                "Key 20",
-                "Key 21",
-                "Key 22",
-                "Key 23",
-                "Key 24",
-                "Key 25",
-                "Key 26",
-                "Key 27",
-                "Key 28"
-        );
+        using namespace GameAPI::Analogs;
+
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            { "Key 1", AnalogType::LinearPositive },
+            { "Key 2", AnalogType::LinearPositive },
+            { "Key 3", AnalogType::LinearPositive },
+            { "Key 4", AnalogType::LinearPositive },
+            { "Key 5", AnalogType::LinearPositive },
+            { "Key 6", AnalogType::LinearPositive },
+            { "Key 7", AnalogType::LinearPositive },
+            { "Key 8", AnalogType::LinearPositive },
+            { "Key 9", AnalogType::LinearPositive },
+            { "Key 10", AnalogType::LinearPositive },
+            { "Key 11", AnalogType::LinearPositive },
+            { "Key 12", AnalogType::LinearPositive },
+            { "Key 13", AnalogType::LinearPositive },
+            { "Key 14", AnalogType::LinearPositive },
+            { "Key 15", AnalogType::LinearPositive },
+            { "Key 16", AnalogType::LinearPositive },
+            { "Key 17", AnalogType::LinearPositive },
+            { "Key 18", AnalogType::LinearPositive },
+            { "Key 19", AnalogType::LinearPositive },
+            { "Key 20", AnalogType::LinearPositive },
+            { "Key 21", AnalogType::LinearPositive },
+            { "Key 22", AnalogType::LinearPositive },
+            { "Key 23", AnalogType::LinearPositive },
+            { "Key 24", AnalogType::LinearPositive },
+            { "Key 25", AnalogType::LinearPositive },
+            { "Key 26", AnalogType::LinearPositive },
+            { "Key 27", AnalogType::LinearPositive },
+            { "Key 28", AnalogType::LinearPositive }
+        });
     }
     return analogs;
 }

src/spice2x/games/nost/poke.cpp

@@ -9,6 +9,7 @@
 #include "misc/eamuse.h"
 #include "touch/touch.h"
 #include "util/logging.h"
+#include "util/precise_timer.h"
 
 namespace games::nost::poke {
 
@@ -63,6 +64,7 @@ namespace games::nost::poke {
         // create new thread
         THREAD_RUNNING = true;
         THREAD = new std::thread([] {
+            timeutils::PreciseSleepTimer timer;
             const DWORD touch_id = (DWORD)(0xFFFFFFFE);
 
             const int swipe_anim_total_frames = 6;
@@ -178,7 +180,7 @@ namespace games::nost::poke {
                 }
 
                 // slow down
-                Sleep(30);
+                timer.sleep(30);
             }
 
             return nullptr;

src/spice2x/games/onpara/touchpanel.cpp

@@ -4,6 +4,7 @@
 #include "touchpanel.h"
 
 #include "util/logging.h"
+#include "util/precise_timer.h"
 #include "touch/touch.h"
 
 using namespace std::chrono_literals;
@@ -44,6 +45,9 @@ bool games::onpara::TouchPanelHandle::open(LPCWSTR lpFileName) {
 }
 
 int games::onpara::TouchPanelHandle::read(LPVOID lpBuffer, DWORD nNumberOfBytesToRead) {
+
+    static thread_local timeutils::PreciseSleepTimer timer;
+
     DWORD i;
     auto buffer = reinterpret_cast<uint8_t *>(lpBuffer);
 
@@ -67,7 +71,7 @@ int games::onpara::TouchPanelHandle::read(LPVOID lpBuffer, DWORD nNumberOfBytesT
         enqueue_packet(report);
 
         // prevent cpu bullying
-        std::this_thread::sleep_for(1ms);
+        timer.sleep(1);
     }
 
     // copy from output queue
@@ -133,4 +137,4 @@ bool games::onpara::TouchPanelHandle::close() {
     log_info("touchpanel", "Closed COM1 (Touch Panel)");
 
     return true;
-}
+}

src/spice2x/games/otoca/otoca.cpp

@@ -4,6 +4,7 @@
 #include "util/detour.h"
 #include "util/libutils.h"
 #include "util/logging.h"
+#include "util/sigscan.h"
 
 #define OTOCA_DEBUG_VERBOSE 0
 #if OTOCA_DEBUG_VERBOSE
@@ -116,6 +117,73 @@ namespace games::otoca {
         return ret;
     }
 
+    static void patch_holo_print_hang() {
+
+        // arkkep.dll's print method sets the printer busy (0x66 / Printing_Busy),
+        // then bails out with -1 WITHOUT printing when its holo parameter is
+        // nonzero (the card path passes 0 and works). The completion callback that
+        // clears busy never fires, so the page's poll loop spins forever -> hang.
+        // The image band is already built, so flip the je that skips the bail-out
+        // into a jmp (74 -> EB) to route holo through the print path like card.
+        //
+        //   je   issue_print     ; <- patched to jmp
+        //   or   eax, -1         ; holo: return without printing
+        //   ret  10h
+        auto arkkep = libutils::try_module("arkkep.dll");
+        if (arkkep == nullptr) {
+            log_warning("otoca", "arkkep.dll not loaded; skipping holo print hang fix");
+            return;
+        }
+
+        auto result = replace_pattern(
+            arkkep,
+            "C7465866000000895DC0C74710080700007416",
+            "C7465866000000895DC0C7471008070000EB16",
+            0, 0);
+        if (result) {
+            log_info("otoca", "patched hologram print hang in arkkep.dll");
+        } else {
+            log_warning("otoca", "could not patch hologram print hang (incompatible arkkep.dll?)");
+        }
+    }
+
+    static void patch_holo_print_search() {
+
+        // before printing, arkkep walks the page's printer list looking for one
+        // whose capability flag (byte at entry+120h) matches the job type
+        // (0 = card, 1 = holo). Our emulated printer is only card-capable, so a
+        // holo job finds no match and aborts without printing ("not found holo
+        // printer"). The emulated printer serves both media, so force the compare
+        // to always match by turning `cmp cl,[edi+120h]` into `cmp cl,cl` + NOPs,
+        // which always sets ZF. Two such compares exist, so patch both.
+        //
+        //   cmp  cl, [edi+120h]  ; <- patched to `cmp cl,cl` (3A 8F.. -> 3A C9..)
+        //   je   printer_found
+        auto arkkep = libutils::try_module("arkkep.dll");
+        if (arkkep == nullptr) {
+            log_warning("otoca", "arkkep.dll not loaded; skipping holo print search fix");
+            return;
+        }
+
+        int patched = 0;
+        while (patched < 2) {
+            auto result = replace_pattern(
+                arkkep,
+                "3A8F20010000",
+                "3AC990909090",
+                0, 0);
+            if (!result) {
+                break;
+            }
+            patched++;
+        }
+        if (patched > 0) {
+            log_info("otoca", "patched hologram printer search in arkkep.dll ({} site(s))", patched);
+        } else {
+            log_warning("otoca", "could not patch hologram printer search (incompatible arkkep.dll?)");
+        }
+    }
+
     void OtocaGame::attach() {
         Game::attach();
 
@@ -129,6 +197,12 @@ namespace games::otoca {
         p4io_hook();
         games::shared::printer_attach();
 
+        // fix hologram print hard hang in arkkep.dll
+        patch_holo_print_hang();
+
+        // make hologram jobs find the emulated printer (else they abort unprinted)
+        patch_holo_print_search();
+
         if (BYPASS_CAMERA) {
             libutils::try_library("libcamera.dll");
             const auto libcamera = "libcamera.dll";

src/spice2x/games/pc/io.cpp

@@ -51,11 +51,12 @@ std::vector<Analog> &games::pc::get_analogs() {
     if (analogs.empty()) {
         analogs = GameAPI::Analogs::getAnalogs("Polaris Chord");
 
-        GameAPI::Analogs::sortAnalogs(
-                &analogs,
-                "Fader-L",
-                "Fader-R"
-        );
+        using GameAPI::Analogs::AnalogType;
+
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            { "Fader-L", AnalogType::LinearCentered },
+            { "Fader-R", AnalogType::LinearCentered }
+        });
     }
 
     return analogs;

src/spice2x/games/popn/popn.cpp

@@ -261,7 +261,7 @@ namespace games::popn {
             } else if (requestPacket->type == DISPLAYCONFIG_DEVICE_INFO_GET_SOURCE_NAME) {
                 const auto source = reinterpret_cast<DISPLAYCONFIG_SOURCE_DEVICE_NAME*>(requestPacket);
                 // value must match WrappedIDirect3D9::GetAdapterIdentifier
-                wcscpy(source->viewGdiDeviceName, L"\\\\.\\DISPLAY_SPICE_FAKE");
+                wcscpy(source->viewGdiDeviceName, L"\\\\.\\DISPLAY_SPICE_FAKE_1");
             } else {
                 log_fatal(
                     "popn",

src/spice2x/games/rb/rb.cpp

@@ -4,29 +4,34 @@
 #include "hooks/devicehook.h"
 #include "hooks/audio/backends/dsound/dsound_backend.h"
 #include "util/detour.h"
+#include "util/precise_timer.h"
+#include "util/logging.h"
+#include "cfg/configurator.h"
 
 #include "touch.h"
 
+namespace games::rb {
+    uint16_t TOUCH_POLL_RATE = 0;
+}
+
 static decltype(SleepEx) *SleepEx_orig;
 
-static DWORD WINAPI SleepEx_hook(DWORD dwMilliseconds, BOOL bAltertable) {
+static DWORD WINAPI SleepEx_hook(DWORD dwMilliseconds, BOOL bAlertable) {
 
-    /*
-     * Increase touch poll from ~110 FPS to ~500 FPS
-     */
+    // increase touch poll rate
     if (dwMilliseconds == 8) {
-        static bool initialized = false;
-        if (!initialized) {
-            initialized = true;
+        dwMilliseconds = 1000 / games::rb::TOUCH_POLL_RATE;
+    }
 
-            // if we only sleep for 1ms we also don't need the high priority RB sets
-            SetThreadPriority(GetCurrentThread(), THREAD_PRIORITY_BELOW_NORMAL);
-        }
-        dwMilliseconds = 1;
+    // most calls from rb are actually non-alertable
+    if (!bAlertable) {
+        static thread_local timeutils::PreciseSleepTimer timer;
+        timer.sleep(dwMilliseconds);
+        return 0;
     }
 
     // call original
-    return SleepEx_orig(dwMilliseconds, bAltertable);
+    return SleepEx_orig(dwMilliseconds, bAlertable);
 }
 
 games::rb::RBGame::RBGame() : Game("Reflec Beat") {
@@ -35,6 +40,10 @@ games::rb::RBGame::RBGame() : Game("Reflec Beat") {
 void games::rb::RBGame::attach() {
     Game::attach();
 
+    if (1000 < games::rb::TOUCH_POLL_RATE) {
+        log_fatal("rb", "-rbtouchsize is set too high; cannot exceed 1000");
+    }
+
     // init stuff
     devicehook_init();
 
@@ -47,7 +56,12 @@ void games::rb::RBGame::attach() {
     }
 
     // hooks
-    SleepEx_orig = detour::iat_try("SleepEx", SleepEx_hook, avs::game::DLL_INSTANCE);
+    if (0 < TOUCH_POLL_RATE) {
+        log_info("rb", "force increasing touch poll rate by hooking SleepEx ({}Hz)", TOUCH_POLL_RATE);
+        SleepEx_orig = detour::iat_try("SleepEx", SleepEx_hook, avs::game::DLL_INSTANCE);
+    } else {
+        log_info("rb", "not changing touch poll rate (120Hz by default)");
+    }
 }
 
 void games::rb::RBGame::detach() {

src/spice2x/games/rb/rb.h

@@ -7,6 +7,8 @@
 namespace games::rb {
 
     extern uint16_t TOUCH_SCALING;
+    extern uint8_t TOUCH_SIZE;
+    extern uint16_t TOUCH_POLL_RATE;
 
     class RBGame : public games::Game {
     public:

src/spice2x/games/rb/touch.cpp

@@ -13,6 +13,7 @@ static std::string WINDOW_TITLE = "REFLEC BEAT";
 
 namespace games::rb {
     uint16_t TOUCH_SCALING = 1000;
+    uint8_t TOUCH_SIZE = 1;
 }
 
 games::rb::ReflecBeatTouchDeviceHandle::ReflecBeatTouchDeviceHandle(bool log_fps) {
@@ -160,16 +161,20 @@ int games::rb::ReflecBeatTouchDeviceHandle::read(LPVOID lpBuffer, DWORD nNumberO
         const auto point_x = (int) ((x - window_width / 2.0) * 48.0 / 54.0 + window_width / 2.0);
         const auto point_y = (int) (y - window_height / 76);
 
-        // insert 9 times with offset to double the precision
+        // center
         grid_insert(data, point_x, point_y);
-        grid_insert(data, point_x - offset_x, point_y);
-        grid_insert(data, point_x - offset_x, point_y - offset_y);
-        grid_insert(data, point_x - offset_x, point_y + offset_y);
-        grid_insert(data, point_x + offset_x, point_y);
-        grid_insert(data, point_x + offset_x, point_y + offset_y);
-        grid_insert(data, point_x + offset_x, point_y - offset_y);
-        grid_insert(data, point_x, point_y - offset_y);
-        grid_insert(data, point_x, point_y + offset_y);
+
+        // surrounding points
+        if (games::rb::TOUCH_SIZE == 3) {
+            grid_insert(data, point_x - offset_x, point_y);            // west
+            grid_insert(data, point_x - offset_x, point_y - offset_y); // northwest
+            grid_insert(data, point_x - offset_x, point_y + offset_y); // southwest
+            grid_insert(data, point_x + offset_x, point_y);            // east
+            grid_insert(data, point_x + offset_x, point_y + offset_y); // southeast
+            grid_insert(data, point_x + offset_x, point_y - offset_y); // northeast
+            grid_insert(data, point_x, point_y - offset_y);            // north
+            grid_insert(data, point_x, point_y + offset_y);            // south
+        }
     }
 
     // copy data to buffer

src/spice2x/games/rf3d/io.cpp

@@ -35,12 +35,13 @@ std::vector<Analog> &games::rf3d::get_analogs() {
     if (analogs.empty()) {
         analogs = GameAPI::Analogs::getAnalogs("Road Fighters 3D");
 
-        GameAPI::Analogs::sortAnalogs(
-                &analogs,
-                "Wheel",
-                "Accelerate",
-                "Brake"
-        );
+        using GameAPI::Analogs::AnalogType;
+
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            { "Wheel", AnalogType::LinearCentered },
+            { "Accelerate", AnalogType::LinearPositive },
+            { "Brake", AnalogType::LinearPositive }
+        });
     }
 
     return analogs;

src/spice2x/games/sc/io.cpp

@@ -31,13 +31,14 @@ std::vector<Analog> &games::sc::get_analogs() {
     if (analogs.empty()) {
         analogs = GameAPI::Analogs::getAnalogs("Steel Chronicle");
 
-        GameAPI::Analogs::sortAnalogs(
-                &analogs,
-                "Left Stick X",
-                "Left Stick Y",
-                "Right Stick X",
-                "Right Stick Y"
-        );
+        using GameAPI::Analogs::AnalogType;
+
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            { "Left Stick X", AnalogType::LinearCentered },
+            { "Left Stick Y", AnalogType::LinearCentered },
+            { "Right Stick X", AnalogType::LinearCentered },
+            { "Right Stick Y", AnalogType::LinearCentered }
+        });
     }
 
     return analogs;

src/spice2x/games/shared/printer.cpp

@@ -7,6 +7,7 @@
 #include "hooks/sleephook.h"
 #include "hooks/libraryhook.h"
 #include "launcher/launcher.h"
+#include "overlay/notifications.h"
 #include "util/detour.h"
 #include "util/fileutils.h"
 #include "util/libutils.h"
@@ -397,8 +398,14 @@ namespace games::shared {
                 // logging
                 if (success) {
                     log_info("printer", "printer emulation has written an image to {}", image_path);
+                    overlay::notifications::add(
+                            overlay::notifications::Severity::Success,
+                            fmt::format("Printer: saved {}", fileutils::basename(image_path)));
                 } else {
                     log_warning("printer", "printer emulation failed to write image to {}", image_path);
+                    overlay::notifications::add(
+                            overlay::notifications::Severity::Error,
+                            fmt::format("Printer: failed to write {}", fileutils::basename(image_path)));
                 }
             }
         }

src/spice2x/games/silentscope/io.cpp

@@ -31,11 +31,12 @@ std::vector<Analog> &games::silentscope::get_analogs() {
     if (analogs.empty()) {
         analogs = GameAPI::Analogs::getAnalogs("Silent Scope: Bone Eater");
 
-        GameAPI::Analogs::sortAnalogs(
-            &analogs,
-            "Gun X",
-            "Gun Y"
-        );
+        using GameAPI::Analogs::AnalogType;
+
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            { "Gun X", AnalogType::LinearCentered },
+            { "Gun Y", AnalogType::LinearCentered }
+        });
     }
 
     return analogs;

src/spice2x/games/we/io.cpp

@@ -50,13 +50,14 @@ std::vector<Analog> &games::we::get_analogs() {
     if (analogs.empty()) {
         analogs = GameAPI::Analogs::getAnalogs("Winning Eleven");
 
-        GameAPI::Analogs::sortAnalogs(
-                &analogs,
-                "Pad Stick Left X",
-                "Pad Stick Left Y",
-                "Pad Stick Right X",
-                "Pad Stick Right Y"
-        );
+        using GameAPI::Analogs::AnalogType;
+
+        GameAPI::Analogs::sortAnalogsWithType(&analogs, {
+            { "Pad Stick Left X", AnalogType::LinearCentered },
+            { "Pad Stick Left Y", AnalogType::LinearCentered },
+            { "Pad Stick Right X", AnalogType::LinearCentered },
+            { "Pad Stick Right Y", AnalogType::LinearCentered }
+        });
     }
 
     return analogs;

src/spice2x/hooks/audio/asio_driver_scan.cpp

@@ -0,0 +1,74 @@
+#include "asio_driver_scan.h"
+
+#include <algorithm>
+
+#include <windows.h>
+
+#include "util/utils.h"
+
+namespace hooks::audio {
+
+    static constexpr char ASIO_REG_PATH[] = "software\\asio";
+    static constexpr char ASIO_REG_DESC[] = "description";
+
+    // enumerate a single registry view, appending to entries while merging
+    // duplicates discovered in another view. Drivers are matched by name (not
+    // CLSID): the game's ASIO loader selects drivers by name, and some vendors
+    // register the same CLSID under different 32-bit/64-bit names (e.g. "XONAR
+    // SOUND CARD" vs "XONAR SOUND CARD(64)"), which are distinct user choices.
+    static void scan_view(
+            REGSAM wow64_flag,
+            bool is_64bit,
+            std::vector<AsioDriverScanEntry> &entries) {
+
+        HKEY hkEnum = nullptr;
+        if (RegOpenKeyExA(HKEY_LOCAL_MACHINE, ASIO_REG_PATH, 0,
+                KEY_READ | wow64_flag, &hkEnum) != ERROR_SUCCESS) {
+            return;
+        }
+
+        char key_name[256];
+        for (DWORD index = 0;
+             RegEnumKeyA(hkEnum, index, key_name, sizeof(key_name)) == ERROR_SUCCESS;
+             index++) {
+
+            // read description (display name), fall back to the key name
+            char desc[256] = { 0 };
+            DWORD size = sizeof(desc);
+            std::string name = key_name;
+            if (RegGetValueA(hkEnum,
+                             key_name,
+                             ASIO_REG_DESC,
+                             RRF_RT_REG_SZ | wow64_flag,
+                             nullptr,
+                             desc,
+                             &size) == ERROR_SUCCESS && desc[0]) {
+                name = desc;
+            }
+
+            // merge with an existing entry from the other view (match by name)
+            const std::string name_lower = strtolower(name);
+            auto it = std::find_if(entries.begin(), entries.end(), [&](const auto &e) {
+                return strtolower(e.name) == name_lower;
+            });
+            if (it == entries.end()) {
+                entries.push_back({ name });
+                it = entries.end() - 1;
+            }
+            it->found_32bit |= !is_64bit;
+            it->found_64bit |= is_64bit;
+        }
+
+        RegCloseKey(hkEnum);
+    }
+
+    std::vector<AsioDriverScanEntry> scan_asio_drivers() {
+        std::vector<AsioDriverScanEntry> entries;
+
+        // 64-bit view first so it wins ordering when present in both
+        scan_view(KEY_WOW64_64KEY, true, entries);
+        scan_view(KEY_WOW64_32KEY, false, entries);
+
+        return entries;
+    }
+}

src/spice2x/hooks/audio/asio_driver_scan.h

@@ -0,0 +1,15 @@
+#pragma once
+
+#include <string>
+#include <vector>
+
+namespace hooks::audio {
+
+    struct AsioDriverScanEntry {
+        std::string name;
+        bool found_32bit = false;
+        bool found_64bit = false;
+    };
+
+    std::vector<AsioDriverScanEntry> scan_asio_drivers();
+}

src/spice2x/hooks/audio/asio_proxy.cpp

@@ -0,0 +1,813 @@
+#include "asio_proxy.h"
+
+#include <algorithm>
+#include <cmath>
+#include <cstdint>
+#include <cstring>
+#include <mutex>
+#include <vector>
+
+#include "external/asio/asiolist.h"
+#include "hooks/audio/audio.h"
+#include "util/logging.h"
+#include "util/utils.h"
+
+namespace {
+
+    // readable name for an ASIO sample type (e.g. "ASIOSTInt32LSB"), falling back to the
+    // numeric value for unknown types
+    const char *asio_sample_type_name(AsioSampleType type) {
+        switch (type) {
+            case ASIOSTInt16MSB: return "ASIOSTInt16MSB";
+            case ASIOSTInt24MSB: return "ASIOSTInt24MSB";
+            case ASIOSTInt32MSB: return "ASIOSTInt32MSB";
+            case ASIOSTFloat32MSB: return "ASIOSTFloat32MSB";
+            case ASIOSTFloat64MSB: return "ASIOSTFloat64MSB";
+            case ASIOSTInt32MSB16: return "ASIOSTInt32MSB16";
+            case ASIOSTInt32MSB18: return "ASIOSTInt32MSB18";
+            case ASIOSTInt32MSB20: return "ASIOSTInt32MSB20";
+            case ASIOSTInt32MSB24: return "ASIOSTInt32MSB24";
+            case ASIOSTInt16LSB: return "ASIOSTInt16LSB";
+            case ASIOSTInt24LSB: return "ASIOSTInt24LSB";
+            case ASIOSTInt32LSB: return "ASIOSTInt32LSB";
+            case ASIOSTFloat32LSB: return "ASIOSTFloat32LSB";
+            case ASIOSTFloat64LSB: return "ASIOSTFloat64LSB";
+            case ASIOSTInt32LSB16: return "ASIOSTInt32LSB16";
+            case ASIOSTInt32LSB18: return "ASIOSTInt32LSB18";
+            case ASIOSTInt32LSB20: return "ASIOSTInt32LSB20";
+            case ASIOSTInt32LSB24: return "ASIOSTInt32LSB24";
+            default: return "unknown";
+        }
+    }
+
+    // duration in milliseconds of a buffer of the given frame count at a sample rate, or a
+    // negative sentinel when the frame count or sample rate is unusable
+    double frames_to_ms(long frames, AsioSampleRate sample_rate) {
+        if (frames < 0 || sample_rate <= 0.0) {
+            return -1.0;
+        }
+        return (frames * 1000.0) / sample_rate;
+    }
+
+    // scales one planar ASIO output buffer (frames samples of the given type) by gain in
+    // place, clamping integer formats so a boost saturates instead of wrapping. unsupported
+    // formats are left untouched. runs on the driver's realtime thread, so no allocation,
+    // locking or logging here
+    void apply_gain_planar(void *buffer, long frames, AsioSampleType type, float gain) {
+        if (buffer == nullptr || frames <= 0) {
+            return;
+        }
+        switch (type) {
+            case ASIOSTFloat32LSB: {
+                auto p = static_cast<float *>(buffer);
+                for (long i = 0; i < frames; i++) {
+                    p[i] = std::clamp(p[i] * gain, -1.0f, 1.0f);
+                }
+                break;
+            }
+            case ASIOSTFloat64LSB: {
+                auto p = static_cast<double *>(buffer);
+                for (long i = 0; i < frames; i++) {
+                    p[i] = std::clamp(p[i] * static_cast<double>(gain), -1.0, 1.0);
+                }
+                break;
+            }
+            case ASIOSTInt16LSB: {
+                auto p = static_cast<int16_t *>(buffer);
+                for (long i = 0; i < frames; i++) {
+                    p[i] = static_cast<int16_t>(
+                        std::clamp(std::lround(p[i] * gain), -32768L, 32767L));
+                }
+                break;
+            }
+            case ASIOSTInt24LSB: {
+                // packed 24-bit little-endian, 3 bytes per sample
+                auto bytes = static_cast<uint8_t *>(buffer);
+                for (long i = 0; i < frames; i++) {
+                    uint8_t *s = bytes + i * 3;
+                    int32_t v = s[0] | (s[1] << 8) | (s[2] << 16);
+                    if (v & 0x800000) {
+                        v |= ~0xFFFFFF; // sign extend
+                    }
+                    int64_t scaled = std::clamp<int64_t>(
+                        std::llround(static_cast<double>(v) * gain), -8388608, 8388607);
+                    s[0] = scaled & 0xFF;
+                    s[1] = (scaled >> 8) & 0xFF;
+                    s[2] = (scaled >> 16) & 0xFF;
+                }
+                break;
+            }
+            case ASIOSTInt32LSB: {
+                auto p = static_cast<int32_t *>(buffer);
+                for (long i = 0; i < frames; i++) {
+                    p[i] = static_cast<int32_t>(std::clamp<int64_t>(
+                        std::llround(static_cast<double>(p[i]) * gain), INT32_MIN, INT32_MAX));
+                }
+                break;
+            }
+            default:
+                // unsupported format (MSB, aligned 32-bit, DSD): leave untouched
+                break;
+        }
+    }
+
+    // live wrappers by CLSID. ASIO drivers are single-instance, so a host that creates a
+    // new instance without releasing the old one has leaked it; we track this to tear the
+    // stale one down
+    std::mutex g_wrappers_mutex;
+    std::vector<std::pair<CLSID, WrappedAsio *>> g_wrappers;
+
+    // register a wrapper as the live instance for its CLSID, returning any stale wrapper it
+    // replaces so the caller can tear it down outside the lock
+    WrappedAsio *register_wrapper(REFCLSID clsid, WrappedAsio *wrapper) {
+        std::lock_guard lock(g_wrappers_mutex);
+        for (auto &entry : g_wrappers) {
+            if (IsEqualCLSID(entry.first, clsid)) {
+                WrappedAsio *stale = entry.second;
+                entry.second = wrapper;
+                return stale;
+            }
+        }
+
+        g_wrappers.emplace_back(clsid, wrapper);
+        return nullptr;
+    }
+
+    void unregister_wrapper(WrappedAsio *wrapper) {
+        std::lock_guard lock(g_wrappers_mutex);
+        for (auto it = g_wrappers.begin(); it != g_wrappers.end(); ++it) {
+            if (it->second == wrapper) {
+                g_wrappers.erase(it);
+                return;
+            }
+        }
+    }
+
+    // ASIO drivers registered on this system (CLSID + registry name), scanned once
+    const std::vector<std::pair<CLSID, std::string>> &registered_asio_drivers() {
+        static const std::vector<std::pair<CLSID, std::string>> drivers = [] {
+            std::vector<std::pair<CLSID, std::string>> result;
+            AsioDriverList driver_list;
+            for (const auto &driver : driver_list.driver_list) {
+                result.emplace_back(driver.clsid, driver.name);
+                log_info(
+                    "audio::wrappedasio",
+                    "registered ASIO driver: name='{}', clsid={}",
+                    driver.name,
+                    guid2s(driver.clsid));
+            }
+
+            log_info("audio::wrappedasio", "discovered {} registered ASIO driver(s)", result.size());
+            return result;
+        }();
+
+        return drivers;
+    }
+
+    std::string registered_asio_name(REFCLSID clsid) {
+        for (const auto &driver : registered_asio_drivers()) {
+            if (IsEqualCLSID(driver.first, clsid)) {
+                return driver.second;
+            }
+        }
+
+        return guid2s(clsid);
+    }
+}
+
+namespace hooks::audio::asio {
+
+    bool is_asio_creation(REFCLSID rclsid, REFIID riid) {
+
+        // ASIO hosts request the driver using its own CLSID as the interface id
+        if (!IsEqualGUID(rclsid, riid)) {
+            return false;
+        }
+
+        for (const auto &driver : registered_asio_drivers()) {
+            if (IsEqualCLSID(driver.first, rclsid)) {
+                return true;
+            }
+        }
+
+        return false;
+    }
+}
+
+#pragma region IUnknown
+bool WrappedAsio::FORCE_TWO_CHANNELS = false;
+std::atomic<WrappedAsio *> WrappedAsio::volume_active_instance {nullptr};
+
+WrappedAsio::~WrappedAsio() {
+    unregister_wrapper(this);
+
+    this->detach_volume();
+
+    // our refcount is decoupled from the host's (see AddRef/Release), so pReal's count is
+    // exactly one here and this releases/unloads it deterministically
+    this->pReal->Release();
+
+    log_info("audio::wrappedasio", "destroying wrapped ASIO driver, clsid={}", guid2s(this->clsid));
+}
+
+HRESULT STDMETHODCALLTYPE WrappedAsio::QueryInterface(REFIID riid, void **ppv) {
+    if (ppv == nullptr) {
+        return E_POINTER;
+    }
+
+    // ASIO hosts query for the driver using its own CLSID as the IID
+    if (IsEqualIID(riid, IID_IUnknown) || IsEqualIID(riid, this->clsid)) {
+        this->AddRef();
+        *ppv = static_cast<IAsio *>(this);
+
+        return S_OK;
+    }
+
+    // the host is asking for some other interface; forward to the real driver. a failure
+    // here is a common reason a host discards a driver and retries
+    const HRESULT ret = this->pReal->QueryInterface(riid, ppv);
+    if (SUCCEEDED(ret)) {
+        log_info("audio::wrappedasio", "QueryInterface({}) -> forwarded to real driver", guid2s(riid));
+    } else {
+        log_info(
+            "audio::wrappedasio",
+            "QueryInterface({}) -> not supported by driver, hr={:#x}",
+            guid2s(riid),
+            static_cast<unsigned long>(ret));
+    }
+
+    return ret;
+}
+
+ULONG STDMETHODCALLTYPE WrappedAsio::AddRef() {
+    // decoupled from the real driver: we count host references on the wrapper and hold a
+    // single reference on pReal for our lifetime. this neutralizes a host bug (iidx32+)
+    // that takes a duplicate AddRef with no matching Release, which would leak the driver
+    return ++this->ref_count;
+}
+
+ULONG STDMETHODCALLTYPE WrappedAsio::Release() {
+    const ULONG refs = --this->ref_count;
+    if (refs == 0) {
+        delete this;
+    }
+
+    return refs;
+}
+#pragma endregion
+
+#pragma region IAsio
+AsioBool __thiscall WrappedAsio::init(void *sys_handle) {
+    const AsioBool result = this->pReal->init(sys_handle);
+    if (result == AsioTrue) {
+        log_info(
+            "audio::wrappedasio",
+            "init succeeded for '{}' (driver version {})",
+            this->driver_name,
+            this->pReal->get_driver_version());
+    } else {
+        char message[128] = {};
+        this->pReal->get_error_message(message);
+        log_warning("audio::wrappedasio", "init failed: {}", message);
+    }
+
+    return result;
+}
+
+void __thiscall WrappedAsio::get_driver_name(char *name) {
+    this->pReal->get_driver_name(name);
+}
+
+long __thiscall WrappedAsio::get_driver_version() {
+    return this->pReal->get_driver_version();
+}
+
+void __thiscall WrappedAsio::get_error_message(char *string) {
+    this->pReal->get_error_message(string);
+}
+
+AsioError __thiscall WrappedAsio::start() {
+    const AsioError result = this->pReal->start();
+    if (result == ASE_OK) {
+        log_info(
+            "audio::wrappedasio",
+            "start succeeded, ASIO stream is now running on '{}'",
+            this->driver_name);
+    } else {
+        log_warning("audio::wrappedasio", "start failed, err={}", static_cast<long>(result));
+    }
+
+    return result;
+}
+
+AsioError __thiscall WrappedAsio::stop() {
+    const AsioError result = this->pReal->stop();
+    if (result == ASE_OK) {
+        log_info("audio::wrappedasio", "stop succeeded, ASIO stream on '{}' halted", this->driver_name);
+    } else {
+        log_warning("audio::wrappedasio", "stop failed, err={}", static_cast<long>(result));
+    }
+
+    return result;
+}
+
+AsioError __thiscall WrappedAsio::get_channels(long *num_input_channels, long *num_output_channels) {
+    const AsioError result = this->pReal->get_channels(num_input_channels, num_output_channels);
+    if (result != ASE_OK) {
+        log_warning("audio::wrappedasio", "get_channels failed, err={}", static_cast<long>(result));
+        return result;
+    }
+
+    if (FORCE_TWO_CHANNELS
+        && num_output_channels != nullptr
+        && *num_output_channels < FORCED_OUTPUT_CHANNELS)
+    {
+        // the device has fewer outputs than the game hardcodes; report the count it
+        // expects so it proceeds to create_buffers, where we forward only the real
+        // front pair and discard the rest
+        log_info(
+            "audio::wrappedasio",
+            "reporting output channel count as {} (device has {}) for forced two-channel",
+            FORCED_OUTPUT_CHANNELS,
+            *num_output_channels);
+        *num_output_channels = FORCED_OUTPUT_CHANNELS;
+    }
+
+    log_info(
+        "audio::wrappedasio",
+        "get_channels -> in={}, out={}",
+        num_input_channels ? *num_input_channels : -1,
+        num_output_channels ? *num_output_channels : -1);
+
+    return result;
+}
+
+AsioError __thiscall WrappedAsio::get_latencies(long *input_latency, long *output_latency) {
+    const AsioError result = this->pReal->get_latencies(input_latency, output_latency);
+    if (result == ASE_OK) {
+        // include millisecond equivalents alongside the frame counts for readability
+        AsioSampleRate sample_rate = 0.0;
+        this->pReal->get_sample_rate(&sample_rate);
+        const long in_frames = input_latency ? *input_latency : -1;
+        const long out_frames = output_latency ? *output_latency : -1;
+        log_info(
+            "audio::wrappedasio",
+            "get_latencies -> in={} frames ({:.2f} ms), out={} frames ({:.2f} ms)",
+            in_frames,
+            frames_to_ms(in_frames, sample_rate),
+            out_frames,
+            frames_to_ms(out_frames, sample_rate));
+    } else {
+        log_warning("audio::wrappedasio", "get_latencies failed, err={}", static_cast<long>(result));
+    }
+
+    return result;
+}
+
+AsioError __thiscall WrappedAsio::get_buffer_size(
+    long *min_size,
+    long *max_size,
+    long *preferred_size,
+    long *granularity)
+{
+    const AsioError result = this->pReal->get_buffer_size(min_size, max_size, preferred_size, granularity);
+    if (result != ASE_OK) {
+        log_warning("audio::wrappedasio", "get_buffer_size failed, err={}", static_cast<long>(result));
+        return result;
+    }
+
+    // include millisecond equivalents alongside the frame counts for readability
+    AsioSampleRate sample_rate = 0.0;
+    this->pReal->get_sample_rate(&sample_rate);
+    const long min_frames = min_size ? *min_size : -1;
+    const long max_frames = max_size ? *max_size : -1;
+    const long preferred_frames = preferred_size ? *preferred_size : -1;
+    log_info(
+        "audio::wrappedasio",
+        "get_buffer_size -> min={} frames ({:.2f} ms), max={} frames ({:.2f} ms), "
+        "preferred={} frames ({:.2f} ms), granularity={}",
+        min_frames,
+        frames_to_ms(min_frames, sample_rate),
+        max_frames,
+        frames_to_ms(max_frames, sample_rate),
+        preferred_frames,
+        frames_to_ms(preferred_frames, sample_rate),
+        granularity ? *granularity : -1);
+
+    return result;
+}
+
+AsioError __thiscall WrappedAsio::can_sample_rate(AsioSampleRate sample_rate) {
+    const AsioError result = this->pReal->can_sample_rate(sample_rate);
+    if (result == ASE_OK) {
+        log_misc("audio::wrappedasio", "can_sample_rate({} Hz) -> supported", sample_rate);
+    } else {
+        log_misc(
+            "audio::wrappedasio",
+            "can_sample_rate({} Hz) -> not supported, err={}",
+            sample_rate,
+            static_cast<long>(result));
+    }
+
+    return result;
+}
+
+AsioError __thiscall WrappedAsio::get_sample_rate(AsioSampleRate *sample_rate) {
+    const AsioError result = this->pReal->get_sample_rate(sample_rate);
+    if (result == ASE_OK) {
+        log_misc("audio::wrappedasio", "get_sample_rate -> {} Hz", sample_rate ? *sample_rate : 0.0);
+    } else {
+        log_warning("audio::wrappedasio", "get_sample_rate failed, err={}", static_cast<long>(result));
+    }
+
+    return result;
+}
+
+AsioError __thiscall WrappedAsio::set_sample_rate(AsioSampleRate sample_rate) {
+    const AsioError result = this->pReal->set_sample_rate(sample_rate);
+    if (result == ASE_OK) {
+        log_info("audio::wrappedasio", "set_sample_rate({} Hz) succeeded", sample_rate);
+    } else {
+        log_warning(
+            "audio::wrappedasio",
+            "set_sample_rate({} Hz) failed, err={}",
+            sample_rate,
+            static_cast<long>(result));
+    }
+
+    return result;
+}
+
+AsioError __thiscall WrappedAsio::get_clock_sources(ASIOClockSource *clocks, long *num_sources) {
+    return this->pReal->get_clock_sources(clocks, num_sources);
+}
+
+AsioError __thiscall WrappedAsio::set_clock_source(long reference) {
+    return this->pReal->set_clock_source(reference);
+}
+
+AsioError __thiscall WrappedAsio::get_sample_position(ASIOSamples *s_pos, ASIOTimeStamp *t_stamp) {
+    return this->pReal->get_sample_position(s_pos, t_stamp);
+}
+
+AsioError __thiscall WrappedAsio::get_channel_info(AsioChannelInfo *info) {
+    // forced two-channel: the game probes all output channels it thinks exist, but the
+    // device only has the real front pair. fabricate a plausible entry for the channels
+    // beyond the device without touching the real driver - they are discarded in
+    // create_buffers anyway
+    long real_in = 0, real_out = 0;
+    if (FORCE_TWO_CHANNELS
+        && info != nullptr
+        && info->is_input == AsioFalse
+        && this->pReal->get_channels(&real_in, &real_out) == ASE_OK
+        && info->channel >= real_out)
+    {
+        const long channel = info->channel;
+        info->is_active = AsioTrue;
+        info->channel_group = 0;
+        info->type = ASIOSTInt32LSB;
+        snprintf(info->name, sizeof(info->name), "Fake ASIO OUT %ld", channel);
+        log_info(
+            "audio::wrappedasio",
+            "get_channel_info(channel={}, dir=output) -> fake channel, type={} ({})",
+            channel,
+            asio_sample_type_name(info->type),
+            static_cast<long>(info->type));
+
+        return ASE_OK;
+    }
+
+    const AsioError result = this->pReal->get_channel_info(info);
+    if (result == ASE_OK && info != nullptr) {
+        log_info(
+            "audio::wrappedasio",
+            "get_channel_info(channel={}, dir={}) -> active={}, group={}, type={} ({}), name='{}'",
+            info->channel,
+            info->is_input == AsioTrue ? "input" : "output",
+            info->is_active == AsioTrue,
+            info->channel_group,
+            asio_sample_type_name(info->type),
+            static_cast<long>(info->type),
+            info->name);
+    } else if (result != ASE_OK) {
+        log_warning("audio::wrappedasio", "get_channel_info failed, err={}", static_cast<long>(result));
+    }
+
+    return result;
+}
+
+AsioCallbacks *WrappedAsio::install_volume_callbacks(AsioCallbacks *game_callbacks) {
+    const float gain = hooks::audio::VOLUME_BOOST;
+
+    // start from a clean slate; a previous buffer set may have left state behind
+    this->volume_channels.clear();
+    this->volume_active = false;
+
+    // no boost configured (or no callbacks to wrap): pass the game's callbacks straight
+    // through and do zero realtime work, exactly as before
+    if (gain == 1.0f || game_callbacks == nullptr) {
+        return game_callbacks;
+    }
+
+    this->volume_active = true;
+    this->volume_gain = gain;
+    this->volume_game_callbacks = *game_callbacks;
+
+    // wrap only the buffer-switch callbacks, where the audio data lives and we apply the
+    // gain. the other two carry no data we touch, so forward the game's own pointers
+    // unchanged - the driver expects them non-null and the game already owns their context
+    this->volume_proxy_callbacks = {};
+    this->volume_proxy_callbacks.buffer_switch = &WrappedAsio::volume_buffer_switch;
+    this->volume_proxy_callbacks.sample_rate_did_change = game_callbacks->sample_rate_did_change;
+    this->volume_proxy_callbacks.asio_message = game_callbacks->asio_message;
+    this->volume_proxy_callbacks.buffer_switch_time_info =
+        game_callbacks->buffer_switch_time_info ? &WrappedAsio::volume_buffer_switch_time_info : nullptr;
+
+    return &this->volume_proxy_callbacks;
+}
+
+void WrappedAsio::record_volume_output_channel(const AsioBufferInfo &info) {
+    if (info.is_input != AsioFalse) {
+        return;
+    }
+
+    // ask the real driver for this channel's sample format so the realtime path knows how
+    // to scale it; fall back to a sentinel that apply_gain_planar leaves untouched
+    AsioChannelInfo ci {};
+    ci.channel = info.channel_num;
+    ci.is_input = AsioFalse;
+    AsioSampleType type = ASIOSTLastEntry;
+    if (this->pReal->get_channel_info(&ci) == ASE_OK) {
+        type = ci.type;
+    }
+
+    VolumeOutputChannel ch;
+    ch.buffers[0] = info.buffers[0];
+    ch.buffers[1] = info.buffers[1];
+    ch.type = type;
+    this->volume_channels.push_back(ch);
+}
+
+void WrappedAsio::publish_volume(long buffer_size) {
+    if (!this->volume_active) {
+        return;
+    }
+
+    // everything the realtime thread reads is now in place; make ourselves reachable
+    this->volume_buffer_size = buffer_size;
+    WrappedAsio::volume_active_instance.store(this, std::memory_order_release);
+    log_info(
+        "audio::wrappedasio",
+        "volume boost active: gain={}, scaling {} output channel(s)",
+        this->volume_gain,
+        this->volume_channels.size());
+}
+
+void WrappedAsio::detach_volume() {
+    // stop our realtime trampolines from reaching this wrapper, but only if we are the
+    // currently published instance
+    WrappedAsio *expected = this;
+    WrappedAsio::volume_active_instance.compare_exchange_strong(expected, nullptr);
+}
+
+void WrappedAsio::apply_output_volume(long double_buffer_index) {
+    if (double_buffer_index != 0 && double_buffer_index != 1) {
+        return;
+    }
+    const float gain = this->volume_gain;
+    const long frames = this->volume_buffer_size;
+    for (const VolumeOutputChannel &ch : this->volume_channels) {
+        apply_gain_planar(ch.buffers[double_buffer_index], frames, ch.type, gain);
+    }
+}
+
+void __cdecl WrappedAsio::volume_buffer_switch(long double_buffer_index, AsioBool direct_process) {
+    WrappedAsio *self = WrappedAsio::volume_active_instance.load(std::memory_order_acquire);
+    if (self == nullptr) {
+        return;
+    }
+
+    // let the game write its samples into the driver buffers first, then scale them before
+    // the driver plays this half on the next switch
+    if (self->volume_game_callbacks.buffer_switch != nullptr) {
+        self->volume_game_callbacks.buffer_switch(double_buffer_index, direct_process);
+    }
+    self->apply_output_volume(double_buffer_index);
+}
+
+AsioTime * __cdecl WrappedAsio::volume_buffer_switch_time_info(
+    AsioTime *params, long double_buffer_index, AsioBool direct_process)
+{
+    WrappedAsio *self = WrappedAsio::volume_active_instance.load(std::memory_order_acquire);
+    if (self == nullptr) {
+        return params;
+    }
+
+    AsioTime *ret = params;
+    if (self->volume_game_callbacks.buffer_switch_time_info != nullptr) {
+        ret = self->volume_game_callbacks.buffer_switch_time_info(
+            params, double_buffer_index, direct_process);
+    } else if (self->volume_game_callbacks.buffer_switch != nullptr) {
+        self->volume_game_callbacks.buffer_switch(double_buffer_index, direct_process);
+    }
+    self->apply_output_volume(double_buffer_index);
+    return ret;
+}
+
+AsioError __thiscall WrappedAsio::create_buffers(
+    AsioBufferInfo *buffer_infos,
+    long num_channels,
+    long buffer_size,
+    AsioCallbacks *callbacks)
+{
+    // swap in our buffer-switch trampolines if a volume boost is configured, so the real
+    // driver calls us and we scale its output after the game fills it (no-op otherwise)
+    AsioCallbacks *effective = this->install_volume_callbacks(callbacks);
+
+    if (FORCE_TWO_CHANNELS) {
+        return this->create_buffers_front_pair(buffer_infos, num_channels, buffer_size, effective);
+    }
+
+    const AsioError result = this->pReal->create_buffers(buffer_infos, num_channels, buffer_size, effective);
+    if (result == ASE_OK) {
+        log_info(
+            "audio::wrappedasio",
+            "create_buffers(channels={}, size={} frames) succeeded",
+            num_channels,
+            buffer_size);
+
+        // capture the device output channels we will scale, then publish ourselves to the
+        // realtime thread once everything is in place
+        if (this->volume_active) {
+            for (long i = 0; i < num_channels; i++) {
+                this->record_volume_output_channel(buffer_infos[i]);
+            }
+            this->publish_volume(buffer_size);
+        }
+    } else {
+        log_warning(
+            "audio::wrappedasio",
+            "create_buffers(channels={}, size={} frames) failed, err={}",
+            num_channels,
+            buffer_size,
+            static_cast<long>(result));
+    }
+
+    return result;
+}
+
+AsioError WrappedAsio::create_buffers_front_pair(
+    AsioBufferInfo *buffer_infos,
+    long num_channels,
+    long buffer_size,
+    AsioCallbacks *callbacks)
+{
+    // front-pair extraction (forced two-channel ASIO): the game asks for more output
+    // channels than the real device has (e.g. 8 vs 2). forward only the channels the
+    // device actually provides (channel 0/1 = front L/R) and hand the game throwaway
+    // buffers for the rest, so its front mix lands on the device and the surround
+    // channels are discarded. the game writes directly into the driver/dummy buffers
+    // from its own bufferSwitch; the only realtime work we do is the volume boost (if
+    // configured), which scales the forwarded device channels via our trampolines
+    long real_in = 0, real_out = 0;
+    const AsioError ch_result = this->pReal->get_channels(&real_in, &real_out);
+    if (ch_result != ASE_OK) {
+        log_warning(
+            "audio::wrappedasio",
+            "create_buffers: get_channels failed, err={}",
+            static_cast<long>(ch_result));
+        return ch_result;
+    }
+
+    // partition the requested channels: those the device can serve are forwarded, the rest
+    // are discarded. record source indices for both so we can patch the game's array after
+    std::vector<AsioBufferInfo> forwarded;
+    std::vector<long> forwarded_src;
+    std::vector<long> discarded_src;
+    forwarded.reserve(num_channels);
+    forwarded_src.reserve(num_channels);
+    discarded_src.reserve(num_channels);
+    for (long i = 0; i < num_channels; i++) {
+        const AsioBufferInfo &bi = buffer_infos[i];
+        const long limit = (bi.is_input == AsioTrue) ? real_in : real_out;
+        if (bi.channel_num < limit) {
+            forwarded.push_back(bi);
+            forwarded_src.push_back(i);
+        } else {
+            discarded_src.push_back(i);
+        }
+    }
+
+    const AsioError result = this->pReal->create_buffers(
+        forwarded.data(), static_cast<long>(forwarded.size()), buffer_size, callbacks);
+    if (result != ASE_OK) {
+        log_warning(
+            "audio::wrappedasio",
+            "create_buffers(forwarded={} of {}, size={} frames) failed, err={}",
+            forwarded.size(),
+            num_channels,
+            buffer_size,
+            static_cast<long>(result));
+        return result;
+    }
+
+    // copy the real driver buffer pointers back into the game's array
+    for (size_t k = 0; k < forwarded.size(); k++) {
+        AsioBufferInfo &dst = buffer_infos[forwarded_src[k]];
+        dst.buffers[0] = forwarded[k].buffers[0];
+        dst.buffers[1] = forwarded[k].buffers[1];
+
+        // only the forwarded channels reach the device, so those are the ones the volume
+        // boost scales (the discarded channels go to throwaway buffers below)
+        if (this->volume_active) {
+            this->record_volume_output_channel(dst);
+        }
+    }
+
+    this->publish_volume(buffer_size);
+
+    // hand throwaway double buffers to the discarded channels. sized generously at
+    // 8 bytes/sample (covers every ASIO sample type) so the game can never overrun them
+    // regardless of the negotiated format
+    this->dummy_buffers.clear();
+    this->dummy_buffers.reserve(discarded_src.size() * 2);
+    const size_t dummy_bytes = static_cast<size_t>(buffer_size) * 8;
+    for (const long i : discarded_src) {
+        for (void *&buffer : buffer_infos[i].buffers) {
+            auto buf = std::make_unique<uint8_t[]>(dummy_bytes);
+            std::memset(buf.get(), 0, dummy_bytes);
+            buffer = buf.get();
+            this->dummy_buffers.push_back(std::move(buf));
+        }
+    }
+
+    log_info(
+        "audio::wrappedasio",
+        "create_buffers: front-pair extraction - forwarded {} channel(s) to device, "
+        "discarded {} (requested {}, size={} frames)",
+        forwarded.size(),
+        discarded_src.size(),
+        num_channels,
+        buffer_size);
+
+    return ASE_OK;
+}
+
+AsioError __thiscall WrappedAsio::dispose_buffers() {
+    // stop our realtime trampolines from touching buffers the driver is about to free
+    this->detach_volume();
+
+    const AsioError result = this->pReal->dispose_buffers();
+    this->dummy_buffers.clear();
+    this->volume_channels.clear();
+    this->volume_active = false;
+    return result;
+}
+
+AsioError __thiscall WrappedAsio::control_panel() {
+    return this->pReal->control_panel();
+}
+
+AsioError __thiscall WrappedAsio::future(long selector, void *opt) {
+    return this->pReal->future(selector, opt);
+}
+
+AsioError __thiscall WrappedAsio::output_ready() {
+    return this->pReal->output_ready();
+}
+#pragma endregion
+
+namespace hooks::audio::asio {
+
+    IUnknown *wrap(REFCLSID clsid, void *real) {
+        log_info("audio::wrappedasio", "wrapping ASIO driver interface, clsid={}", guid2s(clsid));
+
+        auto *wrapper = new WrappedAsio(
+            reinterpret_cast<IAsio *>(real), clsid, registered_asio_name(clsid));
+
+        // if the host already had a live wrapper for this CLSID it leaked the previous
+        // instance (ASIO is single-instance); tear it down now so the real driver is
+        // released before the reinit. works around games (iidx32+) that ref twice but
+        // deref once before re-initializing.
+        //
+        // FlexASIO 1.9 and many DAC ASIO drivers can't handle this; FlexASIO 1.10 and
+        // Xonar AE can
+        if (WrappedAsio *stale = register_wrapper(clsid, wrapper)) {
+            log_info(
+                "audio::wrappedasio",
+                "host did not release previous instance for clsid={}, forcing teardown",
+                guid2s(clsid));
+
+            // the stale instance may still have a running stream on the driver's realtime
+            // thread. stop and dispose it before deleting so no in-flight buffer switch
+            // (e.g. our volume trampoline) touches buffers we are about to free. ASIO
+            // guarantees no further buffer_switch once stop() returns, and dispose_buffers
+            // detaches our trampolines, fully quiescing the realtime path before delete
+            stale->stop();
+            stale->dispose_buffers();
+            delete stale;
+        }
+
+        return static_cast<IAsio *>(wrapper);
+    }
+}

src/spice2x/hooks/audio/asio_proxy.h

@@ -0,0 +1,160 @@
+#pragma once
+
+#include <atomic>
+#include <memory>
+#include <string>
+#include <vector>
+
+#include <windows.h>
+
+#include "external/asio/asio.h"
+#include "external/asio/iasiodrv.h"
+
+namespace hooks::audio::asio {
+
+    // returns true if a CoCreateInstance call is instantiating a registered ASIO driver.
+    // ASIO hosts pass the driver CLSID as both class id and interface id; we also validate
+    // it against the system's registered ASIO drivers to avoid false positives
+    bool is_asio_creation(REFCLSID rclsid, REFIID riid);
+
+    // wrap a real ASIO driver instance, taking ownership of the supplied reference, and
+    // return a proxy that forwards every call to it
+    IUnknown *wrap(REFCLSID clsid, void *real);
+}
+
+// transparent proxy around a real ASIO driver; a single place to intercept ASIO traffic
+struct WrappedAsio final : IAsio {
+    WrappedAsio(IAsio *real, REFCLSID clsid, std::string name)
+        : pReal(real), clsid(clsid), driver_name(std::move(name)) {
+    }
+
+    WrappedAsio(const WrappedAsio &) = delete;
+    WrappedAsio &operator=(const WrappedAsio &) = delete;
+
+    virtual ~WrappedAsio();
+
+    // when set, the proxy presents the game's expected multichannel layout to the host so
+    // it proceeds to create_buffers, then forwards only the device's real front pair and
+    // discards the rest (see create_buffers). set once at boot, before any wrapper exists,
+    // so it needs no synchronization
+    static bool FORCE_TWO_CHANNELS;
+
+    // some games hardcode a multichannel ASIO output and bail before create_buffers if
+    // get_channels reports fewer, so we report at least this many output channels when
+    // FORCE_TWO_CHANNELS is active
+    static constexpr long FORCED_OUTPUT_CHANNELS = 8;
+
+#pragma region IUnknown
+    HRESULT STDMETHODCALLTYPE QueryInterface(REFIID riid, void **ppv) override;
+    ULONG STDMETHODCALLTYPE AddRef() override;
+    ULONG STDMETHODCALLTYPE Release() override;
+#pragma endregion
+
+#pragma region IAsio
+    AsioBool __thiscall init(void *sys_handle) override;
+    void __thiscall get_driver_name(char *name) override;
+    long __thiscall get_driver_version() override;
+    void __thiscall get_error_message(char *string) override;
+    AsioError __thiscall start() override;
+    AsioError __thiscall stop() override;
+    AsioError __thiscall get_channels(long *num_input_channels, long *num_output_channels) override;
+    AsioError __thiscall get_latencies(long *input_latency, long *output_latency) override;
+    AsioError __thiscall get_buffer_size(
+        long *min_size,
+        long *max_size,
+        long *preferred_size,
+        long *granularity) override;
+    AsioError __thiscall can_sample_rate(AsioSampleRate sample_rate) override;
+    AsioError __thiscall get_sample_rate(AsioSampleRate *sample_rate) override;
+    AsioError __thiscall set_sample_rate(AsioSampleRate sample_rate) override;
+    AsioError __thiscall get_clock_sources(ASIOClockSource *clocks, long *num_sources) override;
+    AsioError __thiscall set_clock_source(long reference) override;
+    AsioError __thiscall get_sample_position(ASIOSamples *s_pos, ASIOTimeStamp *t_stamp) override;
+    AsioError __thiscall get_channel_info(AsioChannelInfo *info) override;
+    AsioError __thiscall create_buffers(
+        AsioBufferInfo *buffer_infos,
+        long num_channels,
+        long buffer_size,
+        AsioCallbacks *callbacks) override;
+    AsioError __thiscall dispose_buffers() override;
+    AsioError __thiscall control_panel() override;
+    AsioError __thiscall future(long selector, void *opt) override;
+    AsioError __thiscall output_ready() override;
+#pragma endregion
+
+private:
+    // create_buffers implementation used when FORCE_TWO_CHANNELS is active: forwards only
+    // the channels the real device has and hands the game throwaway buffers for the rest
+    AsioError create_buffers_front_pair(
+        AsioBufferInfo *buffer_infos,
+        long num_channels,
+        long buffer_size,
+        AsioCallbacks *callbacks);
+
+    // if hooks::audio::VOLUME_BOOST is set, saves the game's callbacks and returns a proxy
+    // callback set (our buffer-switch trampolines) to hand the real driver instead, so we
+    // can scale its output buffers after the game fills them. otherwise returns the game's
+    // callbacks unchanged. called at create_buffers time, before the stream starts
+    AsioCallbacks *install_volume_callbacks(AsioCallbacks *game_callbacks);
+
+    // records a device output channel whose buffers we scale by the volume boost. queries
+    // the real driver for the channel's sample format. called at create_buffers time
+    void record_volume_output_channel(const AsioBufferInfo &info);
+
+    // publishes the captured volume state to the realtime thread once the buffers exist,
+    // making our trampolines start scaling. called at the end of either create_buffers path
+    void publish_volume(long buffer_size);
+
+    // detaches this instance from the realtime trampolines so they stop touching its
+    // buffers. called from dispose_buffers and the destructor
+    void detach_volume();
+
+    // multiplies every recorded output channel's buffer for the given double-buffer index
+    // by the volume boost. runs on the driver's realtime thread from our buffer switch
+    void apply_output_volume(long double_buffer_index);
+
+    // realtime-thread trampolines for the buffer-switch callbacks, handed to the real
+    // driver in place of the game's; ASIO callbacks carry no user data, so they reach the
+    // active wrapper through volume_active_instance, call the game's original, then scale.
+    // the other two callbacks (sample_rate_did_change, asio_message) are forwarded as the
+    // game's own pointers, so they need no trampoline
+    static void __cdecl volume_buffer_switch(long double_buffer_index, AsioBool direct_process);
+    static AsioTime * __cdecl volume_buffer_switch_time_info(
+        AsioTime *params, long double_buffer_index, AsioBool direct_process);
+
+    // the single wrapper whose proxy callbacks are installed (ASIO is single-instance with
+    // one running stream); read by the static trampolines to reach the right wrapper
+    static std::atomic<WrappedAsio *> volume_active_instance;
+
+    IAsio *const pReal;
+    const CLSID clsid;
+
+    // registry name of the driver (not get_driver_name), used in our logs as a single
+    // unambiguous name; constant for our lifetime
+    std::string driver_name;
+
+    // our own reference count; we hold one reference on pReal and release it when this
+    // drops to zero
+    std::atomic<ULONG> ref_count {1};
+
+    // throwaway double buffers handed to the channels we discard when FORCE_TWO_CHANNELS
+    // is active (see create_buffers). owned for the lifetime of the buffer set and freed
+    // in dispose_buffers; only read by the game from its own bufferSwitch, never by us
+    std::vector<std::unique_ptr<uint8_t[]>> dummy_buffers;
+
+    // one device output channel scaled by the volume boost in our buffer switch
+    struct VolumeOutputChannel {
+        void *buffers[2];
+        AsioSampleType type;
+    };
+
+    // volume boost state, captured at create_buffers time and published to the realtime
+    // thread via volume_active_instance once fully built; untouched while the stream runs.
+    // volume_active gates whether we install our proxy callbacks at all
+    bool volume_active = false;
+    float volume_gain = 1.0f;
+    long volume_buffer_size = 0;
+    AsioCallbacks volume_game_callbacks {};
+    AsioCallbacks volume_proxy_callbacks {};
+    std::vector<VolumeOutputChannel> volume_channels;
+};

src/spice2x/hooks/audio/audio.cpp

@@ -14,6 +14,7 @@
 
 #include "audio_private.h"
 #include "acm.h"
+#include "asio_proxy.h"
 
 #ifdef _MSC_VER
 DEFINE_GUID(CLSID_MMDeviceEnumerator,
@@ -39,9 +40,14 @@ namespace hooks::audio {
     // public globals
     bool ENABLED = true;
     bool VOLUME_HOOK_ENABLED = true;
+    std::optional<DownmixAlgorithm> DOWNMIX_ALGORITHM = std::nullopt;
+    float VOLUME_BOOST = 1.0f;
+    std::optional<uint32_t> RESAMPLE_RATE = std::nullopt;
+    std::optional<uint32_t> EXCLUSIVE_BUFFER_MS = std::nullopt;
     bool USE_DUMMY = false;
     WAVEFORMATEXTENSIBLE FORMAT {};
     std::optional<Backend> BACKEND = std::nullopt;
+    bool INJECT_FAKE_REALTEK_AUDIO = false;
     std::optional<size_t> ASIO_DRIVER_ID = std::nullopt;
     std::string ASIO_DRIVER_NAME = "";
     bool ASIO_FORCE_UNLOAD_ON_STOP = false;
@@ -66,14 +72,16 @@ static HRESULT STDAPICALLTYPE CoCreateInstance_hook(
     if (FAILED(ret)) {
         if (IsEqualCLSID(rclsid, CLSID_MMDeviceEnumerator)) {
             log_warning("audio", "CoCreateInstance failed for CLSID_MMDeviceEnumerator, hr={}", FMT_HRESULT(ret));
-        }
-        
-        // Windows 11 KB5052093 issue - reverb DMO went missing from dsdmo.dll (fails with 0x80040154)
-        if (IsEqualCLSID(rclsid, CLSID_DirectSoundI3DL2ReverbDMO) && ret == (HRESULT)0x80040154) {
-            log_warning(
-                "audio",
-                "CoCreateInstance for CLSID_DirectSoundI3DL2ReverbDMO failed with 0x80040154, swap with CLSID_DirectSoundWavesReverbDMO "
-                "(workaround for Windows 11 KB5052093 issue); REVERB EX replaced with REVERB");
+        } else if (IsEqualCLSID(rclsid, CLSID_DirectSoundI3DL2ReverbDMO)) {
+            // deal with reverb DMO missing from dsdmo.dll
+            // it usually fails with 0x80040154 (REGDB_E_CLASSNOTREG), but we have also seen 0x8007007e (ERROR_MOD_NOT_FOUND) 
+            static std::once_flag printed;
+            std::call_once(printed, [ret]() {
+                log_warning(
+                    "audio",
+                    "CoCreateInstance for CLSID_DirectSoundI3DL2ReverbDMO failed with {}, swap with CLSID_DirectSoundWavesReverbDMO...",
+                    FMT_HRESULT(ret));
+            });
             ret = CoCreateInstance_orig(CLSID_DirectSoundWavesReverbDMO, pUnkOuter, dwClsContext, riid, ppv);
             if (FAILED(ret)) {
                 log_warning("audio", "CoCreateInstance(CLSID_DirectSoundWavesReverbDMO...) failed, hr={}", FMT_HRESULT(ret));
@@ -89,6 +97,10 @@ static HRESULT STDAPICALLTYPE CoCreateInstance_hook(
         // wrap object
         auto mmde = reinterpret_cast<IMMDeviceEnumerator **>(ppv);
         *mmde = new WrappedIMMDeviceEnumerator(*mmde);
+    } else if (ppv != nullptr && *ppv != nullptr && hooks::audio::asio::is_asio_creation(rclsid, riid)) {
+
+        // wrap every ASIO driver so calls pass through to the real driver
+        *ppv = hooks::audio::asio::wrap(rclsid, *ppv);
     }
 
     // return original result

src/spice2x/hooks/audio/audio.h

@@ -1,5 +1,6 @@
 #pragma once
 
+#include <cstdint>
 #include <optional>
 #include <string>
 
@@ -16,11 +17,34 @@ namespace hooks::audio {
         WaveOut,
     };
 
+    // surround-to-stereo downmix algorithm
+    enum class DownmixAlgorithm {
+        FrontOnly, // keep only the front channels
+        RearOnly,  // keep only the rear/back channels
+        SideOnly,  // keep only the side channels
+        AC4,       // AC-4 stereo downmix coefficients (ETSI TS 103 190-1)
+        Normalize, // all channels equally loud, LFE dropped
+    };
+
     extern bool ENABLED;
     extern bool VOLUME_HOOK_ENABLED;
+    extern std::optional<DownmixAlgorithm> DOWNMIX_ALGORITHM;
+    extern float VOLUME_BOOST;
+
+    // target sample rate the hooked output is resampled to, if set
+    extern std::optional<uint32_t> RESAMPLE_RATE;
+
+    // minimum WASAPI exclusive buffer duration (milliseconds), if set. enlarges the device buffer
+    // to avoid underrun crackle on endpoints that cannot service a tiny buffer in time.
+    extern std::optional<uint32_t> EXCLUSIVE_BUFFER_MS;
     extern bool USE_DUMMY;
     extern WAVEFORMATEXTENSIBLE FORMAT;
     extern std::optional<Backend> BACKEND;
+
+    // when true, a synthetic "Realtek" render endpoint is injected into device enumeration that
+    // discards all audio. used by gitadora arena, whose device search crashes when no render
+    // endpoint reports a "Realtek" friendly name.
+    extern bool INJECT_FAKE_REALTEK_AUDIO;
     extern std::optional<size_t> ASIO_DRIVER_ID;
     extern std::string ASIO_DRIVER_NAME;
     extern bool ASIO_FORCE_UNLOAD_ON_STOP;

src/spice2x/hooks/audio/backends/mmdevice/device_collection.cpp

@@ -1,5 +1,6 @@
 #include "device_collection.h"
 #include "device.h"
+#include "null_device.h"
 #include "util/utils.h"
 #include "util/logging.h"
 
@@ -28,17 +29,48 @@ ULONG STDMETHODCALLTYPE WrappedIMMDeviceCollection::Release() {
     return refs;
 }
 
+bool WrappedIMMDeviceCollection::should_inject_fake_realtek() const {
+    return null_render_device_enabled()
+        && (data_flow == eRender || data_flow == eAll);
+}
+
 HRESULT STDMETHODCALLTYPE WrappedIMMDeviceCollection::GetCount(UINT *pcDevices) {
+    // when active, hide all real devices and present only the synthetic one
+    if (should_inject_fake_realtek()) {
+        if (pcDevices == nullptr) {
+            return E_POINTER;
+        }
+        *pcDevices = 1;
+        return S_OK;
+    }
+
     return pReal->GetCount(pcDevices);
 }
 
 HRESULT STDMETHODCALLTYPE WrappedIMMDeviceCollection::Item(UINT nDevice, IMMDevice **ppDevice) {
+    if (ppDevice == nullptr) {
+        return E_POINTER;
+    }
+
+    // when active, the only device in the collection is the synthetic fake Realtek
+    // render device; all real devices are hidden
+    if (should_inject_fake_realtek()) {
+        if (nDevice != 0) {
+            return E_INVALIDARG;
+        }
+        log_info("audio", "WrappedIMMDeviceCollection::Item[{}] -> synthetic fake Realtek render device", nDevice);
+        *ppDevice = new NullMMDevice();
+        return S_OK;
+    }
+
     log_info("audio", "WrappedIMMDeviceCollection::Item[{}]", nDevice);
 
     // call original
     const auto hr = pReal->Item(nDevice, ppDevice);
 
     // wrap interface
-    *ppDevice = new WrappedIMMDevice(*ppDevice);
+    if (SUCCEEDED(hr) && *ppDevice != nullptr) {
+        *ppDevice = new WrappedIMMDevice(*ppDevice);
+    }
     return hr;
 }

src/spice2x/hooks/audio/backends/mmdevice/device_collection.h

@@ -4,7 +4,8 @@
 #include <mmdeviceapi.h>
 
 struct WrappedIMMDeviceCollection : IMMDeviceCollection {
-    explicit WrappedIMMDeviceCollection(IMMDeviceCollection *orig) : pReal(orig) {
+    WrappedIMMDeviceCollection(IMMDeviceCollection *orig, EDataFlow dataFlow)
+        : pReal(orig), data_flow(dataFlow) {
     }
 
     WrappedIMMDeviceCollection(const WrappedIMMDeviceCollection &) = delete;
@@ -24,5 +25,9 @@ struct WrappedIMMDeviceCollection : IMMDeviceCollection {
 #pragma endregion
 
 private:
+    // whether the synthetic fake Realtek render device should be appended to this collection
+    bool should_inject_fake_realtek() const;
+
     IMMDeviceCollection *const pReal;
+    const EDataFlow data_flow;
 };

src/spice2x/hooks/audio/backends/mmdevice/device_enumerator.cpp

@@ -45,7 +45,7 @@ HRESULT STDMETHODCALLTYPE WrappedIMMDeviceEnumerator::EnumAudioEndpoints(
 {
     const auto hr = pReal->EnumAudioEndpoints(dataFlow, dwStateMask, ppDevices);
     if (SUCCEEDED(hr) && (ppDevices != nullptr) && (*ppDevices != nullptr)) {
-        *ppDevices = new WrappedIMMDeviceCollection(*ppDevices);
+        *ppDevices = new WrappedIMMDeviceCollection(*ppDevices, dataFlow);
     }
     return hr;
 }

src/spice2x/hooks/audio/backends/mmdevice/null_device.cpp

@@ -0,0 +1,195 @@
+#include "null_device.h"
+
+#include <atomic>
+#include <cstring>
+
+#include <audioclient.h>
+
+#include "hooks/audio/audio.h"
+#include "hooks/audio/audio_private.h"
+#include "hooks/audio/backends/wasapi/dummy_audio_client.h"
+#include "util/logging.h"
+#include "util/utils.h"
+
+#include "null_discard_backend.h"
+
+// friendly name reported by the synthetic device. must contain "Realtek" so the
+// gitadora arena device search matches it.
+static const wchar_t NULL_DEVICE_FRIENDLY_NAME[] = L"Realtek High Definition Audio";
+
+// arbitrary identifier reported by the synthetic device.
+static const wchar_t NULL_DEVICE_ID[] = L"{spice2x-null-render-device}";
+
+// PKEY_Device_FriendlyName, hardcoded to avoid pulling in functiondiscoverykeys_devpkey.h
+static const PROPERTYKEY PKEY_DEVICE_FRIENDLY_NAME_LOCAL = {
+    { 0xa45c254e, 0xdf1c, 0x4efd, { 0x80, 0x20, 0x67, 0xd1, 0x46, 0xa8, 0x50, 0xe0 } },
+    14
+};
+
+bool null_render_device_enabled() {
+    return hooks::audio::INJECT_FAKE_REALTEK_AUDIO;
+}
+
+// duplicate a wide string into CoTaskMem so the caller can free it with
+// CoTaskMemFree / PropVariantClear as the COM API contract requires.
+static LPWSTR co_task_wcsdup(const wchar_t *src) {
+    const size_t bytes = (wcslen(src) + 1) * sizeof(wchar_t);
+    auto *dst = static_cast<LPWSTR>(CoTaskMemAlloc(bytes));
+    if (dst != nullptr) {
+        memcpy(dst, src, bytes);
+    }
+    return dst;
+}
+
+namespace {
+
+    // minimal IPropertyStore that only answers PKEY_Device_FriendlyName.
+    struct NullPropertyStore : IPropertyStore {
+        std::atomic<ULONG> ref_cnt = 1;
+
+        virtual ~NullPropertyStore() = default;
+
+        HRESULT STDMETHODCALLTYPE QueryInterface(REFIID riid, void **ppvObj) override {
+            if (ppvObj == nullptr) {
+                return E_POINTER;
+            }
+            if (riid == __uuidof(IUnknown) || riid == __uuidof(IPropertyStore)) {
+                this->AddRef();
+                *ppvObj = this;
+                return S_OK;
+            }
+            *ppvObj = nullptr;
+            return E_NOINTERFACE;
+        }
+        ULONG STDMETHODCALLTYPE AddRef() override {
+            return ++this->ref_cnt;
+        }
+        ULONG STDMETHODCALLTYPE Release() override {
+            const ULONG refs = --this->ref_cnt;
+            if (refs == 0) {
+                delete this;
+            }
+            return refs;
+        }
+        HRESULT STDMETHODCALLTYPE GetCount(DWORD *cProps) override {
+            if (cProps == nullptr) {
+                return E_POINTER;
+            }
+            *cProps = 1;
+            return S_OK;
+        }
+        HRESULT STDMETHODCALLTYPE GetAt(DWORD iProp, PROPERTYKEY *pkey) override {
+            if (pkey == nullptr) {
+                return E_POINTER;
+            }
+            if (iProp != 0) {
+                return E_INVALIDARG;
+            }
+            *pkey = PKEY_DEVICE_FRIENDLY_NAME_LOCAL;
+            return S_OK;
+        }
+        HRESULT STDMETHODCALLTYPE GetValue(REFPROPERTYKEY key, PROPVARIANT *pv) override {
+            if (pv == nullptr) {
+                return E_POINTER;
+            }
+            PropVariantInit(pv);
+            if (key.fmtid == PKEY_DEVICE_FRIENDLY_NAME_LOCAL.fmtid
+                    && key.pid == PKEY_DEVICE_FRIENDLY_NAME_LOCAL.pid) {
+                pv->pwszVal = co_task_wcsdup(NULL_DEVICE_FRIENDLY_NAME);
+                if (pv->pwszVal == nullptr) {
+                    return E_OUTOFMEMORY;
+                }
+                pv->vt = VT_LPWSTR;
+            }
+            // unknown keys are returned as VT_EMPTY / S_OK
+            return S_OK;
+        }
+        HRESULT STDMETHODCALLTYPE SetValue(REFPROPERTYKEY, REFPROPVARIANT) override {
+            return STG_E_ACCESSDENIED;
+        }
+        HRESULT STDMETHODCALLTYPE Commit() override {
+            return S_OK;
+        }
+    };
+}
+
+#pragma region IUnknown
+HRESULT STDMETHODCALLTYPE NullMMDevice::QueryInterface(REFIID riid, void **ppvObj) {
+    if (ppvObj == nullptr) {
+        return E_POINTER;
+    }
+    if (riid == __uuidof(IUnknown) || riid == __uuidof(IMMDevice)) {
+        this->AddRef();
+        *ppvObj = this;
+        return S_OK;
+    }
+    *ppvObj = nullptr;
+    return E_NOINTERFACE;
+}
+ULONG STDMETHODCALLTYPE NullMMDevice::AddRef() {
+    return ++this->ref_cnt;
+}
+ULONG STDMETHODCALLTYPE NullMMDevice::Release() {
+    const ULONG refs = --this->ref_cnt;
+    if (refs == 0) {
+        delete this;
+    }
+    return refs;
+}
+#pragma endregion
+
+#pragma region IMMDevice
+HRESULT STDMETHODCALLTYPE NullMMDevice::Activate(
+    REFIID iid,
+    DWORD,
+    PROPVARIANT *,
+    void **ppInterface)
+{
+    if (ppInterface == nullptr) {
+        return E_POINTER;
+    }
+    *ppInterface = nullptr;
+
+    log_info("audio::null", "NullMMDevice::Activate {}", guid2s(iid));
+
+    if (iid == IID_IAudioClient) {
+
+        // release any previously persisted client
+        if (hooks::audio::CLIENT != nullptr) {
+            hooks::audio::CLIENT->Release();
+        }
+
+        auto *client = static_cast<IAudioClient *>(new DummyIAudioClient(new NullDiscardBackend()));
+        *ppInterface = client;
+
+        // persist the audio client
+        hooks::audio::CLIENT = client;
+        hooks::audio::CLIENT->AddRef();
+
+        return S_OK;
+    }
+
+    return E_NOINTERFACE;
+}
+HRESULT STDMETHODCALLTYPE NullMMDevice::OpenPropertyStore(DWORD, IPropertyStore **ppProperties) {
+    if (ppProperties == nullptr) {
+        return E_POINTER;
+    }
+    *ppProperties = new NullPropertyStore();
+    return S_OK;
+}
+HRESULT STDMETHODCALLTYPE NullMMDevice::GetId(LPWSTR *ppstrId) {
+    if (ppstrId == nullptr) {
+        return E_POINTER;
+    }
+    *ppstrId = co_task_wcsdup(NULL_DEVICE_ID);
+    return *ppstrId != nullptr ? S_OK : E_OUTOFMEMORY;
+}
+HRESULT STDMETHODCALLTYPE NullMMDevice::GetState(DWORD *pdwState) {
+    if (pdwState == nullptr) {
+        return E_POINTER;
+    }
+    *pdwState = DEVICE_STATE_ACTIVE;
+    return S_OK;
+}
+#pragma endregion

src/spice2x/hooks/audio/backends/mmdevice/null_device.h

@@ -0,0 +1,39 @@
+#pragma once
+
+#include <atomic>
+
+#include <mmdeviceapi.h>
+
+// returns true when a synthetic render endpoint should be injected into device
+// enumeration. games like gitadora arena search the render endpoint list for a
+// device whose friendly name contains "Realtek" and crash with a null pointer
+// dereference when no match exists. presenting a fake match that routes to the
+// null audio backend lets the search succeed while discarding the audio.
+bool null_render_device_enabled();
+
+// fake IMMDevice that reports a "Realtek" friendly name and activates straight
+// into the null audio backend, never touching real hardware.
+struct NullMMDevice : IMMDevice {
+    NullMMDevice() = default;
+
+    NullMMDevice(const NullMMDevice &) = delete;
+    NullMMDevice &operator=(const NullMMDevice &) = delete;
+
+    virtual ~NullMMDevice() = default;
+
+#pragma region IUnknown
+    HRESULT STDMETHODCALLTYPE QueryInterface(REFIID riid, void **ppvObj) override;
+    ULONG STDMETHODCALLTYPE AddRef() override;
+    ULONG STDMETHODCALLTYPE Release() override;
+#pragma endregion
+
+#pragma region IMMDevice
+    HRESULT STDMETHODCALLTYPE Activate(REFIID iid, DWORD dwClsCtx, PROPVARIANT *pActivationParams, void **ppInterface) override;
+    HRESULT STDMETHODCALLTYPE OpenPropertyStore(DWORD stgmAccess, IPropertyStore **ppProperties) override;
+    HRESULT STDMETHODCALLTYPE GetId(LPWSTR *ppstrId) override;
+    HRESULT STDMETHODCALLTYPE GetState(DWORD *pdwState) override;
+#pragma endregion
+
+private:
+    std::atomic<ULONG> ref_cnt = 1;
+};

src/spice2x/hooks/audio/backends/mmdevice/null_discard_backend.cpp

@@ -0,0 +1,141 @@
+#include "null_discard_backend.h"
+
+#include <algorithm>
+#include <chrono>
+
+#include "hooks/audio/util.h"
+#include "util/logging.h"
+#include "util/precise_timer.h"
+
+NullDiscardBackend::~NullDiscardBackend() {
+    this->running = false;
+    if (this->pacing_thread.joinable()) {
+        this->pacing_thread.join();
+    }
+}
+
+const WAVEFORMATEXTENSIBLE &NullDiscardBackend::format() const noexcept {
+    return this->format_;
+}
+
+HRESULT NullDiscardBackend::on_initialize(
+    AUDCLNT_SHAREMODE *,
+    DWORD *,
+    REFERENCE_TIME *hnsBufferDuration,
+    REFERENCE_TIME *,
+    const WAVEFORMATEX *pFormat,
+    LPCGUID)
+{
+    copy_wave_format(&this->format_, pFormat);
+
+    // honor the game's requested buffer duration, falling back to 10 ms
+    constexpr REFERENCE_TIME DEFAULT_REFTIME = 100000; // 10 ms in 100-ns units
+    this->period_reftime = (hnsBufferDuration && *hnsBufferDuration > 0)
+            ? *hnsBufferDuration
+            : DEFAULT_REFTIME;
+
+    this->buffer_frames = std::max<uint32_t>(1, static_cast<uint32_t>(
+            static_cast<double>(this->format_.Format.nSamplesPerSec)
+                    * this->period_reftime / 10000000.0 + 0.5));
+
+    log_info("audio::null", "initializing null render device with {} channels, {} Hz, {}-bit",
+             this->format_.Format.nChannels,
+             this->format_.Format.nSamplesPerSec,
+             this->format_.Format.wBitsPerSample);
+
+    return S_OK;
+}
+
+HRESULT NullDiscardBackend::on_get_buffer_size(uint32_t *buffer_frames) {
+    *buffer_frames = this->buffer_frames;
+    return S_OK;
+}
+
+HRESULT NullDiscardBackend::on_get_stream_latency(REFERENCE_TIME *latency) {
+    *latency = this->period_reftime;
+    return S_OK;
+}
+
+HRESULT NullDiscardBackend::on_get_current_padding(std::optional<uint32_t> &padding_frames) {
+    // discarded immediately, so the buffer always reads as fully drained
+    padding_frames = 0;
+    return S_OK;
+}
+
+HRESULT NullDiscardBackend::on_is_format_supported(
+    AUDCLNT_SHAREMODE *,
+    const WAVEFORMATEX *,
+    WAVEFORMATEX **ppClosestMatch)
+{
+    if (ppClosestMatch) {
+        *ppClosestMatch = nullptr;
+    }
+    return S_OK;
+}
+
+HRESULT NullDiscardBackend::on_get_mix_format(WAVEFORMATEX **) {
+    return E_NOTIMPL;
+}
+
+HRESULT NullDiscardBackend::on_get_device_period(
+    REFERENCE_TIME *default_device_period,
+    REFERENCE_TIME *minimum_device_period)
+{
+    if (default_device_period) {
+        *default_device_period = this->period_reftime;
+    }
+    if (minimum_device_period) {
+        *minimum_device_period = this->period_reftime;
+    }
+    return S_OK;
+}
+
+HRESULT NullDiscardBackend::on_start() {
+    if (!this->running.exchange(true)) {
+        this->pacing_thread = std::thread(&NullDiscardBackend::pace_loop, this);
+    }
+    return S_OK;
+}
+
+HRESULT NullDiscardBackend::on_stop() {
+    return S_OK;
+}
+
+HRESULT NullDiscardBackend::on_set_event_handle(HANDLE *event_handle) {
+    // keep the game's event so pace_loop() can wake it; there is no real device behind it
+    this->relay_handle = *event_handle;
+    return S_OK;
+}
+
+HRESULT NullDiscardBackend::on_get_buffer(uint32_t num_frames_requested, BYTE **ppData) {
+    const size_t buffer_size =
+            static_cast<size_t>(this->format_.Format.nBlockAlign) * num_frames_requested;
+    if (this->scratch.size() < buffer_size) {
+        this->scratch.resize(buffer_size);
+    }
+    *ppData = this->scratch.data();
+    return S_OK;
+}
+
+HRESULT NullDiscardBackend::on_release_buffer(uint32_t, DWORD) {
+    // discard the audio entirely
+    return S_OK;
+}
+
+void NullDiscardBackend::pace_loop() {
+    using namespace std::chrono;
+
+    timeutils::PreciseSleepTimer timer;
+
+    // audio is discarded, so timing precision and drift do not matter; just wake the
+    // game once per buffer period to keep its render thread from blocking on the event.
+    const auto period = duration_cast<steady_clock::duration>(
+            duration<double>(this->period_reftime / 10000000.0));
+
+    while (this->running.load()) {
+        if (this->relay_handle) {
+            SetEvent(this->relay_handle);
+        }
+        timer.sleep(period);
+    }
+}

src/spice2x/hooks/audio/backends/mmdevice/null_discard_backend.h

@@ -0,0 +1,54 @@
+#pragma once
+
+#include <atomic>
+#include <optional>
+#include <thread>
+#include <vector>
+
+#include <audioclient.h>
+
+#include "hooks/audio/implementations/backend.h"
+
+// discards all audio while pacing the game's event handle once per buffer period, so the game
+// keeps running normally with nothing output to any real device. routed through the shared
+// DummyIAudioClient, the same plumbing the asio backend uses.
+struct NullDiscardBackend final : AudioBackend {
+    ~NullDiscardBackend() final;
+
+    const WAVEFORMATEXTENSIBLE &format() const noexcept override;
+
+    HRESULT on_initialize(
+        AUDCLNT_SHAREMODE *,
+        DWORD *,
+        REFERENCE_TIME *hnsBufferDuration,
+        REFERENCE_TIME *,
+        const WAVEFORMATEX *pFormat,
+        LPCGUID) override;
+    HRESULT on_get_buffer_size(uint32_t *buffer_frames) override;
+    HRESULT on_get_stream_latency(REFERENCE_TIME *latency) override;
+    HRESULT on_get_current_padding(std::optional<uint32_t> &padding_frames) override;
+    HRESULT on_is_format_supported(
+        AUDCLNT_SHAREMODE *,
+        const WAVEFORMATEX *,
+        WAVEFORMATEX **ppClosestMatch) override;
+    HRESULT on_get_mix_format(WAVEFORMATEX **) override;
+    HRESULT on_get_device_period(
+        REFERENCE_TIME *default_device_period,
+        REFERENCE_TIME *minimum_device_period) override;
+    HRESULT on_start() override;
+    HRESULT on_stop() override;
+    HRESULT on_set_event_handle(HANDLE *event_handle) override;
+    HRESULT on_get_buffer(uint32_t num_frames_requested, BYTE **ppData) override;
+    HRESULT on_release_buffer(uint32_t, DWORD) override;
+
+private:
+    void pace_loop();
+
+    WAVEFORMATEXTENSIBLE format_ {};
+    uint32_t buffer_frames = 0;
+    REFERENCE_TIME period_reftime = 0;
+    HANDLE relay_handle = nullptr;
+    std::vector<BYTE> scratch;
+    std::thread pacing_thread;
+    std::atomic<bool> running = false;
+};

src/spice2x/hooks/audio/backends/wasapi/audio_client.cpp

@@ -4,6 +4,7 @@
 #include <ksmedia.h>
 
 #include "avs/game.h"
+#include "games/gitadora/gitadora.h"
 #include "hooks/audio/audio.h"
 #include "hooks/audio/util.h"
 #include "hooks/audio/backends/wasapi/util.h"
@@ -41,6 +42,27 @@ static void fix_rec_format(WAVEFORMATEX *pFormat) {
     pFormat->nAvgBytesPerSec = pFormat->nSamplesPerSec * pFormat->nBlockAlign;
 }
 
+// decide whether the given multi-channel format should be downmixed to stereo and which algorithm
+// to use. an explicit user selection (-downmix) takes precedence; otherwise gitadora arena
+// two-channel mode defaults to the AC-4 algorithm.
+static std::optional<hooks::audio::DownmixAlgorithm> resolve_downmix(const WAVEFORMATEX *format) {
+    if (format == nullptr
+            || format->nChannels <= 2
+            || format->wFormatTag != WAVE_FORMAT_EXTENSIBLE) {
+        return std::nullopt;
+    }
+
+    if (hooks::audio::DOWNMIX_ALGORITHM.has_value()) {
+        return hooks::audio::DOWNMIX_ALGORITHM;
+    }
+
+    if (games::gitadora::is_arena_model() && games::gitadora::TWOCHANNEL) {
+        return hooks::audio::DownmixAlgorithm::AC4;
+    }
+
+    return std::nullopt;
+}
+
 IAudioClient *wrap_audio_client(IAudioClient *audio_client) {
     log_misc("audio::wasapi", "wrapping IAudioClient");
 
@@ -145,13 +167,36 @@ HRESULT STDMETHODCALLTYPE WrappedIAudioClient::Initialize(
         fix_rec_format(const_cast<WAVEFORMATEX *>(pFormat));
     }
 
+    // when downmixing, open the real device as stereo while the game keeps writing its native
+    // multi-channel format into the scratch buffer.
+    WAVEFORMATEXTENSIBLE stereo_storage = {};
+    const WAVEFORMATEX *device_format = pFormat;
+
+    if (auto algorithm = resolve_downmix(pFormat)) {
+        this->downmix.setup(pFormat, &stereo_storage, *algorithm);
+        device_format = reinterpret_cast<const WAVEFORMATEX *>(&stereo_storage);
+        log_info("audio::wasapi", "downmix enabled: {} channels -> 2 channels ({})",
+                pFormat->nChannels, hooks::audio::Downmix::algorithm_name(*algorithm));
+    } else if (games::gitadora::is_arena_model()) {
+        games::gitadora::fix_audio_channel_mask(const_cast<WAVEFORMATEX *>(pFormat));
+    }
+
+    // when resampling, open the real device at the target rate while the game keeps writing its
+    // native-rate audio into the scratch buffer. this runs on whatever device_format is now: the
+    // game's native format, or the stereo format produced above when downmix is also active, so
+    // the two stages chain as multi-channel -> stereo -> resampled stereo.
+    WAVEFORMATEXTENSIBLE resample_storage = {};
+    if (auto target_rate = hooks::audio::Resampler::resolve(device_format)) {
+        const uint32_t src_rate = device_format->nSamplesPerSec;
+        this->resample.setup(device_format, &resample_storage, *target_rate);
+        device_format = reinterpret_cast<const WAVEFORMATEX *>(&resample_storage);
+        log_info("audio::wasapi", "resample enabled: {} Hz -> {} Hz{}",
+                src_rate, *target_rate, this->downmix.enabled ? " (after downmix)" : "");
+    }
+
     // verbose output
     log_info("audio::wasapi", "IAudioClient::Initialize hook hit");
-    log_info("audio::wasapi", "... ShareMode         : {}", share_mode_str(ShareMode));
-    log_info("audio::wasapi", "... StreamFlags       : {}", stream_flags_str(StreamFlags));
-    log_info("audio::wasapi", "... hnsBufferDuration : {}", hnsBufferDuration);
-    log_info("audio::wasapi", "... hnsPeriodicity    : {}", hnsPeriodicity);
-    print_format(pFormat);
+    print_format(ShareMode, StreamFlags, hnsBufferDuration, hnsPeriodicity, device_format);
 
     if (this->backend) {
         SAFE_CALL("AudioBackend", "on_initialize", this->backend->on_initialize(
@@ -163,27 +208,67 @@ HRESULT STDMETHODCALLTYPE WrappedIAudioClient::Initialize(
                 AudioSessionGuid));
 
         log_info("audio::wasapi", "AudioBackend::on_initialize call finished");
-        log_info("audio::wasapi", "... ShareMode         : {}", share_mode_str(ShareMode));
-        log_info("audio::wasapi", "... StreamFlags       : {}", stream_flags_str(StreamFlags));
-        log_info("audio::wasapi", "... hnsBufferDuration : {}", hnsBufferDuration);
-        log_info("audio::wasapi", "... hnsPeriodicity    : {}", hnsPeriodicity);
-        print_format(pFormat);
+        print_format(ShareMode, StreamFlags, hnsBufferDuration, hnsPeriodicity, pFormat);
     }
 
     // check for exclusive mode
     if (ShareMode == AUDCLNT_SHAREMODE_EXCLUSIVE) {
         this->exclusive_mode = true;
-        this->frame_size = pFormat->nChannels * (pFormat->wBitsPerSample / 8);
+        this->frame_size = device_format->nChannels * (device_format->wBitsPerSample / 8);
+
+        // optionally enlarge the exclusive buffer. games request a very small buffer (e.g. 3 ms)
+        // which some endpoints (notably NVIDIA HDMI/DP display audio) cannot service in time,
+        // underrunning mid-period and crackling. a larger buffer gives the device slack. exclusive
+        // mode requires periodicity == buffer_duration, so raise both together; the initialize
+        // paths below handle any required buffer-size realignment.
+        if (hooks::audio::EXCLUSIVE_BUFFER_MS.has_value()) {
+            const REFERENCE_TIME min_duration =
+                    (REFERENCE_TIME) hooks::audio::EXCLUSIVE_BUFFER_MS.value() * 10000;
+            if (hnsBufferDuration < min_duration) {
+                log_info("audio::wasapi",
+                        "raising exclusive buffer from {} hns to {} hns ({} ms)",
+                        hnsBufferDuration, min_duration, hooks::audio::EXCLUSIVE_BUFFER_MS.value());
+                hnsBufferDuration = min_duration;
+                if (hnsPeriodicity != 0) {
+                    hnsPeriodicity = min_duration;
+                }
+            }
+        }
     }
 
-    // call next
-    HRESULT ret = pReal->Initialize(
-            ShareMode,
-            StreamFlags,
-            hnsBufferDuration,
-            hnsPeriodicity,
-            pFormat,
-            AudioSessionGuid);
+    // call next. the resampler owns the device interaction whenever it is active (including when
+    // chained after the downmix), otherwise the downmix does, otherwise the device is opened
+    // directly.
+    HRESULT ret;
+    if (this->resample.enabled) {
+        ret = this->resample.initialize(
+                pReal,
+                ShareMode,
+                StreamFlags,
+                hnsBufferDuration,
+                hnsPeriodicity,
+                device_format,
+                AudioSessionGuid);
+    } else if (this->downmix.enabled) {
+        ret = this->downmix.initialize(
+                pReal,
+                ShareMode,
+                StreamFlags,
+                hnsBufferDuration,
+                hnsPeriodicity,
+                device_format,
+                AudioSessionGuid);
+    } else {
+        ret = initialize_with_alignment_retry(
+                pReal,
+                "audio::wasapi",
+                ShareMode,
+                StreamFlags,
+                hnsBufferDuration,
+                hnsPeriodicity,
+                device_format,
+                AudioSessionGuid);
+    }
 
     // check for failure
     if (FAILED(ret)) {
@@ -192,7 +277,8 @@ HRESULT STDMETHODCALLTYPE WrappedIAudioClient::Initialize(
     }
 
     log_info("audio::wasapi", "IAudioClient::Initialize success, hr={}", FMT_HRESULT(ret));
-    copy_wave_format(&hooks::audio::FORMAT, pFormat);
+    copy_wave_format(&hooks::audio::FORMAT, device_format);
+    copy_wave_format(&this->device_format, device_format);
 
     return ret;
 }
@@ -214,7 +300,15 @@ HRESULT STDMETHODCALLTYPE WrappedIAudioClient::GetBufferSize(UINT32 *pNumBufferF
         }
     }
 
-    CHECK_RESULT(pReal->GetBufferSize(pNumBufferFrames));
+    HRESULT ret = pReal->GetBufferSize(pNumBufferFrames);
+
+    // report the buffer size at the game's native rate; the real device buffer is at the
+    // resampled rate, so translate it back so the game paces its writes correctly.
+    if (SUCCEEDED(ret) && this->resample.enabled && pNumBufferFrames) {
+        *pNumBufferFrames = this->resample.frames_device_to_game(*pNumBufferFrames);
+    }
+
+    CHECK_RESULT(ret);
 }
 HRESULT STDMETHODCALLTYPE WrappedIAudioClient::GetStreamLatency(REFERENCE_TIME *phnsLatency) {
     static std::once_flag printed;
@@ -256,7 +350,15 @@ HRESULT STDMETHODCALLTYPE WrappedIAudioClient::GetCurrentPadding(UINT32 *pNumPad
         }
     }
 
-    CHECK_RESULT(pReal->GetCurrentPadding(pNumPaddingFrames));
+    HRESULT ret = pReal->GetCurrentPadding(pNumPaddingFrames);
+
+    // the device buffer is at the resampled rate; report padding at the game's native rate so the
+    // game's free-space calculation stays paced correctly.
+    if (SUCCEEDED(ret) && this->resample.enabled && pNumPaddingFrames) {
+        *pNumPaddingFrames = this->resample.padding_device_to_game(*pNumPaddingFrames);
+    }
+
+    CHECK_RESULT(ret);
 }
 HRESULT STDMETHODCALLTYPE WrappedIAudioClient::IsFormatSupported(
     AUDCLNT_SHAREMODE ShareMode,
@@ -274,6 +376,44 @@ HRESULT STDMETHODCALLTYPE WrappedIAudioClient::IsFormatSupported(
         fix_rec_format(const_cast<WAVEFORMATEX *>(pFormat));
     }
 
+    // log the format the game is asking about
+    log_info("audio::wasapi", "IAudioClient::IsFormatSupported hook hit");
+    print_format(ShareMode, pFormat);
+
+    // when downmixing, the real device is opened as stereo, so check whether the equivalent
+    // stereo format is supported instead of the multi-channel one. when resampling is also active
+    // it chains onto that stereo format, so check the resampled stereo format.
+    if (resolve_downmix(pFormat)) {
+        WAVEFORMATEXTENSIBLE stereo_storage = {};
+        hooks::audio::Downmix::make_stereo_format(pFormat, &stereo_storage);
+        const WAVEFORMATEX *check_format = reinterpret_cast<const WAVEFORMATEX *>(&stereo_storage);
+
+        WAVEFORMATEXTENSIBLE resample_storage = {};
+        if (auto target_rate = hooks::audio::Resampler::resolve(check_format)) {
+            hooks::audio::Resampler::make_device_format(check_format, &resample_storage, *target_rate);
+            check_format = reinterpret_cast<const WAVEFORMATEX *>(&resample_storage);
+        }
+
+        log_info("audio::wasapi", "... checking device format instead (after downmix/resample):");
+        print_format(check_format);
+
+        CHECK_RESULT(pReal->IsFormatSupported(ShareMode, check_format, ppClosestMatch));
+    } else if (games::gitadora::is_arena_model()) {
+        games::gitadora::fix_audio_channel_mask(const_cast<WAVEFORMATEX *>(pFormat));
+    } else if (auto target_rate = hooks::audio::Resampler::resolve(pFormat)) {
+
+        // when resampling, the real device is opened at the target rate, so check whether the
+        // equivalent format at that rate is supported instead of the game's native rate.
+        WAVEFORMATEXTENSIBLE resample_storage = {};
+        hooks::audio::Resampler::make_device_format(pFormat, &resample_storage, *target_rate);
+        const auto resample_format = reinterpret_cast<const WAVEFORMATEX *>(&resample_storage);
+
+        log_info("audio::wasapi", "... checking device format instead (after resample):");
+        print_format(resample_format);
+
+        CHECK_RESULT(pReal->IsFormatSupported(ShareMode, resample_format, ppClosestMatch));
+    }
+
     if (this->backend) {
         HRESULT ret = this->backend->on_is_format_supported(&ShareMode, pFormat, ppClosestMatch);
 
@@ -466,5 +606,6 @@ HRESULT STDMETHODCALLTYPE WrappedIAudioClient::InitializeSharedAudioStream(
 
     log_info("audio::wasapi", "IAudioClient3::InitializeSharedAudioStream success, hr={}", FMT_HRESULT(ret));
     copy_wave_format(&hooks::audio::FORMAT, pFormat);
+    copy_wave_format(&this->device_format, pFormat);
     return ret;
 }

src/spice2x/hooks/audio/backends/wasapi/audio_client.h

@@ -8,8 +8,10 @@
 #include "hooks/audio/audio_private.h"
 #include "util/logging.h"
 
-#include "audio_render_client.h"
+#include "downmix.h"
+#include "resample.h"
 
+#include "audio_render_client.h"
 // {1FBC8530-AF3E-4128-B418-115DE72F76B6}
 static const GUID IID_WrappedIAudioClient = {
     0x1fbc8530, 0xaf3e, 0x4128, { 0xb4, 0x18, 0x11, 0x5d, 0xe7, 0x2f, 0x76, 0xb6 }
@@ -93,4 +95,17 @@ struct WrappedIAudioClient : IAudioClient3 {
     AudioBackend *const backend;
     bool exclusive_mode = false;
     int frame_size = 0;
+
+    // the format the real device was opened with (after any downmix). used to scale the final
+    // output buffer for the volume boost.
+    WAVEFORMATEXTENSIBLE device_format = {};
+
+    // surround -> stereo downmix. the real device is opened as stereo while the game keeps
+    // writing multi-channel audio into a scratch buffer that we downmix in the render client.
+    hooks::audio::Downmix downmix;
+
+    // native-rate -> target-rate sample-rate conversion. the real device is opened at the target
+    // rate while the game keeps writing its native-rate audio into a scratch buffer that we
+    // resample in the render client.
+    hooks::audio::Resampler resample;
 };

src/spice2x/hooks/audio/backends/wasapi/audio_render_client.cpp

@@ -1,10 +1,71 @@
 #include "audio_render_client.h"
 
+#include <algorithm>
+#include <cmath>
+#include <cstdint>
+#include <cstring>
+
 #include "audio_client.h"
+#include "hooks/audio/audio.h"
+#include "util.h"
 #include "wasapi_private.h"
 
 const char CLASS_NAME[] = "WrappedIAudioRenderClient";
 
+// scale every sample of an interleaved device buffer by `gain`, clamped to the format's range.
+// supports 16/24/32-bit PCM and 32-bit float; other formats are left untouched.
+static void apply_gain(BYTE *buffer, UINT32 frames, const WAVEFORMATEXTENSIBLE &fmt, float gain) {
+    const WAVEFORMATEX &f = fmt.Format;
+    const size_t samples = (size_t) frames * f.nChannels;
+
+    bool is_float = is_ieee_float(&f);
+
+    if (is_float && f.wBitsPerSample == 32) {
+        auto p = reinterpret_cast<float *>(buffer);
+        for (size_t i = 0; i < samples; i++) {
+            p[i] = std::clamp(p[i] * gain, -1.0f, 1.0f);
+        }
+        return;
+    }
+
+    switch (f.wBitsPerSample) {
+        case 16: {
+            auto p = reinterpret_cast<int16_t *>(buffer);
+            for (size_t i = 0; i < samples; i++) {
+                p[i] = (int16_t) std::clamp((int) std::lround(p[i] * gain), -32768, 32767);
+            }
+            break;
+        }
+        case 24: {
+            // packed 24-bit little-endian
+            for (size_t i = 0; i < samples; i++) {
+                BYTE *s = buffer + i * 3;
+                int32_t v = s[0] | (s[1] << 8) | (s[2] << 16);
+                if (v & 0x800000) {
+                    v |= ~0xFFFFFF; // sign extend
+                }
+                int64_t scaled = std::clamp<int64_t>(
+                    std::llround((double) v * gain), -8388608, 8388607);
+                s[0] = scaled & 0xFF;
+                s[1] = (scaled >> 8) & 0xFF;
+                s[2] = (scaled >> 16) & 0xFF;
+            }
+            break;
+        }
+        case 32: {
+            auto p = reinterpret_cast<int32_t *>(buffer);
+            for (size_t i = 0; i < samples; i++) {
+                p[i] = (int32_t) std::clamp(
+                    std::llround((double) p[i] * gain),
+                    (long long) INT32_MIN, (long long) INT32_MAX);
+            }
+            break;
+        }
+        default:
+            break;
+    }
+}
+
 HRESULT STDMETHODCALLTYPE WrappedIAudioRenderClient::QueryInterface(REFIID riid, void **ppvObj) {
     if (ppvObj == nullptr) {
         return E_POINTER;
@@ -51,6 +112,25 @@ HRESULT STDMETHODCALLTYPE WrappedIAudioRenderClient::GetBuffer(UINT32 NumFramesR
         return S_OK;
     }
 
+    // downmix + resample chained: the game writes its multi-channel native-rate audio into the
+    // downmix scratch, which is downmixed to stereo and then resampled on release. size the
+    // resampler's (stereo) input scratch now and hand the game the multi-channel downmix scratch.
+    if (this->client->downmix.enabled && this->client->resample.enabled) {
+        BYTE *resample_scratch = nullptr;
+        this->client->resample.get_buffer(NumFramesRequested, &resample_scratch);
+        CHECK_RESULT(this->client->downmix.get_scratch(NumFramesRequested, ppData));
+
+    // surround downmix: reserve the real (stereo) device buffer, but hand the game a
+    // multi-channel scratch buffer that we downmix on release
+    } else if (this->client->downmix.enabled) {
+        CHECK_RESULT(this->client->downmix.get_buffer(pReal, NumFramesRequested, ppData));
+
+    // resample: hand the game a native-rate scratch buffer that we convert on release. the real
+    // device buffer is acquired in ReleaseBuffer once the converted frame count is known.
+    } else if (this->client->resample.enabled) {
+        CHECK_RESULT(this->client->resample.get_buffer(NumFramesRequested, ppData));
+    }
+
     // call original
     HRESULT ret = pReal->GetBuffer(NumFramesRequested, ppData);
 
@@ -75,14 +155,72 @@ HRESULT STDMETHODCALLTYPE WrappedIAudioRenderClient::ReleaseBuffer(UINT32 NumFra
         return S_OK;
     }
 
-    // fix for audio pop effect
-    if (this->buffers_to_mute > 0 && this->client->frame_size > 0) {
-
-        // zero out = mute
-        memset(this->audio_buffer, 0, NumFramesWritten * this->client->frame_size);
-
-        this->buffers_to_mute--;
+    // downmix + resample chained: downmix the game's multi-channel scratch into the resampler's
+    // stereo input scratch, then let the resampler convert and push it to the device. a silent
+    // buffer skips the downmix and feeds silence straight through.
+    if (this->client->downmix.enabled && this->client->resample.enabled) {
+        if ((dwFlags & AUDCLNT_BUFFERFLAGS_SILENT) == 0) {
+            this->client->downmix.downmix_into(
+                    this->client->resample.input_data(), NumFramesWritten);
+        }
+        return this->client->resample.flush(
+                pReal,
+                this->client->pReal,
+                NumFramesWritten,
+                dwFlags,
+                hooks::audio::VOLUME_BOOST);
     }
 
-    CHECK_RESULT(pReal->ReleaseBuffer(NumFramesWritten, dwFlags));
+    // resample: convert the game's native-rate scratch and push as many output frames as the
+    // device has room for, applying the volume boost to the converted output. handles acquiring
+    // and releasing the real device buffer itself.
+    if (this->client->resample.enabled) {
+        return this->client->resample.flush(
+                pReal,
+                this->client->pReal,
+                NumFramesWritten,
+                dwFlags,
+                hooks::audio::VOLUME_BOOST);
+    }
+
+    // resolve the real device buffer for whichever path produced the audio
+    BYTE *device_buffer;
+    if (this->client->downmix.enabled) {
+
+        // downmix the game's multi-channel scratch into the real stereo buffer held since GetBuffer
+        this->client->downmix.write_device_buffer(NumFramesWritten, dwFlags);
+        device_buffer = this->client->downmix.current_buffer();
+    } else {
+        device_buffer = this->audio_buffer;
+
+        // mute the first few buffers to avoid a startup pop
+        if (this->buffers_to_mute > 0 && this->client->frame_size > 0) {
+            memset(this->audio_buffer, 0, NumFramesWritten * this->client->frame_size);
+            this->buffers_to_mute--;
+        }
+    }
+
+    // boost the final output volume just before it reaches the device, layout-agnostic
+    if (hooks::audio::VOLUME_BOOST != 1.0f
+            && device_buffer != nullptr
+            && (dwFlags & AUDCLNT_BUFFERFLAGS_SILENT) == 0) {
+        static std::once_flag boost_printed;
+        std::call_once(boost_printed, []() {
+            log_info("audio::wasapi", "volume boost active: gain={}", hooks::audio::VOLUME_BOOST);
+        });
+        apply_gain(device_buffer, NumFramesWritten, this->client->device_format,
+                hooks::audio::VOLUME_BOOST);
+    }
+
+    HRESULT ret = pReal->ReleaseBuffer(NumFramesWritten, dwFlags);
+
+    if (this->client->downmix.enabled) {
+        this->client->downmix.buffer_released();
+    }
+
+    if (FAILED(ret)) {
+        PRINT_FAILED_RESULT(CLASS_NAME, __func__, ret);
+    }
+
+    return ret;
 }

src/spice2x/hooks/audio/backends/wasapi/downmix.cpp

@@ -0,0 +1,264 @@
+#include "downmix.h"
+
+#include <algorithm>
+#include <cmath>
+#include <cstdint>
+#include <cstring>
+
+#include <audioclient.h>
+#include <ks.h>
+#include <ksmedia.h>
+
+#include "util/logging.h"
+
+#include "util.h"
+
+namespace hooks::audio {
+
+    namespace {
+
+        constexpr float ATT_3DB = 0.70710678f;
+
+        // speakers routed to the left/right output; anything else (center) feeds both sides
+        constexpr DWORD LEFT_SPEAKERS = SPEAKER_FRONT_LEFT | SPEAKER_BACK_LEFT | SPEAKER_SIDE_LEFT
+            | SPEAKER_FRONT_LEFT_OF_CENTER | SPEAKER_TOP_FRONT_LEFT | SPEAKER_TOP_BACK_LEFT;
+        constexpr DWORD RIGHT_SPEAKERS = SPEAKER_FRONT_RIGHT | SPEAKER_BACK_RIGHT | SPEAKER_SIDE_RIGHT
+            | SPEAKER_FRONT_RIGHT_OF_CENTER | SPEAKER_TOP_FRONT_RIGHT | SPEAKER_TOP_BACK_RIGHT;
+
+        // the speaker mask is only present on WAVE_FORMAT_EXTENSIBLE formats
+        DWORD read_channel_mask(const WAVEFORMATEX *fmt) {
+            if (fmt->wFormatTag == WAVE_FORMAT_EXTENSIBLE
+                    && fmt->cbSize >= sizeof(WAVEFORMATEXTENSIBLE) - sizeof(WAVEFORMATEX)) {
+                return reinterpret_cast<const WAVEFORMATEXTENSIBLE *>(fmt)->dwChannelMask;
+            }
+            return 0;
+        }
+
+        // call visit(channel_index, speaker_bit) for each present speaker, in channel order
+        template <typename F>
+        void for_each_speaker(DWORD mask, int channels, F &&visit) {
+            int channel = 0;
+            for (int bit = 0; bit < 18 && channel < channels; bit++) {
+                const DWORD speaker = 1u << bit;
+                if (mask & speaker) {
+                    visit(channel++, speaker);
+                }
+            }
+        }
+    }
+
+    void Downmix::setup(const WAVEFORMATEX *game_format, WAVEFORMATEXTENSIBLE *stereo_out,
+            DownmixAlgorithm algorithm) {
+        this->enabled = true;
+        this->algorithm = algorithm;
+        this->bytes_per_sample = game_format->wBitsPerSample / 8;
+        this->game_frame_size = game_format->nChannels * this->bytes_per_sample;
+
+        this->is_float = is_ieee_float(game_format);
+
+        // supported: 16/24/32-bit integer PCM and 32-bit float; anything else mixes to silence
+        const bool supported = this->is_float
+            ? this->bytes_per_sample == 4
+            : (this->bytes_per_sample >= 2 && this->bytes_per_sample <= 4);
+        if (!supported) {
+            log_fatal(
+                "audio::downmix",
+                "unsupported sample format ({}-bit {}), downmix will output silence",
+                game_format->wBitsPerSample, this->is_float ? "float" : "int");
+        }
+
+        this->left_mix.clear();
+        this->right_mix.clear();
+        this->build_layout_mix(game_format);
+
+        make_stereo_format(game_format, stereo_out);
+    }
+
+    void Downmix::make_stereo_format(const WAVEFORMATEX *game_format,
+            WAVEFORMATEXTENSIBLE *stereo_out) {
+        const int bytes_per_sample = game_format->wBitsPerSample / 8;
+
+        memcpy(stereo_out, game_format, sizeof(WAVEFORMATEXTENSIBLE));
+        stereo_out->Format.nChannels = 2;
+        stereo_out->Format.nBlockAlign = 2 * bytes_per_sample;
+        stereo_out->Format.nAvgBytesPerSec =
+            game_format->nSamplesPerSec * stereo_out->Format.nBlockAlign;
+        stereo_out->dwChannelMask = SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT;
+    }
+
+    HRESULT Downmix::initialize(IAudioClient *real, AUDCLNT_SHAREMODE share_mode, DWORD stream_flags,
+            REFERENCE_TIME buffer_duration, REFERENCE_TIME periodicity,
+            const WAVEFORMATEX *device_format, LPCGUID session_guid) {
+
+        // the smaller stereo buffer can end up unaligned for the device when the game sized the
+        // duration for its larger multi-channel format; the helper recovers from that.
+        return initialize_with_alignment_retry(real, "audio::downmix", share_mode, stream_flags,
+                buffer_duration, periodicity, device_format, session_guid);
+    }
+
+    void Downmix::add_channel(int channel, DWORD speaker, float gain) {
+        if (speaker & LEFT_SPEAKERS) {
+            this->left_mix.push_back({ channel, gain });
+        } else if (speaker & RIGHT_SPEAKERS) {
+            this->right_mix.push_back({ channel, gain });
+        } else { // center: feed both sides
+            this->left_mix.push_back({ channel, gain });
+            this->right_mix.push_back({ channel, gain });
+        }
+    }
+
+    // AC-4 stereo downmix (ETSI TS 103 190-1): front pair at unity, everything else -3 dB, LFE dropped
+    void Downmix::build_ac4_mix(DWORD mask, int channels) {
+        for_each_speaker(mask, channels, [&](int ch, DWORD speaker) {
+            if (speaker == SPEAKER_LOW_FREQUENCY) {
+                return;
+            }
+            const bool front_pair = speaker & (SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT);
+            this->add_channel(ch, speaker, front_pair ? 1.0f : ATT_3DB);
+        });
+    }
+
+    // keep only the channels in `keep` (front/rear/side), each at unity gain
+    void Downmix::build_extract_mix(DWORD mask, int channels, DWORD keep) {
+        for_each_speaker(mask, channels, [&](int ch, DWORD speaker) {
+            if (speaker & keep) {
+                this->add_channel(ch, speaker, 1.0f);
+            }
+        });
+    }
+
+    // keep every channel (LFE dropped), then average each side so its gains sum to unity
+    void Downmix::build_normalize_mix(DWORD mask, int channels) {
+        for_each_speaker(mask, channels, [&](int ch, DWORD speaker) {
+            if (speaker != SPEAKER_LOW_FREQUENCY) {
+                this->add_channel(ch, speaker, 1.0f);
+            }
+        });
+
+        for (auto *mix : { &this->left_mix, &this->right_mix }) {
+            if (!mix->empty()) {
+                const float gain = 1.0f / mix->size();
+                for (auto &c : *mix) {
+                    c.gain = gain;
+                }
+            }
+        }
+    }
+
+    // fallback when no speaker mask is present: fold interleaved L/R pairs (even->left, odd->right)
+    void Downmix::build_pairs_mix(int channels, float gain) {
+        for (int ch = 0; ch < channels; ch++) {
+            (((ch & 1) == 0) ? this->left_mix : this->right_mix).push_back({ ch, gain });
+        }
+    }
+
+    void Downmix::build_layout_mix(const WAVEFORMATEX *game_format) {
+        const int channels = game_format->nChannels;
+        const DWORD mask = read_channel_mask(game_format);
+
+        // without a mask the layout is unknown: extract/normalize have nothing to act on, so all
+        // algorithms fall back to folding L/R pairs (AC-4 still attenuates by -3 dB)
+        if (mask == 0) {
+            this->build_pairs_mix(channels,
+                this->algorithm == DownmixAlgorithm::AC4 ? ATT_3DB : 1.0f);
+            return;
+        }
+
+        switch (this->algorithm) {
+            case DownmixAlgorithm::FrontOnly:
+                this->build_extract_mix(mask, channels,
+                    SPEAKER_FRONT_LEFT | SPEAKER_FRONT_RIGHT);
+                break;
+            case DownmixAlgorithm::RearOnly:
+                this->build_extract_mix(mask, channels,
+                    SPEAKER_BACK_LEFT | SPEAKER_BACK_RIGHT | SPEAKER_BACK_CENTER);
+                break;
+            case DownmixAlgorithm::SideOnly:
+                this->build_extract_mix(mask, channels,
+                    SPEAKER_SIDE_LEFT | SPEAKER_SIDE_RIGHT);
+                break;
+            case DownmixAlgorithm::Normalize:
+                this->build_normalize_mix(mask, channels);
+                break;
+            case DownmixAlgorithm::AC4:
+                this->build_ac4_mix(mask, channels);
+                break;
+        }
+    }
+
+    void Downmix::process(BYTE *dst, const BYTE *src, UINT32 frames) const {
+        const int bps = this->bytes_per_sample;
+        const int src_stride = this->game_frame_size;
+        const int dst_stride = 2 * bps;
+
+        if (dst == nullptr || src == nullptr || bps <= 0) {
+            return;
+        }
+
+        // sum each speaker's source channels into the matching stereo output
+        for (UINT32 i = 0; i < frames; i++) {
+            const BYTE *in = src + (size_t) i * src_stride;
+            BYTE *out = dst + (size_t) i * dst_stride;
+
+            float left = 0.0f;
+            float right = 0.0f;
+            for (const auto &c : this->left_mix) {
+                left += read_sample(in + c.channel * bps, bps, this->is_float) * c.gain;
+            }
+            for (const auto &c : this->right_mix) {
+                right += read_sample(in + c.channel * bps, bps, this->is_float) * c.gain;
+            }
+            write_sample(out, bps, this->is_float, left);
+            write_sample(out + bps, bps, this->is_float, right);
+        }
+    }
+
+    HRESULT Downmix::get_buffer(IAudioRenderClient *real, UINT32 frames, BYTE **ppData) {
+        const size_t needed = (size_t) frames * this->game_frame_size;
+        if (this->scratch.size() < needed) {
+            this->scratch.resize(needed);
+        }
+
+        HRESULT ret = real->GetBuffer(frames, &this->device_buffer);
+        if (FAILED(ret)) {
+            this->device_buffer = nullptr;
+            return ret;
+        }
+
+        *ppData = this->scratch.data();
+
+        return S_OK;
+    }
+
+    HRESULT Downmix::get_scratch(UINT32 frames, BYTE **ppData) {
+        const size_t needed = (size_t) frames * this->game_frame_size;
+        if (this->scratch.size() < needed) {
+            this->scratch.resize(needed);
+        }
+
+        *ppData = this->scratch.data();
+
+        return S_OK;
+    }
+
+    void Downmix::downmix_into(BYTE *dst, UINT32 frames) const {
+        this->process(dst, this->scratch.data(), frames);
+    }
+
+    void Downmix::write_device_buffer(UINT32 frames, DWORD flags) {
+        const int bps = this->bytes_per_sample;
+        const int dst_stride = 2 * bps;
+
+        if (this->device_buffer == nullptr || frames == 0 || bps <= 0) {
+            return;
+        }
+
+        // mute the first few buffers to avoid a pop on stream start
+        if (this->buffers_to_mute > 0) {
+            memset(this->device_buffer, 0, (size_t) frames * dst_stride);
+            this->buffers_to_mute--;
+        } else if ((flags & AUDCLNT_BUFFERFLAGS_SILENT) == 0) {
+            this->process(this->device_buffer, this->scratch.data(), frames);
+        }
+    }
+}

src/spice2x/hooks/audio/backends/wasapi/downmix.h

@@ -0,0 +1,150 @@
+#pragma once
+
+#include <optional>
+#include <vector>
+
+#include <windows.h>
+#include <mmreg.h>
+#include <audioclient.h>
+
+#include "hooks/audio/audio.h"
+
+struct IAudioClient;
+struct IAudioRenderClient;
+
+namespace hooks::audio {
+
+    // Generic WASAPI surround-to-stereo downmix. The real device is opened in stereo while the
+    // game keeps writing its native multi-channel audio into a scratch buffer; on release that
+    // buffer is mixed down into the two front channels.
+    //
+    // The mix is derived from the source format's speaker mask according to the selected
+    // DownmixAlgorithm:
+    //   FrontOnly / RearOnly / SideOnly - keep only that group of channels, routed to their side
+    //   AC4       - AC-4 stereo downmix coefficients (ETSI TS 103 190-1 §6.2.17): front left/right
+    //               pass at 0 dB, center and surrounds fold in at -3 dB, LFE dropped
+    //   Normalize - every channel folded in (center to both sides) with each output side averaged
+    //               so its channels are equally loud, LFE dropped
+    struct Downmix {
+
+        // a source channel routed into one output speaker at the given gain
+        struct Contribution {
+            int channel;
+            float gain;
+        };
+
+        // map an option value (front/rear/side/ac4/normalize) to its algorithm.
+        static std::optional<DownmixAlgorithm> name_to_algorithm(const char *value) {
+            if (_stricmp(value, "front") == 0) {
+                return DownmixAlgorithm::FrontOnly;
+            } else if (_stricmp(value, "rear") == 0) {
+                return DownmixAlgorithm::RearOnly;
+            } else if (_stricmp(value, "side") == 0) {
+                return DownmixAlgorithm::SideOnly;
+            } else if (_stricmp(value, "ac4") == 0) {
+                return DownmixAlgorithm::AC4;
+            } else if (_stricmp(value, "normalize") == 0) {
+                return DownmixAlgorithm::Normalize;
+            }
+
+            return std::nullopt;
+        }
+
+        // human-readable name of an algorithm, for logging.
+        static const char *algorithm_name(DownmixAlgorithm algorithm) {
+            switch (algorithm) {
+                case DownmixAlgorithm::FrontOnly: return "front";
+                case DownmixAlgorithm::RearOnly: return "rear";
+                case DownmixAlgorithm::SideOnly: return "side";
+                case DownmixAlgorithm::AC4: return "ac4";
+                case DownmixAlgorithm::Normalize: return "normalize";
+                default: return "unknown";
+            }
+        }
+
+        // whether the downmix is active for the current stream
+        bool enabled = false;
+
+        // algorithm used to fold the multi-channel audio into stereo
+        DownmixAlgorithm algorithm = DownmixAlgorithm::AC4;
+
+        // size in bytes of one frame of the game's multi-channel format
+        int game_frame_size = 0;
+
+        // size in bytes of a single sample (per channel)
+        int bytes_per_sample = 0;
+
+        // whether samples are IEEE floating point rather than integer PCM
+        bool is_float = false;
+
+        // enable the downmix for the given game format and fill stereo_out with the equivalent
+        // stereo format to open the real device with.
+        void setup(const WAVEFORMATEX *game_format, WAVEFORMATEXTENSIBLE *stereo_out,
+                   DownmixAlgorithm algorithm);
+
+        // build the stereo format equivalent to game_format (same sample rate and bit depth).
+        static void make_stereo_format(const WAVEFORMATEX *game_format,
+                                       WAVEFORMATEXTENSIBLE *stereo_out);
+
+        // initialize the real device with the stereo format. downmixing reduces the channel count,
+        // shrinking the buffer's byte size, so the duration the game sized for its multi-channel
+        // format can leave the smaller stereo buffer unaligned. on AUDCLNT_E_BUFFER_SIZE_NOT_ALIGNED
+        // this performs the standard WASAPI realignment and retries.
+        HRESULT initialize(IAudioClient *real, AUDCLNT_SHAREMODE share_mode, DWORD stream_flags,
+                           REFERENCE_TIME buffer_duration, REFERENCE_TIME periodicity,
+                           const WAVEFORMATEX *device_format, LPCGUID session_guid);
+
+        // mix `frames` frames of multi-channel `src` down into stereo `dst`.
+        void process(BYTE *dst, const BYTE *src, UINT32 frames) const;
+
+        // grab the real stereo device buffer and hand the game the scratch buffer to write into.
+        HRESULT get_buffer(IAudioRenderClient *real, UINT32 frames, BYTE **ppData);
+
+        // size the scratch and hand it to the game without acquiring a device buffer. used when a
+        // later stage (the resampler) owns the device interaction.
+        HRESULT get_scratch(UINT32 frames, BYTE **ppData);
+
+        // downmix the scratch the game wrote into the caller's stereo buffer, without touching the
+        // device. used to feed the resampler when the two stages are chained.
+        void downmix_into(BYTE *dst, UINT32 frames) const;
+
+        // mix the scratch buffer into the stereo device buffer held since get_buffer. the caller
+        // owns releasing the device buffer afterwards (see current_buffer / buffer_released).
+        void write_device_buffer(UINT32 frames, DWORD flags);
+
+        // the real device buffer currently held, or null.
+        BYTE *current_buffer() const { return this->device_buffer; }
+
+        // forget the held device buffer once the caller has released it.
+        void buffer_released() { this->device_buffer = nullptr; }
+
+    private:
+
+        // build the mix from the source speaker layout for the selected algorithm
+        void build_layout_mix(const WAVEFORMATEX *game_format);
+
+        // per-algorithm builders, each filling left_mix / right_mix from the speaker mask
+        void build_ac4_mix(DWORD mask, int channels);
+        void build_extract_mix(DWORD mask, int channels, DWORD keep);
+        void build_normalize_mix(DWORD mask, int channels);
+
+        // fallback for streams without a speaker mask: fold interleaved L/R pairs at `gain`
+        void build_pairs_mix(int channels, float gain);
+
+        // append one source channel to the output side(s) matching its speaker, at `gain`
+        void add_channel(int channel, DWORD speaker, float gain);
+
+        // source channels summed into each output speaker
+        std::vector<Contribution> left_mix;
+        std::vector<Contribution> right_mix;
+
+        // buffer the game writes its multi-channel audio into between get/release
+        std::vector<BYTE> scratch;
+
+        // the real stereo device buffer currently held, or null
+        BYTE *device_buffer = nullptr;
+
+        // leading buffers to silence to avoid a pop on stream start
+        int buffers_to_mute = 16;
+    };
+}

src/spice2x/hooks/audio/backends/wasapi/dummy_audio_client.cpp

@@ -73,11 +73,7 @@ HRESULT STDMETHODCALLTYPE DummyIAudioClient::Initialize(
 
     // verbose output
     log_info("audio::wasapi", "IAudioClient::Initialize hook hit");
-    log_info("audio::wasapi", "... ShareMode         : {}", share_mode_str(ShareMode));
-    log_info("audio::wasapi", "... StreamFlags       : {}", stream_flags_str(StreamFlags));
-    log_info("audio::wasapi", "... hnsBufferDuration : {}", hnsBufferDuration);
-    log_info("audio::wasapi", "... hnsPeriodicity    : {}", hnsPeriodicity);
-    print_format(pFormat);
+    print_format(ShareMode, StreamFlags, hnsBufferDuration, hnsPeriodicity, pFormat);
 
     CHECK_RESULT(this->backend->on_initialize(
         &ShareMode,

src/spice2x/hooks/audio/backends/wasapi/resample.cpp

@@ -0,0 +1,437 @@
+#include "resample.h"
+
+#include <algorithm>
+#include <cmath>
+#include <cstdint>
+#include <cstring>
+#include <mutex>
+
+#include <audioclient.h>
+
+#include "util/logging.h"
+
+#include "util.h"
+
+namespace hooks::audio {
+
+    namespace {
+
+        constexpr double PI = 3.14159265358979323846;
+
+        // normalized sinc: sin(pi*x) / (pi*x), with the removable singularity at 0 filled in
+        inline double sinc(double x) {
+            if (x == 0.0) {
+                return 1.0;
+            }
+            const double px = PI * x;
+            return std::sin(px) / px;
+        }
+
+        // Blackman window across the kernel radius; zero at +/- radius
+        inline double blackman(double x, double radius) {
+            const double n = (x + radius) / (2.0 * radius);
+            if (n <= 0.0 || n >= 1.0) {
+                return 0.0;
+            }
+            return 0.42 - 0.5 * std::cos(2.0 * PI * n) + 0.08 * std::cos(4.0 * PI * n);
+        }
+    }
+
+    std::optional<uint32_t> Resampler::resolve(const WAVEFORMATEX *game_format) {
+        if (game_format == nullptr || !RESAMPLE_RATE.has_value()) {
+            return std::nullopt;
+        }
+        if (game_format->nSamplesPerSec == 0
+                || game_format->nSamplesPerSec == RESAMPLE_RATE.value()) {
+            return std::nullopt;
+        }
+        return RESAMPLE_RATE;
+    }
+
+    void Resampler::setup(const WAVEFORMATEX *game_format, WAVEFORMATEXTENSIBLE *device_out,
+            uint32_t target_rate) {
+        this->enabled = true;
+        this->channels = game_format->nChannels;
+        this->bytes_per_sample = game_format->wBitsPerSample / 8;
+        this->game_frame_size = this->channels * this->bytes_per_sample;
+
+        this->is_float = is_ieee_float(game_format);
+
+        const bool supported = this->is_float
+            ? this->bytes_per_sample == 4
+            : (this->bytes_per_sample >= 2 && this->bytes_per_sample <= 4);
+        if (!supported) {
+            log_fatal(
+                "audio::resample",
+                "unsupported sample format ({}-bit {}) for -resample",
+                game_format->wBitsPerSample, this->is_float ? "float" : "int");
+        }
+
+        this->src_rate = game_format->nSamplesPerSec;
+        this->dst_rate = target_rate;
+
+        // anti-alias cutoff: full bandwidth when upsampling, scaled down when decimating
+        this->cutoff = std::min(1.0, (double) this->dst_rate / (double) this->src_rate);
+        this->half_taps = 16;
+
+        // precompute the windowed-sinc kernel now that cutoff is known
+        this->build_kernel();
+
+        // prime the queue with half a window of silence so the first outputs have left history
+        this->in_queue.assign((size_t) this->half_taps * this->channels, 0.0f);
+        this->in_pos = this->half_taps;
+
+        this->make_device_format(game_format, device_out, target_rate);
+    }
+
+    void Resampler::make_device_format(const WAVEFORMATEX *game_format,
+            WAVEFORMATEXTENSIBLE *device_out, uint32_t target_rate) {
+        const size_t src_size = sizeof(WAVEFORMATEX) + game_format->cbSize;
+
+        memset(device_out, 0, sizeof(WAVEFORMATEXTENSIBLE));
+        memcpy(device_out, game_format, std::min(src_size, sizeof(WAVEFORMATEXTENSIBLE)));
+
+        device_out->Format.nSamplesPerSec = target_rate;
+        device_out->Format.nAvgBytesPerSec = target_rate * device_out->Format.nBlockAlign;
+    }
+
+    HRESULT Resampler::initialize(IAudioClient *real, AUDCLNT_SHAREMODE share_mode,
+            DWORD stream_flags, REFERENCE_TIME buffer_duration, REFERENCE_TIME periodicity,
+            const WAVEFORMATEX *device_format, LPCGUID session_guid) {
+
+        // the resampler bypasses the OS mixer and talks to the device directly, so it only makes
+        // sense (and only works) for exclusive streams. shared streams are already resampled by
+        // the Windows audio engine, so refuse loudly rather than silently doing nothing.
+        if (share_mode != AUDCLNT_SHAREMODE_EXCLUSIVE) {
+            log_fatal("audio::resample",
+                    "-resample requires WASAPI exclusive mode, but this stream is shared "
+                    "(Windows already resamples shared streams)");
+        }
+
+        // record the pacing model. event-driven streams fill the whole device buffer each period
+        // (produce_exact); timer-driven streams poll padding and write variable partial chunks, so
+        // they drain the pending output to the device's free space each call (flush_timer).
+        this->event_driven = (stream_flags & AUDCLNT_STREAMFLAGS_EVENTCALLBACK) != 0;
+
+        return initialize_with_alignment_retry(real, "audio::resample", share_mode, stream_flags,
+                buffer_duration, periodicity, device_format, session_guid);
+    }
+
+    UINT32 Resampler::frames_device_to_game(UINT32 device_frames) const {
+        if (this->dst_rate == 0) {
+            return device_frames;
+        }
+        // round down so the game never believes it has more room than the device can hold
+        return (UINT32) (((double) device_frames * this->src_rate) / this->dst_rate);
+    }
+
+    UINT32 Resampler::padding_device_to_game(UINT32 device_padding) const {
+        if (this->dst_rate == 0) {
+            return device_padding;
+        }
+        // round up so the reported free space stays conservative
+        return (UINT32) std::ceil(((double) device_padding * this->src_rate) / this->dst_rate);
+    }
+
+    HRESULT Resampler::get_buffer(UINT32 frames, BYTE **ppData) {
+        const size_t needed = (size_t) frames * this->game_frame_size;
+        if (this->scratch.size() < needed) {
+            this->scratch.resize(needed);
+        }
+
+        *ppData = this->scratch.data();
+
+        return S_OK;
+    }
+
+    void Resampler::enqueue_input(UINT32 frames, bool silent) {
+        const int bps = this->bytes_per_sample;
+        const int ch = this->channels;
+        const size_t base = this->in_queue.size();
+
+        this->in_queue.resize(base + (size_t) frames * ch);
+
+        if (silent || bps <= 0 || ch <= 0) {
+            std::fill(this->in_queue.begin() + base, this->in_queue.end(), 0.0f);
+            return;
+        }
+
+        const BYTE *src = this->scratch.data();
+        for (UINT32 f = 0; f < frames; f++) {
+            for (int c = 0; c < ch; c++) {
+                const size_t s = (size_t) f * ch + c;
+                this->in_queue[base + s] = read_sample(src + s * bps, bps, this->is_float);
+            }
+        }
+    }
+
+    void Resampler::build_kernel() {
+        const int taps = 2 * this->half_taps;
+        const int phases = this->kernel_phases;
+        const double cut = this->cutoff;
+        const double radius = (double) this->half_taps;
+
+        // one extra row at frac == 1.0 so emit_frame can interpolate against row p + 1 safely
+        this->kernel_table.resize((size_t) (phases + 1) * taps);
+
+        for (int p = 0; p <= phases; p++) {
+            const double frac = (double) p / (double) phases;
+            for (int k = 0; k < taps; k++) {
+                // tap k maps to input offset t = k - (half_taps - 1), matching emit_frame
+                const double x = frac - (double) (k - (this->half_taps - 1));
+                this->kernel_table[(size_t) p * taps + k] =
+                        (float) (cut * sinc(cut * x) * blackman(x, radius));
+            }
+        }
+    }
+
+    void Resampler::emit_frame() {
+        const int ch = this->channels;
+        const int radius = this->half_taps;
+        const int taps = 2 * radius;
+        const long avail = (long) (this->in_queue.size() / ch);
+        const long center = (long) std::floor(this->in_pos);
+
+        // pick the two kernel rows bracketing this fractional position and the blend between them
+        const double frac = this->in_pos - (double) center;
+        const double fp = frac * (double) this->kernel_phases;
+        const int p0 = (int) fp;
+        const float blend = (float) (fp - (double) p0);
+        const float *row0 = &this->kernel_table[(size_t) p0 * taps];
+        const float *row1 = &this->kernel_table[(size_t) (p0 + 1) * taps];
+
+        // base input index for tap 0 (t = -(radius - 1))
+        const long base = center - (radius - 1);
+
+        for (int c = 0; c < ch; c++) {
+            double acc = 0.0;
+            for (int k = 0; k < taps; k++) {
+                const long idx = base + k;
+                if (idx < 0 || idx >= avail) {
+                    continue;
+                }
+                const float w = row0[k] + blend * (row1[k] - row0[k]);
+                acc += (double) this->in_queue[(size_t) idx * ch + c] * w;
+            }
+            this->out_float.push_back((float) acc);
+        }
+    }
+
+    void Resampler::drop_consumed() {
+        const int ch = this->channels;
+        const long drop = (long) std::floor(this->in_pos) - this->half_taps;
+        if (drop > 0) {
+            const size_t drop_samples = (size_t) drop * ch;
+            if (drop_samples <= this->in_queue.size()) {
+                this->in_queue.erase(this->in_queue.begin(),
+                        this->in_queue.begin() + drop_samples);
+                this->in_pos -= drop;
+            }
+        }
+    }
+
+    UINT32 Resampler::produce_exact(UINT32 out_frames) {
+        const int ch = this->channels;
+        this->out_float.clear();
+        if (ch <= 0 || out_frames == 0) {
+            return 0;
+        }
+        this->out_float.reserve((size_t) out_frames * ch);
+
+        // resample ratio. drive it from the buffer size actually advertised to the game rather
+        // than the nominal src/dst ratio: GetBufferSize reports floor(dev_buf * src/dst) game
+        // frames, so the game only ever delivers that many input frames per device period.
+        // consuming at the nominal ratio would eat slightly more input than arrives on any device
+        // where dev_buf * src/dst is non-integer (e.g. 144 -> 132.3, floored to 132), slowly
+        // draining the queue until it underruns to permanent silence. using the advertised integer
+        // ratio keeps input and output exactly balanced; the resulting pitch error is below 0.3%
+        // and inaudible, and it collapses to the exact ratio when the division is integer (160 ->
+        // 147 stays 147/160 = 44100/48000).
+        const double step = (double) this->frames_device_to_game(this->device_buffer_frames)
+                / (double) this->device_buffer_frames;
+
+        // input frames the block will touch: from in_pos through the right edge of the sinc kernel
+        // at the final output sample. if the queue is short of this, the kernel tail reads past the
+        // end and distorts every buffer, so buffer one extra block of input before the first output
+        // (emitting silence without consuming) to build a cushion the kernel can always reach into.
+        const long avail = (long) (this->in_queue.size() / ch);
+        const long need = (long) std::ceil(this->in_pos + step * (double) out_frames)
+                + this->half_taps;
+
+        if (this->priming) {
+            if (avail < need + (long) out_frames) {
+                this->out_float.assign((size_t) out_frames * ch, 0.0f);
+                return out_frames;
+            }
+            this->priming = false;
+        }
+
+        for (UINT32 o = 0; o < out_frames; o++) {
+            this->emit_frame();
+            this->in_pos += step;
+        }
+
+        this->drop_consumed();
+        return out_frames;
+    }
+
+    UINT32 Resampler::produce_variable() {
+        const int ch = this->channels;
+        if (ch <= 0) {
+            return 0;
+        }
+
+        // input frames consumed per output frame. timer-driven streams write variable partial
+        // chunks, so produce however many output frames the currently queued input can fully
+        // support and leave the rest for the next call; this keeps input and output balanced at
+        // the exact src/dst ratio over time without depending on the device buffer size.
+        const double step = (double) this->src_rate / (double) this->dst_rate;
+        const long avail = (long) (this->in_queue.size() / ch);
+
+        // emit only while the sinc kernel's right edge stays within the queued input. the kernel
+        // reaches from in_pos out to half_taps frames ahead, so stop once that would read past the
+        // end; the remaining input becomes the next block's lookahead.
+        UINT32 produced = 0;
+        while ((long) std::ceil(this->in_pos) + this->half_taps < avail) {
+            this->emit_frame();
+            this->in_pos += step;
+            produced++;
+        }
+
+        this->drop_consumed();
+        return produced;
+    }
+
+    void Resampler::write_output(BYTE *dst, UINT32 frames, float gain) const {
+        const int bps = this->bytes_per_sample;
+        const int ch = this->channels;
+        const size_t count = (size_t) frames * ch;
+
+        for (size_t i = 0; i < count; i++) {
+            write_sample(dst + i * bps, bps, this->is_float, this->out_float[i] * gain);
+        }
+    }
+
+    HRESULT Resampler::flush(IAudioRenderClient *real, IAudioClient *client, UINT32 frames,
+            DWORD flags, float boost) {
+        if (!this->enabled) {
+            return S_OK;
+        }
+
+        // cache the device buffer size once
+        if (this->device_buffer_frames == 0) {
+            client->GetBufferSize(&this->device_buffer_frames);
+        }
+        if (this->device_buffer_frames == 0) {
+            return S_OK;
+        }
+
+        const bool silent = (flags & AUDCLNT_BUFFERFLAGS_SILENT) != 0;
+        this->enqueue_input(frames, silent);
+
+        // confirm once that conversion actually started producing output
+        static std::once_flag active_printed;
+        std::call_once(active_printed, [this]() {
+            log_info("audio::resample", "resample active: {} Hz -> {} Hz ({} ch, {})",
+                    this->src_rate, this->dst_rate, this->channels,
+                    this->event_driven ? "event-driven" : "timer-driven");
+        });
+
+        // the boost is applied here (inside write_output) rather than in the standard ReleaseBuffer
+        // path, so log it once for parity with that path's "volume boost active" line.
+        if (boost != 1.0f) {
+            static std::once_flag boost_printed;
+            std::call_once(boost_printed, [boost]() {
+                log_info("audio::resample", "volume boost active (resample): gain={}", boost);
+            });
+        }
+
+        return this->event_driven
+                ? this->flush_event(real, boost)
+                : this->flush_timer(real, client, boost);
+    }
+
+    HRESULT Resampler::flush_event(IAudioRenderClient *real, float boost) {
+
+        // event-driven exclusive streams must hand the device a full buffer every period and may
+        // not push partial counts. resample the whole input block into exactly the device buffer
+        // size.
+        const UINT32 produced = this->produce_exact(this->device_buffer_frames);
+        if (produced == 0) {
+            return S_OK;
+        }
+
+        BYTE *dev = nullptr;
+        HRESULT ret = real->GetBuffer(produced, &dev);
+        if (FAILED(ret) || dev == nullptr) {
+            return ret;
+        }
+
+        // mute the first few buffers to avoid a pop on stream start
+        float gain = boost;
+        if (this->buffers_to_mute > 0) {
+            gain = 0.0f;
+            this->buffers_to_mute--;
+        }
+
+        this->write_output(dev, produced, gain);
+
+        return real->ReleaseBuffer(produced, 0);
+    }
+
+    HRESULT Resampler::flush_timer(IAudioRenderClient *real, IAudioClient *client, float boost) {
+
+        // convert everything currently queued into the pending output FIFO (out_float). timer-
+        // driven games write variable partial chunks, so produce only what the queued input can
+        // fully support and keep the remainder for the next call.
+        this->produce_variable();
+
+        const int ch = this->channels;
+        if (ch <= 0) {
+            return S_OK;
+        }
+
+        const UINT32 pending = (UINT32) (this->out_float.size() / ch);
+        if (pending == 0) {
+            return S_OK;
+        }
+
+        // push as many frames as the device currently has free, keeping the rest queued for the
+        // next call. timer-driven games poll padding and write whenever there is room, so matching
+        // the device's free space here avoids overflowing the ring while staying device-paced.
+        UINT32 padding = 0;
+        if (FAILED(client->GetCurrentPadding(&padding))) {
+            return S_OK;
+        }
+        const UINT32 device_free = this->device_buffer_frames > padding
+                ? this->device_buffer_frames - padding
+                : 0;
+        if (device_free == 0) {
+            return S_OK;
+        }
+
+        const UINT32 to_write = std::min(pending, device_free);
+
+        BYTE *dev = nullptr;
+        HRESULT ret = real->GetBuffer(to_write, &dev);
+        if (FAILED(ret) || dev == nullptr) {
+            return ret;
+        }
+
+        // mute the first few buffers to avoid a pop on stream start
+        float gain = boost;
+        if (this->buffers_to_mute > 0) {
+            gain = 0.0f;
+            this->buffers_to_mute--;
+        }
+
+        this->write_output(dev, to_write, gain);
+        ret = real->ReleaseBuffer(to_write, 0);
+
+        // drop the frames just written from the front of the pending FIFO
+        this->out_float.erase(this->out_float.begin(),
+                this->out_float.begin() + (size_t) to_write * ch);
+
+        return ret;
+    }
+}

src/spice2x/hooks/audio/backends/wasapi/resample.h

@@ -0,0 +1,149 @@
+#pragma once
+
+#include <cstdint>
+#include <optional>
+#include <vector>
+
+#include <windows.h>
+#include <mmreg.h>
+#include <audioclient.h>
+
+#include "hooks/audio/audio.h"
+
+struct IAudioClient;
+struct IAudioRenderClient;
+
+namespace hooks::audio {
+
+    // Streaming sample-rate converter for the WASAPI render path. The real device is opened at the
+    // target rate while the game keeps writing its native-rate audio into a scratch buffer; on
+    // release that buffer is converted with a windowed-sinc kernel and pushed to the device.
+    // Channel count and sample format are preserved; only the sample rate changes.
+    //
+    // Frame counts differ between the two rates, so unlike the per-frame downmix this is stateful:
+    // a fractional read position and a window of input history carry across ReleaseBuffer calls,
+    // and the device buffer is only filled up to the space the device currently has free.
+    struct Resampler {
+
+        // whether the resampler is active for the current stream
+        bool enabled = false;
+
+        // whether the stream is event-driven (AUDCLNT_STREAMFLAGS_EVENTCALLBACK). timer-driven
+        // streams instead poll padding and write variable partial chunks, so they drain the
+        // pending output to the device's free space rather than pushing a full buffer per period.
+        bool event_driven = true;
+
+        // decide whether the stream should be resampled and to which rate. returns the target rate
+        // when RESAMPLE_RATE is set and differs from the game's rate, otherwise nullopt.
+        static std::optional<uint32_t> resolve(const WAVEFORMATEX *game_format);
+
+        // enable resampling for game_format and fill device_out with the equivalent format at the
+        // target rate to open the real device with.
+        void setup(const WAVEFORMATEX *game_format, WAVEFORMATEXTENSIBLE *device_out,
+                   uint32_t target_rate);
+
+        // build the device format equivalent to game_format at target_rate (same channels/depth).
+        static void make_device_format(const WAVEFORMATEX *game_format,
+                                       WAVEFORMATEXTENSIBLE *device_out, uint32_t target_rate);
+
+        // initialize the real device at the target rate, performing the standard WASAPI buffer
+        // realignment retry on AUDCLNT_E_BUFFER_SIZE_NOT_ALIGNED.
+        HRESULT initialize(IAudioClient *real, AUDCLNT_SHAREMODE share_mode, DWORD stream_flags,
+                           REFERENCE_TIME buffer_duration, REFERENCE_TIME periodicity,
+                           const WAVEFORMATEX *device_format, LPCGUID session_guid);
+
+        // translate a device-rate frame count to the equivalent game-rate count, so the buffer-size
+        // and padding values reported to the game stay paced at the game's native rate.
+        UINT32 frames_device_to_game(UINT32 device_frames) const;
+        UINT32 padding_device_to_game(UINT32 device_padding) const;
+
+        // hand the game a scratch buffer sized for `frames` of its native format to write into.
+        HRESULT get_buffer(UINT32 frames, BYTE **ppData);
+
+        // pointer to the input scratch (sized by get_buffer). when chained after the downmix, the
+        // downmix writes its stereo output here for the resampler to consume on the next flush.
+        BYTE *input_data() { return this->scratch.data(); }
+
+        // convert the `frames` the game wrote and push output to the real render client. `boost`
+        // is applied to the converted output. event-driven streams fill exactly one device buffer
+        // per period; timer-driven streams push as many converted frames as the device has free.
+        HRESULT flush(IAudioRenderClient *real, IAudioClient *client, UINT32 frames, DWORD flags,
+                      float boost);
+
+    private:
+
+        // append `frames` of the scratch buffer (native format), or silence, to the input queue
+        void enqueue_input(UINT32 frames, bool silent);
+
+        // event-driven path: produce exactly one full device buffer and push it.
+        HRESULT flush_event(IAudioRenderClient *real, float boost);
+
+        // timer-driven path: convert all queued input into the pending output FIFO, then push as
+        // many frames as the device currently has free, keeping the remainder for the next call.
+        HRESULT flush_timer(IAudioRenderClient *real, IAudioClient *client, float boost);
+
+        // produce exactly out_frames output frames using the fixed src/dst ratio. event-driven
+        // exclusive streams must fill the whole device buffer every period; a small input cushion
+        // is buffered first (see priming) so the sinc kernel always has lookahead.
+        UINT32 produce_exact(UINT32 out_frames);
+
+        // convert all input the kernel can fully support into the pending output FIFO (out_float),
+        // appending without clearing. returns the number of frames produced. used by the
+        // timer-driven path where output is drained to the device in device-paced chunks.
+        UINT32 produce_variable();
+
+        // convolve the windowed-sinc kernel at the current in_pos and append the resulting frame
+        // (one sample per channel) to out_float
+        void emit_frame();
+
+        // precompute the windowed-sinc kernel sampled at kernel_phases sub-sample positions, so
+        // emit_frame is a table lookup instead of recomputing sin/cos per tap (which is far too
+        // expensive to run per sample on the audio callback thread and causes underrun crackle).
+        void build_kernel();
+
+        // drop input frames that in_pos has advanced past, keeping a window of history for the
+        // next block's left context
+        void drop_consumed();
+
+        // convert the first `frames` of out_float to the device format, scaled by `gain`
+        void write_output(BYTE *dst, UINT32 frames, float gain) const;
+
+        // sample format of the stream
+        int channels = 0;
+        int bytes_per_sample = 0;
+        bool is_float = false;
+        int game_frame_size = 0;
+
+        uint32_t src_rate = 0;
+        uint32_t dst_rate = 0;
+
+        // sinc low-pass cutoff (1.0 when upsampling, dst/src when downsampling) and window radius
+        double cutoff = 1.0;
+        int half_taps = 16;
+
+        // precomputed kernel: (kernel_phases + 1) rows of 2*half_taps weights, indexed by the
+        // fractional sample position (linearly interpolated between adjacent rows in emit_frame)
+        std::vector<float> kernel_table;
+        int kernel_phases = 1024;
+
+        // interleaved float input queue and the fractional read position within it (in frames)
+        std::vector<float> in_queue;
+        double in_pos = 0.0;
+
+        // emit silence until a full block of input lookahead has accumulated, so the sinc kernel
+        // never reads past the end of the queue (which would distort the tail of every buffer)
+        bool priming = true;
+
+        // interleaved float scratch for produced output
+        std::vector<float> out_float;
+
+        // buffer the game writes its native-rate audio into between get_buffer / flush
+        std::vector<BYTE> scratch;
+
+        // cached device buffer size (frames); a full buffer is produced every period
+        UINT32 device_buffer_frames = 0;
+
+        // leading buffers to silence to avoid a pop on stream start
+        int buffers_to_mute = 16;
+    };
+}