diff options
Diffstat (limited to 'src')
121 files changed, 3175 insertions, 2014 deletions
diff --git a/src/audio_core/audio_core.cpp b/src/audio_core/audio_core.cpp index ba6acf28e..84f9c03a7 100644 --- a/src/audio_core/audio_core.cpp +++ b/src/audio_core/audio_core.cpp @@ -56,20 +56,8 @@ void AddAddressSpace(Kernel::VMManager& address_space) { } void SelectSink(std::string sink_id) { - auto iter = - std::find_if(g_sink_details.begin(), g_sink_details.end(), - [sink_id](const auto& sink_detail) { return sink_detail.id == sink_id; }); - - if (sink_id == "auto" || iter == g_sink_details.end()) { - if (sink_id != "auto") { - LOG_ERROR(Audio, "AudioCore::SelectSink given invalid sink_id %s", sink_id.c_str()); - } - // Auto-select. - // g_sink_details is ordered in terms of desirability, with the best choice at the front. - iter = g_sink_details.begin(); - } - - DSP::HLE::SetSink(iter->factory()); + const SinkDetails& sink_details = GetSinkDetails(sink_id); + DSP::HLE::SetSink(sink_details.factory()); } void EnableStretching(bool enable) { diff --git a/src/audio_core/null_sink.h b/src/audio_core/null_sink.h index e7668438c..c732926a2 100644 --- a/src/audio_core/null_sink.h +++ b/src/audio_core/null_sink.h @@ -23,6 +23,12 @@ public: size_t SamplesInQueue() const override { return 0; } + + void SetDevice(int device_id) override {} + + std::vector<std::string> GetDeviceList() const override { + return {}; + } }; } // namespace AudioCore diff --git a/src/audio_core/sdl2_sink.cpp b/src/audio_core/sdl2_sink.cpp index 4b66cd826..933c5f16d 100644 --- a/src/audio_core/sdl2_sink.cpp +++ b/src/audio_core/sdl2_sink.cpp @@ -4,12 +4,12 @@ #include <list> #include <numeric> -#include <vector> #include <SDL.h> #include "audio_core/audio_core.h" #include "audio_core/sdl2_sink.h" #include "common/assert.h" #include "common/logging/log.h" +#include "core/settings.h" namespace AudioCore { @@ -42,10 +42,24 @@ SDL2Sink::SDL2Sink() : impl(std::make_unique<Impl>()) { SDL_AudioSpec obtained_audiospec; SDL_zero(obtained_audiospec); - impl->audio_device_id = - SDL_OpenAudioDevice(nullptr, false, &desired_audiospec, &obtained_audiospec, 0); + int device_count = SDL_GetNumAudioDevices(0); + device_list.clear(); + for (int i = 0; i < device_count; ++i) { + device_list.push_back(SDL_GetAudioDeviceName(i, 0)); + } + + const char* device = nullptr; + + if (device_count >= 1 && Settings::values.audio_device_id != "auto" && + !Settings::values.audio_device_id.empty()) { + device = Settings::values.audio_device_id.c_str(); + } + + impl->audio_device_id = SDL_OpenAudioDevice(device, false, &desired_audiospec, + &obtained_audiospec, SDL_AUDIO_ALLOW_ANY_CHANGE); if (impl->audio_device_id <= 0) { - LOG_CRITICAL(Audio_Sink, "SDL_OpenAudioDevice failed with: %s", SDL_GetError()); + LOG_CRITICAL(Audio_Sink, "SDL_OpenAudioDevice failed with code %d for device \"%s\"", + impl->audio_device_id, Settings::values.audio_device_id.c_str()); return; } @@ -69,6 +83,10 @@ unsigned int SDL2Sink::GetNativeSampleRate() const { return impl->sample_rate; } +std::vector<std::string> SDL2Sink::GetDeviceList() const { + return device_list; +} + void SDL2Sink::EnqueueSamples(const s16* samples, size_t sample_count) { if (impl->audio_device_id <= 0) return; @@ -96,6 +114,10 @@ size_t SDL2Sink::SamplesInQueue() const { return total_size; } +void SDL2Sink::SetDevice(int device_id) { + this->device_id = device_id; +} + void SDL2Sink::Impl::Callback(void* impl_, u8* buffer, int buffer_size_in_bytes) { Impl* impl = reinterpret_cast<Impl*>(impl_); diff --git a/src/audio_core/sdl2_sink.h b/src/audio_core/sdl2_sink.h index ccd0f7c7e..bcc725369 100644 --- a/src/audio_core/sdl2_sink.h +++ b/src/audio_core/sdl2_sink.h @@ -21,9 +21,14 @@ public: size_t SamplesInQueue() const override; + std::vector<std::string> GetDeviceList() const override; + void SetDevice(int device_id) override; + private: struct Impl; std::unique_ptr<Impl> impl; + int device_id; + std::vector<std::string> device_list; }; } // namespace AudioCore diff --git a/src/audio_core/sink.h b/src/audio_core/sink.h index 08f3bab5b..558c8c0fe 100644 --- a/src/audio_core/sink.h +++ b/src/audio_core/sink.h @@ -31,6 +31,15 @@ public: /// Samples enqueued that have not been played yet. virtual std::size_t SamplesInQueue() const = 0; + + /** + * Sets the desired output device. + * @paran device_id Id of the desired device. + */ + virtual void SetDevice(int device_id) = 0; + + /// Returns the list of available devices. + virtual std::vector<std::string> GetDeviceList() const = 0; }; } // namespace diff --git a/src/audio_core/sink_details.cpp b/src/audio_core/sink_details.cpp index 95ccc9e9d..6972395af 100644 --- a/src/audio_core/sink_details.cpp +++ b/src/audio_core/sink_details.cpp @@ -2,6 +2,7 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include <algorithm> #include <memory> #include <vector> #include "audio_core/null_sink.h" @@ -9,6 +10,7 @@ #ifdef HAVE_SDL2 #include "audio_core/sdl2_sink.h" #endif +#include "common/logging/log.h" namespace AudioCore { @@ -20,4 +22,21 @@ const std::vector<SinkDetails> g_sink_details = { {"null", []() { return std::make_unique<NullSink>(); }}, }; +const SinkDetails& GetSinkDetails(std::string sink_id) { + auto iter = + std::find_if(g_sink_details.begin(), g_sink_details.end(), + [sink_id](const auto& sink_detail) { return sink_detail.id == sink_id; }); + + if (sink_id == "auto" || iter == g_sink_details.end()) { + if (sink_id != "auto") { + LOG_ERROR(Audio, "AudioCore::SelectSink given invalid sink_id %s", sink_id.c_str()); + } + // Auto-select. + // g_sink_details is ordered in terms of desirability, with the best choice at the front. + iter = g_sink_details.begin(); + } + + return *iter; +} + } // namespace AudioCore diff --git a/src/audio_core/sink_details.h b/src/audio_core/sink_details.h index 4b30cf835..9d3735171 100644 --- a/src/audio_core/sink_details.h +++ b/src/audio_core/sink_details.h @@ -24,4 +24,6 @@ struct SinkDetails { extern const std::vector<SinkDetails> g_sink_details; +const SinkDetails& GetSinkDetails(std::string sink_id); + } // namespace AudioCore diff --git a/src/citra/citra.cpp b/src/citra/citra.cpp index 99c096ac7..76f5caeb1 100644 --- a/src/citra/citra.cpp +++ b/src/citra/citra.cpp @@ -141,6 +141,26 @@ int main(int argc, char** argv) { case Core::System::ResultStatus::ErrorLoader: LOG_CRITICAL(Frontend, "Failed to load ROM!"); return -1; + case Core::System::ResultStatus::ErrorLoader_ErrorEncrypted: + LOG_CRITICAL(Frontend, "The game that you are trying to load must be decrypted before " + "being used with Citra. \n\n For more information on dumping and " + "decrypting games, please refer to: " + "https://citra-emu.org/wiki/Dumping-Game-Cartridges"); + return -1; + case Core::System::ResultStatus::ErrorLoader_ErrorInvalidFormat: + LOG_CRITICAL(Frontend, "Error while loading ROM: The ROM format is not supported."); + return -1; + case Core::System::ResultStatus::ErrorNotInitialized: + LOG_CRITICAL(Frontend, "CPUCore not initialized"); + return -1; + case Core::System::ResultStatus::ErrorSystemMode: + LOG_CRITICAL(Frontend, "Failed to determine system mode!"); + return -1; + case Core::System::ResultStatus::ErrorVideoCore: + LOG_CRITICAL(Frontend, "VideoCore not initialized"); + return -1; + case Core::System::ResultStatus::Success: + break; // Expected case } while (emu_window->IsOpen()) { diff --git a/src/citra/config.cpp b/src/citra/config.cpp index 2223bf0a0..fac1c9a0e 100644 --- a/src/citra/config.cpp +++ b/src/citra/config.cpp @@ -63,8 +63,8 @@ void Config::ReadValues() { // Renderer Settings::values.use_hw_renderer = sdl2_config->GetBoolean("Renderer", "use_hw_renderer", true); Settings::values.use_shader_jit = sdl2_config->GetBoolean("Renderer", "use_shader_jit", true); - Settings::values.use_scaled_resolution = - sdl2_config->GetBoolean("Renderer", "use_scaled_resolution", false); + Settings::values.resolution_factor = + (float)sdl2_config->GetReal("Renderer", "resolution_factor", 1.0); Settings::values.use_vsync = sdl2_config->GetBoolean("Renderer", "use_vsync", false); Settings::values.toggle_framelimit = sdl2_config->GetBoolean("Renderer", "toggle_framelimit", true); @@ -82,6 +82,7 @@ void Config::ReadValues() { Settings::values.sink_id = sdl2_config->Get("Audio", "output_engine", "auto"); Settings::values.enable_audio_stretching = sdl2_config->GetBoolean("Audio", "enable_audio_stretching", true); + Settings::values.audio_device_id = sdl2_config->Get("Audio", "output_device", "auto"); // Data Storage Settings::values.use_virtual_sd = @@ -89,7 +90,8 @@ void Config::ReadValues() { // System Settings::values.is_new_3ds = sdl2_config->GetBoolean("System", "is_new_3ds", false); - Settings::values.region_value = sdl2_config->GetInteger("System", "region_value", 1); + Settings::values.region_value = + sdl2_config->GetInteger("System", "region_value", Settings::REGION_VALUE_AUTO_SELECT); // Camera using namespace Service::CAM; diff --git a/src/citra/default_ini.h b/src/citra/default_ini.h index 618f1aeaa..435ba6f00 100644 --- a/src/citra/default_ini.h +++ b/src/citra/default_ini.h @@ -51,14 +51,21 @@ use_hw_renderer = # 0: Interpreter (slow), 1 (default): JIT (fast) use_shader_jit = -# Whether to use native 3DS screen resolution or to scale rendering resolution to the displayed screen size. -# 0 (default): Native, 1: Scaled -use_scaled_resolution = +# Resolution scale factor +# 0: Auto (scales resolution to window size), 1: Native 3DS screen resolution, Otherwise a scale +# factor for the 3DS resolution +resolution_factor = # Whether to enable V-Sync (caps the framerate at 60FPS) or not. # 0 (default): Off, 1: On use_vsync = +# The clear color for the renderer. What shows up on the sides of the bottom screen. +# Must be in range of 0.0-1.0. Defaults to 1.0 for all. +bg_red = +bg_blue = +bg_green = + [Layout] # Layout for the screen inside the render window. # 0 (default): Default Top Bottom Screen, 1: Single Screen Only, 2: Large Screen Small Screen @@ -73,12 +80,6 @@ toggle_framelimit = # 0 (default): Top Screen is prominent, 1: Bottom Screen is prominent swap_screen = -# The clear color for the renderer. What shows up on the sides of the bottom screen. -# Must be in range of 0.0-1.0. Defaults to 1.0 for all. -bg_red = -bg_blue = -bg_green = - [Audio] # Which audio output engine to use. # auto (default): Auto-select, null: No audio output, sdl2: SDL2 (if available) @@ -90,6 +91,10 @@ output_engine = # 0: No, 1 (default): Yes enable_audio_stretching = +# Which audio device to use. +# auto (default): Auto-select +output_device = + [Data Storage] # Whether to create a virtual SD card. # 1 (default): Yes, 0: No @@ -101,7 +106,7 @@ use_virtual_sd = is_new_3ds = # The system region that Citra will use during emulation -# 0: Japan, 1: USA (default), 2: Europe, 3: Australia, 4: China, 5: Korea, 6: Taiwan +# -1: Auto-select (default), 0: Japan, 1: USA, 2: Europe, 3: Australia, 4: China, 5: Korea, 6: Taiwan region_value = [Camera] diff --git a/src/citra/emu_window/emu_window_sdl2.cpp b/src/citra/emu_window/emu_window_sdl2.cpp index b0d82b670..81a3abe3f 100644 --- a/src/citra/emu_window/emu_window_sdl2.cpp +++ b/src/citra/emu_window/emu_window_sdl2.cpp @@ -19,16 +19,22 @@ void EmuWindow_SDL2::OnMouseMotion(s32 x, s32 y) { TouchMoved((unsigned)std::max(x, 0), (unsigned)std::max(y, 0)); + motion_emu->Tilt(x, y); } void EmuWindow_SDL2::OnMouseButton(u32 button, u8 state, s32 x, s32 y) { - if (button != SDL_BUTTON_LEFT) - return; - - if (state == SDL_PRESSED) { - TouchPressed((unsigned)std::max(x, 0), (unsigned)std::max(y, 0)); - } else { - TouchReleased(); + if (button == SDL_BUTTON_LEFT) { + if (state == SDL_PRESSED) { + TouchPressed((unsigned)std::max(x, 0), (unsigned)std::max(y, 0)); + } else { + TouchReleased(); + } + } else if (button == SDL_BUTTON_RIGHT) { + if (state == SDL_PRESSED) { + motion_emu->BeginTilt(x, y); + } else { + motion_emu->EndTilt(); + } } } @@ -54,6 +60,7 @@ EmuWindow_SDL2::EmuWindow_SDL2() { keyboard_id = KeyMap::NewDeviceId(); ReloadSetKeymaps(); + motion_emu = std::make_unique<Motion::MotionEmu>(*this); SDL_SetMainReady(); @@ -109,6 +116,7 @@ EmuWindow_SDL2::EmuWindow_SDL2() { EmuWindow_SDL2::~EmuWindow_SDL2() { SDL_GL_DeleteContext(gl_context); SDL_Quit(); + motion_emu = nullptr; } void EmuWindow_SDL2::SwapBuffers() { diff --git a/src/citra/emu_window/emu_window_sdl2.h b/src/citra/emu_window/emu_window_sdl2.h index c8cd919c6..b1cbf16d7 100644 --- a/src/citra/emu_window/emu_window_sdl2.h +++ b/src/citra/emu_window/emu_window_sdl2.h @@ -4,8 +4,10 @@ #pragma once +#include <memory> #include <utility> #include "core/frontend/emu_window.h" +#include "core/frontend/motion_emu.h" struct SDL_Window; @@ -61,4 +63,7 @@ private: /// Device id of keyboard for use with KeyMap int keyboard_id; + + /// Motion sensors emulation + std::unique_ptr<Motion::MotionEmu> motion_emu; }; diff --git a/src/citra_qt/CMakeLists.txt b/src/citra_qt/CMakeLists.txt index 93f1c339d..d4460bf01 100644 --- a/src/citra_qt/CMakeLists.txt +++ b/src/citra_qt/CMakeLists.txt @@ -14,7 +14,6 @@ set(SRCS debugger/graphics/graphics_tracing.cpp debugger/graphics/graphics_vertex_shader.cpp debugger/profiler.cpp - debugger/ramview.cpp debugger/registers.cpp debugger/wait_tree.cpp util/spinbox.cpp @@ -48,7 +47,6 @@ set(HEADERS debugger/graphics/graphics_tracing.h debugger/graphics/graphics_vertex_shader.h debugger/profiler.h - debugger/ramview.h debugger/registers.h debugger/wait_tree.h util/spinbox.h @@ -100,7 +98,7 @@ if (APPLE) else() add_executable(citra-qt ${SRCS} ${HEADERS} ${UI_HDRS}) endif() -target_link_libraries(citra-qt core video_core audio_core common qhexedit) +target_link_libraries(citra-qt core video_core audio_core common) target_link_libraries(citra-qt ${OPENGL_gl_LIBRARY} ${CITRA_QT_LIBS}) target_link_libraries(citra-qt ${PLATFORM_LIBRARIES} Threads::Threads) diff --git a/src/citra_qt/bootmanager.cpp b/src/citra_qt/bootmanager.cpp index 57fde6caa..948db384d 100644 --- a/src/citra_qt/bootmanager.cpp +++ b/src/citra_qt/bootmanager.cpp @@ -99,7 +99,7 @@ private: }; GRenderWindow::GRenderWindow(QWidget* parent, EmuThread* emu_thread) - : QWidget(parent), keyboard_id(0), emu_thread(emu_thread), child(nullptr) { + : QWidget(parent), child(nullptr), keyboard_id(0), emu_thread(emu_thread) { std::string window_title = Common::StringFromFormat("Citra | %s-%s", Common::g_scm_branch, Common::g_scm_desc); @@ -191,6 +191,7 @@ qreal GRenderWindow::windowPixelRatio() { } void GRenderWindow::closeEvent(QCloseEvent* event) { + motion_emu = nullptr; emit Closed(); QWidget::closeEvent(event); } @@ -204,11 +205,13 @@ void GRenderWindow::keyReleaseEvent(QKeyEvent* event) { } void GRenderWindow::mousePressEvent(QMouseEvent* event) { + auto pos = event->pos(); if (event->button() == Qt::LeftButton) { - auto pos = event->pos(); qreal pixelRatio = windowPixelRatio(); this->TouchPressed(static_cast<unsigned>(pos.x() * pixelRatio), static_cast<unsigned>(pos.y() * pixelRatio)); + } else if (event->button() == Qt::RightButton) { + motion_emu->BeginTilt(pos.x(), pos.y()); } } @@ -217,11 +220,14 @@ void GRenderWindow::mouseMoveEvent(QMouseEvent* event) { qreal pixelRatio = windowPixelRatio(); this->TouchMoved(std::max(static_cast<unsigned>(pos.x() * pixelRatio), 0u), std::max(static_cast<unsigned>(pos.y() * pixelRatio), 0u)); + motion_emu->Tilt(pos.x(), pos.y()); } void GRenderWindow::mouseReleaseEvent(QMouseEvent* event) { if (event->button() == Qt::LeftButton) this->TouchReleased(); + else if (event->button() == Qt::RightButton) + motion_emu->EndTilt(); } void GRenderWindow::ReloadSetKeymaps() { @@ -279,11 +285,13 @@ void GRenderWindow::OnMinimalClientAreaChangeRequest( } void GRenderWindow::OnEmulationStarting(EmuThread* emu_thread) { + motion_emu = std::make_unique<Motion::MotionEmu>(*this); this->emu_thread = emu_thread; child->DisablePainting(); } void GRenderWindow::OnEmulationStopping() { + motion_emu = nullptr; emu_thread = nullptr; child->EnablePainting(); } diff --git a/src/citra_qt/bootmanager.h b/src/citra_qt/bootmanager.h index 43015390b..7dac1c480 100644 --- a/src/citra_qt/bootmanager.h +++ b/src/citra_qt/bootmanager.h @@ -11,6 +11,7 @@ #include <QThread> #include "common/thread.h" #include "core/frontend/emu_window.h" +#include "core/frontend/motion_emu.h" class QKeyEvent; class QScreen; @@ -156,6 +157,9 @@ private: EmuThread* emu_thread; + /// Motion sensors emulation + std::unique_ptr<Motion::MotionEmu> motion_emu; + protected: void showEvent(QShowEvent* event) override; }; diff --git a/src/citra_qt/config.cpp b/src/citra_qt/config.cpp index 58eb5a96e..b65f57fdc 100644 --- a/src/citra_qt/config.cpp +++ b/src/citra_qt/config.cpp @@ -44,8 +44,7 @@ void Config::ReadValues() { qt_config->beginGroup("Renderer"); Settings::values.use_hw_renderer = qt_config->value("use_hw_renderer", true).toBool(); Settings::values.use_shader_jit = qt_config->value("use_shader_jit", true).toBool(); - Settings::values.use_scaled_resolution = - qt_config->value("use_scaled_resolution", false).toBool(); + Settings::values.resolution_factor = qt_config->value("resolution_factor", 1.0).toFloat(); Settings::values.use_vsync = qt_config->value("use_vsync", false).toBool(); Settings::values.toggle_framelimit = qt_config->value("toggle_framelimit", true).toBool(); @@ -64,6 +63,8 @@ void Config::ReadValues() { Settings::values.sink_id = qt_config->value("output_engine", "auto").toString().toStdString(); Settings::values.enable_audio_stretching = qt_config->value("enable_audio_stretching", true).toBool(); + Settings::values.audio_device_id = + qt_config->value("output_device", "auto").toString().toStdString(); qt_config->endGroup(); using namespace Service::CAM; @@ -88,7 +89,8 @@ void Config::ReadValues() { qt_config->beginGroup("System"); Settings::values.is_new_3ds = qt_config->value("is_new_3ds", false).toBool(); - Settings::values.region_value = qt_config->value("region_value", 1).toInt(); + Settings::values.region_value = + qt_config->value("region_value", Settings::REGION_VALUE_AUTO_SELECT).toInt(); qt_config->endGroup(); qt_config->beginGroup("Miscellaneous"); @@ -167,7 +169,7 @@ void Config::SaveValues() { qt_config->beginGroup("Renderer"); qt_config->setValue("use_hw_renderer", Settings::values.use_hw_renderer); qt_config->setValue("use_shader_jit", Settings::values.use_shader_jit); - qt_config->setValue("use_scaled_resolution", Settings::values.use_scaled_resolution); + qt_config->setValue("resolution_factor", (double)Settings::values.resolution_factor); qt_config->setValue("use_vsync", Settings::values.use_vsync); qt_config->setValue("toggle_framelimit", Settings::values.toggle_framelimit); @@ -185,6 +187,7 @@ void Config::SaveValues() { qt_config->beginGroup("Audio"); qt_config->setValue("output_engine", QString::fromStdString(Settings::values.sink_id)); qt_config->setValue("enable_audio_stretching", Settings::values.enable_audio_stretching); + qt_config->setValue("output_device", QString::fromStdString(Settings::values.audio_device_id)); qt_config->endGroup(); using namespace Service::CAM; diff --git a/src/citra_qt/configure_audio.cpp b/src/citra_qt/configure_audio.cpp index 3cdd4c780..3ddcf9232 100644 --- a/src/citra_qt/configure_audio.cpp +++ b/src/citra_qt/configure_audio.cpp @@ -2,6 +2,9 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include <memory> +#include "audio_core/audio_core.h" +#include "audio_core/sink.h" #include "audio_core/sink_details.h" #include "citra_qt/configure_audio.h" #include "core/settings.h" @@ -18,6 +21,8 @@ ConfigureAudio::ConfigureAudio(QWidget* parent) } this->setConfiguration(); + connect(ui->output_sink_combo_box, SIGNAL(currentIndexChanged(int)), this, + SLOT(updateAudioDevices(int))); } ConfigureAudio::~ConfigureAudio() {} @@ -33,6 +38,19 @@ void ConfigureAudio::setConfiguration() { ui->output_sink_combo_box->setCurrentIndex(new_sink_index); ui->toggle_audio_stretching->setChecked(Settings::values.enable_audio_stretching); + + // The device list cannot be pre-populated (nor listed) until the output sink is known. + updateAudioDevices(new_sink_index); + + int new_device_index = -1; + for (int index = 0; index < ui->audio_device_combo_box->count(); index++) { + if (ui->audio_device_combo_box->itemText(index).toStdString() == + Settings::values.audio_device_id) { + new_device_index = index; + break; + } + } + ui->audio_device_combo_box->setCurrentIndex(new_device_index); } void ConfigureAudio::applyConfiguration() { @@ -40,5 +58,20 @@ void ConfigureAudio::applyConfiguration() { ui->output_sink_combo_box->itemText(ui->output_sink_combo_box->currentIndex()) .toStdString(); Settings::values.enable_audio_stretching = ui->toggle_audio_stretching->isChecked(); + Settings::values.audio_device_id = + ui->audio_device_combo_box->itemText(ui->audio_device_combo_box->currentIndex()) + .toStdString(); Settings::Apply(); } + +void ConfigureAudio::updateAudioDevices(int sink_index) { + ui->audio_device_combo_box->clear(); + ui->audio_device_combo_box->addItem("auto"); + + std::string sink_id = ui->output_sink_combo_box->itemText(sink_index).toStdString(); + std::vector<std::string> device_list = + AudioCore::GetSinkDetails(sink_id).factory()->GetDeviceList(); + for (const auto& device : device_list) { + ui->audio_device_combo_box->addItem(device.c_str()); + } +} diff --git a/src/citra_qt/configure_audio.h b/src/citra_qt/configure_audio.h index 51df2e27b..8190e694f 100644 --- a/src/citra_qt/configure_audio.h +++ b/src/citra_qt/configure_audio.h @@ -20,6 +20,9 @@ public: void applyConfiguration(); +public slots: + void updateAudioDevices(int sink_index); + private: void setConfiguration(); diff --git a/src/citra_qt/configure_audio.ui b/src/citra_qt/configure_audio.ui index 3e2b4635f..dd870eb61 100644 --- a/src/citra_qt/configure_audio.ui +++ b/src/citra_qt/configure_audio.ui @@ -35,6 +35,21 @@ </property> </widget> </item> + <item> + <layout class="QHBoxLayout"> + <item> + <widget class="QLabel"> + <property name="text"> + <string>Audio Device:</string> + </property> + </widget> + </item> + <item> + <widget class="QComboBox" name="audio_device_combo_box"> + </widget> + </item> + </layout> + </item> </layout> </widget> </item> diff --git a/src/citra_qt/configure_general.cpp b/src/citra_qt/configure_general.cpp index 03cd8835b..ac90a6df4 100644 --- a/src/citra_qt/configure_general.cpp +++ b/src/citra_qt/configure_general.cpp @@ -23,13 +23,15 @@ void ConfigureGeneral::setConfiguration() { ui->toggle_deepscan->setChecked(UISettings::values.gamedir_deepscan); ui->toggle_check_exit->setChecked(UISettings::values.confirm_before_closing); ui->toggle_cpu_jit->setChecked(Settings::values.use_cpu_jit); - ui->region_combobox->setCurrentIndex(Settings::values.region_value); + + // The first item is "auto-select" with actual value -1, so plus one here will do the trick + ui->region_combobox->setCurrentIndex(Settings::values.region_value + 1); } void ConfigureGeneral::applyConfiguration() { UISettings::values.gamedir_deepscan = ui->toggle_deepscan->isChecked(); UISettings::values.confirm_before_closing = ui->toggle_check_exit->isChecked(); - Settings::values.region_value = ui->region_combobox->currentIndex(); + Settings::values.region_value = ui->region_combobox->currentIndex() - 1; Settings::values.use_cpu_jit = ui->toggle_cpu_jit->isChecked(); Settings::Apply(); } diff --git a/src/citra_qt/configure_general.ui b/src/citra_qt/configure_general.ui index 81688113f..0f3352a1d 100644 --- a/src/citra_qt/configure_general.ui +++ b/src/citra_qt/configure_general.ui @@ -43,26 +43,26 @@ </layout> </widget> </item> - <item> - <widget class="QGroupBox" name="groupBox_2"> - <property name="title"> - <string>Performance</string> - </property> - <layout class="QHBoxLayout" name="horizontalLayout_7"> - <item> - <layout class="QVBoxLayout" name="verticalLayout_5"> - <item> - <widget class="QCheckBox" name="toggle_cpu_jit"> - <property name="text"> - <string>Enable CPU JIT</string> - </property> - </widget> - </item> - </layout> - </item> - </layout> - </widget> - </item> + <item> + <widget class="QGroupBox" name="groupBox_2"> + <property name="title"> + <string>Performance</string> + </property> + <layout class="QHBoxLayout" name="horizontalLayout_7"> + <item> + <layout class="QVBoxLayout" name="verticalLayout_5"> + <item> + <widget class="QCheckBox" name="toggle_cpu_jit"> + <property name="text"> + <string>Enable CPU JIT</string> + </property> + </widget> + </item> + </layout> + </item> + </layout> + </widget> + </item> <item> <widget class="QGroupBox" name="groupBox_4"> <property name="title"> @@ -84,6 +84,11 @@ <widget class="QComboBox" name="region_combobox"> <item> <property name="text"> + <string>Auto-select</string> + </property> + </item> + <item> + <property name="text"> <string notr="true">JPN</string> </property> </item> diff --git a/src/citra_qt/configure_graphics.cpp b/src/citra_qt/configure_graphics.cpp index cea7db388..54f799b47 100644 --- a/src/citra_qt/configure_graphics.cpp +++ b/src/citra_qt/configure_graphics.cpp @@ -18,10 +18,81 @@ ConfigureGraphics::ConfigureGraphics(QWidget* parent) ConfigureGraphics::~ConfigureGraphics() {} +enum class Resolution : int { + Auto, + Scale1x, + Scale2x, + Scale3x, + Scale4x, + Scale5x, + Scale6x, + Scale7x, + Scale8x, + Scale9x, + Scale10x, +}; + +float ToResolutionFactor(Resolution option) { + switch (option) { + case Resolution::Auto: + return 0.f; + case Resolution::Scale1x: + return 1.f; + case Resolution::Scale2x: + return 2.f; + case Resolution::Scale3x: + return 3.f; + case Resolution::Scale4x: + return 4.f; + case Resolution::Scale5x: + return 5.f; + case Resolution::Scale6x: + return 6.f; + case Resolution::Scale7x: + return 7.f; + case Resolution::Scale8x: + return 8.f; + case Resolution::Scale9x: + return 9.f; + case Resolution::Scale10x: + return 10.f; + } + return 0.f; +} + +Resolution FromResolutionFactor(float factor) { + if (factor == 0.f) { + return Resolution::Auto; + } else if (factor == 1.f) { + return Resolution::Scale1x; + } else if (factor == 2.f) { + return Resolution::Scale2x; + } else if (factor == 3.f) { + return Resolution::Scale3x; + } else if (factor == 4.f) { + return Resolution::Scale4x; + } else if (factor == 5.f) { + return Resolution::Scale5x; + } else if (factor == 6.f) { + return Resolution::Scale6x; + } else if (factor == 7.f) { + return Resolution::Scale7x; + } else if (factor == 8.f) { + return Resolution::Scale8x; + } else if (factor == 9.f) { + return Resolution::Scale9x; + } else if (factor == 10.f) { + return Resolution::Scale10x; + } + return Resolution::Auto; +} + void ConfigureGraphics::setConfiguration() { ui->toggle_hw_renderer->setChecked(Settings::values.use_hw_renderer); + ui->resolution_factor_combobox->setEnabled(Settings::values.use_hw_renderer); ui->toggle_shader_jit->setChecked(Settings::values.use_shader_jit); - ui->toggle_scaled_resolution->setChecked(Settings::values.use_scaled_resolution); + ui->resolution_factor_combobox->setCurrentIndex( + static_cast<int>(FromResolutionFactor(Settings::values.resolution_factor))); ui->toggle_vsync->setChecked(Settings::values.use_vsync); ui->toggle_framelimit->setChecked(Settings::values.toggle_framelimit); ui->layout_combobox->setCurrentIndex(static_cast<int>(Settings::values.layout_option)); @@ -31,7 +102,8 @@ void ConfigureGraphics::setConfiguration() { void ConfigureGraphics::applyConfiguration() { Settings::values.use_hw_renderer = ui->toggle_hw_renderer->isChecked(); Settings::values.use_shader_jit = ui->toggle_shader_jit->isChecked(); - Settings::values.use_scaled_resolution = ui->toggle_scaled_resolution->isChecked(); + Settings::values.resolution_factor = + ToResolutionFactor(static_cast<Resolution>(ui->resolution_factor_combobox->currentIndex())); Settings::values.use_vsync = ui->toggle_vsync->isChecked(); Settings::values.toggle_framelimit = ui->toggle_framelimit->isChecked(); Settings::values.layout_option = diff --git a/src/citra_qt/configure_graphics.ui b/src/citra_qt/configure_graphics.ui index 964aa0bbd..a091f4c60 100644 --- a/src/citra_qt/configure_graphics.ui +++ b/src/citra_qt/configure_graphics.ui @@ -37,13 +37,6 @@ </widget> </item> <item> - <widget class="QCheckBox" name="toggle_scaled_resolution"> - <property name="text"> - <string>Enable scaled resolution</string> - </property> - </widget> - </item> - <item> <widget class="QCheckBox" name="toggle_vsync"> <property name="text"> <string>Enable V-Sync</string> @@ -57,6 +50,76 @@ </property> </widget> </item> + <item> + <layout class="QHBoxLayout" name="horizontalLayout"> + <item> + <widget class="QLabel" name="label"> + <property name="text"> + <string>Internal Resolution:</string> + </property> + </widget> + </item> + <item> + <widget class="QComboBox" name="resolution_factor_combobox"> + <item> + <property name="text"> + <string notr="true">Auto (Window Size)</string> + </property> + </item> + <item> + <property name="text"> + <string notr="true">Native (400x240)</string> + </property> + </item> + <item> + <property name="text"> + <string notr="true">2x Native (800x480)</string> + </property> + </item> + <item> + <property name="text"> + <string notr="true">3x Native (1200x720)</string> + </property> + </item> + <item> + <property name="text"> + <string notr="true">4x Native (1600x960)</string> + </property> + </item> + <item> + <property name="text"> + <string notr="true">5x Native (2000x1200)</string> + </property> + </item> + <item> + <property name="text"> + <string notr="true">6x Native (2400x1440)</string> + </property> + </item> + <item> + <property name="text"> + <string notr="true">7x Native (2800x1680)</string> + </property> + </item> + <item> + <property name="text"> + <string notr="true">8x Native (3200x1920)</string> + </property> + </item> + <item> + <property name="text"> + <string notr="true">9x Native (3600x2160)</string> + </property> + </item> + <item> + <property name="text"> + <string notr="true">10x Native (4000x2400)</string> + </property> + </item> + </widget> + </item> + </layout> + </item> </layout> </widget> </item> @@ -69,9 +132,9 @@ </property> <layout class="QHBoxLayout" name="horizontalLayout_3"> <item> - <layout class="QVBoxLayout" name="verticalLayout_2"> + <layout class="QVBoxLayout" name="verticalLayout_4"> <item> - <layout class="QHBoxLayout" name="horizontalLayout_3"> + <layout class="QHBoxLayout" name="horizontalLayout_4"> <item> <widget class="QLabel" name="label1"> <property name="text"> @@ -128,5 +191,12 @@ </layout> </widget> <resources/> - <connections/> + <connections> + <connection> + <sender>toggle_hw_renderer</sender> + <signal>toggled(bool)</signal> + <receiver>resolution_factor_combobox</receiver> + <slot>setEnabled(bool)</slot> + </connection> + </connections> </ui> diff --git a/src/citra_qt/configure_input.ui b/src/citra_qt/configure_input.ui index a040d4df4..2760787e5 100644 --- a/src/citra_qt/configure_input.ui +++ b/src/citra_qt/configure_input.ui @@ -275,7 +275,6 @@ </layout> </item> </layout> - <zorder></zorder> </widget> </item> <item row="1" column="1"> @@ -521,7 +520,7 @@ <item row="1" column="1"> <layout class="QVBoxLayout" name="verticalLayout_28"> <item> - <widget class="QLabel" name="label_34"> + <widget class="QLabel" name="label_36"> <property name="text"> <string>Circle Mod:</string> </property> diff --git a/src/citra_qt/configure_system.ui b/src/citra_qt/configure_system.ui index 6a906b61b..cc54fa37f 100644 --- a/src/citra_qt/configure_system.ui +++ b/src/citra_qt/configure_system.ui @@ -129,6 +129,9 @@ </item> <item row="2" column="1"> <widget class="QComboBox" name="combo_language"> + <property name="toolTip"> + <string>Note: this can be overridden when region setting is auto-select</string> + </property> <item> <property name="text"> <string>Japanese (日本語)</string> diff --git a/src/citra_qt/debugger/callstack.cpp b/src/citra_qt/debugger/callstack.cpp index c1db93583..08d2e7a22 100644 --- a/src/citra_qt/debugger/callstack.cpp +++ b/src/citra_qt/debugger/callstack.cpp @@ -45,7 +45,6 @@ void CallstackWidget::OnDebugModeEntered() { if (ARM_Disasm::Decode(insn) == OP_BL) { std::string name; // ripped from disasm - u8 cond = (insn >> 28) & 0xf; u32 i_offset = insn & 0xffffff; // Sign-extend the 24-bit offset if ((i_offset >> 23) & 1) diff --git a/src/citra_qt/debugger/graphics/graphics_cmdlists.cpp b/src/citra_qt/debugger/graphics/graphics_cmdlists.cpp index dab529e3a..f5a2ec761 100644 --- a/src/citra_qt/debugger/graphics/graphics_cmdlists.cpp +++ b/src/citra_qt/debugger/graphics/graphics_cmdlists.cpp @@ -135,11 +135,6 @@ void GPUCommandListWidget::OnCommandDoubleClicked(const QModelIndex& index) { UNREACHABLE_MSG("Unknown texture command"); } - const auto texture = Pica::g_state.regs.GetTextures()[texture_index]; - const auto config = texture.config; - const auto format = texture.format; - const auto info = Pica::DebugUtils::TextureInfo::FromPicaRegister(config, format); - // TODO: Open a surface debugger } } diff --git a/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp b/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp index b75b94ef8..f37524190 100644 --- a/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp +++ b/src/citra_qt/debugger/graphics/graphics_vertex_shader.cpp @@ -18,7 +18,9 @@ #include "citra_qt/util/util.h" #include "video_core/pica.h" #include "video_core/pica_state.h" +#include "video_core/shader/debug_data.h" #include "video_core/shader/shader.h" +#include "video_core/shader/shader_interpreter.h" using nihstro::OpCode; using nihstro::Instruction; @@ -276,9 +278,6 @@ QVariant GraphicsVertexShaderModel::data(const QModelIndex& index, int role) con output << 'b' << instr.flow_control.bool_uniform_id << ' '; } - u32 target_addr = instr.flow_control.dest_offset; - u32 target_addr_else = instr.flow_control.dest_offset; - if (opcode_info.subtype & OpCode::Info::HasAlternative) { output << "else jump to 0x" << std::setw(4) << std::right << std::setfill('0') << std::hex @@ -473,7 +472,6 @@ GraphicsVertexShaderWidget::GraphicsVertexShaderWidget( } void GraphicsVertexShaderWidget::OnBreakPointHit(Pica::DebugContext::Event event, void* data) { - auto input = static_cast<Pica::Shader::InputVertex*>(data); if (event == Pica::DebugContext::Event::VertexShaderInvocation) { Reload(true, data); } else { @@ -522,8 +520,9 @@ void GraphicsVertexShaderWidget::Reload(bool replace_vertex_data, void* vertex_d info.labels.insert({entry_point, "main"}); // Generate debug information - debug_data = Pica::g_state.vs.ProduceDebugInfo(input_vertex, num_attributes, shader_config, - shader_setup); + Pica::Shader::InterpreterEngine shader_engine; + shader_engine.SetupBatch(shader_setup, entry_point); + debug_data = shader_engine.ProduceDebugInfo(shader_setup, input_vertex, num_attributes); // Reload widget state for (int attr = 0; attr < num_attributes; ++attr) { diff --git a/src/citra_qt/debugger/graphics/graphics_vertex_shader.h b/src/citra_qt/debugger/graphics/graphics_vertex_shader.h index bedea0bed..3292573f3 100644 --- a/src/citra_qt/debugger/graphics/graphics_vertex_shader.h +++ b/src/citra_qt/debugger/graphics/graphics_vertex_shader.h @@ -8,6 +8,7 @@ #include <QTreeView> #include "citra_qt/debugger/graphics/graphics_breakpoint_observer.h" #include "nihstro/parser_shbin.h" +#include "video_core/shader/debug_data.h" #include "video_core/shader/shader.h" class QLabel; diff --git a/src/citra_qt/debugger/ramview.cpp b/src/citra_qt/debugger/ramview.cpp deleted file mode 100644 index 10a09dda8..000000000 --- a/src/citra_qt/debugger/ramview.cpp +++ /dev/null @@ -1,12 +0,0 @@ -// Copyright 2014 Citra Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. - -#include "citra_qt/debugger/ramview.h" - -GRamView::GRamView(QWidget* parent) : QHexEdit(parent) {} - -void GRamView::OnCPUStepped() { - // TODO: QHexEdit doesn't show vertical scroll bars for > 10MB data streams... - // setData(QByteArray((const char*)Mem_RAM,sizeof(Mem_RAM)/8)); -} diff --git a/src/citra_qt/debugger/ramview.h b/src/citra_qt/debugger/ramview.h deleted file mode 100644 index d01cea93b..000000000 --- a/src/citra_qt/debugger/ramview.h +++ /dev/null @@ -1,17 +0,0 @@ -// Copyright 2014 Citra Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. - -#pragma once - -#include "qhexedit.h" - -class GRamView : public QHexEdit { - Q_OBJECT - -public: - explicit GRamView(QWidget* parent = nullptr); - -public slots: - void OnCPUStepped(); -}; diff --git a/src/citra_qt/debugger/wait_tree.cpp b/src/citra_qt/debugger/wait_tree.cpp index 1d2de5185..b6ecf3819 100644 --- a/src/citra_qt/debugger/wait_tree.cpp +++ b/src/citra_qt/debugger/wait_tree.cpp @@ -153,7 +153,8 @@ QString WaitTreeThread::GetText() const { case THREADSTATUS_WAIT_SLEEP: status = tr("sleeping"); break; - case THREADSTATUS_WAIT_SYNCH: + case THREADSTATUS_WAIT_SYNCH_ALL: + case THREADSTATUS_WAIT_SYNCH_ANY: status = tr("waiting for objects"); break; case THREADSTATUS_DORMANT: @@ -180,7 +181,8 @@ QColor WaitTreeThread::GetColor() const { return QColor(Qt::GlobalColor::darkRed); case THREADSTATUS_WAIT_SLEEP: return QColor(Qt::GlobalColor::darkYellow); - case THREADSTATUS_WAIT_SYNCH: + case THREADSTATUS_WAIT_SYNCH_ALL: + case THREADSTATUS_WAIT_SYNCH_ANY: return QColor(Qt::GlobalColor::red); case THREADSTATUS_DORMANT: return QColor(Qt::GlobalColor::darkCyan); @@ -228,7 +230,8 @@ std::vector<std::unique_ptr<WaitTreeItem>> WaitTreeThread::GetChildren() const { } else { list.push_back(std::make_unique<WaitTreeMutexList>(thread.held_mutexes)); } - if (thread.status == THREADSTATUS_WAIT_SYNCH) { + if (thread.status == THREADSTATUS_WAIT_SYNCH_ANY || + thread.status == THREADSTATUS_WAIT_SYNCH_ALL) { list.push_back(std::make_unique<WaitTreeObjectList>(thread.wait_objects, thread.IsSleepingOnWaitAll())); } diff --git a/src/citra_qt/main.cpp b/src/citra_qt/main.cpp index 6d59cf640..f765c0147 100644 --- a/src/citra_qt/main.cpp +++ b/src/citra_qt/main.cpp @@ -12,6 +12,7 @@ #include <QFileDialog> #include <QMessageBox> #include <QtGui> +#include <QtWidgets> #include "citra_qt/bootmanager.h" #include "citra_qt/config.h" #include "citra_qt/configure_dialog.h" @@ -24,7 +25,6 @@ #include "citra_qt/debugger/graphics/graphics_tracing.h" #include "citra_qt/debugger/graphics/graphics_vertex_shader.h" #include "citra_qt/debugger/profiler.h" -#include "citra_qt/debugger/ramview.h" #include "citra_qt/debugger/registers.h" #include "citra_qt/debugger/wait_tree.h" #include "citra_qt/game_list.h" @@ -46,7 +46,6 @@ #include "core/gdbstub/gdbstub.h" #include "core/loader/loader.h" #include "core/settings.h" -#include "qhexedit.h" #include "video_core/video_core.h" #ifdef QT_STATICPLUGIN diff --git a/src/citra_qt/util/spinbox.cpp b/src/citra_qt/util/spinbox.cpp index feb0ea1b3..212709007 100644 --- a/src/citra_qt/util/spinbox.cpp +++ b/src/citra_qt/util/spinbox.cpp @@ -165,13 +165,6 @@ void CSpinBox::UpdateText() { // Uppercase digits greater than 9. mask += ">"; - // The greatest signed 64-bit number has 19 decimal digits. - // TODO: Could probably make this more generic with some logarithms. - // For reference, unsigned 64-bit can have up to 20 decimal digits. - int digits = (num_digits != 0) - ? num_digits - : (base == 16) ? 16 : (base == 10) ? 19 : 0xFF; // fallback case... - // Match num_digits digits // Digits irrelevant to the chosen number base are filtered in the validator mask += QString("H").repeated(std::max(num_digits, 1)); diff --git a/src/common/CMakeLists.txt b/src/common/CMakeLists.txt index 5aecf6e6e..a7a4a688c 100644 --- a/src/common/CMakeLists.txt +++ b/src/common/CMakeLists.txt @@ -46,6 +46,7 @@ set(HEADERS microprofileui.h platform.h profiler_reporting.h + quaternion.h scm_rev.h scope_exit.h string_util.h diff --git a/src/common/hash.cpp b/src/common/hash.cpp index 2309320bb..f3d390dc5 100644 --- a/src/common/hash.cpp +++ b/src/common/hash.cpp @@ -16,7 +16,7 @@ namespace Common { // Block read - if your platform needs to do endian-swapping or can only handle aligned reads, do // the conversion here -static FORCE_INLINE u64 getblock64(const u64* p, int i) { +static FORCE_INLINE u64 getblock64(const u64* p, size_t i) { return p[i]; } @@ -34,9 +34,9 @@ static FORCE_INLINE u64 fmix64(u64 k) { // This is the 128-bit variant of the MurmurHash3 hash function that is targeted for 64-bit // platforms (MurmurHash3_x64_128). It was taken from: // https://code.google.com/p/smhasher/source/browse/trunk/MurmurHash3.cpp -void MurmurHash3_128(const void* key, int len, u32 seed, void* out) { +void MurmurHash3_128(const void* key, size_t len, u32 seed, void* out) { const u8* data = (const u8*)key; - const int nblocks = len / 16; + const size_t nblocks = len / 16; u64 h1 = seed; u64 h2 = seed; @@ -48,7 +48,7 @@ void MurmurHash3_128(const void* key, int len, u32 seed, void* out) { const u64* blocks = (const u64*)(data); - for (int i = 0; i < nblocks; i++) { + for (size_t i = 0; i < nblocks; i++) { u64 k1 = getblock64(blocks, i * 2 + 0); u64 k2 = getblock64(blocks, i * 2 + 1); diff --git a/src/common/hash.h b/src/common/hash.h index a3850be68..ee2560dad 100644 --- a/src/common/hash.h +++ b/src/common/hash.h @@ -4,11 +4,12 @@ #pragma once +#include <cstddef> #include "common/common_types.h" namespace Common { -void MurmurHash3_128(const void* key, int len, u32 seed, void* out); +void MurmurHash3_128(const void* key, size_t len, u32 seed, void* out); /** * Computes a 64-bit hash over the specified block of data @@ -16,7 +17,7 @@ void MurmurHash3_128(const void* key, int len, u32 seed, void* out); * @param len Length of data (in bytes) to compute hash over * @returns 64-bit hash value that was computed over the data block */ -static inline u64 ComputeHash64(const void* data, int len) { +static inline u64 ComputeHash64(const void* data, size_t len) { u64 res[2]; MurmurHash3_128(data, len, 0, res); return res[0]; diff --git a/src/common/logging/backend.cpp b/src/common/logging/backend.cpp index 3ea102229..2ef3e6b05 100644 --- a/src/common/logging/backend.cpp +++ b/src/common/logging/backend.cpp @@ -45,6 +45,7 @@ namespace Log { SUB(Service, LDR) \ SUB(Service, MIC) \ SUB(Service, NDM) \ + SUB(Service, NFC) \ SUB(Service, NIM) \ SUB(Service, NWM) \ SUB(Service, CAM) \ diff --git a/src/common/logging/log.h b/src/common/logging/log.h index 9d8c18d8e..4330ef879 100644 --- a/src/common/logging/log.h +++ b/src/common/logging/log.h @@ -62,6 +62,7 @@ enum class Class : ClassType { Service_LDR, ///< The LDR (3ds dll loader) service Service_MIC, ///< The MIC (Microphone) service Service_NDM, ///< The NDM (Network daemon manager) service + Service_NFC, ///< The NFC service Service_NIM, ///< The NIM (Network interface manager) service Service_NWM, ///< The NWM (Network wlan manager) service Service_CAM, ///< The CAM (Camera) service diff --git a/src/common/math_util.h b/src/common/math_util.h index cdeaeb733..45a1ed367 100644 --- a/src/common/math_util.h +++ b/src/common/math_util.h @@ -10,6 +10,8 @@ namespace MathUtil { +static constexpr float PI = 3.14159265f; + inline bool IntervalsIntersect(unsigned start0, unsigned length0, unsigned start1, unsigned length1) { return (std::max(start0, start1) < std::min(start0 + length0, start1 + length1)); diff --git a/src/common/quaternion.h b/src/common/quaternion.h new file mode 100644 index 000000000..84ac82ed3 --- /dev/null +++ b/src/common/quaternion.h @@ -0,0 +1,44 @@ +// Copyright 2016 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#include "common/vector_math.h" + +namespace Math { + +template <typename T> +class Quaternion { +public: + Math::Vec3<T> xyz; + T w; + + Quaternion<decltype(-T{})> Inverse() const { + return {-xyz, w}; + } + + Quaternion<decltype(T{} + T{})> operator+(const Quaternion& other) const { + return {xyz + other.xyz, w + other.w}; + } + + Quaternion<decltype(T{} - T{})> operator-(const Quaternion& other) const { + return {xyz - other.xyz, w - other.w}; + } + + Quaternion<decltype(T{} * T{} - T{} * T{})> operator*(const Quaternion& other) const { + return {xyz * other.w + other.xyz * w + Cross(xyz, other.xyz), + w * other.w - Dot(xyz, other.xyz)}; + } +}; + +template <typename T> +auto QuaternionRotate(const Quaternion<T>& q, const Math::Vec3<T>& v) { + return v + 2 * Cross(q.xyz, Cross(q.xyz, v) + v * q.w); +} + +inline Quaternion<float> MakeQuaternion(const Math::Vec3<float>& axis, float angle) { + return {axis * std::sin(angle / 2), std::cos(angle / 2)}; +} + +} // namspace Math diff --git a/src/common/thread.h b/src/common/thread.h index 9c08be7e3..fa475ab51 100644 --- a/src/common/thread.h +++ b/src/common/thread.h @@ -4,6 +4,7 @@ #pragma once +#include <chrono> #include <condition_variable> #include <cstddef> #include <mutex> @@ -54,6 +55,15 @@ public: is_set = false; } + template <class Clock, class Duration> + bool WaitUntil(const std::chrono::time_point<Clock, Duration>& time) { + std::unique_lock<std::mutex> lk(mutex); + if (!condvar.wait_until(lk, time, [this] { return is_set; })) + return false; + is_set = false; + return true; + } + void Reset() { std::unique_lock<std::mutex> lk(mutex); // no other action required, since wait loops on the predicate and any lingering signal will diff --git a/src/common/vector_math.h b/src/common/vector_math.h index a57d86d88..7ca8e15f5 100644 --- a/src/common/vector_math.h +++ b/src/common/vector_math.h @@ -186,6 +186,18 @@ Vec2<T> operator*(const V& f, const Vec2<T>& vec) { typedef Vec2<float> Vec2f; +template <> +inline float Vec2<float>::Length() const { + return std::sqrt(x * x + y * y); +} + +template <> +inline float Vec2<float>::Normalize() { + float length = Length(); + *this /= length; + return length; +} + template <typename T> class Vec3 { public: @@ -388,6 +400,13 @@ inline Vec3<float> Vec3<float>::Normalized() const { return *this / Length(); } +template <> +inline float Vec3<float>::Normalize() { + float length = Length(); + *this /= length; + return length; +} + typedef Vec3<float> Vec3f; template <typename T> diff --git a/src/core/CMakeLists.txt b/src/core/CMakeLists.txt index 2dd4158e5..bd0e3c595 100644 --- a/src/core/CMakeLists.txt +++ b/src/core/CMakeLists.txt @@ -34,6 +34,7 @@ set(SRCS frontend/camera/interface.cpp frontend/emu_window.cpp frontend/key_map.cpp + frontend/motion_emu.cpp gdbstub/gdbstub.cpp hle/config_mem.cpp hle/applets/applet.cpp @@ -56,7 +57,9 @@ set(SRCS hle/kernel/thread.cpp hle/kernel/timer.cpp hle/kernel/vm_manager.cpp - hle/service/ac_u.cpp + hle/service/ac/ac.cpp + hle/service/ac/ac_i.cpp + hle/service/ac/ac_u.cpp hle/service/act/act.cpp hle/service/act/act_a.cpp hle/service/act/act_u.cpp @@ -208,6 +211,7 @@ set(HEADERS frontend/camera/interface.h frontend/emu_window.h frontend/key_map.h + frontend/motion_emu.h gdbstub/gdbstub.h hle/config_mem.h hle/function_wrappers.h @@ -233,7 +237,9 @@ set(HEADERS hle/kernel/timer.h hle/kernel/vm_manager.h hle/result.h - hle/service/ac_u.h + hle/service/ac/ac.h + hle/service/ac/ac_i.h + hle/service/ac/ac_u.h hle/service/act/act.h hle/service/act/act_a.h hle/service/act/act_u.h diff --git a/src/core/core.cpp b/src/core/core.cpp index ee5237096..202cd332b 100644 --- a/src/core/core.cpp +++ b/src/core/core.cpp @@ -159,6 +159,7 @@ void System::Shutdown() { Kernel::Shutdown(); HW::Shutdown(); CoreTiming::Shutdown(); + cpu_core.reset(); LOG_DEBUG(Core, "Shutdown OK"); } diff --git a/src/core/core.h b/src/core/core.h index 1015e8847..17572a74f 100644 --- a/src/core/core.h +++ b/src/core/core.h @@ -115,7 +115,7 @@ private: static System s_instance; }; -static ARM_Interface& CPU() { +inline ARM_Interface& CPU() { return System::GetInstance().CPU(); } diff --git a/src/core/core_timing.cpp b/src/core/core_timing.cpp index a437d0823..276ecfdf6 100644 --- a/src/core/core_timing.cpp +++ b/src/core/core_timing.cpp @@ -13,7 +13,7 @@ #include "core/core.h" #include "core/core_timing.h" -int g_clock_rate_arm11 = 268123480; +int g_clock_rate_arm11 = BASE_CLOCK_RATE_ARM11; // is this really necessary? #define INITIAL_SLICE_LENGTH 20000 diff --git a/src/core/core_timing.h b/src/core/core_timing.h index b72a1b500..d2f85cd4d 100644 --- a/src/core/core_timing.h +++ b/src/core/core_timing.h @@ -21,6 +21,7 @@ // inside callback: // ScheduleEvent(periodInCycles - cycles_late, callback, "whatever") +constexpr int BASE_CLOCK_RATE_ARM11 = 268123480; extern int g_clock_rate_arm11; inline s64 msToCycles(int ms) { diff --git a/src/core/file_sys/archive_extsavedata.cpp b/src/core/file_sys/archive_extsavedata.cpp index 51ce78435..dd2fb167f 100644 --- a/src/core/file_sys/archive_extsavedata.cpp +++ b/src/core/file_sys/archive_extsavedata.cpp @@ -107,6 +107,8 @@ public: case PathParser::NotFound: LOG_ERROR(Service_FS, "%s not found", full_path.c_str()); return ERROR_FILE_NOT_FOUND; + case PathParser::FileFound: + break; // Expected 'success' case } FileUtil::IOFile file(full_path, "r+b"); diff --git a/src/core/file_sys/archive_sdmc.cpp b/src/core/file_sys/archive_sdmc.cpp index 333dfb92e..72ff05c65 100644 --- a/src/core/file_sys/archive_sdmc.cpp +++ b/src/core/file_sys/archive_sdmc.cpp @@ -72,6 +72,8 @@ ResultVal<std::unique_ptr<FileBackend>> SDMCArchive::OpenFileBase(const Path& pa FileUtil::CreateEmptyFile(full_path); } break; + case PathParser::FileFound: + break; // Expected 'success' case } FileUtil::IOFile file(full_path, mode.write_flag ? "r+b" : "rb"); @@ -106,6 +108,8 @@ ResultCode SDMCArchive::DeleteFile(const Path& path) const { case PathParser::DirectoryFound: LOG_ERROR(Service_FS, "%s is not a file", full_path.c_str()); return ERROR_UNEXPECTED_FILE_OR_DIRECTORY_SDMC; + case PathParser::FileFound: + break; // Expected 'success' case } if (FileUtil::Delete(full_path)) { @@ -154,6 +158,8 @@ static ResultCode DeleteDirectoryHelper(const Path& path, const std::string& mou case PathParser::FileFound: LOG_ERROR(Service_FS, "Unexpected file in path %s", full_path.c_str()); return ERROR_UNEXPECTED_FILE_OR_DIRECTORY_SDMC; + case PathParser::DirectoryFound: + break; // Expected 'success' case } if (deleter(full_path)) { @@ -197,6 +203,8 @@ ResultCode SDMCArchive::CreateFile(const FileSys::Path& path, u64 size) const { case PathParser::FileFound: LOG_ERROR(Service_FS, "%s already exists", full_path.c_str()); return ERROR_ALREADY_EXISTS; + case PathParser::NotFound: + break; // Expected 'success' case } if (size == 0) { @@ -238,6 +246,8 @@ ResultCode SDMCArchive::CreateDirectory(const Path& path) const { case PathParser::FileFound: LOG_ERROR(Service_FS, "%s already exists", full_path.c_str()); return ERROR_ALREADY_EXISTS; + case PathParser::NotFound: + break; // Expected 'success' case } if (FileUtil::CreateDir(mount_point + path.AsString())) { @@ -281,6 +291,8 @@ ResultVal<std::unique_ptr<DirectoryBackend>> SDMCArchive::OpenDirectory(const Pa case PathParser::FileInPath: LOG_ERROR(Service_FS, "Unexpected file in path %s", full_path.c_str()); return ERROR_UNEXPECTED_FILE_OR_DIRECTORY_SDMC; + case PathParser::DirectoryFound: + break; // Expected 'success' case } auto directory = std::make_unique<DiskDirectory>(full_path); diff --git a/src/core/file_sys/savedata_archive.cpp b/src/core/file_sys/savedata_archive.cpp index f2e6a06bc..f540c4a93 100644 --- a/src/core/file_sys/savedata_archive.cpp +++ b/src/core/file_sys/savedata_archive.cpp @@ -57,6 +57,8 @@ ResultVal<std::unique_ptr<FileBackend>> SaveDataArchive::OpenFile(const Path& pa FileUtil::CreateEmptyFile(full_path); } break; + case PathParser::FileFound: + break; // Expected 'success' case } FileUtil::IOFile file(full_path, mode.write_flag ? "r+b" : "rb"); @@ -91,6 +93,8 @@ ResultCode SaveDataArchive::DeleteFile(const Path& path) const { case PathParser::NotFound: LOG_ERROR(Service_FS, "File not found %s", full_path.c_str()); return ERROR_FILE_NOT_FOUND; + case PathParser::FileFound: + break; // Expected 'success' case } if (FileUtil::Delete(full_path)) { @@ -139,6 +143,8 @@ static ResultCode DeleteDirectoryHelper(const Path& path, const std::string& mou case PathParser::FileFound: LOG_ERROR(Service_FS, "Unexpected file or directory %s", full_path.c_str()); return ERROR_UNEXPECTED_FILE_OR_DIRECTORY; + case PathParser::DirectoryFound: + break; // Expected 'success' case } if (deleter(full_path)) { @@ -182,6 +188,8 @@ ResultCode SaveDataArchive::CreateFile(const FileSys::Path& path, u64 size) cons case PathParser::FileFound: LOG_ERROR(Service_FS, "%s already exists", full_path.c_str()); return ERROR_FILE_ALREADY_EXISTS; + case PathParser::NotFound: + break; // Expected 'success' case } if (size == 0) { @@ -225,6 +233,8 @@ ResultCode SaveDataArchive::CreateDirectory(const Path& path) const { case PathParser::FileFound: LOG_ERROR(Service_FS, "%s already exists", full_path.c_str()); return ERROR_DIRECTORY_ALREADY_EXISTS; + case PathParser::NotFound: + break; // Expected 'success' case } if (FileUtil::CreateDir(mount_point + path.AsString())) { @@ -269,6 +279,8 @@ ResultVal<std::unique_ptr<DirectoryBackend>> SaveDataArchive::OpenDirectory( case PathParser::FileFound: LOG_ERROR(Service_FS, "Unexpected file in path %s", full_path.c_str()); return ERROR_UNEXPECTED_FILE_OR_DIRECTORY; + case PathParser::DirectoryFound: + break; // Expected 'success' case } auto directory = std::make_unique<DiskDirectory>(full_path); diff --git a/src/core/frontend/emu_window.cpp b/src/core/frontend/emu_window.cpp index f6f90f9e1..4f0f786ce 100644 --- a/src/core/frontend/emu_window.cpp +++ b/src/core/frontend/emu_window.cpp @@ -5,6 +5,7 @@ #include <algorithm> #include <cmath> #include "common/assert.h" +#include "common/profiler_reporting.h" #include "core/frontend/emu_window.h" #include "core/frontend/key_map.h" #include "video_core/video_core.h" @@ -89,6 +90,30 @@ void EmuWindow::TouchMoved(unsigned framebuffer_x, unsigned framebuffer_y) { TouchPressed(framebuffer_x, framebuffer_y); } +void EmuWindow::AccelerometerChanged(float x, float y, float z) { + constexpr float coef = 512; + + std::lock_guard<std::mutex> lock(accel_mutex); + + // TODO(wwylele): do a time stretch as it in GyroscopeChanged + // The time stretch formula should be like + // stretched_vector = (raw_vector - gravity) * stretch_ratio + gravity + accel_x = static_cast<s16>(x * coef); + accel_y = static_cast<s16>(y * coef); + accel_z = static_cast<s16>(z * coef); +} + +void EmuWindow::GyroscopeChanged(float x, float y, float z) { + constexpr float FULL_FPS = 60; + float coef = GetGyroscopeRawToDpsCoefficient(); + float stretch = + FULL_FPS / Common::Profiling::GetTimingResultsAggregator()->GetAggregatedResults().fps; + std::lock_guard<std::mutex> lock(gyro_mutex); + gyro_x = static_cast<s16>(x * coef * stretch); + gyro_y = static_cast<s16>(y * coef * stretch); + gyro_z = static_cast<s16>(z * coef * stretch); +} + void EmuWindow::UpdateCurrentFramebufferLayout(unsigned width, unsigned height) { Layout::FramebufferLayout layout; switch (Settings::values.layout_option) { diff --git a/src/core/frontend/emu_window.h b/src/core/frontend/emu_window.h index 835c4d500..1ba64c92b 100644 --- a/src/core/frontend/emu_window.h +++ b/src/core/frontend/emu_window.h @@ -4,6 +4,7 @@ #pragma once +#include <mutex> #include <tuple> #include <utility> #include "common/common_types.h" @@ -93,6 +94,27 @@ public: void TouchMoved(unsigned framebuffer_x, unsigned framebuffer_y); /** + * Signal accelerometer state has changed. + * @param x X-axis accelerometer value + * @param y Y-axis accelerometer value + * @param z Z-axis accelerometer value + * @note all values are in unit of g (gravitational acceleration). + * e.g. x = 1.0 means 9.8m/s^2 in x direction. + * @see GetAccelerometerState for axis explanation. + */ + void AccelerometerChanged(float x, float y, float z); + + /** + * Signal gyroscope state has changed. + * @param x X-axis accelerometer value + * @param y Y-axis accelerometer value + * @param z Z-axis accelerometer value + * @note all values are in deg/sec. + * @see GetGyroscopeState for axis explanation. + */ + void GyroscopeChanged(float x, float y, float z); + + /** * Gets the current pad state (which buttons are pressed). * @note This should be called by the core emu thread to get a state set by the window thread. * @note This doesn't include analog input like circle pad direction @@ -134,12 +156,11 @@ public: * 1 unit of return value = 1/512 g (measured by hw test), * where g is the gravitational acceleration (9.8 m/sec2). * @note This should be called by the core emu thread to get a state set by the window thread. - * @todo Implement accelerometer input in front-end. * @return std::tuple of (x, y, z) */ - std::tuple<s16, s16, s16> GetAccelerometerState() const { - // stubbed - return std::make_tuple(0, -512, 0); + std::tuple<s16, s16, s16> GetAccelerometerState() { + std::lock_guard<std::mutex> lock(accel_mutex); + return std::make_tuple(accel_x, accel_y, accel_z); } /** @@ -153,12 +174,11 @@ public: * 1 unit of return value = (1/coef) deg/sec, * where coef is the return value of GetGyroscopeRawToDpsCoefficient(). * @note This should be called by the core emu thread to get a state set by the window thread. - * @todo Implement gyroscope input in front-end. * @return std::tuple of (x, y, z) */ - std::tuple<s16, s16, s16> GetGyroscopeState() const { - // stubbed - return std::make_tuple(0, 0, 0); + std::tuple<s16, s16, s16> GetGyroscopeState() { + std::lock_guard<std::mutex> lock(gyro_mutex); + return std::make_tuple(gyro_x, gyro_y, gyro_z); } /** @@ -216,6 +236,12 @@ protected: circle_pad_x = 0; circle_pad_y = 0; touch_pressed = false; + accel_x = 0; + accel_y = -512; + accel_z = 0; + gyro_x = 0; + gyro_y = 0; + gyro_z = 0; } virtual ~EmuWindow() {} @@ -281,6 +307,16 @@ private: s16 circle_pad_x; ///< Circle pad X-position in native 3DS pixel coordinates (-156 - 156) s16 circle_pad_y; ///< Circle pad Y-position in native 3DS pixel coordinates (-156 - 156) + std::mutex accel_mutex; + s16 accel_x; ///< Accelerometer X-axis value in native 3DS units + s16 accel_y; ///< Accelerometer Y-axis value in native 3DS units + s16 accel_z; ///< Accelerometer Z-axis value in native 3DS units + + std::mutex gyro_mutex; + s16 gyro_x; ///< Gyroscope X-axis value in native 3DS units + s16 gyro_y; ///< Gyroscope Y-axis value in native 3DS units + s16 gyro_z; ///< Gyroscope Z-axis value in native 3DS units + /** * Clip the provided coordinates to be inside the touchscreen area. */ diff --git a/src/core/frontend/motion_emu.cpp b/src/core/frontend/motion_emu.cpp new file mode 100644 index 000000000..9a5b3185d --- /dev/null +++ b/src/core/frontend/motion_emu.cpp @@ -0,0 +1,89 @@ +// Copyright 2016 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include "common/math_util.h" +#include "common/quaternion.h" +#include "core/frontend/emu_window.h" +#include "core/frontend/motion_emu.h" + +namespace Motion { + +static constexpr int update_millisecond = 100; +static constexpr auto update_duration = + std::chrono::duration_cast<std::chrono::steady_clock::duration>( + std::chrono::milliseconds(update_millisecond)); + +MotionEmu::MotionEmu(EmuWindow& emu_window) + : motion_emu_thread(&MotionEmu::MotionEmuThread, this, std::ref(emu_window)) {} + +MotionEmu::~MotionEmu() { + if (motion_emu_thread.joinable()) { + shutdown_event.Set(); + motion_emu_thread.join(); + } +} + +void MotionEmu::MotionEmuThread(EmuWindow& emu_window) { + auto update_time = std::chrono::steady_clock::now(); + Math::Quaternion<float> q = MakeQuaternion(Math::Vec3<float>(), 0); + Math::Quaternion<float> old_q; + + while (!shutdown_event.WaitUntil(update_time)) { + update_time += update_duration; + old_q = q; + + { + std::lock_guard<std::mutex> guard(tilt_mutex); + + // Find the quaternion describing current 3DS tilting + q = MakeQuaternion(Math::MakeVec(-tilt_direction.y, 0.0f, tilt_direction.x), + tilt_angle); + } + + auto inv_q = q.Inverse(); + + // Set the gravity vector in world space + auto gravity = Math::MakeVec(0.0f, -1.0f, 0.0f); + + // Find the angular rate vector in world space + auto angular_rate = ((q - old_q) * inv_q).xyz * 2; + angular_rate *= 1000 / update_millisecond / MathUtil::PI * 180; + + // Transform the two vectors from world space to 3DS space + gravity = QuaternionRotate(inv_q, gravity); + angular_rate = QuaternionRotate(inv_q, angular_rate); + + // Update the sensor state + emu_window.AccelerometerChanged(gravity.x, gravity.y, gravity.z); + emu_window.GyroscopeChanged(angular_rate.x, angular_rate.y, angular_rate.z); + } +} + +void MotionEmu::BeginTilt(int x, int y) { + mouse_origin = Math::MakeVec(x, y); + is_tilting = true; +} + +void MotionEmu::Tilt(int x, int y) { + constexpr float SENSITIVITY = 0.01f; + auto mouse_move = Math::MakeVec(x, y) - mouse_origin; + if (is_tilting) { + std::lock_guard<std::mutex> guard(tilt_mutex); + if (mouse_move.x == 0 && mouse_move.y == 0) { + tilt_angle = 0; + } else { + tilt_direction = mouse_move.Cast<float>(); + tilt_angle = MathUtil::Clamp(tilt_direction.Normalize() * SENSITIVITY, 0.0f, + MathUtil::PI * 0.5f); + } + } +} + +void MotionEmu::EndTilt() { + std::lock_guard<std::mutex> guard(tilt_mutex); + tilt_angle = 0; + is_tilting = false; +} + +} // namespace Motion diff --git a/src/core/frontend/motion_emu.h b/src/core/frontend/motion_emu.h new file mode 100644 index 000000000..99d41a726 --- /dev/null +++ b/src/core/frontend/motion_emu.h @@ -0,0 +1,52 @@ +// Copyright 2016 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once +#include "common/thread.h" +#include "common/vector_math.h" + +class EmuWindow; + +namespace Motion { + +class MotionEmu final { +public: + MotionEmu(EmuWindow& emu_window); + ~MotionEmu(); + + /** + * Signals that a motion sensor tilt has begun. + * @param x the x-coordinate of the cursor + * @param y the y-coordinate of the cursor + */ + void BeginTilt(int x, int y); + + /** + * Signals that a motion sensor tilt is occurring. + * @param x the x-coordinate of the cursor + * @param y the y-coordinate of the cursor + */ + void Tilt(int x, int y); + + /** + * Signals that a motion sensor tilt has ended. + */ + void EndTilt(); + +private: + Math::Vec2<int> mouse_origin; + + std::mutex tilt_mutex; + Math::Vec2<float> tilt_direction; + float tilt_angle = 0; + + bool is_tilting = false; + + Common::Event shutdown_event; + std::thread motion_emu_thread; + + void MotionEmuThread(EmuWindow& emu_window); +}; + +} // namespace Motion diff --git a/src/core/gdbstub/gdbstub.cpp b/src/core/gdbstub/gdbstub.cpp index d88e25073..5cf45ada5 100644 --- a/src/core/gdbstub/gdbstub.cpp +++ b/src/core/gdbstub/gdbstub.cpp @@ -57,7 +57,6 @@ const u32 SIGTERM = 15; const u32 MSG_WAITALL = 8; #endif -const u32 R0_REGISTER = 0; const u32 R15_REGISTER = 15; const u32 CPSR_REGISTER = 25; const u32 FPSCR_REGISTER = 58; @@ -816,10 +815,6 @@ static void RemoveBreakpoint() { auto addr_pos = std::find(start_offset, command_buffer + command_length, ','); PAddr addr = HexToInt(start_offset, static_cast<u32>(addr_pos - start_offset)); - start_offset = addr_pos + 1; - u32 len = - HexToInt(start_offset, static_cast<u32>((command_buffer + command_length) - start_offset)); - if (type == BreakpointType::Access) { // Access is made up of Read and Write types, so add both breakpoints type = BreakpointType::Read; diff --git a/src/core/hle/hle.cpp b/src/core/hle/hle.cpp deleted file mode 100644 index d73d98a70..000000000 --- a/src/core/hle/hle.cpp +++ /dev/null @@ -1,58 +0,0 @@ -// Copyright 2014 Citra Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. - -#include "common/assert.h" -#include "common/logging/log.h" -#include "core/arm/arm_interface.h" -#include "core/core.h" -#include "core/hle/hle.h" -#include "core/hle/service/service.h" - -//////////////////////////////////////////////////////////////////////////////////////////////////// - -namespace { - -bool reschedule; ///< If true, immediately reschedules the CPU to a new thread -} - -namespace HLE { - -void Reschedule(const char* reason) { - DEBUG_ASSERT_MSG(reason != nullptr && strlen(reason) < 256, - "Reschedule: Invalid or too long reason."); - - // TODO(bunnei): It seems that games depend on some CPU execution time elapsing during HLE - // routines. This simulates that time by artificially advancing the number of CPU "ticks". - // The value was chosen empirically, it seems to work well enough for everything tested, but - // is likely not ideal. We should find a more accurate way to simulate timing with HLE. - Core::AppCore().AddTicks(4000); - - Core::AppCore().PrepareReschedule(); - - reschedule = true; -} - -bool IsReschedulePending() { - return reschedule; -} - -void DoneRescheduling() { - reschedule = false; -} - -void Init() { - Service::Init(); - - reschedule = false; - - LOG_DEBUG(Kernel, "initialized OK"); -} - -void Shutdown() { - Service::Shutdown(); - - LOG_DEBUG(Kernel, "shutdown OK"); -} - -} // namespace diff --git a/src/core/hle/kernel/event.cpp b/src/core/hle/kernel/event.cpp index 3e116e3df..23f9df0d6 100644 --- a/src/core/hle/kernel/event.cpp +++ b/src/core/hle/kernel/event.cpp @@ -22,23 +22,17 @@ SharedPtr<Event> Event::Create(ResetType reset_type, std::string name) { evt->reset_type = reset_type; evt->name = std::move(name); - if (reset_type == ResetType::Pulse) { - LOG_ERROR(Kernel, "Unimplemented event reset type Pulse"); - UNIMPLEMENTED(); - } - return evt; } -bool Event::ShouldWait() { +bool Event::ShouldWait(Thread* thread) const { return !signaled; } -void Event::Acquire() { - ASSERT_MSG(!ShouldWait(), "object unavailable!"); +void Event::Acquire(Thread* thread) { + ASSERT_MSG(!ShouldWait(thread), "object unavailable!"); - // Release the event if it's not sticky... - if (reset_type != ResetType::Sticky) + if (reset_type == ResetType::OneShot) signaled = false; } @@ -51,4 +45,11 @@ void Event::Clear() { signaled = false; } +void Event::WakeupAllWaitingThreads() { + WaitObject::WakeupAllWaitingThreads(); + + if (reset_type == ResetType::Pulse) + signaled = false; +} + } // namespace diff --git a/src/core/hle/kernel/event.h b/src/core/hle/kernel/event.h index 8dcd23edb..3e3673508 100644 --- a/src/core/hle/kernel/event.h +++ b/src/core/hle/kernel/event.h @@ -35,8 +35,10 @@ public: bool signaled; ///< Whether the event has already been signaled std::string name; ///< Name of event (optional) - bool ShouldWait() override; - void Acquire() override; + bool ShouldWait(Thread* thread) const override; + void Acquire(Thread* thread) override; + + void WakeupAllWaitingThreads() override; void Signal(); void Clear(); diff --git a/src/core/hle/kernel/kernel.cpp b/src/core/hle/kernel/kernel.cpp index 1db8e102f..f599916f0 100644 --- a/src/core/hle/kernel/kernel.cpp +++ b/src/core/hle/kernel/kernel.cpp @@ -3,7 +3,6 @@ // Refer to the license.txt file included. #include <algorithm> -#include <boost/range/algorithm_ext/erase.hpp> #include "common/assert.h" #include "common/logging/log.h" #include "core/hle/config_mem.h" @@ -28,32 +27,39 @@ void WaitObject::AddWaitingThread(SharedPtr<Thread> thread) { void WaitObject::RemoveWaitingThread(Thread* thread) { auto itr = std::find(waiting_threads.begin(), waiting_threads.end(), thread); + // If a thread passed multiple handles to the same object, + // the kernel might attempt to remove the thread from the object's + // waiting threads list multiple times. if (itr != waiting_threads.end()) waiting_threads.erase(itr); } SharedPtr<Thread> WaitObject::GetHighestPriorityReadyThread() { - // Remove the threads that are ready or already running from our waitlist - boost::range::remove_erase_if(waiting_threads, [](const SharedPtr<Thread>& thread) { - return thread->status == THREADSTATUS_RUNNING || thread->status == THREADSTATUS_READY || - thread->status == THREADSTATUS_DEAD; - }); - - // TODO(Subv): This call should be performed inside the loop below to check if an object can be - // acquired by a particular thread. This is useful for things like recursive locking of Mutexes. - if (ShouldWait()) - return nullptr; - Thread* candidate = nullptr; s32 candidate_priority = THREADPRIO_LOWEST + 1; for (const auto& thread : waiting_threads) { + // The list of waiting threads must not contain threads that are not waiting to be awakened. + ASSERT_MSG(thread->status == THREADSTATUS_WAIT_SYNCH_ANY || + thread->status == THREADSTATUS_WAIT_SYNCH_ALL, + "Inconsistent thread statuses in waiting_threads"); + if (thread->current_priority >= candidate_priority) continue; - bool ready_to_run = - std::none_of(thread->wait_objects.begin(), thread->wait_objects.end(), - [](const SharedPtr<WaitObject>& object) { return object->ShouldWait(); }); + if (ShouldWait(thread.get())) + continue; + + // A thread is ready to run if it's either in THREADSTATUS_WAIT_SYNCH_ANY or + // in THREADSTATUS_WAIT_SYNCH_ALL and the rest of the objects it is waiting on are ready. + bool ready_to_run = true; + if (thread->status == THREADSTATUS_WAIT_SYNCH_ALL) { + ready_to_run = std::none_of(thread->wait_objects.begin(), thread->wait_objects.end(), + [&thread](const SharedPtr<WaitObject>& object) { + return object->ShouldWait(thread.get()); + }); + } + if (ready_to_run) { candidate = thread.get(); candidate_priority = thread->current_priority; @@ -66,7 +72,7 @@ SharedPtr<Thread> WaitObject::GetHighestPriorityReadyThread() { void WaitObject::WakeupAllWaitingThreads() { while (auto thread = GetHighestPriorityReadyThread()) { if (!thread->IsSleepingOnWaitAll()) { - Acquire(); + Acquire(thread.get()); // Set the output index of the WaitSynchronizationN call to the index of this object. if (thread->wait_set_output) { thread->SetWaitSynchronizationOutput(thread->GetWaitObjectIndex(this)); @@ -74,18 +80,17 @@ void WaitObject::WakeupAllWaitingThreads() { } } else { for (auto& object : thread->wait_objects) { - object->Acquire(); - object->RemoveWaitingThread(thread.get()); + object->Acquire(thread.get()); } // Note: This case doesn't update the output index of WaitSynchronizationN. - // Clear the thread's waitlist - thread->wait_objects.clear(); } + for (auto& object : thread->wait_objects) + object->RemoveWaitingThread(thread.get()); + thread->wait_objects.clear(); + thread->SetWaitSynchronizationResult(RESULT_SUCCESS); thread->ResumeFromWait(); - // Note: Removing the thread from the object's waitlist will be - // done by GetHighestPriorityReadyThread. } } diff --git a/src/core/hle/kernel/kernel.h b/src/core/hle/kernel/kernel.h index 9503e7d04..bb8b99bb5 100644 --- a/src/core/hle/kernel/kernel.h +++ b/src/core/hle/kernel/kernel.h @@ -132,31 +132,32 @@ using SharedPtr = boost::intrusive_ptr<T>; class WaitObject : public Object { public: /** - * Check if the current thread should wait until the object is available + * Check if the specified thread should wait until the object is available + * @param thread The thread about which we're deciding. * @return True if the current thread should wait due to this object being unavailable */ - virtual bool ShouldWait() = 0; + virtual bool ShouldWait(Thread* thread) const = 0; - /// Acquire/lock the object if it is available - virtual void Acquire() = 0; + /// Acquire/lock the object for the specified thread if it is available + virtual void Acquire(Thread* thread) = 0; /** * Add a thread to wait on this object * @param thread Pointer to thread to add */ - void AddWaitingThread(SharedPtr<Thread> thread); + virtual void AddWaitingThread(SharedPtr<Thread> thread); /** * Removes a thread from waiting on this object (e.g. if it was resumed already) * @param thread Pointer to thread to remove */ - void RemoveWaitingThread(Thread* thread); + virtual void RemoveWaitingThread(Thread* thread); /** * Wake up all threads waiting on this object that can be awoken, in priority order, * and set the synchronization result and output of the thread. */ - void WakeupAllWaitingThreads(); + virtual void WakeupAllWaitingThreads(); /// Obtains the highest priority thread that is ready to run from this object's waiting list. SharedPtr<Thread> GetHighestPriorityReadyThread(); diff --git a/src/core/hle/kernel/mutex.cpp b/src/core/hle/kernel/mutex.cpp index 736944bae..cef961289 100644 --- a/src/core/hle/kernel/mutex.cpp +++ b/src/core/hle/kernel/mutex.cpp @@ -6,26 +6,18 @@ #include <vector> #include <boost/range/algorithm_ext/erase.hpp> #include "common/assert.h" +#include "core/core.h" #include "core/hle/kernel/kernel.h" #include "core/hle/kernel/mutex.h" #include "core/hle/kernel/thread.h" namespace Kernel { -/** - * Resumes a thread waiting for the specified mutex - * @param mutex The mutex that some thread is waiting on - */ -static void ResumeWaitingThread(Mutex* mutex) { - // Reset mutex lock thread handle, nothing is waiting - mutex->lock_count = 0; - mutex->holding_thread = nullptr; - mutex->WakeupAllWaitingThreads(); -} - void ReleaseThreadMutexes(Thread* thread) { for (auto& mtx : thread->held_mutexes) { - ResumeWaitingThread(mtx.get()); + mtx->lock_count = 0; + mtx->holding_thread = nullptr; + mtx->WakeupAllWaitingThreads(); } thread->held_mutexes.clear(); } @@ -40,52 +32,74 @@ SharedPtr<Mutex> Mutex::Create(bool initial_locked, std::string name) { mutex->name = std::move(name); mutex->holding_thread = nullptr; - // Acquire mutex with current thread if initialized as locked... + // Acquire mutex with current thread if initialized as locked if (initial_locked) - mutex->Acquire(); + mutex->Acquire(GetCurrentThread()); return mutex; } -bool Mutex::ShouldWait() { - auto thread = GetCurrentThread(); - bool wait = lock_count > 0 && holding_thread != thread; - - // If the holding thread of the mutex is lower priority than this thread, that thread should - // temporarily inherit this thread's priority - if (wait && thread->current_priority < holding_thread->current_priority) - holding_thread->BoostPriority(thread->current_priority); - - return wait; -} - -void Mutex::Acquire() { - Acquire(GetCurrentThread()); +bool Mutex::ShouldWait(Thread* thread) const { + return lock_count > 0 && thread != holding_thread; } -void Mutex::Acquire(SharedPtr<Thread> thread) { - ASSERT_MSG(!ShouldWait(), "object unavailable!"); +void Mutex::Acquire(Thread* thread) { + ASSERT_MSG(!ShouldWait(thread), "object unavailable!"); - // Actually "acquire" the mutex only if we don't already have it... + // Actually "acquire" the mutex only if we don't already have it if (lock_count == 0) { + priority = thread->current_priority; thread->held_mutexes.insert(this); - holding_thread = std::move(thread); + holding_thread = thread; + thread->UpdatePriority(); + Core::System::GetInstance().PrepareReschedule(); } lock_count++; } void Mutex::Release() { - // Only release if the mutex is held... + // Only release if the mutex is held if (lock_count > 0) { lock_count--; - // Yield to the next thread only if we've fully released the mutex... + // Yield to the next thread only if we've fully released the mutex if (lock_count == 0) { holding_thread->held_mutexes.erase(this); - ResumeWaitingThread(this); + holding_thread->UpdatePriority(); + holding_thread = nullptr; + WakeupAllWaitingThreads(); + Core::System::GetInstance().PrepareReschedule(); } } } +void Mutex::AddWaitingThread(SharedPtr<Thread> thread) { + WaitObject::AddWaitingThread(thread); + thread->pending_mutexes.insert(this); + UpdatePriority(); +} + +void Mutex::RemoveWaitingThread(Thread* thread) { + WaitObject::RemoveWaitingThread(thread); + thread->pending_mutexes.erase(this); + UpdatePriority(); +} + +void Mutex::UpdatePriority() { + if (!holding_thread) + return; + + s32 best_priority = THREADPRIO_LOWEST; + for (auto& waiter : GetWaitingThreads()) { + if (waiter->current_priority < best_priority) + best_priority = waiter->current_priority; + } + + if (best_priority != priority) { + priority = best_priority; + holding_thread->UpdatePriority(); + } +} + } // namespace diff --git a/src/core/hle/kernel/mutex.h b/src/core/hle/kernel/mutex.h index 53c3dc1f1..c57adf400 100644 --- a/src/core/hle/kernel/mutex.h +++ b/src/core/hle/kernel/mutex.h @@ -35,17 +35,22 @@ public: } int lock_count; ///< Number of times the mutex has been acquired + u32 priority; ///< The priority of the mutex, used for priority inheritance. std::string name; ///< Name of mutex (optional) SharedPtr<Thread> holding_thread; ///< Thread that has acquired the mutex - bool ShouldWait() override; - void Acquire() override; - /** - * Acquires the specified mutex for the specified thread - * @param thread Thread that will acquire the mutex + * Elevate the mutex priority to the best priority + * among the priorities of all its waiting threads. */ - void Acquire(SharedPtr<Thread> thread); + void UpdatePriority(); + + bool ShouldWait(Thread* thread) const override; + void Acquire(Thread* thread) override; + + void AddWaitingThread(SharedPtr<Thread> thread) override; + void RemoveWaitingThread(Thread* thread) override; + void Release(); private: diff --git a/src/core/hle/kernel/resource_limit.cpp b/src/core/hle/kernel/resource_limit.cpp index 253ab7045..3f51bc5de 100644 --- a/src/core/hle/kernel/resource_limit.cpp +++ b/src/core/hle/kernel/resource_limit.cpp @@ -62,6 +62,8 @@ s32 ResourceLimit::GetCurrentResourceValue(u32 resource) const { s32 ResourceLimit::GetMaxResourceValue(u32 resource) const { switch (resource) { + case PRIORITY: + return max_priority; case COMMIT: return max_commit; case THREAD: diff --git a/src/core/hle/kernel/semaphore.cpp b/src/core/hle/kernel/semaphore.cpp index bf7600780..8bda2f75d 100644 --- a/src/core/hle/kernel/semaphore.cpp +++ b/src/core/hle/kernel/semaphore.cpp @@ -30,12 +30,13 @@ ResultVal<SharedPtr<Semaphore>> Semaphore::Create(s32 initial_count, s32 max_cou return MakeResult<SharedPtr<Semaphore>>(std::move(semaphore)); } -bool Semaphore::ShouldWait() { +bool Semaphore::ShouldWait(Thread* thread) const { return available_count <= 0; } -void Semaphore::Acquire() { - ASSERT_MSG(!ShouldWait(), "object unavailable!"); +void Semaphore::Acquire(Thread* thread) { + if (available_count <= 0) + return; --available_count; } diff --git a/src/core/hle/kernel/semaphore.h b/src/core/hle/kernel/semaphore.h index e01908a25..cde94f7cc 100644 --- a/src/core/hle/kernel/semaphore.h +++ b/src/core/hle/kernel/semaphore.h @@ -39,8 +39,8 @@ public: s32 available_count; ///< Number of free slots left in the semaphore std::string name; ///< Name of semaphore (optional) - bool ShouldWait() override; - void Acquire() override; + bool ShouldWait(Thread* thread) const override; + void Acquire(Thread* thread) override; /** * Releases a certain number of slots from a semaphore. diff --git a/src/core/hle/kernel/server_port.cpp b/src/core/hle/kernel/server_port.cpp index 6c19aa7c0..fd3bbbcad 100644 --- a/src/core/hle/kernel/server_port.cpp +++ b/src/core/hle/kernel/server_port.cpp @@ -14,13 +14,13 @@ namespace Kernel { ServerPort::ServerPort() {} ServerPort::~ServerPort() {} -bool ServerPort::ShouldWait() { +bool ServerPort::ShouldWait(Thread* thread) const { // If there are no pending sessions, we wait until a new one is added. return pending_sessions.size() == 0; } -void ServerPort::Acquire() { - ASSERT_MSG(!ShouldWait(), "object unavailable!"); +void ServerPort::Acquire(Thread* thread) { + ASSERT_MSG(!ShouldWait(thread), "object unavailable!"); } std::tuple<SharedPtr<ServerPort>, SharedPtr<ClientPort>> ServerPort::CreatePortPair( diff --git a/src/core/hle/kernel/server_port.h b/src/core/hle/kernel/server_port.h index b0f8df62c..6f8bdb6a9 100644 --- a/src/core/hle/kernel/server_port.h +++ b/src/core/hle/kernel/server_port.h @@ -53,8 +53,8 @@ public: /// ServerSessions created from this port inherit a reference to this handler. std::shared_ptr<Service::SessionRequestHandler> hle_handler; - bool ShouldWait() override; - void Acquire() override; + bool ShouldWait(Thread* thread) const override; + void Acquire(Thread* thread) override; private: ServerPort(); diff --git a/src/core/hle/kernel/server_session.cpp b/src/core/hle/kernel/server_session.cpp index 146458c1c..9447ff236 100644 --- a/src/core/hle/kernel/server_session.cpp +++ b/src/core/hle/kernel/server_session.cpp @@ -29,12 +29,12 @@ ResultVal<SharedPtr<ServerSession>> ServerSession::Create( return MakeResult<SharedPtr<ServerSession>>(std::move(server_session)); } -bool ServerSession::ShouldWait() { +bool ServerSession::ShouldWait(Thread* thread) const { return !signaled; } -void ServerSession::Acquire() { - ASSERT_MSG(!ShouldWait(), "object unavailable!"); +void ServerSession::Acquire(Thread* thread) { + ASSERT_MSG(!ShouldWait(thread), "object unavailable!"); signaled = false; } diff --git a/src/core/hle/kernel/server_session.h b/src/core/hle/kernel/server_session.h index 458284a5d..c088b9a19 100644 --- a/src/core/hle/kernel/server_session.h +++ b/src/core/hle/kernel/server_session.h @@ -57,9 +57,9 @@ public: */ ResultCode HandleSyncRequest(); - bool ShouldWait() override; + bool ShouldWait(Thread* thread) const override; - void Acquire() override; + void Acquire(Thread* thread) override; std::string name; ///< The name of this session (optional) bool signaled; ///< Whether there's new data available to this ServerSession diff --git a/src/core/hle/kernel/thread.cpp b/src/core/hle/kernel/thread.cpp index 5fb95dada..3b7555d87 100644 --- a/src/core/hle/kernel/thread.cpp +++ b/src/core/hle/kernel/thread.cpp @@ -27,12 +27,12 @@ namespace Kernel { /// Event type for the thread wake up event static int ThreadWakeupEventType; -bool Thread::ShouldWait() { +bool Thread::ShouldWait(Thread* thread) const { return status != THREADSTATUS_DEAD; } -void Thread::Acquire() { - ASSERT_MSG(!ShouldWait(), "object unavailable!"); +void Thread::Acquire(Thread* thread) { + ASSERT_MSG(!ShouldWait(thread), "object unavailable!"); } // TODO(yuriks): This can be removed if Thread objects are explicitly pooled in the future, allowing @@ -66,20 +66,6 @@ Thread* GetCurrentThread() { } /** - * Check if a thread is waiting on the specified wait object - * @param thread The thread to test - * @param wait_object The object to test against - * @return True if the thread is waiting, false otherwise - */ -static bool CheckWait_WaitObject(const Thread* thread, WaitObject* wait_object) { - if (thread->status != THREADSTATUS_WAIT_SYNCH) - return false; - - auto itr = std::find(thread->wait_objects.begin(), thread->wait_objects.end(), wait_object); - return itr != thread->wait_objects.end(); -} - -/** * Check if the specified thread is waiting on the specified address to be arbitrated * @param thread The thread to test * @param wait_address The address to test against @@ -90,9 +76,6 @@ static bool CheckWait_AddressArbiter(const Thread* thread, VAddr wait_address) { } void Thread::Stop() { - // Release all the mutexes that this thread holds - ReleaseThreadMutexes(this); - // Cancel any outstanding wakeup events for this thread CoreTiming::UnscheduleEvent(ThreadWakeupEventType, callback_handle); wakeup_callback_handle_table.Close(callback_handle); @@ -114,6 +97,9 @@ void Thread::Stop() { } wait_objects.clear(); + // Release all the mutexes that this thread holds + ReleaseThreadMutexes(this); + // Mark the TLS slot in the thread's page as free. u32 tls_page = (tls_address - Memory::TLS_AREA_VADDR) / Memory::PAGE_SIZE; u32 tls_slot = @@ -155,28 +141,6 @@ void ArbitrateAllThreads(u32 address) { } } -/// Boost low priority threads (temporarily) that have been starved -static void PriorityBoostStarvedThreads() { - u64 current_ticks = CoreTiming::GetTicks(); - - for (auto& thread : thread_list) { - // TODO(bunnei): Threads that have been waiting to be scheduled for `boost_ticks` (or - // longer) will have their priority temporarily adjusted to 1 higher than the highest - // priority thread to prevent thread starvation. This general behavior has been verified - // on hardware. However, this is almost certainly not perfect, and the real CTR OS scheduler - // should probably be reversed to verify this. - - const u64 boost_timeout = 2000000; // Boost threads that have been ready for > this long - - u64 delta = current_ticks - thread->last_running_ticks; - - if (thread->status == THREADSTATUS_READY && delta > boost_timeout) { - const s32 priority = std::max(ready_queue.get_first()->current_priority - 1, 0); - thread->BoostPriority(priority); - } - } -} - /** * Switches the CPU's active thread context to that of the specified thread * @param new_thread The thread to switch to @@ -199,8 +163,8 @@ static void SwitchContext(Thread* new_thread) { // Load context of new thread if (new_thread) { - DEBUG_ASSERT_MSG(new_thread->status == THREADSTATUS_READY, - "Thread must be ready to become running."); + ASSERT_MSG(new_thread->status == THREADSTATUS_READY, + "Thread must be ready to become running."); // Cancel any outstanding wakeup events for this thread CoreTiming::UnscheduleEvent(ThreadWakeupEventType, new_thread->callback_handle); @@ -210,9 +174,6 @@ static void SwitchContext(Thread* new_thread) { ready_queue.remove(new_thread->current_priority, new_thread); new_thread->status = THREADSTATUS_RUNNING; - // Restores thread to its nominal priority if it has been temporarily changed - new_thread->current_priority = new_thread->nominal_priority; - Core::CPU().LoadContext(new_thread->context); Core::CPU().SetCP15Register(CP15_THREAD_URO, new_thread->GetTLSAddress()); } else { @@ -248,14 +209,6 @@ void WaitCurrentThread_Sleep() { thread->status = THREADSTATUS_WAIT_SLEEP; } -void WaitCurrentThread_WaitSynchronization(std::vector<SharedPtr<WaitObject>> wait_objects, - bool wait_set_output) { - Thread* thread = GetCurrentThread(); - thread->wait_set_output = wait_set_output; - thread->wait_objects = std::move(wait_objects); - thread->status = THREADSTATUS_WAIT_SYNCH; -} - void WaitCurrentThread_ArbitrateAddress(VAddr wait_address) { Thread* thread = GetCurrentThread(); thread->wait_address = wait_address; @@ -281,7 +234,8 @@ static void ThreadWakeupCallback(u64 thread_handle, int cycles_late) { return; } - if (thread->status == THREADSTATUS_WAIT_SYNCH || thread->status == THREADSTATUS_WAIT_ARB) { + if (thread->status == THREADSTATUS_WAIT_SYNCH_ANY || + thread->status == THREADSTATUS_WAIT_SYNCH_ALL || thread->status == THREADSTATUS_WAIT_ARB) { thread->wait_set_output = false; // Remove the thread from each of its waiting objects' waitlists for (auto& object : thread->wait_objects) @@ -305,8 +259,11 @@ void Thread::WakeAfterDelay(s64 nanoseconds) { } void Thread::ResumeFromWait() { + ASSERT_MSG(wait_objects.empty(), "Thread is waking up while waiting for objects"); + switch (status) { - case THREADSTATUS_WAIT_SYNCH: + case THREADSTATUS_WAIT_SYNCH_ALL: + case THREADSTATUS_WAIT_SYNCH_ANY: case THREADSTATUS_WAIT_ARB: case THREADSTATUS_WAIT_SLEEP: break; @@ -396,14 +353,8 @@ static void ResetThreadContext(ARM_Interface::ThreadContext& context, u32 stack_ ResultVal<SharedPtr<Thread>> Thread::Create(std::string name, VAddr entry_point, s32 priority, u32 arg, s32 processor_id, VAddr stack_top) { - if (priority < THREADPRIO_HIGHEST || priority > THREADPRIO_LOWEST) { - s32 new_priority = MathUtil::Clamp<s32>(priority, THREADPRIO_HIGHEST, THREADPRIO_LOWEST); - LOG_WARNING(Kernel_SVC, "(name=%s): invalid priority=%d, clamping to %d", name.c_str(), - priority, new_priority); - // TODO(bunnei): Clamping to a valid priority is not necessarily correct behavior... Confirm - // validity of this - priority = new_priority; - } + ASSERT_MSG(priority >= THREADPRIO_HIGHEST && priority <= THREADPRIO_LOWEST, + "Invalid thread priority"); if (!Memory::IsValidVirtualAddress(entry_point)) { LOG_ERROR(Kernel_SVC, "(name=%s): invalid entry %08x", name.c_str(), entry_point); @@ -487,25 +438,9 @@ ResultVal<SharedPtr<Thread>> Thread::Create(std::string name, VAddr entry_point, return MakeResult<SharedPtr<Thread>>(std::move(thread)); } -// TODO(peachum): Remove this. Range checking should be done, and an appropriate error should be -// returned. -static void ClampPriority(const Thread* thread, s32* priority) { - if (*priority < THREADPRIO_HIGHEST || *priority > THREADPRIO_LOWEST) { - DEBUG_ASSERT_MSG( - false, "Application passed an out of range priority. An error should be returned."); - - s32 new_priority = MathUtil::Clamp<s32>(*priority, THREADPRIO_HIGHEST, THREADPRIO_LOWEST); - LOG_WARNING(Kernel_SVC, "(name=%s): invalid priority=%d, clamping to %d", - thread->name.c_str(), *priority, new_priority); - // TODO(bunnei): Clamping to a valid priority is not necessarily correct behavior... Confirm - // validity of this - *priority = new_priority; - } -} - void Thread::SetPriority(s32 priority) { - ClampPriority(this, &priority); - + ASSERT_MSG(priority <= THREADPRIO_LOWEST && priority >= THREADPRIO_HIGHEST, + "Invalid priority value."); // If thread was ready, adjust queues if (status == THREADSTATUS_READY) ready_queue.move(this, current_priority, priority); @@ -515,8 +450,21 @@ void Thread::SetPriority(s32 priority) { nominal_priority = current_priority = priority; } +void Thread::UpdatePriority() { + s32 best_priority = nominal_priority; + for (auto& mutex : held_mutexes) { + if (mutex->priority < best_priority) + best_priority = mutex->priority; + } + BoostPriority(best_priority); +} + void Thread::BoostPriority(s32 priority) { - ready_queue.move(this, current_priority, priority); + // If thread was ready, adjust queues + if (status == THREADSTATUS_READY) + ready_queue.move(this, current_priority, priority); + else + ready_queue.prepare(priority); current_priority = priority; } @@ -538,9 +486,11 @@ SharedPtr<Thread> SetupMainThread(u32 entry_point, s32 priority) { return thread; } -void Reschedule() { - PriorityBoostStarvedThreads(); +bool HaveReadyThreads() { + return ready_queue.get_first() != nullptr; +} +void Reschedule() { Thread* cur = GetCurrentThread(); Thread* next = PopNextReadyThread(); @@ -563,6 +513,12 @@ void Thread::SetWaitSynchronizationOutput(s32 output) { context.cpu_registers[1] = output; } +s32 Thread::GetWaitObjectIndex(WaitObject* object) const { + ASSERT_MSG(!wait_objects.empty(), "Thread is not waiting for anything"); + auto match = std::find(wait_objects.rbegin(), wait_objects.rend(), object); + return std::distance(match, wait_objects.rend()) - 1; +} + //////////////////////////////////////////////////////////////////////////////////////////////////// void ThreadingInit() { diff --git a/src/core/hle/kernel/thread.h b/src/core/hle/kernel/thread.h index c77ac644d..c557a2279 100644 --- a/src/core/hle/kernel/thread.h +++ b/src/core/hle/kernel/thread.h @@ -31,13 +31,14 @@ enum ThreadProcessorId : s32 { }; enum ThreadStatus { - THREADSTATUS_RUNNING, ///< Currently running - THREADSTATUS_READY, ///< Ready to run - THREADSTATUS_WAIT_ARB, ///< Waiting on an address arbiter - THREADSTATUS_WAIT_SLEEP, ///< Waiting due to a SleepThread SVC - THREADSTATUS_WAIT_SYNCH, ///< Waiting due to a WaitSynchronization SVC - THREADSTATUS_DORMANT, ///< Created but not yet made ready - THREADSTATUS_DEAD ///< Run to completion, or forcefully terminated + THREADSTATUS_RUNNING, ///< Currently running + THREADSTATUS_READY, ///< Ready to run + THREADSTATUS_WAIT_ARB, ///< Waiting on an address arbiter + THREADSTATUS_WAIT_SLEEP, ///< Waiting due to a SleepThread SVC + THREADSTATUS_WAIT_SYNCH_ANY, ///< Waiting due to WaitSynch1 or WaitSynchN with wait_all = false + THREADSTATUS_WAIT_SYNCH_ALL, ///< Waiting due to WaitSynchronizationN with wait_all = true + THREADSTATUS_DORMANT, ///< Created but not yet made ready + THREADSTATUS_DEAD ///< Run to completion, or forcefully terminated }; namespace Kernel { @@ -72,8 +73,8 @@ public: return HANDLE_TYPE; } - bool ShouldWait() override; - void Acquire() override; + bool ShouldWait(Thread* thread) const override; + void Acquire(Thread* thread) override; /** * Gets the thread's current priority @@ -90,6 +91,12 @@ public: void SetPriority(s32 priority); /** + * Boost's a thread's priority to the best priority among the thread's held mutexes. + * This prevents priority inversion via priority inheritance. + */ + void UpdatePriority(); + + /** * Temporarily boosts the thread's priority until the next time it is scheduled * @param priority The new priority */ @@ -128,13 +135,14 @@ public: /** * Retrieves the index that this particular object occupies in the list of objects - * that the thread passed to WaitSynchronizationN. + * that the thread passed to WaitSynchronizationN, starting the search from the last element. * It is used to set the output value of WaitSynchronizationN when the thread is awakened. + * When a thread wakes up due to an object signal, the kernel will use the index of the last + * matching object in the wait objects list in case of having multiple instances of the same + * object in the list. * @param object Object to query the index of. */ - s32 GetWaitObjectIndex(const WaitObject* object) const { - return wait_objects_index.at(object->GetObjectId()); - } + s32 GetWaitObjectIndex(WaitObject* object) const; /** * Stops a thread, invalidating it from further use @@ -152,10 +160,10 @@ public: /** * Returns whether this thread is waiting for all the objects in * its wait list to become ready, as a result of a WaitSynchronizationN call - * with wait_all = true, or a ReplyAndReceive call. + * with wait_all = true. */ bool IsSleepingOnWaitAll() const { - return !wait_objects.empty(); + return status == THREADSTATUS_WAIT_SYNCH_ALL; } ARM_Interface::ThreadContext context; @@ -178,15 +186,15 @@ public: /// Mutexes currently held by this thread, which will be released when it exits. boost::container::flat_set<SharedPtr<Mutex>> held_mutexes; + /// Mutexes that this thread is currently waiting for. + boost::container::flat_set<SharedPtr<Mutex>> pending_mutexes; + SharedPtr<Process> owner_process; ///< Process that owns this thread - /// Objects that the thread is waiting on. - /// This is only populated when the thread should wait for all the objects to become ready. + /// Objects that the thread is waiting on, in the same order as they were + // passed to WaitSynchronization1/N. std::vector<SharedPtr<WaitObject>> wait_objects; - /// Mapping of Object ids to their position in the last waitlist that this object waited on. - boost::container::flat_map<int, s32> wait_objects_index; - VAddr wait_address; ///< If waiting on an AddressArbiter, this is the arbitration address /// True if the WaitSynchronizationN output parameter should be set on thread wakeup. @@ -211,6 +219,11 @@ private: SharedPtr<Thread> SetupMainThread(u32 entry_point, s32 priority); /** + * Returns whether there are any threads that are ready to run. + */ +bool HaveReadyThreads(); + +/** * Reschedules to the next available thread (call after current thread is suspended) */ void Reschedule(); @@ -238,15 +251,6 @@ Thread* GetCurrentThread(); void WaitCurrentThread_Sleep(); /** - * Waits the current thread from a WaitSynchronization call - * @param wait_objects Kernel objects that we are waiting on - * @param wait_set_output If true, set the output parameter on thread wakeup (for - * WaitSynchronizationN only) - */ -void WaitCurrentThread_WaitSynchronization(std::vector<SharedPtr<WaitObject>> wait_objects, - bool wait_set_output); - -/** * Waits the current thread from an ArbitrateAddress call * @param wait_address Arbitration address used to resume from wait */ diff --git a/src/core/hle/kernel/timer.cpp b/src/core/hle/kernel/timer.cpp index b50cf520d..60537f355 100644 --- a/src/core/hle/kernel/timer.cpp +++ b/src/core/hle/kernel/timer.cpp @@ -31,20 +31,15 @@ SharedPtr<Timer> Timer::Create(ResetType reset_type, std::string name) { timer->interval_delay = 0; timer->callback_handle = timer_callback_handle_table.Create(timer).MoveFrom(); - if (reset_type == ResetType::Pulse) { - LOG_ERROR(Kernel, "Unimplemented timer reset type Pulse"); - UNIMPLEMENTED(); - } - return timer; } -bool Timer::ShouldWait() { +bool Timer::ShouldWait(Thread* thread) const { return !signaled; } -void Timer::Acquire() { - ASSERT_MSG(!ShouldWait(), "object unavailable!"); +void Timer::Acquire(Thread* thread) { + ASSERT_MSG(!ShouldWait(thread), "object unavailable!"); if (reset_type == ResetType::OneShot) signaled = false; @@ -70,6 +65,13 @@ void Timer::Clear() { signaled = false; } +void Timer::WakeupAllWaitingThreads() { + WaitObject::WakeupAllWaitingThreads(); + + if (reset_type == ResetType::Pulse) + signaled = false; +} + /// The timer callback event, called when a timer is fired static void TimerCallback(u64 timer_handle, int cycles_late) { SharedPtr<Timer> timer = diff --git a/src/core/hle/kernel/timer.h b/src/core/hle/kernel/timer.h index 18ea0236b..c174f5664 100644 --- a/src/core/hle/kernel/timer.h +++ b/src/core/hle/kernel/timer.h @@ -39,8 +39,10 @@ public: u64 initial_delay; ///< The delay until the timer fires for the first time u64 interval_delay; ///< The delay until the timer fires after the first time - bool ShouldWait() override; - void Acquire() override; + bool ShouldWait(Thread* thread) const override; + void Acquire(Thread* thread) override; + + void WakeupAllWaitingThreads() override; /** * Starts the timer, with the specified initial delay and interval. diff --git a/src/core/hle/service/ac/ac.cpp b/src/core/hle/service/ac/ac.cpp new file mode 100644 index 000000000..aa270a2c3 --- /dev/null +++ b/src/core/hle/service/ac/ac.cpp @@ -0,0 +1,181 @@ +// Copyright 2016 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include <array> + +#include "common/logging/log.h" +#include "core/hle/kernel/event.h" +#include "core/hle/service/ac/ac.h" +#include "core/hle/service/ac/ac_i.h" +#include "core/hle/service/ac/ac_u.h" + +namespace Service { +namespace AC { + +struct ACConfig { + std::array<u8, 0x200> data; +}; + +static ACConfig default_config{}; + +static bool ac_connected = false; + +static Kernel::SharedPtr<Kernel::Event> close_event; +static Kernel::SharedPtr<Kernel::Event> connect_event; +static Kernel::SharedPtr<Kernel::Event> disconnect_event; + +void CreateDefaultConfig(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + u32 ac_config_addr = cmd_buff[65]; + + ASSERT_MSG(cmd_buff[64] == (sizeof(ACConfig) << 14 | 2), + "Output buffer size not equal ACConfig size"); + + Memory::WriteBlock(ac_config_addr, &default_config, sizeof(ACConfig)); + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + + LOG_WARNING(Service_AC, "(STUBBED) called"); +} + +void ConnectAsync(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + connect_event = Kernel::g_handle_table.Get<Kernel::Event>(cmd_buff[4]); + if (connect_event) { + connect_event->name = "AC:connect_event"; + connect_event->Signal(); + ac_connected = true; + } + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + + LOG_WARNING(Service_AC, "(STUBBED) called"); +} + +void GetConnectResult(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + + LOG_WARNING(Service_AC, "(STUBBED) called"); +} + +void CloseAsync(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + if (ac_connected && disconnect_event) { + disconnect_event->Signal(); + } + + close_event = Kernel::g_handle_table.Get<Kernel::Event>(cmd_buff[4]); + if (close_event) { + close_event->name = "AC:close_event"; + close_event->Signal(); + } + + ac_connected = false; + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + LOG_WARNING(Service_AC, "(STUBBED) called"); +} + +void GetCloseResult(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + + LOG_WARNING(Service_AC, "(STUBBED) called"); +} + +void GetWifiStatus(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + // TODO(purpasmart96): This function is only a stub, + // it returns a valid result without implementing full functionality. + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + cmd_buff[2] = 0; // Connection type set to none + + LOG_WARNING(Service_AC, "(STUBBED) called"); +} + +void GetInfraPriority(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + cmd_buff[2] = 0; // Infra Priority, default 0 + + LOG_WARNING(Service_AC, "(STUBBED) called"); +} + +void SetRequestEulaVersion(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + u32 major = cmd_buff[1] & 0xFF; + u32 minor = cmd_buff[2] & 0xFF; + + ASSERT_MSG(cmd_buff[3] == (sizeof(ACConfig) << 14 | 2), + "Input buffer size not equal ACConfig size"); + ASSERT_MSG(cmd_buff[64] == (sizeof(ACConfig) << 14 | 2), + "Output buffer size not equal ACConfig size"); + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + cmd_buff[2] = 0; // Infra Priority + + LOG_WARNING(Service_AC, "(STUBBED) called, major=%u, minor=%u", major, minor); +} + +void RegisterDisconnectEvent(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + disconnect_event = Kernel::g_handle_table.Get<Kernel::Event>(cmd_buff[4]); + if (disconnect_event) { + disconnect_event->name = "AC:disconnect_event"; + } + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + + LOG_WARNING(Service_AC, "(STUBBED) called"); +} + +void IsConnected(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + cmd_buff[2] = ac_connected; + + LOG_WARNING(Service_AC, "(STUBBED) called"); +} + +void SetClientVersion(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + const u32 version = cmd_buff[1]; + self->SetVersion(version); + + LOG_WARNING(Service_AC, "(STUBBED) called, version: 0x%08X", version); + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error +} + +void Init() { + AddService(new AC_I); + AddService(new AC_U); + + ac_connected = false; + + close_event = nullptr; + connect_event = nullptr; + disconnect_event = nullptr; +} + +void Shutdown() { + ac_connected = false; + + close_event = nullptr; + connect_event = nullptr; + disconnect_event = nullptr; +} + +} // namespace AC +} // namespace Service diff --git a/src/core/hle/service/ac/ac.h b/src/core/hle/service/ac/ac.h new file mode 100644 index 000000000..6185faf9b --- /dev/null +++ b/src/core/hle/service/ac/ac.h @@ -0,0 +1,134 @@ +// Copyright 2016 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +namespace Service { + +class Interface; + +namespace AC { + +/** + * AC::CreateDefaultConfig service function + * Inputs: + * 64 : ACConfig size << 14 | 2 + * 65 : pointer to ACConfig struct + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void CreateDefaultConfig(Interface* self); + +/** + * AC::ConnectAsync service function + * Inputs: + * 1 : ProcessId Header + * 3 : Copy Handle Header + * 4 : Connection Event handle + * 5 : ACConfig size << 14 | 2 + * 6 : pointer to ACConfig struct + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void ConnectAsync(Interface* self); + +/** + * AC::GetConnectResult service function + * Inputs: + * 1 : ProcessId Header + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void GetConnectResult(Interface* self); + +/** + * AC::CloseAsync service function + * Inputs: + * 1 : ProcessId Header + * 3 : Copy Handle Header + * 4 : Event handle, should be signaled when AC connection is closed + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void CloseAsync(Interface* self); + +/** + * AC::GetCloseResult service function + * Inputs: + * 1 : ProcessId Header + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void GetCloseResult(Interface* self); + +/** + * AC::GetWifiStatus service function + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + * 2 : Output connection type, 0 = none, 1 = Old3DS Internet, 2 = New3DS Internet. + */ +void GetWifiStatus(Interface* self); + +/** + * AC::GetInfraPriority service function + * Inputs: + * 1 : ACConfig size << 14 | 2 + * 2 : pointer to ACConfig struct + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + * 2 : Infra Priority + */ +void GetInfraPriority(Interface* self); + +/** + * AC::SetRequestEulaVersion service function + * Inputs: + * 1 : Eula Version major + * 2 : Eula Version minor + * 3 : ACConfig size << 14 | 2 + * 4 : Input pointer to ACConfig struct + * 64 : ACConfig size << 14 | 2 + * 65 : Output pointer to ACConfig struct + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + * 2 : Infra Priority + */ +void SetRequestEulaVersion(Interface* self); + +/** + * AC::RegisterDisconnectEvent service function + * Inputs: + * 1 : ProcessId Header + * 3 : Copy Handle Header + * 4 : Event handle, should be signaled when AC connection is closed + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void RegisterDisconnectEvent(Interface* self); + +/** + * AC::IsConnected service function + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + * 2 : bool, is connected + */ +void IsConnected(Interface* self); + +/** + * AC::SetClientVersion service function + * Inputs: + * 1 : Used SDK Version + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void SetClientVersion(Interface* self); + +/// Initialize AC service +void Init(); + +/// Shutdown AC service +void Shutdown(); + +} // namespace AC +} // namespace Service diff --git a/src/core/hle/service/ac/ac_i.cpp b/src/core/hle/service/ac/ac_i.cpp new file mode 100644 index 000000000..b22fe3698 --- /dev/null +++ b/src/core/hle/service/ac/ac_i.cpp @@ -0,0 +1,39 @@ +// Copyright 2016 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include "core/hle/service/ac/ac.h" +#include "core/hle/service/ac/ac_i.h" + +namespace Service { +namespace AC { + +const Interface::FunctionInfo FunctionTable[] = { + {0x00010000, CreateDefaultConfig, "CreateDefaultConfig"}, + {0x00040006, ConnectAsync, "ConnectAsync"}, + {0x00050002, GetConnectResult, "GetConnectResult"}, + {0x00070002, nullptr, "CancelConnectAsync"}, + {0x00080004, CloseAsync, "CloseAsync"}, + {0x00090002, GetCloseResult, "GetCloseResult"}, + {0x000A0000, nullptr, "GetLastErrorCode"}, + {0x000C0000, nullptr, "GetStatus"}, + {0x000D0000, GetWifiStatus, "GetWifiStatus"}, + {0x000E0042, nullptr, "GetCurrentAPInfo"}, + {0x00100042, nullptr, "GetCurrentNZoneInfo"}, + {0x00110042, nullptr, "GetNZoneApNumService"}, + {0x001D0042, nullptr, "ScanAPs"}, + {0x00240042, nullptr, "AddDenyApType"}, + {0x00270002, GetInfraPriority, "GetInfraPriority"}, + {0x002D0082, SetRequestEulaVersion, "SetRequestEulaVersion"}, + {0x00300004, RegisterDisconnectEvent, "RegisterDisconnectEvent"}, + {0x003C0042, nullptr, "GetAPSSIDList"}, + {0x003E0042, IsConnected, "IsConnected"}, + {0x00400042, SetClientVersion, "SetClientVersion"}, +}; + +AC_I::AC_I() { + Register(FunctionTable); +} + +} // namespace AC +} // namespace Service diff --git a/src/core/hle/service/ac/ac_i.h b/src/core/hle/service/ac/ac_i.h new file mode 100644 index 000000000..465bba59c --- /dev/null +++ b/src/core/hle/service/ac/ac_i.h @@ -0,0 +1,22 @@ +// Copyright 2016 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#include "core/hle/service/service.h" + +namespace Service { +namespace AC { + +class AC_I final : public Interface { +public: + AC_I(); + + std::string GetPortName() const override { + return "ac:i"; + } +}; + +} // namespace AC +} // namespace Service diff --git a/src/core/hle/service/ac/ac_u.cpp b/src/core/hle/service/ac/ac_u.cpp new file mode 100644 index 000000000..346671b4a --- /dev/null +++ b/src/core/hle/service/ac/ac_u.cpp @@ -0,0 +1,39 @@ +// Copyright 2014 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include "core/hle/service/ac/ac.h" +#include "core/hle/service/ac/ac_u.h" + +namespace Service { +namespace AC { + +const Interface::FunctionInfo FunctionTable[] = { + {0x00010000, CreateDefaultConfig, "CreateDefaultConfig"}, + {0x00040006, ConnectAsync, "ConnectAsync"}, + {0x00050002, GetConnectResult, "GetConnectResult"}, + {0x00070002, nullptr, "CancelConnectAsync"}, + {0x00080004, CloseAsync, "CloseAsync"}, + {0x00090002, GetCloseResult, "GetCloseResult"}, + {0x000A0000, nullptr, "GetLastErrorCode"}, + {0x000C0000, nullptr, "GetStatus"}, + {0x000D0000, GetWifiStatus, "GetWifiStatus"}, + {0x000E0042, nullptr, "GetCurrentAPInfo"}, + {0x00100042, nullptr, "GetCurrentNZoneInfo"}, + {0x00110042, nullptr, "GetNZoneApNumService"}, + {0x001D0042, nullptr, "ScanAPs"}, + {0x00240042, nullptr, "AddDenyApType"}, + {0x00270002, GetInfraPriority, "GetInfraPriority"}, + {0x002D0082, SetRequestEulaVersion, "SetRequestEulaVersion"}, + {0x00300004, RegisterDisconnectEvent, "RegisterDisconnectEvent"}, + {0x003C0042, nullptr, "GetAPSSIDList"}, + {0x003E0042, IsConnected, "IsConnected"}, + {0x00400042, SetClientVersion, "SetClientVersion"}, +}; + +AC_U::AC_U() { + Register(FunctionTable); +} + +} // namespace AC +} // namespace Service diff --git a/src/core/hle/service/ac_u.h b/src/core/hle/service/ac/ac_u.h index 573c32d7e..f9d21e112 100644 --- a/src/core/hle/service/ac_u.h +++ b/src/core/hle/service/ac/ac_u.h @@ -12,7 +12,6 @@ namespace AC { class AC_U final : public Interface { public: AC_U(); - ~AC_U(); std::string GetPortName() const override { return "ac:u"; diff --git a/src/core/hle/service/ac_u.cpp b/src/core/hle/service/ac_u.cpp deleted file mode 100644 index 36204db4d..000000000 --- a/src/core/hle/service/ac_u.cpp +++ /dev/null @@ -1,291 +0,0 @@ -// Copyright 2014 Citra Emulator Project -// Licensed under GPLv2 or any later version -// Refer to the license.txt file included. - -#include <array> - -#include "common/logging/log.h" -#include "core/hle/kernel/event.h" -#include "core/hle/service/ac_u.h" - -namespace Service { -namespace AC { - -struct ACConfig { - std::array<u8, 0x200> data; -}; - -static ACConfig default_config{}; - -static bool ac_connected = false; - -static Kernel::SharedPtr<Kernel::Event> close_event; -static Kernel::SharedPtr<Kernel::Event> connect_event; -static Kernel::SharedPtr<Kernel::Event> disconnect_event; - -/** - * AC_U::CreateDefaultConfig service function - * Inputs: - * 64 : ACConfig size << 14 | 2 - * 65 : pointer to ACConfig struct - * Outputs: - * 1 : Result of function, 0 on success, otherwise error code - */ -static void CreateDefaultConfig(Interface* self) { - u32* cmd_buff = Kernel::GetCommandBuffer(); - - u32 ac_config_addr = cmd_buff[65]; - - ASSERT_MSG(cmd_buff[64] == (sizeof(ACConfig) << 14 | 2), - "Output buffer size not equal ACConfig size"); - - Memory::WriteBlock(ac_config_addr, &default_config, sizeof(ACConfig)); - cmd_buff[1] = RESULT_SUCCESS.raw; // No error - - LOG_WARNING(Service_AC, "(STUBBED) called"); -} - -/** - * AC_U::ConnectAsync service function - * Inputs: - * 1 : ProcessId Header - * 3 : Copy Handle Header - * 4 : Connection Event handle - * 5 : ACConfig size << 14 | 2 - * 6 : pointer to ACConfig struct - * Outputs: - * 1 : Result of function, 0 on success, otherwise error code - */ -static void ConnectAsync(Interface* self) { - u32* cmd_buff = Kernel::GetCommandBuffer(); - - connect_event = Kernel::g_handle_table.Get<Kernel::Event>(cmd_buff[4]); - if (connect_event) { - connect_event->name = "AC_U:connect_event"; - connect_event->Signal(); - ac_connected = true; - } - cmd_buff[1] = RESULT_SUCCESS.raw; // No error - - LOG_WARNING(Service_AC, "(STUBBED) called"); -} - -/** - * AC_U::GetConnectResult service function - * Inputs: - * 1 : ProcessId Header - * Outputs: - * 1 : Result of function, 0 on success, otherwise error code - */ -static void GetConnectResult(Interface* self) { - u32* cmd_buff = Kernel::GetCommandBuffer(); - - cmd_buff[1] = RESULT_SUCCESS.raw; // No error - - LOG_WARNING(Service_AC, "(STUBBED) called"); -} - -/** - * AC_U::CloseAsync service function - * Inputs: - * 1 : ProcessId Header - * 3 : Copy Handle Header - * 4 : Event handle, should be signaled when AC connection is closed - * Outputs: - * 1 : Result of function, 0 on success, otherwise error code - */ -static void CloseAsync(Interface* self) { - u32* cmd_buff = Kernel::GetCommandBuffer(); - - if (ac_connected && disconnect_event) { - disconnect_event->Signal(); - } - - close_event = Kernel::g_handle_table.Get<Kernel::Event>(cmd_buff[4]); - if (close_event) { - close_event->name = "AC_U:close_event"; - close_event->Signal(); - } - - ac_connected = false; - - cmd_buff[1] = RESULT_SUCCESS.raw; // No error - LOG_WARNING(Service_AC, "(STUBBED) called"); -} - -/** - * AC_U::GetCloseResult service function - * Inputs: - * 1 : ProcessId Header - * Outputs: - * 1 : Result of function, 0 on success, otherwise error code - */ -static void GetCloseResult(Interface* self) { - u32* cmd_buff = Kernel::GetCommandBuffer(); - - cmd_buff[1] = RESULT_SUCCESS.raw; // No error - - LOG_WARNING(Service_AC, "(STUBBED) called"); -} - -/** - * AC_U::GetWifiStatus service function - * Outputs: - * 1 : Result of function, 0 on success, otherwise error code - * 2 : Output connection type, 0 = none, 1 = Old3DS Internet, 2 = New3DS Internet. - */ -static void GetWifiStatus(Interface* self) { - u32* cmd_buff = Kernel::GetCommandBuffer(); - - // TODO(purpasmart96): This function is only a stub, - // it returns a valid result without implementing full functionality. - - cmd_buff[1] = RESULT_SUCCESS.raw; // No error - cmd_buff[2] = 0; // Connection type set to none - - LOG_WARNING(Service_AC, "(STUBBED) called"); -} - -/** - * AC_U::GetInfraPriority service function - * Inputs: - * 1 : ACConfig size << 14 | 2 - * 2 : pointer to ACConfig struct - * Outputs: - * 1 : Result of function, 0 on success, otherwise error code - * 2 : Infra Priority - */ -static void GetInfraPriority(Interface* self) { - u32* cmd_buff = Kernel::GetCommandBuffer(); - - cmd_buff[1] = RESULT_SUCCESS.raw; // No error - cmd_buff[2] = 0; // Infra Priority, default 0 - - LOG_WARNING(Service_AC, "(STUBBED) called"); -} - -/** - * AC_U::SetRequestEulaVersion service function - * Inputs: - * 1 : Eula Version major - * 2 : Eula Version minor - * 3 : ACConfig size << 14 | 2 - * 4 : Input pointer to ACConfig struct - * 64 : ACConfig size << 14 | 2 - * 65 : Output pointer to ACConfig struct - * Outputs: - * 1 : Result of function, 0 on success, otherwise error code - * 2 : Infra Priority - */ -static void SetRequestEulaVersion(Interface* self) { - u32* cmd_buff = Kernel::GetCommandBuffer(); - - u32 major = cmd_buff[1] & 0xFF; - u32 minor = cmd_buff[2] & 0xFF; - - ASSERT_MSG(cmd_buff[3] == (sizeof(ACConfig) << 14 | 2), - "Input buffer size not equal ACConfig size"); - ASSERT_MSG(cmd_buff[64] == (sizeof(ACConfig) << 14 | 2), - "Output buffer size not equal ACConfig size"); - - cmd_buff[1] = RESULT_SUCCESS.raw; // No error - cmd_buff[2] = 0; // Infra Priority - - LOG_WARNING(Service_AC, "(STUBBED) called, major=%u, minor=%u", major, minor); -} - -/** - * AC_U::RegisterDisconnectEvent service function - * Inputs: - * 1 : ProcessId Header - * 3 : Copy Handle Header - * 4 : Event handle, should be signaled when AC connection is closed - * Outputs: - * 1 : Result of function, 0 on success, otherwise error code - */ -static void RegisterDisconnectEvent(Interface* self) { - u32* cmd_buff = Kernel::GetCommandBuffer(); - - disconnect_event = Kernel::g_handle_table.Get<Kernel::Event>(cmd_buff[4]); - if (disconnect_event) { - disconnect_event->name = "AC_U:disconnect_event"; - } - cmd_buff[1] = RESULT_SUCCESS.raw; // No error - - LOG_WARNING(Service_AC, "(STUBBED) called"); -} - -/** - * AC_U::IsConnected service function - * Outputs: - * 1 : Result of function, 0 on success, otherwise error code - * 2 : bool, is connected - */ -static void IsConnected(Interface* self) { - u32* cmd_buff = Kernel::GetCommandBuffer(); - - cmd_buff[1] = RESULT_SUCCESS.raw; // No error - cmd_buff[2] = ac_connected; - - LOG_WARNING(Service_AC, "(STUBBED) called"); -} - -/** - * AC_U::SetClientVersion service function - * Inputs: - * 1 : Used SDK Version - * Outputs: - * 1 : Result of function, 0 on success, otherwise error code - */ -static void SetClientVersion(Interface* self) { - u32* cmd_buff = Kernel::GetCommandBuffer(); - - const u32 version = cmd_buff[1]; - self->SetVersion(version); - - LOG_WARNING(Service_AC, "(STUBBED) called, version: 0x%08X", version); - - cmd_buff[1] = RESULT_SUCCESS.raw; // No error -} - -const Interface::FunctionInfo FunctionTable[] = { - {0x00010000, CreateDefaultConfig, "CreateDefaultConfig"}, - {0x00040006, ConnectAsync, "ConnectAsync"}, - {0x00050002, GetConnectResult, "GetConnectResult"}, - {0x00070002, nullptr, "CancelConnectAsync"}, - {0x00080004, CloseAsync, "CloseAsync"}, - {0x00090002, GetCloseResult, "GetCloseResult"}, - {0x000A0000, nullptr, "GetLastErrorCode"}, - {0x000C0000, nullptr, "GetStatus"}, - {0x000D0000, GetWifiStatus, "GetWifiStatus"}, - {0x000E0042, nullptr, "GetCurrentAPInfo"}, - {0x00100042, nullptr, "GetCurrentNZoneInfo"}, - {0x00110042, nullptr, "GetNZoneApNumService"}, - {0x001D0042, nullptr, "ScanAPs"}, - {0x00240042, nullptr, "AddDenyApType"}, - {0x00270002, GetInfraPriority, "GetInfraPriority"}, - {0x002D0082, SetRequestEulaVersion, "SetRequestEulaVersion"}, - {0x00300004, RegisterDisconnectEvent, "RegisterDisconnectEvent"}, - {0x003C0042, nullptr, "GetAPSSIDList"}, - {0x003E0042, IsConnected, "IsConnected"}, - {0x00400042, SetClientVersion, "SetClientVersion"}, -}; - -AC_U::AC_U() { - Register(FunctionTable); - - ac_connected = false; - - close_event = nullptr; - connect_event = nullptr; - disconnect_event = nullptr; -} - -AC_U::~AC_U() { - close_event = nullptr; - connect_event = nullptr; - disconnect_event = nullptr; -} - -} // namespace AC -} // namespace Service diff --git a/src/core/hle/service/boss/boss.cpp b/src/core/hle/service/boss/boss.cpp index 6ab16ccd5..e0de037f8 100644 --- a/src/core/hle/service/boss/boss.cpp +++ b/src/core/hle/service/boss/boss.cpp @@ -2,6 +2,7 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include <cinttypes> #include "core/hle/service/boss/boss.h" #include "core/hle/service/boss/boss_p.h" #include "core/hle/service/boss/boss_u.h" @@ -33,7 +34,8 @@ void InitializeSession(Service::Interface* self) { cmd_buff[0] = IPC::MakeHeader(0x1, 0x1, 0); cmd_buff[1] = RESULT_SUCCESS.raw; - LOG_WARNING(Service_BOSS, "(STUBBED) unk_param=0x%016X, translation=0x%08X, unk_param4=0x%08X", + LOG_WARNING(Service_BOSS, + "(STUBBED) unk_param=0x%016" PRIX64 ", translation=0x%08X, unk_param4=0x%08X", unk_param, translation, unk_param4); } diff --git a/src/core/hle/service/cfg/cfg.cpp b/src/core/hle/service/cfg/cfg.cpp index 65655f45d..6f13cde27 100644 --- a/src/core/hle/service/cfg/cfg.cpp +++ b/src/core/hle/service/cfg/cfg.cpp @@ -84,7 +84,6 @@ struct ConsoleCountryInfo { static_assert(sizeof(ConsoleCountryInfo) == 4, "ConsoleCountryInfo must be exactly 4 bytes"); } -static const u64 CFG_SAVE_ID = 0x00010017; static const u64 CONSOLE_UNIQUE_ID = 0xDEADC0DE; static const ConsoleModelInfo CONSOLE_MODEL = {NINTENDO_3DS_XL, {0, 0, 0}}; static const u8 CONSOLE_LANGUAGE = LANGUAGE_EN; @@ -115,6 +114,8 @@ static const std::vector<u8> cfg_system_savedata_id = { 0x00, 0x00, 0x00, 0x00, 0x17, 0x00, 0x01, 0x00, }; +static u32 preferred_region_code = 0; + void GetCountryCodeString(Service::Interface* self) { u32* cmd_buff = Kernel::GetCommandBuffer(); u32 country_code_id = cmd_buff[1]; @@ -160,11 +161,18 @@ void GetCountryCodeID(Service::Interface* self) { cmd_buff[2] = country_code_id; } +static u32 GetRegionValue() { + if (Settings::values.region_value == Settings::REGION_VALUE_AUTO_SELECT) + return preferred_region_code; + + return Settings::values.region_value; +} + void SecureInfoGetRegion(Service::Interface* self) { u32* cmd_buff = Kernel::GetCommandBuffer(); cmd_buff[1] = RESULT_SUCCESS.raw; - cmd_buff[2] = Settings::values.region_value; + cmd_buff[2] = GetRegionValue(); } void GenHashConsoleUnique(Service::Interface* self) { @@ -184,7 +192,7 @@ void GetRegionCanadaUSA(Service::Interface* self) { cmd_buff[1] = RESULT_SUCCESS.raw; u8 canada_or_usa = 1; - if (canada_or_usa == Settings::values.region_value) { + if (canada_or_usa == GetRegionValue()) { cmd_buff[2] = 1; } else { cmd_buff[2] = 0; @@ -318,6 +326,7 @@ ResultCode GetConfigInfoBlock(u32 block_id, u32 size, u32 flag, void* output) { void* pointer; CASCADE_RESULT(pointer, GetConfigInfoBlockPointer(block_id, size, flag)); memcpy(output, pointer, size); + return RESULT_SUCCESS; } @@ -535,10 +544,55 @@ void Init() { AddService(new CFG_U); LoadConfigNANDSaveFile(); + + preferred_region_code = 0; } void Shutdown() {} +/// Checks if the language is available in the chosen region, and returns a proper one +static SystemLanguage AdjustLanguageInfoBlock(u32 region, SystemLanguage language) { + static const std::array<std::vector<SystemLanguage>, 7> region_languages{{ + // JPN + {LANGUAGE_JP}, + // USA + {LANGUAGE_EN, LANGUAGE_FR, LANGUAGE_ES, LANGUAGE_PT}, + // EUR + {LANGUAGE_EN, LANGUAGE_FR, LANGUAGE_DE, LANGUAGE_IT, LANGUAGE_ES, LANGUAGE_NL, LANGUAGE_PT, + LANGUAGE_RU}, + // AUS + {LANGUAGE_EN, LANGUAGE_FR, LANGUAGE_DE, LANGUAGE_IT, LANGUAGE_ES, LANGUAGE_NL, LANGUAGE_PT, + LANGUAGE_RU}, + // CHN + {LANGUAGE_ZH}, + // KOR + {LANGUAGE_KO}, + // TWN + {LANGUAGE_TW}, + }}; + const auto& available = region_languages[region]; + if (std::find(available.begin(), available.end(), language) == available.end()) { + return available[0]; + } + return language; +} + +void SetPreferredRegionCode(u32 region_code) { + preferred_region_code = region_code; + LOG_INFO(Service_CFG, "Preferred region code set to %u", preferred_region_code); + + if (Settings::values.region_value == Settings::REGION_VALUE_AUTO_SELECT) { + const SystemLanguage current_language = GetSystemLanguage(); + const SystemLanguage adjusted_language = + AdjustLanguageInfoBlock(region_code, current_language); + if (current_language != adjusted_language) { + LOG_WARNING(Service_CFG, "System language %d does not fit the region. Adjusted to %d", + static_cast<int>(current_language), static_cast<int>(adjusted_language)); + SetSystemLanguage(adjusted_language); + } + } +} + void SetUsername(const std::u16string& name) { ASSERT(name.size() <= 10); UsernameBlock block{}; diff --git a/src/core/hle/service/cfg/cfg.h b/src/core/hle/service/cfg/cfg.h index fb47c2aa5..618c9647e 100644 --- a/src/core/hle/service/cfg/cfg.h +++ b/src/core/hle/service/cfg/cfg.h @@ -282,6 +282,13 @@ void Init(); /// Shutdown the config service void Shutdown(); +/** + * Set the region code preferred by the game so that CFG will adjust to it when the region setting + * is auto. + * @param region_code the preferred region code to set + */ +void SetPreferredRegionCode(u32 region_code); + // Utilities for frontend to set config data. // Note: before calling these functions, LoadConfigNANDSaveFile should be called, // and UpdateConfigNANDSavegame should be called after making changes to config data. diff --git a/src/core/hle/service/err_f.cpp b/src/core/hle/service/err_f.cpp index cd0a1a598..9da55f328 100644 --- a/src/core/hle/service/err_f.cpp +++ b/src/core/hle/service/err_f.cpp @@ -227,6 +227,8 @@ static void ThrowFatalError(Interface* self) { LOG_CRITICAL(Service_ERR, "FINST2: 0x%08X", errtype.exception_data.exception_info.fpinst2); break; + case ExceptionType::Undefined: + break; // Not logging exception_info for this case } LOG_CRITICAL(Service_ERR, "Datetime: %s", GetCurrentSystemTime().c_str()); break; diff --git a/src/core/hle/service/gsp_gpu.cpp b/src/core/hle/service/gsp_gpu.cpp index 947958703..a8c1331ed 100644 --- a/src/core/hle/service/gsp_gpu.cpp +++ b/src/core/hle/service/gsp_gpu.cpp @@ -149,7 +149,7 @@ static ResultCode WriteHWRegsWithMask(u32 base_address, u32 size_in_bytes, VAddr u32 mask = Memory::Read32(masks_vaddr); // Update the current value of the register only for set mask bits - reg_value = (reg_value & ~mask) | (data | mask); + reg_value = (reg_value & ~mask) | (data & mask); WriteSingleHWReg(base_address, reg_value); diff --git a/src/core/hle/service/hid/hid.cpp b/src/core/hle/service/hid/hid.cpp index 676154bd4..f14ab3811 100644 --- a/src/core/hle/service/hid/hid.cpp +++ b/src/core/hle/service/hid/hid.cpp @@ -35,6 +35,15 @@ static u32 next_gyroscope_index; static int enable_accelerometer_count = 0; // positive means enabled static int enable_gyroscope_count = 0; // positive means enabled +static int pad_update_event; +static int accelerometer_update_event; +static int gyroscope_update_event; + +// Updating period for each HID device. These empirical values are measured from a 11.2 3DS. +constexpr u64 pad_update_ticks = BASE_CLOCK_RATE_ARM11 / 234; +constexpr u64 accelerometer_update_ticks = BASE_CLOCK_RATE_ARM11 / 104; +constexpr u64 gyroscope_update_ticks = BASE_CLOCK_RATE_ARM11 / 101; + static PadState GetCirclePadDirectionState(s16 circle_pad_x, s16 circle_pad_y) { // 30 degree and 60 degree are angular thresholds for directions constexpr float TAN30 = 0.577350269f; @@ -65,14 +74,9 @@ static PadState GetCirclePadDirectionState(s16 circle_pad_x, s16 circle_pad_y) { return state; } -void Update() { +static void UpdatePadCallback(u64 userdata, int cycles_late) { SharedMem* mem = reinterpret_cast<SharedMem*>(shared_mem->GetPointer()); - if (mem == nullptr) { - LOG_DEBUG(Service_HID, "Cannot update HID prior to mapping shared memory!"); - return; - } - PadState state = VideoCore::g_emu_window->GetPadState(); // Get current circle pad position and update circle pad direction @@ -131,59 +135,68 @@ void Update() { event_pad_or_touch_1->Signal(); event_pad_or_touch_2->Signal(); - // Update accelerometer - if (enable_accelerometer_count > 0) { - mem->accelerometer.index = next_accelerometer_index; - next_accelerometer_index = - (next_accelerometer_index + 1) % mem->accelerometer.entries.size(); - - AccelerometerDataEntry& accelerometer_entry = - mem->accelerometer.entries[mem->accelerometer.index]; - std::tie(accelerometer_entry.x, accelerometer_entry.y, accelerometer_entry.z) = - VideoCore::g_emu_window->GetAccelerometerState(); - - // Make up "raw" entry - // TODO(wwylele): - // From hardware testing, the raw_entry values are approximately, - // but not exactly, as twice as corresponding entries (or with a minus sign). - // It may caused by system calibration to the accelerometer. - // Figure out how it works, or, if no game reads raw_entry, - // the following three lines can be removed and leave raw_entry unimplemented. - mem->accelerometer.raw_entry.x = -2 * accelerometer_entry.x; - mem->accelerometer.raw_entry.z = 2 * accelerometer_entry.y; - mem->accelerometer.raw_entry.y = -2 * accelerometer_entry.z; - - // If we just updated index 0, provide a new timestamp - if (mem->accelerometer.index == 0) { - mem->accelerometer.index_reset_ticks_previous = mem->accelerometer.index_reset_ticks; - mem->accelerometer.index_reset_ticks = (s64)CoreTiming::GetTicks(); - } + // Reschedule recurrent event + CoreTiming::ScheduleEvent(pad_update_ticks - cycles_late, pad_update_event); +} + +static void UpdateAccelerometerCallback(u64 userdata, int cycles_late) { + SharedMem* mem = reinterpret_cast<SharedMem*>(shared_mem->GetPointer()); + + mem->accelerometer.index = next_accelerometer_index; + next_accelerometer_index = (next_accelerometer_index + 1) % mem->accelerometer.entries.size(); - event_accelerometer->Signal(); + AccelerometerDataEntry& accelerometer_entry = + mem->accelerometer.entries[mem->accelerometer.index]; + std::tie(accelerometer_entry.x, accelerometer_entry.y, accelerometer_entry.z) = + VideoCore::g_emu_window->GetAccelerometerState(); + + // Make up "raw" entry + // TODO(wwylele): + // From hardware testing, the raw_entry values are approximately, but not exactly, as twice as + // corresponding entries (or with a minus sign). It may caused by system calibration to the + // accelerometer. Figure out how it works, or, if no game reads raw_entry, the following three + // lines can be removed and leave raw_entry unimplemented. + mem->accelerometer.raw_entry.x = -2 * accelerometer_entry.x; + mem->accelerometer.raw_entry.z = 2 * accelerometer_entry.y; + mem->accelerometer.raw_entry.y = -2 * accelerometer_entry.z; + + // If we just updated index 0, provide a new timestamp + if (mem->accelerometer.index == 0) { + mem->accelerometer.index_reset_ticks_previous = mem->accelerometer.index_reset_ticks; + mem->accelerometer.index_reset_ticks = (s64)CoreTiming::GetTicks(); } - // Update gyroscope - if (enable_gyroscope_count > 0) { - mem->gyroscope.index = next_gyroscope_index; - next_gyroscope_index = (next_gyroscope_index + 1) % mem->gyroscope.entries.size(); + event_accelerometer->Signal(); - GyroscopeDataEntry& gyroscope_entry = mem->gyroscope.entries[mem->gyroscope.index]; - std::tie(gyroscope_entry.x, gyroscope_entry.y, gyroscope_entry.z) = - VideoCore::g_emu_window->GetGyroscopeState(); + // Reschedule recurrent event + CoreTiming::ScheduleEvent(accelerometer_update_ticks - cycles_late, accelerometer_update_event); +} - // Make up "raw" entry - mem->gyroscope.raw_entry.x = gyroscope_entry.x; - mem->gyroscope.raw_entry.z = -gyroscope_entry.y; - mem->gyroscope.raw_entry.y = gyroscope_entry.z; +static void UpdateGyroscopeCallback(u64 userdata, int cycles_late) { + SharedMem* mem = reinterpret_cast<SharedMem*>(shared_mem->GetPointer()); - // If we just updated index 0, provide a new timestamp - if (mem->gyroscope.index == 0) { - mem->gyroscope.index_reset_ticks_previous = mem->gyroscope.index_reset_ticks; - mem->gyroscope.index_reset_ticks = (s64)CoreTiming::GetTicks(); - } + mem->gyroscope.index = next_gyroscope_index; + next_gyroscope_index = (next_gyroscope_index + 1) % mem->gyroscope.entries.size(); + + GyroscopeDataEntry& gyroscope_entry = mem->gyroscope.entries[mem->gyroscope.index]; + std::tie(gyroscope_entry.x, gyroscope_entry.y, gyroscope_entry.z) = + VideoCore::g_emu_window->GetGyroscopeState(); + + // Make up "raw" entry + mem->gyroscope.raw_entry.x = gyroscope_entry.x; + mem->gyroscope.raw_entry.z = -gyroscope_entry.y; + mem->gyroscope.raw_entry.y = gyroscope_entry.z; - event_gyroscope->Signal(); + // If we just updated index 0, provide a new timestamp + if (mem->gyroscope.index == 0) { + mem->gyroscope.index_reset_ticks_previous = mem->gyroscope.index_reset_ticks; + mem->gyroscope.index_reset_ticks = (s64)CoreTiming::GetTicks(); } + + event_gyroscope->Signal(); + + // Reschedule recurrent event + CoreTiming::ScheduleEvent(gyroscope_update_ticks - cycles_late, gyroscope_update_event); } void GetIPCHandles(Service::Interface* self) { @@ -204,7 +217,11 @@ void EnableAccelerometer(Service::Interface* self) { u32* cmd_buff = Kernel::GetCommandBuffer(); ++enable_accelerometer_count; - event_accelerometer->Signal(); + + // Schedules the accelerometer update event if the accelerometer was just enabled + if (enable_accelerometer_count == 1) { + CoreTiming::ScheduleEvent(accelerometer_update_ticks, accelerometer_update_event); + } cmd_buff[1] = RESULT_SUCCESS.raw; @@ -215,7 +232,11 @@ void DisableAccelerometer(Service::Interface* self) { u32* cmd_buff = Kernel::GetCommandBuffer(); --enable_accelerometer_count; - event_accelerometer->Signal(); + + // Unschedules the accelerometer update event if the accelerometer was just disabled + if (enable_accelerometer_count == 0) { + CoreTiming::UnscheduleEvent(accelerometer_update_event, 0); + } cmd_buff[1] = RESULT_SUCCESS.raw; @@ -226,7 +247,11 @@ void EnableGyroscopeLow(Service::Interface* self) { u32* cmd_buff = Kernel::GetCommandBuffer(); ++enable_gyroscope_count; - event_gyroscope->Signal(); + + // Schedules the gyroscope update event if the gyroscope was just enabled + if (enable_gyroscope_count == 1) { + CoreTiming::ScheduleEvent(gyroscope_update_ticks, gyroscope_update_event); + } cmd_buff[1] = RESULT_SUCCESS.raw; @@ -237,7 +262,11 @@ void DisableGyroscopeLow(Service::Interface* self) { u32* cmd_buff = Kernel::GetCommandBuffer(); --enable_gyroscope_count; - event_gyroscope->Signal(); + + // Unschedules the gyroscope update event if the gyroscope was just disabled + if (enable_gyroscope_count == 0) { + CoreTiming::UnscheduleEvent(gyroscope_update_event, 0); + } cmd_buff[1] = RESULT_SUCCESS.raw; @@ -291,6 +320,8 @@ void Init() { next_pad_index = 0; next_touch_index = 0; + next_accelerometer_index = 0; + next_gyroscope_index = 0; // Create event handles event_pad_or_touch_1 = Event::Create(ResetType::OneShot, "HID:EventPadOrTouch1"); @@ -298,6 +329,15 @@ void Init() { event_accelerometer = Event::Create(ResetType::OneShot, "HID:EventAccelerometer"); event_gyroscope = Event::Create(ResetType::OneShot, "HID:EventGyroscope"); event_debug_pad = Event::Create(ResetType::OneShot, "HID:EventDebugPad"); + + // Register update callbacks + pad_update_event = CoreTiming::RegisterEvent("HID::UpdatePadCallback", UpdatePadCallback); + accelerometer_update_event = + CoreTiming::RegisterEvent("HID::UpdateAccelerometerCallback", UpdateAccelerometerCallback); + gyroscope_update_event = + CoreTiming::RegisterEvent("HID::UpdateGyroscopeCallback", UpdateGyroscopeCallback); + + CoreTiming::ScheduleEvent(pad_update_ticks, pad_update_event); } void Shutdown() { diff --git a/src/core/hle/service/hid/hid.h b/src/core/hle/service/hid/hid.h index 7904e7355..21e66dfe0 100644 --- a/src/core/hle/service/hid/hid.h +++ b/src/core/hle/service/hid/hid.h @@ -296,9 +296,6 @@ void GetGyroscopeLowRawToDpsCoefficient(Service::Interface* self); */ void GetGyroscopeLowCalibrateParam(Service::Interface* self); -/// Checks for user input updates -void Update(); - /// Initialize HID service void Init(); diff --git a/src/core/hle/service/mic_u.cpp b/src/core/hle/service/mic_u.cpp index 4f1dd2fce..e98388560 100644 --- a/src/core/hle/service/mic_u.cpp +++ b/src/core/hle/service/mic_u.cpp @@ -93,13 +93,14 @@ static void StartSampling(Interface* self) { sample_rate = static_cast<SampleRate>(cmd_buff[2] & 0xFF); audio_buffer_offset = cmd_buff[3]; audio_buffer_size = cmd_buff[4]; - audio_buffer_loop = static_cast<bool>(cmd_buff[5] & 0xFF); + audio_buffer_loop = (cmd_buff[5] & 0xFF) != 0; cmd_buff[1] = RESULT_SUCCESS.raw; // No error is_sampling = true; LOG_WARNING(Service_MIC, "(STUBBED) called, encoding=%u, sample_rate=%u, " "audio_buffer_offset=%d, audio_buffer_size=%u, audio_buffer_loop=%u", - encoding, sample_rate, audio_buffer_offset, audio_buffer_size, audio_buffer_loop); + static_cast<u32>(encoding), static_cast<u32>(sample_rate), audio_buffer_offset, + audio_buffer_size, audio_buffer_loop); } /** @@ -114,7 +115,7 @@ static void AdjustSampling(Interface* self) { u32* cmd_buff = Kernel::GetCommandBuffer(); sample_rate = static_cast<SampleRate>(cmd_buff[1] & 0xFF); cmd_buff[1] = RESULT_SUCCESS.raw; // No error - LOG_WARNING(Service_MIC, "(STUBBED) called, sample_rate=%u", sample_rate); + LOG_WARNING(Service_MIC, "(STUBBED) called, sample_rate=%u", static_cast<u32>(sample_rate)); } /** @@ -201,7 +202,7 @@ static void GetGain(Interface* self) { */ static void SetPower(Interface* self) { u32* cmd_buff = Kernel::GetCommandBuffer(); - mic_power = static_cast<bool>(cmd_buff[1] & 0xFF); + mic_power = (cmd_buff[1] & 0xFF) != 0; cmd_buff[1] = RESULT_SUCCESS.raw; // No error LOG_WARNING(Service_MIC, "(STUBBED) called, mic_power=%u", mic_power); } @@ -251,7 +252,7 @@ static void SetIirFilterMic(Interface* self) { */ static void SetClamp(Interface* self) { u32* cmd_buff = Kernel::GetCommandBuffer(); - clamp = static_cast<bool>(cmd_buff[1] & 0xFF); + clamp = (cmd_buff[1] & 0xFF) != 0; cmd_buff[1] = RESULT_SUCCESS.raw; // No error LOG_WARNING(Service_MIC, "(STUBBED) called, clamp=%u", clamp); } @@ -281,7 +282,7 @@ static void GetClamp(Interface* self) { */ static void SetAllowShellClosed(Interface* self) { u32* cmd_buff = Kernel::GetCommandBuffer(); - allow_shell_closed = static_cast<bool>(cmd_buff[1] & 0xFF); + allow_shell_closed = (cmd_buff[1] & 0xFF) != 0; cmd_buff[1] = RESULT_SUCCESS.raw; // No error LOG_WARNING(Service_MIC, "(STUBBED) called, allow_shell_closed=%u", allow_shell_closed); } diff --git a/src/core/hle/service/nfc/nfc.cpp b/src/core/hle/service/nfc/nfc.cpp index d9738c6a1..fd3c7d9c2 100644 --- a/src/core/hle/service/nfc/nfc.cpp +++ b/src/core/hle/service/nfc/nfc.cpp @@ -2,6 +2,7 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include "core/hle/kernel/event.h" #include "core/hle/service/nfc/nfc.h" #include "core/hle/service/nfc/nfc_m.h" #include "core/hle/service/nfc/nfc_u.h" @@ -9,9 +10,133 @@ namespace Service { namespace NFC { +static Kernel::SharedPtr<Kernel::Event> tag_in_range_event; +static Kernel::SharedPtr<Kernel::Event> tag_out_of_range_event; +static TagState nfc_tag_state = TagState::NotInitialized; +static CommunicationStatus nfc_status = CommunicationStatus::NfcInitialized; + +void Initialize(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + u8 param = static_cast<u8>(cmd_buff[1] & 0xFF); + + nfc_tag_state = TagState::NotScanning; + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + LOG_WARNING(Service_NFC, "(STUBBED) called, param=%u", param); +} + +void Shutdown(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + u8 param = static_cast<u8>(cmd_buff[1] & 0xFF); + nfc_tag_state = TagState::NotInitialized; + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + LOG_WARNING(Service_NFC, "(STUBBED) called, param=%u", param); +} + +void StartCommunication(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + LOG_WARNING(Service_NFC, "(STUBBED) called"); +} + +void StopCommunication(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + LOG_WARNING(Service_NFC, "(STUBBED) called"); +} + +void StartTagScanning(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + nfc_tag_state = TagState::TagInRange; + tag_in_range_event->Signal(); + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + LOG_WARNING(Service_NFC, "(STUBBED) called"); +} + +void StopTagScanning(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + nfc_tag_state = TagState::NotScanning; + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + LOG_WARNING(Service_NFC, "(STUBBED) called"); +} + +void LoadAmiiboData(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + nfc_tag_state = TagState::TagDataLoaded; + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + LOG_WARNING(Service_NFC, "(STUBBED) called"); +} + +void ResetTagScanState(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + nfc_tag_state = TagState::NotScanning; + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + LOG_WARNING(Service_NFC, "(STUBBED) called"); +} + +void GetTagInRangeEvent(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + cmd_buff[0] = IPC::MakeHeader(0xB, 1, 2); + cmd_buff[1] = RESULT_SUCCESS.raw; + cmd_buff[2] = IPC::CopyHandleDesc(); + cmd_buff[3] = Kernel::g_handle_table.Create(tag_in_range_event).MoveFrom(); + LOG_WARNING(Service_NFC, "(STUBBED) called"); +} + +void GetTagOutOfRangeEvent(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + cmd_buff[0] = IPC::MakeHeader(0xC, 1, 2); + cmd_buff[1] = RESULT_SUCCESS.raw; + cmd_buff[2] = IPC::CopyHandleDesc(); + cmd_buff[3] = Kernel::g_handle_table.Create(tag_out_of_range_event).MoveFrom(); + LOG_WARNING(Service_NFC, "(STUBBED) called"); +} + +void GetTagState(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + cmd_buff[2] = static_cast<u8>(nfc_tag_state); + LOG_DEBUG(Service_NFC, "(STUBBED) called"); +} + +void CommunicationGetStatus(Interface* self) { + u32* cmd_buff = Kernel::GetCommandBuffer(); + + cmd_buff[1] = RESULT_SUCCESS.raw; // No error + cmd_buff[2] = static_cast<u8>(nfc_status); + LOG_DEBUG(Service_NFC, "(STUBBED) called"); +} + void Init() { AddService(new NFC_M()); AddService(new NFC_U()); + + tag_in_range_event = + Kernel::Event::Create(Kernel::ResetType::OneShot, "NFC::tag_in_range_event"); + tag_out_of_range_event = + Kernel::Event::Create(Kernel::ResetType::OneShot, "NFC::tag_out_range_event"); + nfc_tag_state = TagState::NotInitialized; +} + +void Shutdown() { + tag_in_range_event = nullptr; + tag_out_of_range_event = nullptr; } } // namespace NFC diff --git a/src/core/hle/service/nfc/nfc.h b/src/core/hle/service/nfc/nfc.h index cd65a5fdc..a013bdae7 100644 --- a/src/core/hle/service/nfc/nfc.h +++ b/src/core/hle/service/nfc/nfc.h @@ -4,11 +4,150 @@ #pragma once +#include "common/common_types.h" + namespace Service { + +class Interface; + namespace NFC { +enum class TagState : u8 { + NotInitialized = 0, + NotScanning = 1, + Scanning = 2, + TagInRange = 3, + TagOutOfRange = 4, + TagDataLoaded = 5, +}; + +enum class CommunicationStatus : u8 { + AttemptInitialize = 1, + NfcInitialized = 2, +}; + +/** + * NFC::Initialize service function + * Inputs: + * 0 : Header code [0x00010040] + * 1 : (u8) unknown parameter. Can be either value 0x1 or 0x2 + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void Initialize(Interface* self); + +/** + * NFC::Shutdown service function + * Inputs: + * 0 : Header code [0x00020040] + * 1 : (u8) unknown parameter + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void Shutdown(Interface* self); + +/** + * NFC::StartCommunication service function + * Inputs: + * 0 : Header code [0x00030000] + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void StartCommunication(Interface* self); + +/** + * NFC::StopCommunication service function + * Inputs: + * 0 : Header code [0x00040000] + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void StopCommunication(Interface* self); + +/** + * NFC::StartTagScanning service function + * Inputs: + * 0 : Header code [0x00050040] + * 1 : (u16) unknown. This is normally 0x0 + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void StartTagScanning(Interface* self); + +/** + * NFC::StopTagScanning service function + * Inputs: + * 0 : Header code [0x00060000] + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void StopTagScanning(Interface* self); + +/** + * NFC::LoadAmiiboData service function + * Inputs: + * 0 : Header code [0x00070000] + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void LoadAmiiboData(Interface* self); + +/** + * NFC::ResetTagScanState service function + * Inputs: + * 0 : Header code [0x00080000] + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + */ +void ResetTagScanState(Interface* self); + +/** + * NFC::GetTagInRangeEvent service function + * Inputs: + * 0 : Header code [0x000B0000] + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + * 2 : Copy handle descriptor + * 3 : Event Handle + */ +void GetTagInRangeEvent(Interface* self); + +/** + * NFC::GetTagOutOfRangeEvent service function + * Inputs: + * 0 : Header code [0x000C0000] + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + * 2 : Copy handle descriptor + * 3 : Event Handle + */ +void GetTagOutOfRangeEvent(Interface* self); + +/** + * NFC::GetTagState service function + * Inputs: + * 0 : Header code [0x000D0000] + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + * 2 : (u8) Tag state + */ +void GetTagState(Interface* self); + +/** + * NFC::CommunicationGetStatus service function + * Inputs: + * 0 : Header code [0x000F0000] + * Outputs: + * 1 : Result of function, 0 on success, otherwise error code + * 2 : (u8) Communication state + */ +void CommunicationGetStatus(Interface* self); + /// Initialize all NFC services. void Init(); +/// Shutdown all NFC services. +void Shutdown(); + } // namespace NFC } // namespace Service diff --git a/src/core/hle/service/nfc/nfc_m.cpp b/src/core/hle/service/nfc/nfc_m.cpp index 717335c11..ebe637650 100644 --- a/src/core/hle/service/nfc/nfc_m.cpp +++ b/src/core/hle/service/nfc/nfc_m.cpp @@ -2,6 +2,7 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include "core/hle/service/nfc/nfc.h" #include "core/hle/service/nfc/nfc_m.h" namespace Service { @@ -10,17 +11,19 @@ namespace NFC { const Interface::FunctionInfo FunctionTable[] = { // clang-format off // nfc:u shared commands - {0x00010040, nullptr, "Initialize"}, - {0x00020040, nullptr, "Shutdown"}, - {0x00030000, nullptr, "StartCommunication"}, - {0x00040000, nullptr, "StopCommunication"}, - {0x00050040, nullptr, "StartTagScanning"}, - {0x00060000, nullptr, "StopTagScanning"}, - {0x00070000, nullptr, "LoadAmiiboData"}, - {0x00080000, nullptr, "ResetTagScanState"}, + {0x00010040, Initialize, "Initialize"}, + {0x00020040, Shutdown, "Shutdown"}, + {0x00030000, StartCommunication, "StartCommunication"}, + {0x00040000, StopCommunication, "StopCommunication"}, + {0x00050040, StartTagScanning, "StartTagScanning"}, + {0x00060000, StopTagScanning, "StopTagScanning"}, + {0x00070000, LoadAmiiboData, "LoadAmiiboData"}, + {0x00080000, ResetTagScanState, "ResetTagScanState"}, {0x00090002, nullptr, "UpdateStoredAmiiboData"}, - {0x000D0000, nullptr, "GetTagState"}, - {0x000F0000, nullptr, "CommunicationGetStatus"}, + {0x000B0000, GetTagInRangeEvent, "GetTagInRangeEvent"}, + {0x000C0000, GetTagOutOfRangeEvent, "GetTagOutOfRangeEvent"}, + {0x000D0000, GetTagState, "GetTagState"}, + {0x000F0000, CommunicationGetStatus, "CommunicationGetStatus"}, {0x00100000, nullptr, "GetTagInfo2"}, {0x00110000, nullptr, "GetTagInfo"}, {0x00120000, nullptr, "CommunicationGetResult"}, diff --git a/src/core/hle/service/nfc/nfc_u.cpp b/src/core/hle/service/nfc/nfc_u.cpp index deffb0b4f..5a40c7874 100644 --- a/src/core/hle/service/nfc/nfc_u.cpp +++ b/src/core/hle/service/nfc/nfc_u.cpp @@ -2,6 +2,7 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. +#include "core/hle/service/nfc/nfc.h" #include "core/hle/service/nfc/nfc_u.h" namespace Service { @@ -9,17 +10,19 @@ namespace NFC { const Interface::FunctionInfo FunctionTable[] = { // clang-format off - {0x00010040, nullptr, "Initialize"}, - {0x00020040, nullptr, "Shutdown"}, - {0x00030000, nullptr, "StartCommunication"}, - {0x00040000, nullptr, "StopCommunication"}, - {0x00050040, nullptr, "StartTagScanning"}, - {0x00060000, nullptr, "StopTagScanning"}, - {0x00070000, nullptr, "LoadAmiiboData"}, - {0x00080000, nullptr, "ResetTagScanState"}, + {0x00010040, Initialize, "Initialize"}, + {0x00020040, Shutdown, "Shutdown"}, + {0x00030000, StartCommunication, "StartCommunication"}, + {0x00040000, StopCommunication, "StopCommunication"}, + {0x00050040, StartTagScanning, "StartTagScanning"}, + {0x00060000, StopTagScanning, "StopTagScanning"}, + {0x00070000, LoadAmiiboData, "LoadAmiiboData"}, + {0x00080000, ResetTagScanState, "ResetTagScanState"}, {0x00090002, nullptr, "UpdateStoredAmiiboData"}, - {0x000D0000, nullptr, "GetTagState"}, - {0x000F0000, nullptr, "CommunicationGetStatus"}, + {0x000B0000, GetTagInRangeEvent, "GetTagInRangeEvent"}, + {0x000C0000, GetTagOutOfRangeEvent, "GetTagOutOfRangeEvent"}, + {0x000D0000, GetTagState, "GetTagState"}, + {0x000F0000, CommunicationGetStatus, "CommunicationGetStatus"}, {0x00100000, nullptr, "GetTagInfo2"}, {0x00110000, nullptr, "GetTagInfo"}, {0x00120000, nullptr, "CommunicationGetResult"}, diff --git a/src/core/hle/service/service.cpp b/src/core/hle/service/service.cpp index 7e52a05d9..0672ac2e3 100644 --- a/src/core/hle/service/service.cpp +++ b/src/core/hle/service/service.cpp @@ -6,9 +6,8 @@ #include "common/logging/log.h" #include "common/string_util.h" - #include "core/hle/kernel/server_port.h" -#include "core/hle/service/ac_u.h" +#include "core/hle/service/ac/ac.h" #include "core/hle/service/act/act.h" #include "core/hle/service/am/am.h" #include "core/hle/service/apt/apt.h" @@ -138,6 +137,7 @@ void Init() { AddNamedPort(new ERR::ERR_F); FS::ArchiveInit(); + AC::Init(); ACT::Init(); AM::Init(); APT::Init(); @@ -158,7 +158,6 @@ void Init() { PTM::Init(); QTM::Init(); - AddService(new AC::AC_U); AddService(new CSND::CSND_SND); AddService(new DSP_DSP::Interface); AddService(new GSP::GSP_GPU); @@ -178,6 +177,7 @@ void Init() { /// Shutdown ServiceManager void Shutdown() { PTM::Shutdown(); + NFC::Shutdown(); NIM::Shutdown(); NEWS::Shutdown(); NDM::Shutdown(); @@ -191,6 +191,7 @@ void Shutdown() { BOSS::Shutdown(); APT::Shutdown(); AM::Shutdown(); + AC::Shutdown(); FS::ArchiveShutdown(); g_srv_services.clear(); diff --git a/src/core/hle/service/soc_u.cpp b/src/core/hle/service/soc_u.cpp index c3918cdd0..dcc5c3c90 100644 --- a/src/core/hle/service/soc_u.cpp +++ b/src/core/hle/service/soc_u.cpp @@ -603,7 +603,6 @@ static void RecvFrom(Interface* self) { u32 socket_handle = cmd_buffer[1]; u32 len = cmd_buffer[2]; u32 flags = cmd_buffer[3]; - socklen_t addr_len = static_cast<socklen_t>(cmd_buffer[4]); struct { u32 output_buffer_descriptor; @@ -693,7 +692,6 @@ static void Poll(Interface* self) { static void GetSockName(Interface* self) { u32* cmd_buffer = Kernel::GetCommandBuffer(); u32 socket_handle = cmd_buffer[1]; - socklen_t ctr_len = cmd_buffer[2]; // Memory address of the ctr_dest_addr structure VAddr ctr_dest_addr_addr = cmd_buffer[0x104 >> 2]; @@ -734,7 +732,6 @@ static void Shutdown(Interface* self) { static void GetPeerName(Interface* self) { u32* cmd_buffer = Kernel::GetCommandBuffer(); u32 socket_handle = cmd_buffer[1]; - socklen_t len = cmd_buffer[2]; // Memory address of the ctr_dest_addr structure VAddr ctr_dest_addr_addr = cmd_buffer[0x104 >> 2]; @@ -765,7 +762,6 @@ static void Connect(Interface* self) { // performing nonblocking operations and spinlock until the data is available u32* cmd_buffer = Kernel::GetCommandBuffer(); u32 socket_handle = cmd_buffer[1]; - socklen_t len = cmd_buffer[2]; // Memory address of the ctr_input_addr structure VAddr ctr_input_addr_addr = cmd_buffer[6]; diff --git a/src/core/hle/service/y2r_u.cpp b/src/core/hle/service/y2r_u.cpp index a20194107..31bb466fc 100644 --- a/src/core/hle/service/y2r_u.cpp +++ b/src/core/hle/service/y2r_u.cpp @@ -531,7 +531,9 @@ static void GetStandardCoefficient(Interface* self) { LOG_DEBUG(Service_Y2R, "called standard_coefficient=%u ", index); } else { cmd_buff[0] = IPC::MakeHeader(0x21, 1, 0); - cmd_buff[1] = -1; // TODO(bunnei): Identify the correct error code for this + cmd_buff[1] = ResultCode(ErrorDescription::InvalidEnumValue, ErrorModule::CAM, + ErrorSummary::InvalidArgument, ErrorLevel::Usage) + .raw; LOG_ERROR(Service_Y2R, "called standard_coefficient=%u The argument is invalid!", index); } diff --git a/src/core/hle/svc.cpp b/src/core/hle/svc.cpp index 2ca270de3..96db39ad9 100644 --- a/src/core/hle/svc.cpp +++ b/src/core/hle/svc.cpp @@ -248,6 +248,8 @@ static ResultCode SendSyncRequest(Kernel::Handle handle) { LOG_TRACE(Kernel_SVC, "called handle=0x%08X(%s)", handle, session->GetName().c_str()); + Core::System::GetInstance().PrepareReschedule(); + // TODO(Subv): svcSendSyncRequest should put the caller thread to sleep while the server // responds and cause a reschedule. return session->SendSyncRequest(); @@ -270,27 +272,27 @@ static ResultCode WaitSynchronization1(Kernel::Handle handle, s64 nano_seconds) LOG_TRACE(Kernel_SVC, "called handle=0x%08X(%s:%s), nanoseconds=%lld", handle, object->GetTypeName().c_str(), object->GetName().c_str(), nano_seconds); - if (object->ShouldWait()) { + if (object->ShouldWait(thread)) { if (nano_seconds == 0) return ERR_SYNC_TIMEOUT; + thread->wait_objects = {object}; object->AddWaitingThread(thread); - // TODO(Subv): Perform things like update the mutex lock owner's priority to - // prevent priority inversion. Currently this is done in Mutex::ShouldWait, - // but it should be moved to a function that is called from here. - thread->status = THREADSTATUS_WAIT_SYNCH; + thread->status = THREADSTATUS_WAIT_SYNCH_ANY; // Create an event to wake the thread up after the specified nanosecond delay has passed thread->WakeAfterDelay(nano_seconds); + Core::System::GetInstance().PrepareReschedule(); + // Note: The output of this SVC will be set to RESULT_SUCCESS if the thread // resumes due to a signal in its wait objects. // Otherwise we retain the default value of timeout. return ERR_SYNC_TIMEOUT; } - object->Acquire(); + object->Acquire(thread); return RESULT_SUCCESS; } @@ -324,19 +326,14 @@ static ResultCode WaitSynchronizationN(s32* out, Kernel::Handle* handles, s32 ha objects[i] = object; } - // Clear the mapping of wait object indices. - // We don't want any lingering state in this map. - // It will be repopulated later in the wait_all = false case. - thread->wait_objects_index.clear(); - if (wait_all) { bool all_available = std::all_of(objects.begin(), objects.end(), - [](const ObjectPtr& object) { return !object->ShouldWait(); }); + [thread](const ObjectPtr& object) { return !object->ShouldWait(thread); }); if (all_available) { // We can acquire all objects right now, do so. for (auto& object : objects) - object->Acquire(); + object->Acquire(thread); // Note: In this case, the `out` parameter is not set, // and retains whatever value it had before. return RESULT_SUCCESS; @@ -350,22 +347,20 @@ static ResultCode WaitSynchronizationN(s32* out, Kernel::Handle* handles, s32 ha return ERR_SYNC_TIMEOUT; // Put the thread to sleep - thread->status = THREADSTATUS_WAIT_SYNCH; + thread->status = THREADSTATUS_WAIT_SYNCH_ALL; // Add the thread to each of the objects' waiting threads. for (auto& object : objects) { object->AddWaitingThread(thread); - // TODO(Subv): Perform things like update the mutex lock owner's priority to - // prevent priority inversion. Currently this is done in Mutex::ShouldWait, - // but it should be moved to a function that is called from here. } - // Set the thread's waitlist to the list of objects passed to WaitSynchronizationN thread->wait_objects = std::move(objects); // Create an event to wake the thread up after the specified nanosecond delay has passed thread->WakeAfterDelay(nano_seconds); + Core::System::GetInstance().PrepareReschedule(); + // This value gets set to -1 by default in this case, it is not modified after this. *out = -1; // Note: The output of this SVC will be set to RESULT_SUCCESS if the thread resumes due to @@ -373,13 +368,14 @@ static ResultCode WaitSynchronizationN(s32* out, Kernel::Handle* handles, s32 ha return ERR_SYNC_TIMEOUT; } else { // Find the first object that is acquirable in the provided list of objects - auto itr = std::find_if(objects.begin(), objects.end(), - [](const ObjectPtr& object) { return !object->ShouldWait(); }); + auto itr = std::find_if(objects.begin(), objects.end(), [thread](const ObjectPtr& object) { + return !object->ShouldWait(thread); + }); if (itr != objects.end()) { // We found a ready object, acquire it and set the result value Kernel::WaitObject* object = itr->get(); - object->Acquire(); + object->Acquire(thread); *out = std::distance(objects.begin(), itr); return RESULT_SUCCESS; } @@ -392,28 +388,24 @@ static ResultCode WaitSynchronizationN(s32* out, Kernel::Handle* handles, s32 ha return ERR_SYNC_TIMEOUT; // Put the thread to sleep - thread->status = THREADSTATUS_WAIT_SYNCH; - - // Clear the thread's waitlist, we won't use it for wait_all = false - thread->wait_objects.clear(); + thread->status = THREADSTATUS_WAIT_SYNCH_ANY; // Add the thread to each of the objects' waiting threads. for (size_t i = 0; i < objects.size(); ++i) { Kernel::WaitObject* object = objects[i].get(); - // Set the index of this object in the mapping of Objects -> index for this thread. - thread->wait_objects_index[object->GetObjectId()] = static_cast<int>(i); object->AddWaitingThread(thread); - // TODO(Subv): Perform things like update the mutex lock owner's priority to - // prevent priority inversion. Currently this is done in Mutex::ShouldWait, - // but it should be moved to a function that is called from here. } + thread->wait_objects = std::move(objects); + // Note: If no handles and no timeout were given, then the thread will deadlock, this is // consistent with hardware behavior. // Create an event to wake the thread up after the specified nanosecond delay has passed thread->WakeAfterDelay(nano_seconds); + Core::System::GetInstance().PrepareReschedule(); + // Note: The output of this SVC will be set to RESULT_SUCCESS if the thread resumes due to a // signal in one of its wait objects. // Otherwise we retain the default value of timeout, and -1 in the out parameter @@ -448,6 +440,9 @@ static ResultCode ArbitrateAddress(Kernel::Handle handle, u32 address, u32 type, auto res = arbiter->ArbitrateAddress(static_cast<Kernel::ArbitrationType>(type), address, value, nanoseconds); + // TODO(Subv): Identify in which specific cases this call should cause a reschedule. + Core::System::GetInstance().PrepareReschedule(); + return res; } @@ -537,16 +532,18 @@ static ResultCode CreateThread(Kernel::Handle* out_handle, s32 priority, u32 ent name = Common::StringFromFormat("unknown-%08x", entry_point); } - // TODO(bunnei): Implement resource limits to return an error code instead of the below assert. - // The error code should be: Description::NotAuthorized, Module::OS, Summary::WrongArgument, - // Level::Permanent - ASSERT_MSG(priority >= THREADPRIO_USERLAND_MAX, "Unexpected thread priority!"); - if (priority > THREADPRIO_LOWEST) { return ResultCode(ErrorDescription::OutOfRange, ErrorModule::OS, ErrorSummary::InvalidArgument, ErrorLevel::Usage); } + using Kernel::ResourceLimit; + Kernel::SharedPtr<ResourceLimit>& resource_limit = Kernel::g_current_process->resource_limit; + if (resource_limit->GetMaxResourceValue(Kernel::ResourceTypes::PRIORITY) > priority) { + return ResultCode(ErrorDescription::NotAuthorized, ErrorModule::OS, + ErrorSummary::WrongArgument, ErrorLevel::Permanent); + } + switch (processor_id) { case THREADPROCESSORID_ALL: case THREADPROCESSORID_DEFAULT: @@ -574,6 +571,8 @@ static ResultCode CreateThread(Kernel::Handle* out_handle, s32 priority, u32 ent CASCADE_RESULT(*out_handle, Kernel::g_handle_table.Create(std::move(thread))); + Core::System::GetInstance().PrepareReschedule(); + LOG_TRACE(Kernel_SVC, "called entrypoint=0x%08X (%s), arg=0x%08X, stacktop=0x%08X, " "threadpriority=0x%08X, processorid=0x%08X : created handle=0x%08X", entry_point, name.c_str(), arg, stack_top, priority, processor_id, *out_handle); @@ -586,6 +585,7 @@ static void ExitThread() { LOG_TRACE(Kernel_SVC, "called, pc=0x%08X", Core::CPU().GetPC()); Kernel::ExitCurrentThread(); + Core::System::GetInstance().PrepareReschedule(); } /// Gets the priority for the specified thread @@ -600,11 +600,32 @@ static ResultCode GetThreadPriority(s32* priority, Kernel::Handle handle) { /// Sets the priority for the specified thread static ResultCode SetThreadPriority(Kernel::Handle handle, s32 priority) { + if (priority > THREADPRIO_LOWEST) { + return ResultCode(ErrorDescription::OutOfRange, ErrorModule::OS, + ErrorSummary::InvalidArgument, ErrorLevel::Usage); + } + SharedPtr<Kernel::Thread> thread = Kernel::g_handle_table.Get<Kernel::Thread>(handle); if (thread == nullptr) return ERR_INVALID_HANDLE; + using Kernel::ResourceLimit; + // Note: The kernel uses the current process's resource limit instead of + // the one from the thread owner's resource limit. + Kernel::SharedPtr<ResourceLimit>& resource_limit = Kernel::g_current_process->resource_limit; + if (resource_limit->GetMaxResourceValue(Kernel::ResourceTypes::PRIORITY) > priority) { + return ResultCode(ErrorDescription::NotAuthorized, ErrorModule::OS, + ErrorSummary::WrongArgument, ErrorLevel::Permanent); + } + thread->SetPriority(priority); + thread->UpdatePriority(); + + // Update the mutexes that this thread is waiting for + for (auto& mutex : thread->pending_mutexes) + mutex->UpdatePriority(); + + Core::System::GetInstance().PrepareReschedule(); return RESULT_SUCCESS; } @@ -844,11 +865,18 @@ static ResultCode CancelTimer(Kernel::Handle handle) { static void SleepThread(s64 nanoseconds) { LOG_TRACE(Kernel_SVC, "called nanoseconds=%lld", nanoseconds); + // Don't attempt to yield execution if there are no available threads to run, + // this way we avoid a useless reschedule to the idle thread. + if (nanoseconds == 0 && !Kernel::HaveReadyThreads()) + return; + // Sleep current thread and check for next thread to schedule Kernel::WaitCurrentThread_Sleep(); // Create an event to wake the thread up after the specified nanosecond delay has passed Kernel::GetCurrentThread()->WakeAfterDelay(nanoseconds); + + Core::System::GetInstance().PrepareReschedule(); } /// This returns the total CPU ticks elapsed since the CPU was powered-on @@ -890,7 +918,11 @@ static ResultCode CreateMemoryBlock(Kernel::Handle* out_handle, u32 addr, u32 si return ResultCode(ErrorDescription::InvalidCombination, ErrorModule::OS, ErrorSummary::InvalidArgument, ErrorLevel::Usage); - if (addr < Memory::PROCESS_IMAGE_VADDR || addr + size > Memory::SHARED_MEMORY_VADDR_END) { + // TODO(Subv): Processes with memory type APPLICATION are not allowed + // to create memory blocks with addr = 0, any attempts to do so + // should return error 0xD92007EA. + if ((addr < Memory::PROCESS_IMAGE_VADDR || addr + size > Memory::SHARED_MEMORY_VADDR_END) && + addr != 0) { return ResultCode(ErrorDescription::InvalidAddress, ErrorModule::OS, ErrorSummary::InvalidArgument, ErrorLevel::Usage); } @@ -1184,8 +1216,6 @@ void CallSVC(u32 immediate) { if (info) { if (info->func) { info->func(); - // TODO(Subv): Not all service functions should cause a reschedule in all cases. - Core::System::GetInstance().PrepareReschedule(); } else { LOG_ERROR(Kernel_SVC, "unimplemented SVC function %s(..)", info->name); } diff --git a/src/core/hw/gpu.cpp b/src/core/hw/gpu.cpp index 1a1ee90b2..fa8c13d36 100644 --- a/src/core/hw/gpu.cpp +++ b/src/core/hw/gpu.cpp @@ -15,7 +15,6 @@ #include "common/vector_math.h" #include "core/core_timing.h" #include "core/hle/service/gsp_gpu.h" -#include "core/hle/service/hid/hid.h" #include "core/hw/gpu.h" #include "core/hw/hw.h" #include "core/memory.h" @@ -33,7 +32,7 @@ namespace GPU { Regs g_regs; /// 268MHz CPU clocks / 60Hz frames per second -const u64 frame_ticks = 268123480ull / 60; +const u64 frame_ticks = BASE_CLOCK_RATE_ARM11 / 60; /// Event id for CoreTiming static int vblank_event; /// Total number of frames drawn @@ -551,9 +550,6 @@ static void VBlankCallback(u64 userdata, int cycles_late) { Service::GSP::SignalInterrupt(Service::GSP::InterruptId::PDC0); Service::GSP::SignalInterrupt(Service::GSP::InterruptId::PDC1); - // Check for user input updates - Service::HID::Update(); - if (!Settings::values.use_vsync && Settings::values.toggle_framelimit) { FrameLimiter(); } diff --git a/src/core/loader/3dsx.cpp b/src/core/loader/3dsx.cpp index 1c10740a0..09266e8b0 100644 --- a/src/core/loader/3dsx.cpp +++ b/src/core/loader/3dsx.cpp @@ -177,18 +177,34 @@ static THREEDSX_Error Load3DSXFile(FileUtil::IOFile& file, u32 base_addr, pos += table.skip; s32 num_patches = table.patch; while (0 < num_patches && pos < end_pos) { - u32 in_addr = - static_cast<u32>(reinterpret_cast<u8*>(pos) - program_image.data()); - u32 addr = TranslateAddr(*pos, &loadinfo, offsets); - LOG_TRACE(Loader, "Patching %08X <-- rel(%08X,%d) (%08X)", - base_addr + in_addr, addr, current_segment_reloc_table, *pos); + u32 in_addr = base_addr + static_cast<u32>(reinterpret_cast<u8*>(pos) - + program_image.data()); + u32 orig_data = *pos; + u32 sub_type = orig_data >> (32 - 4); + u32 addr = TranslateAddr(orig_data & ~0xF0000000, &loadinfo, offsets); + LOG_TRACE(Loader, "Patching %08X <-- rel(%08X,%d) (%08X)", in_addr, addr, + current_segment_reloc_table, *pos); switch (current_segment_reloc_table) { - case 0: - *pos = (addr); + case 0: { + if (sub_type != 0) + return ERROR_READ; + *pos = addr; break; - case 1: - *pos = static_cast<u32>(addr - in_addr); + } + case 1: { + u32 data = addr - in_addr; + switch (sub_type) { + case 0: // 32-bit signed offset + *pos = data; + break; + case 1: // 31-bit signed offset + *pos = data & ~(1U << 31); + break; + default: + return ERROR_READ; + } break; + } default: break; // this should never happen } diff --git a/src/core/loader/ncch.cpp b/src/core/loader/ncch.cpp index 6f2164428..5df33f6d2 100644 --- a/src/core/loader/ncch.cpp +++ b/src/core/loader/ncch.cpp @@ -11,8 +11,10 @@ #include "core/file_sys/archive_romfs.h" #include "core/hle/kernel/process.h" #include "core/hle/kernel/resource_limit.h" +#include "core/hle/service/cfg/cfg.h" #include "core/hle/service/fs/archive.h" #include "core/loader/ncch.h" +#include "core/loader/smdh.h" #include "core/memory.h" //////////////////////////////////////////////////////////////////////////////////////////////////// @@ -286,7 +288,7 @@ ResultStatus AppLoader_NCCH::LoadExeFS() { LOG_DEBUG(Loader, "Thread priority: 0x%X", priority); LOG_DEBUG(Loader, "Resource limit category: %d", resource_limit_category); LOG_DEBUG(Loader, "System Mode: %d", - exheader_header.arm11_system_local_caps.system_mode); + static_cast<int>(exheader_header.arm11_system_local_caps.system_mode)); if (exheader_header.arm11_system_local_caps.program_id != ncch_header.program_id) { LOG_ERROR(Loader, "ExHeader Program ID mismatch: the ROM is probably encrypted."); @@ -309,6 +311,23 @@ ResultStatus AppLoader_NCCH::LoadExeFS() { return ResultStatus::Success; } +void AppLoader_NCCH::ParseRegionLockoutInfo() { + std::vector<u8> smdh_buffer; + if (ReadIcon(smdh_buffer) == ResultStatus::Success && smdh_buffer.size() >= sizeof(SMDH)) { + SMDH smdh; + memcpy(&smdh, smdh_buffer.data(), sizeof(SMDH)); + u32 region_lockout = smdh.region_lockout; + constexpr u32 REGION_COUNT = 7; + for (u32 region = 0; region < REGION_COUNT; ++region) { + if (region_lockout & 1) { + Service::CFG::SetPreferredRegionCode(region); + break; + } + region_lockout >>= 1; + } + } +} + ResultStatus AppLoader_NCCH::Load() { if (is_loaded) return ResultStatus::ErrorAlreadyLoaded; @@ -325,6 +344,9 @@ ResultStatus AppLoader_NCCH::Load() { Service::FS::RegisterArchiveType(std::make_unique<FileSys::ArchiveFactory_RomFS>(*this), Service::FS::ArchiveIdCode::RomFS); + + ParseRegionLockoutInfo(); + return ResultStatus::Success; } diff --git a/src/core/loader/ncch.h b/src/core/loader/ncch.h index 6afc171a5..4ef95b5c6 100644 --- a/src/core/loader/ncch.h +++ b/src/core/loader/ncch.h @@ -181,7 +181,7 @@ public: * Loads the Exheader and returns the system mode for this application. * @return Optional with the kernel system mode */ - boost::optional<u32> LoadKernelSystemMode(); + boost::optional<u32> LoadKernelSystemMode() override; ResultStatus ReadCode(std::vector<u8>& buffer) override; @@ -229,6 +229,9 @@ private: */ ResultStatus LoadExeFS(); + /// Reads the region lockout info in the SMDH and send it to CFG service + void ParseRegionLockoutInfo(); + bool is_exefs_loaded = false; bool is_compressed = false; diff --git a/src/core/settings.cpp b/src/core/settings.cpp index 5d23c52f9..9afaf79ec 100644 --- a/src/core/settings.cpp +++ b/src/core/settings.cpp @@ -20,7 +20,6 @@ void Apply() { VideoCore::g_hw_renderer_enabled = values.use_hw_renderer; VideoCore::g_shader_jit_enabled = values.use_shader_jit; - VideoCore::g_scaled_resolution_enabled = values.use_scaled_resolution; VideoCore::g_toggle_framelimit_enabled = values.toggle_framelimit; if (VideoCore::g_emu_window) { diff --git a/src/core/settings.h b/src/core/settings.h index fe47c364f..b6c75531f 100644 --- a/src/core/settings.h +++ b/src/core/settings.h @@ -89,7 +89,7 @@ struct Values { // Renderer bool use_hw_renderer; bool use_shader_jit; - bool use_scaled_resolution; + float resolution_factor; bool use_vsync; bool toggle_framelimit; @@ -105,6 +105,7 @@ struct Values { // Audio std::string sink_id; bool enable_audio_stretching; + std::string audio_device_id; // Camera std::array<std::string, Service::CAM::NumCameras> camera_name; @@ -115,5 +116,9 @@ struct Values { u16 gdbstub_port; } extern values; +// a special value for Values::region_value indicating that citra will automatically select a region +// value to fit the region lockout info of the game +static constexpr int REGION_VALUE_AUTO_SELECT = -1; + void Apply(); } diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt index 6ca319b59..d55b84ce0 100644 --- a/src/video_core/CMakeLists.txt +++ b/src/video_core/CMakeLists.txt @@ -50,10 +50,12 @@ set(HEADERS if(ARCHITECTURE_x86_64) set(SRCS ${SRCS} - shader/shader_jit_x64.cpp) + shader/shader_jit_x64.cpp + shader/shader_jit_x64_compiler.cpp) set(HEADERS ${HEADERS} - shader/shader_jit_x64.h) + shader/shader_jit_x64.h + shader/shader_jit_x64_compiler.h) endif() create_directory_groups(${SRCS} ${HEADERS}) diff --git a/src/video_core/command_processor.cpp b/src/video_core/command_processor.cpp index ea58e9f54..eb79974a8 100644 --- a/src/video_core/command_processor.cpp +++ b/src/video_core/command_processor.cpp @@ -142,16 +142,18 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { MICROPROFILE_SCOPE(GPU_Drawing); immediate_attribute_id = 0; - Shader::UnitState shader_unit; - g_state.vs.Setup(); + auto* shader_engine = Shader::GetEngine(); + shader_engine->SetupBatch(g_state.vs, regs.vs.main_offset); // Send to vertex shader if (g_debug_context) g_debug_context->OnEvent(DebugContext::Event::VertexShaderInvocation, static_cast<void*>(&immediate_input)); - g_state.vs.Run(shader_unit, immediate_input, regs.vs.num_input_attributes + 1); - Shader::OutputVertex output_vertex = - shader_unit.output_registers.ToVertex(regs.vs); + Shader::UnitState shader_unit; + shader_unit.LoadInputVertex(immediate_input, regs.vs.num_input_attributes + 1); + shader_engine->Run(g_state.vs, shader_unit); + auto output_vertex = Shader::OutputVertex::FromRegisters( + shader_unit.registers.output, regs, regs.vs.output_mask); // Send to renderer using Pica::Shader::OutputVertex; @@ -243,8 +245,10 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { unsigned int vertex_cache_pos = 0; vertex_cache_ids.fill(-1); + auto* shader_engine = Shader::GetEngine(); Shader::UnitState shader_unit; - g_state.vs.Setup(); + + shader_engine->SetupBatch(g_state.vs, regs.vs.main_offset); for (unsigned int index = 0; index < regs.num_vertices; ++index) { // Indexed rendering doesn't use the start offset @@ -283,10 +287,12 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) { if (g_debug_context) g_debug_context->OnEvent(DebugContext::Event::VertexShaderInvocation, (void*)&input); - g_state.vs.Run(shader_unit, input, loader.GetNumTotalAttributes()); + shader_unit.LoadInputVertex(input, loader.GetNumTotalAttributes()); + shader_engine->Run(g_state.vs, shader_unit); // Retrieve vertex from register data - output_vertex = shader_unit.output_registers.ToVertex(regs.vs); + output_vertex = Shader::OutputVertex::FromRegisters(shader_unit.registers.output, + regs, regs.vs.output_mask); if (is_indexed) { vertex_cache[vertex_cache_pos] = output_vertex; diff --git a/src/video_core/pica.cpp b/src/video_core/pica.cpp index ce2bd455e..b4a77c632 100644 --- a/src/video_core/pica.cpp +++ b/src/video_core/pica.cpp @@ -499,7 +499,7 @@ void Init() { } void Shutdown() { - Shader::ClearCache(); + Shader::Shutdown(); } template <typename T> diff --git a/src/video_core/renderer_opengl/gl_rasterizer.cpp b/src/video_core/renderer_opengl/gl_rasterizer.cpp index 5a306a5c8..f3674e965 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer.cpp @@ -716,8 +716,6 @@ void RasterizerOpenGL::FlushAndInvalidateRegion(PAddr addr, u32 size) { bool RasterizerOpenGL::AccelerateDisplayTransfer(const GPU::Regs::DisplayTransferConfig& config) { MICROPROFILE_SCOPE(OpenGL_Blits); - using PixelFormat = CachedSurface::PixelFormat; - using SurfaceType = CachedSurface::SurfaceType; CachedSurface src_params; src_params.addr = config.GetPhysicalInputAddress(); @@ -748,7 +746,8 @@ bool RasterizerOpenGL::AccelerateDisplayTransfer(const GPU::Regs::DisplayTransfe // Adjust the source rectangle to take into account parts of the input lines being cropped if (config.input_width > config.output_width) { - src_rect.right -= (config.input_width - config.output_width) * src_surface->res_scale_width; + src_rect.right -= static_cast<int>((config.input_width - config.output_width) * + src_surface->res_scale_width); } // Require destination surface to have same resolution scale as source to preserve scaling diff --git a/src/video_core/renderer_opengl/gl_rasterizer.h b/src/video_core/renderer_opengl/gl_rasterizer.h index e1a9cb361..cc3e4bed5 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer.h +++ b/src/video_core/renderer_opengl/gl_rasterizer.h @@ -76,7 +76,7 @@ union PicaShaderConfig { } state.fog_mode = regs.fog_mode; - state.fog_flip = regs.fog_flip; + state.fog_flip = regs.fog_flip != 0; state.combiner_buffer_input = regs.tev_combiner_buffer_input.update_mask_rgb.Value() | regs.tev_combiner_buffer_input.update_mask_a.Value() << 4; diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp index 85aa06cd5..1e7eedecb 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp @@ -172,7 +172,6 @@ bool RasterizerCacheOpenGL::TryBlitSurfaces(CachedSurface* src_surface, const MathUtil::Rectangle<int>& src_rect, CachedSurface* dst_surface, const MathUtil::Rectangle<int>& dst_rect) { - using SurfaceType = CachedSurface::SurfaceType; if (!CachedSurface::CheckFormatsBlittable(src_surface->pixel_format, dst_surface->pixel_format)) { @@ -556,14 +555,21 @@ RasterizerCacheOpenGL::GetFramebufferSurfaces(const Pica::Regs::FramebufferConfi color_params.width = depth_params.width = config.GetWidth(); color_params.height = depth_params.height = config.GetHeight(); color_params.is_tiled = depth_params.is_tiled = true; - if (VideoCore::g_scaled_resolution_enabled) { - auto layout = VideoCore::g_emu_window->GetFramebufferLayout(); - // Assume same scaling factor for top and bottom screens + // Set the internal resolution, assume the same scaling factor for top and bottom screens + const Layout::FramebufferLayout& layout = VideoCore::g_emu_window->GetFramebufferLayout(); + if (Settings::values.resolution_factor == 0.0f) { + // Auto - scale resolution to the window size color_params.res_scale_width = depth_params.res_scale_width = (float)layout.top_screen.GetWidth() / VideoCore::kScreenTopWidth; color_params.res_scale_height = depth_params.res_scale_height = (float)layout.top_screen.GetHeight() / VideoCore::kScreenTopHeight; + } else { + // Otherwise, scale the resolution by the specified factor + color_params.res_scale_width = Settings::values.resolution_factor; + depth_params.res_scale_width = Settings::values.resolution_factor; + color_params.res_scale_height = Settings::values.resolution_factor; + depth_params.res_scale_height = Settings::values.resolution_factor; } color_params.addr = config.GetColorBufferPhysicalAddress(); diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.h b/src/video_core/renderer_opengl/gl_rasterizer_cache.h index b50e8292b..f57fdb3cc 100644 --- a/src/video_core/renderer_opengl/gl_rasterizer_cache.h +++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.h @@ -8,7 +8,14 @@ #include <memory> #include <set> #include <tuple> +#ifdef __GNUC__ +#pragma GCC diagnostic push +#pragma GCC diagnostic ignored "-Wunused-local-typedef" +#endif #include <boost/icl/interval_map.hpp> +#ifdef __GNUC__ +#pragma GCC diagnostic pop +#endif #include <glad/glad.h> #include "common/assert.h" #include "common/common_funcs.h" diff --git a/src/video_core/renderer_opengl/gl_shader_gen.cpp b/src/video_core/renderer_opengl/gl_shader_gen.cpp index 8f278722d..4c4f98ac9 100644 --- a/src/video_core/renderer_opengl/gl_shader_gen.cpp +++ b/src/video_core/renderer_opengl/gl_shader_gen.cpp @@ -293,7 +293,7 @@ static void AppendAlphaTestCondition(std::string& out, Regs::CompareFunc func) { case CompareFunc::GreaterThanOrEqual: { static const char* op[] = {"!=", "==", ">=", ">", "<=", "<"}; unsigned index = (unsigned)func - (unsigned)CompareFunc::Equal; - out += "int(last_tex_env_out.a * 255.0f) " + std::string(op[index]) + " alphatest_ref"; + out += "int(last_tex_env_out.a * 255.0) " + std::string(op[index]) + " alphatest_ref"; break; } @@ -422,16 +422,13 @@ static void WriteLighting(std::string& out, const PicaShaderConfig& config) { if (abs) { // LUT index is in the range of (0.0, 1.0) index = lighting.light[light_num].two_sided_diffuse ? "abs(" + index + ")" - : "max(" + index + ", 0.f)"; - return "(FLOAT_255 * clamp(" + index + ", 0.0, 1.0))"; + : "max(" + index + ", 0.0)"; } else { // LUT index is in the range of (-1.0, 1.0) - index = "clamp(" + index + ", -1.0, 1.0)"; - return "(FLOAT_255 * ((" + index + " < 0) ? " + index + " + 2.0 : " + index + - ") / 2.0)"; + index = "((" + index + " < 0) ? " + index + " + 2.0 : " + index + ") / 2.0"; } - return std::string(); + return "(OFFSET_256 + SCALE_256 * clamp(" + index + ", 0.0, 1.0))"; }; // Gets the lighting lookup table value given the specified sampler and index @@ -462,7 +459,7 @@ static void WriteLighting(std::string& out, const PicaShaderConfig& config) { if (light_config.dist_atten_enable) { std::string index = "(" + light_src + ".dist_atten_scale * length(-view - " + light_src + ".position) + " + light_src + ".dist_atten_bias)"; - index = "((clamp(" + index + ", 0.0, FLOAT_255)))"; + index = "(OFFSET_256 + SCALE_256 * clamp(" + index + ", 0.0, 1.0))"; const unsigned lut_num = ((unsigned)Regs::LightingSampler::DistanceAttenuation + light_config.num); dist_atten = GetLutValue((Regs::LightingSampler)lut_num, index); @@ -580,8 +577,10 @@ std::string GenerateFragmentShader(const PicaShaderConfig& config) { #version 330 core #define NUM_TEV_STAGES 6 #define NUM_LIGHTS 8 -#define LIGHTING_LUT_SIZE 256 -#define FLOAT_255 (255.0 / 256.0) + +// Texture coordinate offsets and scales +#define OFFSET_256 (0.5 / 256.0) +#define SCALE_256 (255.0 / 256.0) in vec4 primary_color; in vec2 texcoord[3]; diff --git a/src/video_core/shader/shader.cpp b/src/video_core/shader/shader.cpp index a4aa3c9e0..2da50bd62 100644 --- a/src/video_core/shader/shader.cpp +++ b/src/video_core/shader/shader.cpp @@ -2,14 +2,8 @@ // Licensed under GPLv2 or any later version // Refer to the license.txt file included. -#include <atomic> #include <cmath> #include <cstring> -#include <unordered_map> -#include <utility> -#include <boost/range/algorithm/fill.hpp> -#include "common/bit_field.h" -#include "common/hash.h" #include "common/logging/log.h" #include "common/microprofile.h" #include "video_core/pica.h" @@ -25,7 +19,8 @@ namespace Pica { namespace Shader { -OutputVertex OutputRegisters::ToVertex(const Regs::ShaderConfig& config) const { +OutputVertex OutputVertex::FromRegisters(Math::Vec4<float24> output_regs[16], const Regs& regs, + u32 output_mask) { // Setup output data OutputVertex ret; // TODO(neobrain): Under some circumstances, up to 16 attributes may be output. We need to @@ -33,13 +28,13 @@ OutputVertex OutputRegisters::ToVertex(const Regs::ShaderConfig& config) const { unsigned index = 0; for (unsigned i = 0; i < 7; ++i) { - if (index >= g_state.regs.vs_output_total) + if (index >= regs.vs_output_total) break; - if ((config.output_mask & (1 << i)) == 0) + if ((output_mask & (1 << i)) == 0) continue; - const auto& output_register_map = g_state.regs.vs_output_attributes[index]; + const auto& output_register_map = regs.vs_output_attributes[index]; u32 semantics[4] = {output_register_map.map_x, output_register_map.map_y, output_register_map.map_z, output_register_map.map_w}; @@ -47,7 +42,7 @@ OutputVertex OutputRegisters::ToVertex(const Regs::ShaderConfig& config) const { for (unsigned comp = 0; comp < 4; ++comp) { float24* out = ((float24*)&ret) + semantics[comp]; if (semantics[comp] != Regs::VSOutputAttributes::INVALID) { - *out = value[i][comp]; + *out = output_regs[i][comp]; } else { // Zero output so that attributes which aren't output won't have denormals in them, // which would slow us down later. @@ -76,86 +71,41 @@ OutputVertex OutputRegisters::ToVertex(const Regs::ShaderConfig& config) const { return ret; } -#ifdef ARCHITECTURE_x86_64 -static std::unordered_map<u64, std::unique_ptr<JitShader>> shader_map; -static const JitShader* jit_shader; -#endif // ARCHITECTURE_x86_64 +void UnitState::LoadInputVertex(const InputVertex& input, int num_attributes) { + // Setup input register table + const auto& attribute_register_map = g_state.regs.vs.input_register_map; + + for (int i = 0; i < num_attributes; i++) + registers.input[attribute_register_map.GetRegisterForAttribute(i)] = input.attr[i]; +} + +MICROPROFILE_DEFINE(GPU_Shader, "GPU", "Shader", MP_RGB(50, 50, 240)); -void ClearCache() { #ifdef ARCHITECTURE_x86_64 - shader_map.clear(); +static std::unique_ptr<JitX64Engine> jit_engine; #endif // ARCHITECTURE_x86_64 -} +static InterpreterEngine interpreter_engine; -void ShaderSetup::Setup() { +ShaderEngine* GetEngine() { #ifdef ARCHITECTURE_x86_64 + // TODO(yuriks): Re-initialize on each change rather than being persistent if (VideoCore::g_shader_jit_enabled) { - u64 cache_key = - Common::ComputeHash64(&g_state.vs.program_code, sizeof(g_state.vs.program_code)) ^ - Common::ComputeHash64(&g_state.vs.swizzle_data, sizeof(g_state.vs.swizzle_data)); - - auto iter = shader_map.find(cache_key); - if (iter != shader_map.end()) { - jit_shader = iter->second.get(); - } else { - auto shader = std::make_unique<JitShader>(); - shader->Compile(); - jit_shader = shader.get(); - shader_map[cache_key] = std::move(shader); + if (jit_engine == nullptr) { + jit_engine = std::make_unique<JitX64Engine>(); } + return jit_engine.get(); } #endif // ARCHITECTURE_x86_64 -} - -MICROPROFILE_DEFINE(GPU_Shader, "GPU", "Shader", MP_RGB(50, 50, 240)); - -void ShaderSetup::Run(UnitState& state, const InputVertex& input, int num_attributes) { - auto& config = g_state.regs.vs; - auto& setup = g_state.vs; - - MICROPROFILE_SCOPE(GPU_Shader); - // Setup input register table - const auto& attribute_register_map = config.input_register_map; - - for (unsigned i = 0; i < num_attributes; i++) - state.registers.input[attribute_register_map.GetRegisterForAttribute(i)] = input.attr[i]; - - state.conditional_code[0] = false; - state.conditional_code[1] = false; + return &interpreter_engine; +} +void Shutdown() { #ifdef ARCHITECTURE_x86_64 - if (VideoCore::g_shader_jit_enabled) { - jit_shader->Run(setup, state, config.main_offset); - } else { - DebugData<false> dummy_debug_data; - RunInterpreter(setup, state, dummy_debug_data, config.main_offset); - } -#else - DebugData<false> dummy_debug_data; - RunInterpreter(setup, state, dummy_debug_data, config.main_offset); + jit_engine = nullptr; #endif // ARCHITECTURE_x86_64 } -DebugData<true> ShaderSetup::ProduceDebugInfo(const InputVertex& input, int num_attributes, - const Regs::ShaderConfig& config, - const ShaderSetup& setup) { - UnitState state; - DebugData<true> debug_data; - - // Setup input register table - boost::fill(state.registers.input, Math::Vec4<float24>::AssignToAll(float24::Zero())); - const auto& attribute_register_map = config.input_register_map; - for (unsigned i = 0; i < num_attributes; i++) - state.registers.input[attribute_register_map.GetRegisterForAttribute(i)] = input.attr[i]; - - state.conditional_code[0] = false; - state.conditional_code[1] = false; - - RunInterpreter(setup, state, debug_data, config.main_offset); - return debug_data; -} - } // namespace Shader } // namespace Pica diff --git a/src/video_core/shader/shader.h b/src/video_core/shader/shader.h index 2b07759b9..44d9f76c3 100644 --- a/src/video_core/shader/shader.h +++ b/src/video_core/shader/shader.h @@ -6,7 +6,6 @@ #include <array> #include <cstddef> -#include <memory> #include <type_traits> #include <nihstro/shader_bytecode.h> #include "common/assert.h" @@ -15,7 +14,6 @@ #include "common/vector_math.h" #include "video_core/pica.h" #include "video_core/pica_types.h" -#include "video_core/shader/debug_data.h" using nihstro::RegisterType; using nihstro::SourceRegister; @@ -75,19 +73,13 @@ struct OutputVertex { ret.Lerp(factor, v1); return ret; } + + static OutputVertex FromRegisters(Math::Vec4<float24> output_regs[16], const Regs& regs, + u32 output_mask); }; static_assert(std::is_pod<OutputVertex>::value, "Structure is not POD"); static_assert(sizeof(OutputVertex) == 32 * sizeof(float), "OutputVertex has invalid size"); -struct OutputRegisters { - OutputRegisters() = default; - - alignas(16) Math::Vec4<float24> value[16]; - - OutputVertex ToVertex(const Regs::ShaderConfig& config) const; -}; -static_assert(std::is_pod<OutputRegisters>::value, "Structure is not POD"); - /** * This structure contains the state information that needs to be unique for a shader unit. The 3DS * has four shader units that process shaders in parallel. At the present, Citra only implements a @@ -100,11 +92,10 @@ struct UnitState { // required to be 16-byte aligned. alignas(16) Math::Vec4<float24> input[16]; alignas(16) Math::Vec4<float24> temporary[16]; + alignas(16) Math::Vec4<float24> output[16]; } registers; static_assert(std::is_pod<Registers>::value, "Structure is not POD"); - OutputRegisters output_registers; - bool conditional_code[2]; // Two Address registers and one loop counter @@ -130,7 +121,7 @@ struct UnitState { static size_t OutputOffset(const DestRegister& reg) { switch (reg.GetRegisterType()) { case RegisterType::Output: - return offsetof(UnitState, output_registers.value) + + return offsetof(UnitState, registers.output) + reg.GetIndex() * sizeof(Math::Vec4<float24>); case RegisterType::Temporary: @@ -142,13 +133,17 @@ struct UnitState { return 0; } } -}; -/// Clears the shader cache -void ClearCache(); + /** + * Loads the unit state with an input vertex. + * + * @param input Input vertex into the shader + * @param num_attributes The number of vertex shader attributes to load + */ + void LoadInputVertex(const InputVertex& input, int num_attributes); +}; struct ShaderSetup { - struct { // The float uniforms are accessed by the shader JIT using SSE instructions, and are // therefore required to be 16-byte aligned. @@ -173,32 +168,37 @@ struct ShaderSetup { std::array<u32, 1024> program_code; std::array<u32, 1024> swizzle_data; + /// Data private to ShaderEngines + struct EngineData { + unsigned int entry_point; + /// Used by the JIT, points to a compiled shader object. + const void* cached_shader = nullptr; + } engine_data; +}; + +class ShaderEngine { +public: + virtual ~ShaderEngine() = default; + /** * Performs any shader unit setup that only needs to happen once per shader (as opposed to once * per vertex, which would happen within the `Run` function). */ - void Setup(); - - /** - * Runs the currently setup shader - * @param state Shader unit state, must be setup per shader and per shader unit - * @param input Input vertex into the shader - * @param num_attributes The number of vertex shader attributes - */ - void Run(UnitState& state, const InputVertex& input, int num_attributes); + virtual void SetupBatch(ShaderSetup& setup, unsigned int entry_point) = 0; /** - * Produce debug information based on the given shader and input vertex - * @param input Input vertex into the shader - * @param num_attributes The number of vertex shader attributes - * @param config Configuration object for the shader pipeline - * @param setup Setup object for the shader pipeline - * @return Debug information for this shader with regards to the given vertex + * Runs the currently setup shader. + * + * @param setup Shader engine state, must be setup with SetupBatch on each shader change. + * @param state Shader unit state, must be setup with input data before each shader invocation. */ - DebugData<true> ProduceDebugInfo(const InputVertex& input, int num_attributes, - const Regs::ShaderConfig& config, const ShaderSetup& setup); + virtual void Run(const ShaderSetup& setup, UnitState& state) const = 0; }; +// TODO(yuriks): Remove and make it non-global state somewhere +ShaderEngine* GetEngine(); +void Shutdown(); + } // namespace Shader } // namespace Pica diff --git a/src/video_core/shader/shader_interpreter.cpp b/src/video_core/shader/shader_interpreter.cpp index 70db4167e..c0c89b857 100644 --- a/src/video_core/shader/shader_interpreter.cpp +++ b/src/video_core/shader/shader_interpreter.cpp @@ -7,10 +7,12 @@ #include <cmath> #include <numeric> #include <boost/container/static_vector.hpp> +#include <boost/range/algorithm/fill.hpp> #include <nihstro/shader_bytecode.h> #include "common/assert.h" #include "common/common_types.h" #include "common/logging/log.h" +#include "common/microprofile.h" #include "common/vector_math.h" #include "video_core/pica_state.h" #include "video_core/pica_types.h" @@ -27,8 +29,6 @@ namespace Pica { namespace Shader { -constexpr u32 INVALID_ADDRESS = 0xFFFFFFFF; - struct CallStackElement { u32 final_address; // Address upon which we jump to return_address u32 return_address; // Where to jump when leaving scope @@ -39,12 +39,15 @@ struct CallStackElement { }; template <bool Debug> -void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData<Debug>& debug_data, - unsigned offset) { +static void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData<Debug>& debug_data, + unsigned offset) { // TODO: Is there a maximal size for this? boost::container::static_vector<CallStackElement, 16> call_stack; u32 program_counter = offset; + state.conditional_code[0] = false; + state.conditional_code[1] = false; + auto call = [&program_counter, &call_stack](u32 offset, u32 num_instructions, u32 return_offset, u8 repeat_count, u8 loop_increment) { // -1 to make sure when incrementing the PC we end up at the correct offset @@ -75,9 +78,9 @@ void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData<Debug> } }; - const auto& uniforms = g_state.vs.uniforms; - const auto& swizzle_data = g_state.vs.swizzle_data; - const auto& program_code = g_state.vs.program_code; + const auto& uniforms = setup.uniforms; + const auto& swizzle_data = setup.swizzle_data; + const auto& program_code = setup.program_code; // Placeholder for invalid inputs static float24 dummy_vec4_float24[4]; @@ -172,7 +175,7 @@ void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData<Debug> float24* dest = (instr.common.dest.Value() < 0x10) - ? &state.output_registers.value[instr.common.dest.Value().GetIndex()][0] + ? &state.registers.output[instr.common.dest.Value().GetIndex()][0] : (instr.common.dest.Value() < 0x20) ? &state.registers.temporary[instr.common.dest.Value().GetIndex()][0] : dummy_vec4_float24; @@ -515,7 +518,7 @@ void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData<Debug> float24* dest = (instr.mad.dest.Value() < 0x10) - ? &state.output_registers.value[instr.mad.dest.Value().GetIndex()][0] + ? &state.registers.output[instr.mad.dest.Value().GetIndex()][0] : (instr.mad.dest.Value() < 0x20) ? &state.registers.temporary[instr.mad.dest.Value().GetIndex()][0] : dummy_vec4_float24; @@ -649,9 +652,33 @@ void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData<Debug> } } -// Explicit instantiation -template void RunInterpreter(const ShaderSetup&, UnitState&, DebugData<false>&, unsigned offset); -template void RunInterpreter(const ShaderSetup&, UnitState&, DebugData<true>&, unsigned offset); +void InterpreterEngine::SetupBatch(ShaderSetup& setup, unsigned int entry_point) { + ASSERT(entry_point < 1024); + setup.engine_data.entry_point = entry_point; +} + +MICROPROFILE_DECLARE(GPU_Shader); + +void InterpreterEngine::Run(const ShaderSetup& setup, UnitState& state) const { + + MICROPROFILE_SCOPE(GPU_Shader); + + DebugData<false> dummy_debug_data; + RunInterpreter(setup, state, dummy_debug_data, setup.engine_data.entry_point); +} + +DebugData<true> InterpreterEngine::ProduceDebugInfo(const ShaderSetup& setup, + const InputVertex& input, + int num_attributes) const { + UnitState state; + DebugData<true> debug_data; + + // Setup input register table + boost::fill(state.registers.input, Math::Vec4<float24>::AssignToAll(float24::Zero())); + state.LoadInputVertex(input, num_attributes); + RunInterpreter(setup, state, debug_data, setup.engine_data.entry_point); + return debug_data; +} } // namespace diff --git a/src/video_core/shader/shader_interpreter.h b/src/video_core/shader/shader_interpreter.h index d31dcd7a6..d6c0e2d8c 100644 --- a/src/video_core/shader/shader_interpreter.h +++ b/src/video_core/shader/shader_interpreter.h @@ -4,18 +4,28 @@ #pragma once +#include "video_core/shader/debug_data.h" +#include "video_core/shader/shader.h" + namespace Pica { namespace Shader { -struct UnitState; - -template <bool Debug> -struct DebugData; - -template <bool Debug> -void RunInterpreter(const ShaderSetup& setup, UnitState& state, DebugData<Debug>& debug_data, - unsigned offset); +class InterpreterEngine final : public ShaderEngine { +public: + void SetupBatch(ShaderSetup& setup, unsigned int entry_point) override; + void Run(const ShaderSetup& setup, UnitState& state) const override; + + /** + * Produce debug information based on the given shader and input vertex + * @param input Input vertex into the shader + * @param num_attributes The number of vertex shader attributes + * @param config Configuration object for the shader pipeline + * @return Debug information for this shader with regards to the given vertex + */ + DebugData<true> ProduceDebugInfo(const ShaderSetup& setup, const InputVertex& input, + int num_attributes) const; +}; } // namespace diff --git a/src/video_core/shader/shader_jit_x64.cpp b/src/video_core/shader/shader_jit_x64.cpp index c588b778b..0ee0dd9ef 100644 --- a/src/video_core/shader/shader_jit_x64.cpp +++ b/src/video_core/shader/shader_jit_x64.cpp @@ -1,888 +1,48 @@ -// Copyright 2015 Citra Emulator Project +// Copyright 2016 Citra Emulator Project // Licensed under GPLv2 or any later version // Refer to the license.txt file included. -#include <algorithm> -#include <cmath> -#include <cstdint> -#include <nihstro/shader_bytecode.h> -#include <smmintrin.h> -#include <xmmintrin.h> -#include "common/assert.h" -#include "common/logging/log.h" -#include "common/vector_math.h" -#include "common/x64/cpu_detect.h" -#include "common/x64/xbyak_abi.h" -#include "common/x64/xbyak_util.h" -#include "video_core/pica_state.h" -#include "video_core/pica_types.h" +#include "common/hash.h" +#include "common/microprofile.h" #include "video_core/shader/shader.h" #include "video_core/shader/shader_jit_x64.h" - -using namespace Common::X64; -using namespace Xbyak::util; -using Xbyak::Label; -using Xbyak::Reg32; -using Xbyak::Reg64; -using Xbyak::Xmm; +#include "video_core/shader/shader_jit_x64_compiler.h" namespace Pica { - namespace Shader { -typedef void (JitShader::*JitFunction)(Instruction instr); - -const JitFunction instr_table[64] = { - &JitShader::Compile_ADD, // add - &JitShader::Compile_DP3, // dp3 - &JitShader::Compile_DP4, // dp4 - &JitShader::Compile_DPH, // dph - nullptr, // unknown - &JitShader::Compile_EX2, // ex2 - &JitShader::Compile_LG2, // lg2 - nullptr, // unknown - &JitShader::Compile_MUL, // mul - &JitShader::Compile_SGE, // sge - &JitShader::Compile_SLT, // slt - &JitShader::Compile_FLR, // flr - &JitShader::Compile_MAX, // max - &JitShader::Compile_MIN, // min - &JitShader::Compile_RCP, // rcp - &JitShader::Compile_RSQ, // rsq - nullptr, // unknown - nullptr, // unknown - &JitShader::Compile_MOVA, // mova - &JitShader::Compile_MOV, // mov - nullptr, // unknown - nullptr, // unknown - nullptr, // unknown - nullptr, // unknown - &JitShader::Compile_DPH, // dphi - nullptr, // unknown - &JitShader::Compile_SGE, // sgei - &JitShader::Compile_SLT, // slti - nullptr, // unknown - nullptr, // unknown - nullptr, // unknown - nullptr, // unknown - nullptr, // unknown - &JitShader::Compile_NOP, // nop - &JitShader::Compile_END, // end - nullptr, // break - &JitShader::Compile_CALL, // call - &JitShader::Compile_CALLC, // callc - &JitShader::Compile_CALLU, // callu - &JitShader::Compile_IF, // ifu - &JitShader::Compile_IF, // ifc - &JitShader::Compile_LOOP, // loop - nullptr, // emit - nullptr, // sete - &JitShader::Compile_JMP, // jmpc - &JitShader::Compile_JMP, // jmpu - &JitShader::Compile_CMP, // cmp - &JitShader::Compile_CMP, // cmp - &JitShader::Compile_MAD, // madi - &JitShader::Compile_MAD, // madi - &JitShader::Compile_MAD, // madi - &JitShader::Compile_MAD, // madi - &JitShader::Compile_MAD, // madi - &JitShader::Compile_MAD, // madi - &JitShader::Compile_MAD, // madi - &JitShader::Compile_MAD, // madi - &JitShader::Compile_MAD, // mad - &JitShader::Compile_MAD, // mad - &JitShader::Compile_MAD, // mad - &JitShader::Compile_MAD, // mad - &JitShader::Compile_MAD, // mad - &JitShader::Compile_MAD, // mad - &JitShader::Compile_MAD, // mad - &JitShader::Compile_MAD, // mad -}; - -// The following is used to alias some commonly used registers. Generally, RAX-RDX and XMM0-XMM3 can -// be used as scratch registers within a compiler function. The other registers have designated -// purposes, as documented below: +JitX64Engine::JitX64Engine() = default; +JitX64Engine::~JitX64Engine() = default; -/// Pointer to the uniform memory -static const Reg64 SETUP = r9; -/// The two 32-bit VS address offset registers set by the MOVA instruction -static const Reg64 ADDROFFS_REG_0 = r10; -static const Reg64 ADDROFFS_REG_1 = r11; -/// VS loop count register (Multiplied by 16) -static const Reg32 LOOPCOUNT_REG = r12d; -/// Current VS loop iteration number (we could probably use LOOPCOUNT_REG, but this quicker) -static const Reg32 LOOPCOUNT = esi; -/// Number to increment LOOPCOUNT_REG by on each loop iteration (Multiplied by 16) -static const Reg32 LOOPINC = edi; -/// Result of the previous CMP instruction for the X-component comparison -static const Reg64 COND0 = r13; -/// Result of the previous CMP instruction for the Y-component comparison -static const Reg64 COND1 = r14; -/// Pointer to the UnitState instance for the current VS unit -static const Reg64 STATE = r15; -/// SIMD scratch register -static const Xmm SCRATCH = xmm0; -/// Loaded with the first swizzled source register, otherwise can be used as a scratch register -static const Xmm SRC1 = xmm1; -/// Loaded with the second swizzled source register, otherwise can be used as a scratch register -static const Xmm SRC2 = xmm2; -/// Loaded with the third swizzled source register, otherwise can be used as a scratch register -static const Xmm SRC3 = xmm3; -/// Additional scratch register -static const Xmm SCRATCH2 = xmm4; -/// Constant vector of [1.0f, 1.0f, 1.0f, 1.0f], used to efficiently set a vector to one -static const Xmm ONE = xmm14; -/// Constant vector of [-0.f, -0.f, -0.f, -0.f], used to efficiently negate a vector with XOR -static const Xmm NEGBIT = xmm15; +void JitX64Engine::SetupBatch(ShaderSetup& setup, unsigned int entry_point) { + ASSERT(entry_point < 1024); + setup.engine_data.entry_point = entry_point; -// State registers that must not be modified by external functions calls -// Scratch registers, e.g., SRC1 and SCRATCH, have to be saved on the side if needed -static const BitSet32 persistent_regs = BuildRegSet({ - // Pointers to register blocks - SETUP, STATE, - // Cached registers - ADDROFFS_REG_0, ADDROFFS_REG_1, LOOPCOUNT_REG, COND0, COND1, - // Constants - ONE, NEGBIT, -}); + u64 code_hash = Common::ComputeHash64(&setup.program_code, sizeof(setup.program_code)); + u64 swizzle_hash = Common::ComputeHash64(&setup.swizzle_data, sizeof(setup.swizzle_data)); -/// Raw constant for the source register selector that indicates no swizzling is performed -static const u8 NO_SRC_REG_SWIZZLE = 0x1b; -/// Raw constant for the destination register enable mask that indicates all components are enabled -static const u8 NO_DEST_REG_MASK = 0xf; - -/** - * Get the vertex shader instruction for a given offset in the current shader program - * @param offset Offset in the current shader program of the instruction - * @return Instruction at the specified offset - */ -static Instruction GetVertexShaderInstruction(size_t offset) { - return {g_state.vs.program_code[offset]}; -} - -static void LogCritical(const char* msg) { - LOG_CRITICAL(HW_GPU, "%s", msg); -} - -void JitShader::Compile_Assert(bool condition, const char* msg) { - if (!condition) { - mov(ABI_PARAM1, reinterpret_cast<size_t>(msg)); - CallFarFunction(*this, LogCritical); - } -} - -/** - * Loads and swizzles a source register into the specified XMM register. - * @param instr VS instruction, used for determining how to load the source register - * @param src_num Number indicating which source register to load (1 = src1, 2 = src2, 3 = src3) - * @param src_reg SourceRegister object corresponding to the source register to load - * @param dest Destination XMM register to store the loaded, swizzled source register - */ -void JitShader::Compile_SwizzleSrc(Instruction instr, unsigned src_num, SourceRegister src_reg, - Xmm dest) { - Reg64 src_ptr; - size_t src_offset; - - if (src_reg.GetRegisterType() == RegisterType::FloatUniform) { - src_ptr = SETUP; - src_offset = ShaderSetup::GetFloatUniformOffset(src_reg.GetIndex()); + u64 cache_key = code_hash ^ swizzle_hash; + auto iter = cache.find(cache_key); + if (iter != cache.end()) { + setup.engine_data.cached_shader = iter->second.get(); } else { - src_ptr = STATE; - src_offset = UnitState::InputOffset(src_reg); - } - - int src_offset_disp = (int)src_offset; - ASSERT_MSG(src_offset == src_offset_disp, "Source register offset too large for int type"); - - unsigned operand_desc_id; - - const bool is_inverted = - (0 != (instr.opcode.Value().GetInfo().subtype & OpCode::Info::SrcInversed)); - - unsigned address_register_index; - unsigned offset_src; - - if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MAD || - instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MADI) { - operand_desc_id = instr.mad.operand_desc_id; - offset_src = is_inverted ? 3 : 2; - address_register_index = instr.mad.address_register_index; - } else { - operand_desc_id = instr.common.operand_desc_id; - offset_src = is_inverted ? 2 : 1; - address_register_index = instr.common.address_register_index; - } - - if (src_num == offset_src && address_register_index != 0) { - switch (address_register_index) { - case 1: // address offset 1 - movaps(dest, xword[src_ptr + ADDROFFS_REG_0 + src_offset_disp]); - break; - case 2: // address offset 2 - movaps(dest, xword[src_ptr + ADDROFFS_REG_1 + src_offset_disp]); - break; - case 3: // address offset 3 - movaps(dest, xword[src_ptr + LOOPCOUNT_REG.cvt64() + src_offset_disp]); - break; - default: - UNREACHABLE(); - break; - } - } else { - // Load the source - movaps(dest, xword[src_ptr + src_offset_disp]); - } - - SwizzlePattern swiz = {g_state.vs.swizzle_data[operand_desc_id]}; - - // Generate instructions for source register swizzling as needed - u8 sel = swiz.GetRawSelector(src_num); - if (sel != NO_SRC_REG_SWIZZLE) { - // Selector component order needs to be reversed for the SHUFPS instruction - sel = ((sel & 0xc0) >> 6) | ((sel & 3) << 6) | ((sel & 0xc) << 2) | ((sel & 0x30) >> 2); - - // Shuffle inputs for swizzle - shufps(dest, dest, sel); - } - - // If the source register should be negated, flip the negative bit using XOR - const bool negate[] = {swiz.negate_src1, swiz.negate_src2, swiz.negate_src3}; - if (negate[src_num - 1]) { - xorps(dest, NEGBIT); + auto shader = std::make_unique<JitShader>(); + shader->Compile(&setup.program_code, &setup.swizzle_data); + setup.engine_data.cached_shader = shader.get(); + cache.emplace_hint(iter, cache_key, std::move(shader)); } } -void JitShader::Compile_DestEnable(Instruction instr, Xmm src) { - DestRegister dest; - unsigned operand_desc_id; - if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MAD || - instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MADI) { - operand_desc_id = instr.mad.operand_desc_id; - dest = instr.mad.dest.Value(); - } else { - operand_desc_id = instr.common.operand_desc_id; - dest = instr.common.dest.Value(); - } - - SwizzlePattern swiz = {g_state.vs.swizzle_data[operand_desc_id]}; - - size_t dest_offset_disp = UnitState::OutputOffset(dest); - - // If all components are enabled, write the result to the destination register - if (swiz.dest_mask == NO_DEST_REG_MASK) { - // Store dest back to memory - movaps(xword[STATE + dest_offset_disp], src); - - } else { - // Not all components are enabled, so mask the result when storing to the destination - // register... - movaps(SCRATCH, xword[STATE + dest_offset_disp]); - - if (Common::GetCPUCaps().sse4_1) { - u8 mask = ((swiz.dest_mask & 1) << 3) | ((swiz.dest_mask & 8) >> 3) | - ((swiz.dest_mask & 2) << 1) | ((swiz.dest_mask & 4) >> 1); - blendps(SCRATCH, src, mask); - } else { - movaps(SCRATCH2, src); - unpckhps(SCRATCH2, SCRATCH); // Unpack X/Y components of source and destination - unpcklps(SCRATCH, src); // Unpack Z/W components of source and destination - - // Compute selector to selectively copy source components to destination for SHUFPS - // instruction - u8 sel = ((swiz.DestComponentEnabled(0) ? 1 : 0) << 0) | - ((swiz.DestComponentEnabled(1) ? 3 : 2) << 2) | - ((swiz.DestComponentEnabled(2) ? 0 : 1) << 4) | - ((swiz.DestComponentEnabled(3) ? 2 : 3) << 6); - shufps(SCRATCH, SCRATCH2, sel); - } - - // Store dest back to memory - movaps(xword[STATE + dest_offset_disp], SCRATCH); - } -} - -void JitShader::Compile_SanitizedMul(Xmm src1, Xmm src2, Xmm scratch) { - movaps(scratch, src1); - cmpordps(scratch, src2); - - mulps(src1, src2); +MICROPROFILE_DECLARE(GPU_Shader); - movaps(src2, src1); - cmpunordps(src2, src2); +void JitX64Engine::Run(const ShaderSetup& setup, UnitState& state) const { + ASSERT(setup.engine_data.cached_shader != nullptr); - xorps(scratch, src2); - andps(src1, scratch); -} - -void JitShader::Compile_EvaluateCondition(Instruction instr) { - // Note: NXOR is used below to check for equality - switch (instr.flow_control.op) { - case Instruction::FlowControlType::Or: - mov(eax, COND0); - mov(ebx, COND1); - xor(eax, (instr.flow_control.refx.Value() ^ 1)); - xor(ebx, (instr.flow_control.refy.Value() ^ 1)); - or (eax, ebx); - break; - - case Instruction::FlowControlType::And: - mov(eax, COND0); - mov(ebx, COND1); - xor(eax, (instr.flow_control.refx.Value() ^ 1)); - xor(ebx, (instr.flow_control.refy.Value() ^ 1)); - and(eax, ebx); - break; - - case Instruction::FlowControlType::JustX: - mov(eax, COND0); - xor(eax, (instr.flow_control.refx.Value() ^ 1)); - break; - - case Instruction::FlowControlType::JustY: - mov(eax, COND1); - xor(eax, (instr.flow_control.refy.Value() ^ 1)); - break; - } -} + MICROPROFILE_SCOPE(GPU_Shader); -void JitShader::Compile_UniformCondition(Instruction instr) { - size_t offset = ShaderSetup::GetBoolUniformOffset(instr.flow_control.bool_uniform_id); - cmp(byte[SETUP + offset], 0); + const JitShader* shader = static_cast<const JitShader*>(setup.engine_data.cached_shader); + shader->Run(setup, state, setup.engine_data.entry_point); } -BitSet32 JitShader::PersistentCallerSavedRegs() { - return persistent_regs & ABI_ALL_CALLER_SAVED; -} - -void JitShader::Compile_ADD(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); - addps(SRC1, SRC2); - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_DP3(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); - - Compile_SanitizedMul(SRC1, SRC2, SCRATCH); - - movaps(SRC2, SRC1); - shufps(SRC2, SRC2, _MM_SHUFFLE(1, 1, 1, 1)); - - movaps(SRC3, SRC1); - shufps(SRC3, SRC3, _MM_SHUFFLE(2, 2, 2, 2)); - - shufps(SRC1, SRC1, _MM_SHUFFLE(0, 0, 0, 0)); - addps(SRC1, SRC2); - addps(SRC1, SRC3); - - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_DP4(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); - - Compile_SanitizedMul(SRC1, SRC2, SCRATCH); - - movaps(SRC2, SRC1); - shufps(SRC1, SRC1, _MM_SHUFFLE(2, 3, 0, 1)); // XYZW -> ZWXY - addps(SRC1, SRC2); - - movaps(SRC2, SRC1); - shufps(SRC1, SRC1, _MM_SHUFFLE(0, 1, 2, 3)); // XYZW -> WZYX - addps(SRC1, SRC2); - - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_DPH(Instruction instr) { - if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::DPHI) { - Compile_SwizzleSrc(instr, 1, instr.common.src1i, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2i, SRC2); - } else { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); - } - - if (Common::GetCPUCaps().sse4_1) { - // Set 4th component to 1.0 - blendps(SRC1, ONE, 0b1000); - } else { - // Set 4th component to 1.0 - movaps(SCRATCH, SRC1); - unpckhps(SCRATCH, ONE); // XYZW, 1111 -> Z1__ - unpcklpd(SRC1, SCRATCH); // XYZW, Z1__ -> XYZ1 - } - - Compile_SanitizedMul(SRC1, SRC2, SCRATCH); - - movaps(SRC2, SRC1); - shufps(SRC1, SRC1, _MM_SHUFFLE(2, 3, 0, 1)); // XYZW -> ZWXY - addps(SRC1, SRC2); - - movaps(SRC2, SRC1); - shufps(SRC1, SRC1, _MM_SHUFFLE(0, 1, 2, 3)); // XYZW -> WZYX - addps(SRC1, SRC2); - - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_EX2(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - movss(xmm0, SRC1); // ABI_PARAM1 - - ABI_PushRegistersAndAdjustStack(*this, PersistentCallerSavedRegs(), 0); - CallFarFunction(*this, exp2f); - ABI_PopRegistersAndAdjustStack(*this, PersistentCallerSavedRegs(), 0); - - shufps(xmm0, xmm0, _MM_SHUFFLE(0, 0, 0, 0)); // ABI_RETURN - movaps(SRC1, xmm0); - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_LG2(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - movss(xmm0, SRC1); // ABI_PARAM1 - - ABI_PushRegistersAndAdjustStack(*this, PersistentCallerSavedRegs(), 0); - CallFarFunction(*this, log2f); - ABI_PopRegistersAndAdjustStack(*this, PersistentCallerSavedRegs(), 0); - - shufps(xmm0, xmm0, _MM_SHUFFLE(0, 0, 0, 0)); // ABI_RETURN - movaps(SRC1, xmm0); - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_MUL(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); - Compile_SanitizedMul(SRC1, SRC2, SCRATCH); - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_SGE(Instruction instr) { - if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::SGEI) { - Compile_SwizzleSrc(instr, 1, instr.common.src1i, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2i, SRC2); - } else { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); - } - - cmpleps(SRC2, SRC1); - andps(SRC2, ONE); - - Compile_DestEnable(instr, SRC2); -} - -void JitShader::Compile_SLT(Instruction instr) { - if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::SLTI) { - Compile_SwizzleSrc(instr, 1, instr.common.src1i, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2i, SRC2); - } else { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); - } - - cmpltps(SRC1, SRC2); - andps(SRC1, ONE); - - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_FLR(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - - if (Common::GetCPUCaps().sse4_1) { - roundps(SRC1, SRC1, _MM_FROUND_FLOOR); - } else { - cvttps2dq(SRC1, SRC1); - cvtdq2ps(SRC1, SRC1); - } - - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_MAX(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); - // SSE semantics match PICA200 ones: In case of NaN, SRC2 is returned. - maxps(SRC1, SRC2); - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_MIN(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); - // SSE semantics match PICA200 ones: In case of NaN, SRC2 is returned. - minps(SRC1, SRC2); - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_MOVA(Instruction instr) { - SwizzlePattern swiz = {g_state.vs.swizzle_data[instr.common.operand_desc_id]}; - - if (!swiz.DestComponentEnabled(0) && !swiz.DestComponentEnabled(1)) { - return; // NoOp - } - - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - - // Convert floats to integers using truncation (only care about X and Y components) - cvttps2dq(SRC1, SRC1); - - // Get result - movq(rax, SRC1); - - // Handle destination enable - if (swiz.DestComponentEnabled(0) && swiz.DestComponentEnabled(1)) { - // Move and sign-extend low 32 bits - movsxd(ADDROFFS_REG_0, eax); - - // Move and sign-extend high 32 bits - shr(rax, 32); - movsxd(ADDROFFS_REG_1, eax); - - // Multiply by 16 to be used as an offset later - shl(ADDROFFS_REG_0, 4); - shl(ADDROFFS_REG_1, 4); - } else { - if (swiz.DestComponentEnabled(0)) { - // Move and sign-extend low 32 bits - movsxd(ADDROFFS_REG_0, eax); - - // Multiply by 16 to be used as an offset later - shl(ADDROFFS_REG_0, 4); - } else if (swiz.DestComponentEnabled(1)) { - // Move and sign-extend high 32 bits - shr(rax, 32); - movsxd(ADDROFFS_REG_1, eax); - - // Multiply by 16 to be used as an offset later - shl(ADDROFFS_REG_1, 4); - } - } -} - -void JitShader::Compile_MOV(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_RCP(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - - // TODO(bunnei): RCPSS is a pretty rough approximation, this might cause problems if Pica - // performs this operation more accurately. This should be checked on hardware. - rcpss(SRC1, SRC1); - shufps(SRC1, SRC1, _MM_SHUFFLE(0, 0, 0, 0)); // XYWZ -> XXXX - - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_RSQ(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - - // TODO(bunnei): RSQRTSS is a pretty rough approximation, this might cause problems if Pica - // performs this operation more accurately. This should be checked on hardware. - rsqrtss(SRC1, SRC1); - shufps(SRC1, SRC1, _MM_SHUFFLE(0, 0, 0, 0)); // XYWZ -> XXXX - - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_NOP(Instruction instr) {} - -void JitShader::Compile_END(Instruction instr) { - ABI_PopRegistersAndAdjustStack(*this, ABI_ALL_CALLEE_SAVED, 8); - ret(); -} - -void JitShader::Compile_CALL(Instruction instr) { - // Push offset of the return - push(qword, (instr.flow_control.dest_offset + instr.flow_control.num_instructions)); - - // Call the subroutine - call(instruction_labels[instr.flow_control.dest_offset]); - - // Skip over the return offset that's on the stack - add(rsp, 8); -} - -void JitShader::Compile_CALLC(Instruction instr) { - Compile_EvaluateCondition(instr); - Label b; - jz(b); - Compile_CALL(instr); - L(b); -} - -void JitShader::Compile_CALLU(Instruction instr) { - Compile_UniformCondition(instr); - Label b; - jz(b); - Compile_CALL(instr); - L(b); -} - -void JitShader::Compile_CMP(Instruction instr) { - using Op = Instruction::Common::CompareOpType::Op; - Op op_x = instr.common.compare_op.x; - Op op_y = instr.common.compare_op.y; - - Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); - Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); - - // SSE doesn't have greater-than (GT) or greater-equal (GE) comparison operators. You need to - // emulate them by swapping the lhs and rhs and using LT and LE. NLT and NLE can't be used here - // because they don't match when used with NaNs. - static const u8 cmp[] = {CMP_EQ, CMP_NEQ, CMP_LT, CMP_LE, CMP_LT, CMP_LE}; - - bool invert_op_x = (op_x == Op::GreaterThan || op_x == Op::GreaterEqual); - Xmm lhs_x = invert_op_x ? SRC2 : SRC1; - Xmm rhs_x = invert_op_x ? SRC1 : SRC2; - - if (op_x == op_y) { - // Compare X-component and Y-component together - cmpps(lhs_x, rhs_x, cmp[op_x]); - movq(COND0, lhs_x); - - mov(COND1, COND0); - } else { - bool invert_op_y = (op_y == Op::GreaterThan || op_y == Op::GreaterEqual); - Xmm lhs_y = invert_op_y ? SRC2 : SRC1; - Xmm rhs_y = invert_op_y ? SRC1 : SRC2; - - // Compare X-component - movaps(SCRATCH, lhs_x); - cmpss(SCRATCH, rhs_x, cmp[op_x]); - - // Compare Y-component - cmpps(lhs_y, rhs_y, cmp[op_y]); - - movq(COND0, SCRATCH); - movq(COND1, lhs_y); - } - - shr(COND0.cvt32(), 31); // ignores upper 32 bits in source - shr(COND1, 63); -} - -void JitShader::Compile_MAD(Instruction instr) { - Compile_SwizzleSrc(instr, 1, instr.mad.src1, SRC1); - - if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MADI) { - Compile_SwizzleSrc(instr, 2, instr.mad.src2i, SRC2); - Compile_SwizzleSrc(instr, 3, instr.mad.src3i, SRC3); - } else { - Compile_SwizzleSrc(instr, 2, instr.mad.src2, SRC2); - Compile_SwizzleSrc(instr, 3, instr.mad.src3, SRC3); - } - - Compile_SanitizedMul(SRC1, SRC2, SCRATCH); - addps(SRC1, SRC3); - - Compile_DestEnable(instr, SRC1); -} - -void JitShader::Compile_IF(Instruction instr) { - Compile_Assert(instr.flow_control.dest_offset >= program_counter, - "Backwards if-statements not supported"); - Label l_else, l_endif; - - // Evaluate the "IF" condition - if (instr.opcode.Value() == OpCode::Id::IFU) { - Compile_UniformCondition(instr); - } else if (instr.opcode.Value() == OpCode::Id::IFC) { - Compile_EvaluateCondition(instr); - } - jz(l_else, T_NEAR); - - // Compile the code that corresponds to the condition evaluating as true - Compile_Block(instr.flow_control.dest_offset); - - // If there isn't an "ELSE" condition, we are done here - if (instr.flow_control.num_instructions == 0) { - L(l_else); - return; - } - - jmp(l_endif, T_NEAR); - - L(l_else); - // This code corresponds to the "ELSE" condition - // Comple the code that corresponds to the condition evaluating as false - Compile_Block(instr.flow_control.dest_offset + instr.flow_control.num_instructions); - - L(l_endif); -} - -void JitShader::Compile_LOOP(Instruction instr) { - Compile_Assert(instr.flow_control.dest_offset >= program_counter, - "Backwards loops not supported"); - Compile_Assert(!looping, "Nested loops not supported"); - - looping = true; - - // This decodes the fields from the integer uniform at index instr.flow_control.int_uniform_id. - // The Y (LOOPCOUNT_REG) and Z (LOOPINC) component are kept multiplied by 16 (Left shifted by - // 4 bits) to be used as an offset into the 16-byte vector registers later - size_t offset = ShaderSetup::GetIntUniformOffset(instr.flow_control.int_uniform_id); - mov(LOOPCOUNT, dword[SETUP + offset]); - mov(LOOPCOUNT_REG, LOOPCOUNT); - shr(LOOPCOUNT_REG, 4); - and(LOOPCOUNT_REG, 0xFF0); // Y-component is the start - mov(LOOPINC, LOOPCOUNT); - shr(LOOPINC, 12); - and(LOOPINC, 0xFF0); // Z-component is the incrementer - movzx(LOOPCOUNT, LOOPCOUNT.cvt8()); // X-component is iteration count - add(LOOPCOUNT, 1); // Iteration count is X-component + 1 - - Label l_loop_start; - L(l_loop_start); - - Compile_Block(instr.flow_control.dest_offset + 1); - - add(LOOPCOUNT_REG, LOOPINC); // Increment LOOPCOUNT_REG by Z-component - sub(LOOPCOUNT, 1); // Increment loop count by 1 - jnz(l_loop_start); // Loop if not equal - - looping = false; -} - -void JitShader::Compile_JMP(Instruction instr) { - if (instr.opcode.Value() == OpCode::Id::JMPC) - Compile_EvaluateCondition(instr); - else if (instr.opcode.Value() == OpCode::Id::JMPU) - Compile_UniformCondition(instr); - else - UNREACHABLE(); - - bool inverted_condition = - (instr.opcode.Value() == OpCode::Id::JMPU) && (instr.flow_control.num_instructions & 1); - - Label& b = instruction_labels[instr.flow_control.dest_offset]; - if (inverted_condition) { - jz(b, T_NEAR); - } else { - jnz(b, T_NEAR); - } -} - -void JitShader::Compile_Block(unsigned end) { - while (program_counter < end) { - Compile_NextInstr(); - } -} - -void JitShader::Compile_Return() { - // Peek return offset on the stack and check if we're at that offset - mov(rax, qword[rsp + 8]); - cmp(eax, (program_counter)); - - // If so, jump back to before CALL - Label b; - jnz(b); - ret(); - L(b); -} - -void JitShader::Compile_NextInstr() { - if (std::binary_search(return_offsets.begin(), return_offsets.end(), program_counter)) { - Compile_Return(); - } - - L(instruction_labels[program_counter]); - - Instruction instr = GetVertexShaderInstruction(program_counter++); - - OpCode::Id opcode = instr.opcode.Value(); - auto instr_func = instr_table[static_cast<unsigned>(opcode)]; - - if (instr_func) { - // JIT the instruction! - ((*this).*instr_func)(instr); - } else { - // Unhandled instruction - LOG_CRITICAL(HW_GPU, "Unhandled instruction: 0x%02x (0x%08x)", - instr.opcode.Value().EffectiveOpCode(), instr.hex); - } -} - -void JitShader::FindReturnOffsets() { - return_offsets.clear(); - - for (size_t offset = 0; offset < g_state.vs.program_code.size(); ++offset) { - Instruction instr = GetVertexShaderInstruction(offset); - - switch (instr.opcode.Value()) { - case OpCode::Id::CALL: - case OpCode::Id::CALLC: - case OpCode::Id::CALLU: - return_offsets.push_back(instr.flow_control.dest_offset + - instr.flow_control.num_instructions); - break; - default: - break; - } - } - - // Sort for efficient binary search later - std::sort(return_offsets.begin(), return_offsets.end()); -} - -void JitShader::Compile() { - // Reset flow control state - program = (CompiledShader*)getCurr(); - program_counter = 0; - looping = false; - instruction_labels.fill(Xbyak::Label()); - - // Find all `CALL` instructions and identify return locations - FindReturnOffsets(); - - // The stack pointer is 8 modulo 16 at the entry of a procedure - ABI_PushRegistersAndAdjustStack(*this, ABI_ALL_CALLEE_SAVED, 8); - - mov(SETUP, ABI_PARAM1); - mov(STATE, ABI_PARAM2); - - // Zero address/loop registers - xor(ADDROFFS_REG_0.cvt32(), ADDROFFS_REG_0.cvt32()); - xor(ADDROFFS_REG_1.cvt32(), ADDROFFS_REG_1.cvt32()); - xor(LOOPCOUNT_REG, LOOPCOUNT_REG); - - // Used to set a register to one - static const __m128 one = {1.f, 1.f, 1.f, 1.f}; - mov(rax, reinterpret_cast<size_t>(&one)); - movaps(ONE, xword[rax]); - - // Used to negate registers - static const __m128 neg = {-0.f, -0.f, -0.f, -0.f}; - mov(rax, reinterpret_cast<size_t>(&neg)); - movaps(NEGBIT, xword[rax]); - - // Jump to start of the shader program - jmp(ABI_PARAM3); - - // Compile entire program - Compile_Block(static_cast<unsigned>(g_state.vs.program_code.size())); - - // Free memory that's no longer needed - return_offsets.clear(); - return_offsets.shrink_to_fit(); - - ready(); - - uintptr_t size = reinterpret_cast<uintptr_t>(getCurr()) - reinterpret_cast<uintptr_t>(program); - ASSERT_MSG(size <= MAX_SHADER_SIZE, "Compiled a shader that exceeds the allocated size!"); - LOG_DEBUG(HW_GPU, "Compiled shader size=%lu", size); -} - -JitShader::JitShader() : Xbyak::CodeGenerator(MAX_SHADER_SIZE) {} - } // namespace Shader - } // namespace Pica diff --git a/src/video_core/shader/shader_jit_x64.h b/src/video_core/shader/shader_jit_x64.h index f37548306..078b2cba5 100644 --- a/src/video_core/shader/shader_jit_x64.h +++ b/src/video_core/shader/shader_jit_x64.h @@ -1,121 +1,30 @@ -// Copyright 2015 Citra Emulator Project +// Copyright 2016 Citra Emulator Project // Licensed under GPLv2 or any later version // Refer to the license.txt file included. #pragma once -#include <array> -#include <cstddef> -#include <utility> -#include <vector> -#include <nihstro/shader_bytecode.h> -#include <xbyak.h> -#include "common/bit_set.h" +#include <memory> +#include <unordered_map> #include "common/common_types.h" -#include "common/x64/emitter.h" #include "video_core/shader/shader.h" -using nihstro::Instruction; -using nihstro::OpCode; -using nihstro::SwizzlePattern; - namespace Pica { - namespace Shader { -/// Memory allocated for each compiled shader (64Kb) -constexpr size_t MAX_SHADER_SIZE = 1024 * 64; +class JitShader; -/** - * This class implements the shader JIT compiler. It recompiles a Pica shader program into x86_64 - * code that can be executed on the host machine directly. - */ -class JitShader : public Xbyak::CodeGenerator { +class JitX64Engine final : public ShaderEngine { public: - JitShader(); - - void Run(const ShaderSetup& setup, UnitState& state, unsigned offset) const { - program(&setup, &state, instruction_labels[offset].getAddress()); - } - - void Compile(); + JitX64Engine(); + ~JitX64Engine() override; - void Compile_ADD(Instruction instr); - void Compile_DP3(Instruction instr); - void Compile_DP4(Instruction instr); - void Compile_DPH(Instruction instr); - void Compile_EX2(Instruction instr); - void Compile_LG2(Instruction instr); - void Compile_MUL(Instruction instr); - void Compile_SGE(Instruction instr); - void Compile_SLT(Instruction instr); - void Compile_FLR(Instruction instr); - void Compile_MAX(Instruction instr); - void Compile_MIN(Instruction instr); - void Compile_RCP(Instruction instr); - void Compile_RSQ(Instruction instr); - void Compile_MOVA(Instruction instr); - void Compile_MOV(Instruction instr); - void Compile_NOP(Instruction instr); - void Compile_END(Instruction instr); - void Compile_CALL(Instruction instr); - void Compile_CALLC(Instruction instr); - void Compile_CALLU(Instruction instr); - void Compile_IF(Instruction instr); - void Compile_LOOP(Instruction instr); - void Compile_JMP(Instruction instr); - void Compile_CMP(Instruction instr); - void Compile_MAD(Instruction instr); + void SetupBatch(ShaderSetup& setup, unsigned int entry_point) override; + void Run(const ShaderSetup& setup, UnitState& state) const override; private: - void Compile_Block(unsigned end); - void Compile_NextInstr(); - - void Compile_SwizzleSrc(Instruction instr, unsigned src_num, SourceRegister src_reg, - Xbyak::Xmm dest); - void Compile_DestEnable(Instruction instr, Xbyak::Xmm dest); - - /** - * Compiles a `MUL src1, src2` operation, properly handling the PICA semantics when multiplying - * zero by inf. Clobbers `src2` and `scratch`. - */ - void Compile_SanitizedMul(Xbyak::Xmm src1, Xbyak::Xmm src2, Xbyak::Xmm scratch); - - void Compile_EvaluateCondition(Instruction instr); - void Compile_UniformCondition(Instruction instr); - - /** - * Emits the code to conditionally return from a subroutine envoked by the `CALL` instruction. - */ - void Compile_Return(); - - BitSet32 PersistentCallerSavedRegs(); - - /** - * Assertion evaluated at compile-time, but only triggered if executed at runtime. - * @param msg Message to be logged if the assertion fails. - */ - void Compile_Assert(bool condition, const char* msg); - - /** - * Analyzes the entire shader program for `CALL` instructions before emitting any code, - * identifying the locations where a return needs to be inserted. - */ - void FindReturnOffsets(); - - /// Mapping of Pica VS instructions to pointers in the emitted code - std::array<Xbyak::Label, 1024> instruction_labels; - - /// Offsets in code where a return needs to be inserted - std::vector<unsigned> return_offsets; - - unsigned program_counter = 0; ///< Offset of the next instruction to decode - bool looping = false; ///< True if compiling a loop, used to check for nested loops - - using CompiledShader = void(const void* setup, void* state, const u8* start_addr); - CompiledShader* program = nullptr; + std::unordered_map<u64, std::unique_ptr<JitShader>> cache; }; -} // Shader - -} // Pica +} // namespace Shader +} // namespace Pica diff --git a/src/video_core/shader/shader_jit_x64_compiler.cpp b/src/video_core/shader/shader_jit_x64_compiler.cpp new file mode 100644 index 000000000..49806e8c9 --- /dev/null +++ b/src/video_core/shader/shader_jit_x64_compiler.cpp @@ -0,0 +1,884 @@ +// Copyright 2015 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#include <algorithm> +#include <cmath> +#include <cstdint> +#include <nihstro/shader_bytecode.h> +#include <smmintrin.h> +#include <xmmintrin.h> +#include "common/assert.h" +#include "common/logging/log.h" +#include "common/vector_math.h" +#include "common/x64/cpu_detect.h" +#include "common/x64/xbyak_abi.h" +#include "common/x64/xbyak_util.h" +#include "video_core/pica_state.h" +#include "video_core/pica_types.h" +#include "video_core/shader/shader.h" +#include "video_core/shader/shader_jit_x64_compiler.h" + +using namespace Common::X64; +using namespace Xbyak::util; +using Xbyak::Label; +using Xbyak::Reg32; +using Xbyak::Reg64; +using Xbyak::Xmm; + +namespace Pica { + +namespace Shader { + +typedef void (JitShader::*JitFunction)(Instruction instr); + +const JitFunction instr_table[64] = { + &JitShader::Compile_ADD, // add + &JitShader::Compile_DP3, // dp3 + &JitShader::Compile_DP4, // dp4 + &JitShader::Compile_DPH, // dph + nullptr, // unknown + &JitShader::Compile_EX2, // ex2 + &JitShader::Compile_LG2, // lg2 + nullptr, // unknown + &JitShader::Compile_MUL, // mul + &JitShader::Compile_SGE, // sge + &JitShader::Compile_SLT, // slt + &JitShader::Compile_FLR, // flr + &JitShader::Compile_MAX, // max + &JitShader::Compile_MIN, // min + &JitShader::Compile_RCP, // rcp + &JitShader::Compile_RSQ, // rsq + nullptr, // unknown + nullptr, // unknown + &JitShader::Compile_MOVA, // mova + &JitShader::Compile_MOV, // mov + nullptr, // unknown + nullptr, // unknown + nullptr, // unknown + nullptr, // unknown + &JitShader::Compile_DPH, // dphi + nullptr, // unknown + &JitShader::Compile_SGE, // sgei + &JitShader::Compile_SLT, // slti + nullptr, // unknown + nullptr, // unknown + nullptr, // unknown + nullptr, // unknown + nullptr, // unknown + &JitShader::Compile_NOP, // nop + &JitShader::Compile_END, // end + nullptr, // break + &JitShader::Compile_CALL, // call + &JitShader::Compile_CALLC, // callc + &JitShader::Compile_CALLU, // callu + &JitShader::Compile_IF, // ifu + &JitShader::Compile_IF, // ifc + &JitShader::Compile_LOOP, // loop + nullptr, // emit + nullptr, // sete + &JitShader::Compile_JMP, // jmpc + &JitShader::Compile_JMP, // jmpu + &JitShader::Compile_CMP, // cmp + &JitShader::Compile_CMP, // cmp + &JitShader::Compile_MAD, // madi + &JitShader::Compile_MAD, // madi + &JitShader::Compile_MAD, // madi + &JitShader::Compile_MAD, // madi + &JitShader::Compile_MAD, // madi + &JitShader::Compile_MAD, // madi + &JitShader::Compile_MAD, // madi + &JitShader::Compile_MAD, // madi + &JitShader::Compile_MAD, // mad + &JitShader::Compile_MAD, // mad + &JitShader::Compile_MAD, // mad + &JitShader::Compile_MAD, // mad + &JitShader::Compile_MAD, // mad + &JitShader::Compile_MAD, // mad + &JitShader::Compile_MAD, // mad + &JitShader::Compile_MAD, // mad +}; + +// The following is used to alias some commonly used registers. Generally, RAX-RDX and XMM0-XMM3 can +// be used as scratch registers within a compiler function. The other registers have designated +// purposes, as documented below: + +/// Pointer to the uniform memory +static const Reg64 SETUP = r9; +/// The two 32-bit VS address offset registers set by the MOVA instruction +static const Reg64 ADDROFFS_REG_0 = r10; +static const Reg64 ADDROFFS_REG_1 = r11; +/// VS loop count register (Multiplied by 16) +static const Reg32 LOOPCOUNT_REG = r12d; +/// Current VS loop iteration number (we could probably use LOOPCOUNT_REG, but this quicker) +static const Reg32 LOOPCOUNT = esi; +/// Number to increment LOOPCOUNT_REG by on each loop iteration (Multiplied by 16) +static const Reg32 LOOPINC = edi; +/// Result of the previous CMP instruction for the X-component comparison +static const Reg64 COND0 = r13; +/// Result of the previous CMP instruction for the Y-component comparison +static const Reg64 COND1 = r14; +/// Pointer to the UnitState instance for the current VS unit +static const Reg64 STATE = r15; +/// SIMD scratch register +static const Xmm SCRATCH = xmm0; +/// Loaded with the first swizzled source register, otherwise can be used as a scratch register +static const Xmm SRC1 = xmm1; +/// Loaded with the second swizzled source register, otherwise can be used as a scratch register +static const Xmm SRC2 = xmm2; +/// Loaded with the third swizzled source register, otherwise can be used as a scratch register +static const Xmm SRC3 = xmm3; +/// Additional scratch register +static const Xmm SCRATCH2 = xmm4; +/// Constant vector of [1.0f, 1.0f, 1.0f, 1.0f], used to efficiently set a vector to one +static const Xmm ONE = xmm14; +/// Constant vector of [-0.f, -0.f, -0.f, -0.f], used to efficiently negate a vector with XOR +static const Xmm NEGBIT = xmm15; + +// State registers that must not be modified by external functions calls +// Scratch registers, e.g., SRC1 and SCRATCH, have to be saved on the side if needed +static const BitSet32 persistent_regs = BuildRegSet({ + // Pointers to register blocks + SETUP, STATE, + // Cached registers + ADDROFFS_REG_0, ADDROFFS_REG_1, LOOPCOUNT_REG, COND0, COND1, + // Constants + ONE, NEGBIT, +}); + +/// Raw constant for the source register selector that indicates no swizzling is performed +static const u8 NO_SRC_REG_SWIZZLE = 0x1b; +/// Raw constant for the destination register enable mask that indicates all components are enabled +static const u8 NO_DEST_REG_MASK = 0xf; + +static void LogCritical(const char* msg) { + LOG_CRITICAL(HW_GPU, "%s", msg); +} + +void JitShader::Compile_Assert(bool condition, const char* msg) { + if (!condition) { + mov(ABI_PARAM1, reinterpret_cast<size_t>(msg)); + CallFarFunction(*this, LogCritical); + } +} + +/** + * Loads and swizzles a source register into the specified XMM register. + * @param instr VS instruction, used for determining how to load the source register + * @param src_num Number indicating which source register to load (1 = src1, 2 = src2, 3 = src3) + * @param src_reg SourceRegister object corresponding to the source register to load + * @param dest Destination XMM register to store the loaded, swizzled source register + */ +void JitShader::Compile_SwizzleSrc(Instruction instr, unsigned src_num, SourceRegister src_reg, + Xmm dest) { + Reg64 src_ptr; + size_t src_offset; + + if (src_reg.GetRegisterType() == RegisterType::FloatUniform) { + src_ptr = SETUP; + src_offset = ShaderSetup::GetFloatUniformOffset(src_reg.GetIndex()); + } else { + src_ptr = STATE; + src_offset = UnitState::InputOffset(src_reg); + } + + int src_offset_disp = (int)src_offset; + ASSERT_MSG(src_offset == src_offset_disp, "Source register offset too large for int type"); + + unsigned operand_desc_id; + + const bool is_inverted = + (0 != (instr.opcode.Value().GetInfo().subtype & OpCode::Info::SrcInversed)); + + unsigned address_register_index; + unsigned offset_src; + + if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MAD || + instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MADI) { + operand_desc_id = instr.mad.operand_desc_id; + offset_src = is_inverted ? 3 : 2; + address_register_index = instr.mad.address_register_index; + } else { + operand_desc_id = instr.common.operand_desc_id; + offset_src = is_inverted ? 2 : 1; + address_register_index = instr.common.address_register_index; + } + + if (src_num == offset_src && address_register_index != 0) { + switch (address_register_index) { + case 1: // address offset 1 + movaps(dest, xword[src_ptr + ADDROFFS_REG_0 + src_offset_disp]); + break; + case 2: // address offset 2 + movaps(dest, xword[src_ptr + ADDROFFS_REG_1 + src_offset_disp]); + break; + case 3: // address offset 3 + movaps(dest, xword[src_ptr + LOOPCOUNT_REG.cvt64() + src_offset_disp]); + break; + default: + UNREACHABLE(); + break; + } + } else { + // Load the source + movaps(dest, xword[src_ptr + src_offset_disp]); + } + + SwizzlePattern swiz = {(*swizzle_data)[operand_desc_id]}; + + // Generate instructions for source register swizzling as needed + u8 sel = swiz.GetRawSelector(src_num); + if (sel != NO_SRC_REG_SWIZZLE) { + // Selector component order needs to be reversed for the SHUFPS instruction + sel = ((sel & 0xc0) >> 6) | ((sel & 3) << 6) | ((sel & 0xc) << 2) | ((sel & 0x30) >> 2); + + // Shuffle inputs for swizzle + shufps(dest, dest, sel); + } + + // If the source register should be negated, flip the negative bit using XOR + const bool negate[] = {swiz.negate_src1, swiz.negate_src2, swiz.negate_src3}; + if (negate[src_num - 1]) { + xorps(dest, NEGBIT); + } +} + +void JitShader::Compile_DestEnable(Instruction instr, Xmm src) { + DestRegister dest; + unsigned operand_desc_id; + if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MAD || + instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MADI) { + operand_desc_id = instr.mad.operand_desc_id; + dest = instr.mad.dest.Value(); + } else { + operand_desc_id = instr.common.operand_desc_id; + dest = instr.common.dest.Value(); + } + + SwizzlePattern swiz = {(*swizzle_data)[operand_desc_id]}; + + size_t dest_offset_disp = UnitState::OutputOffset(dest); + + // If all components are enabled, write the result to the destination register + if (swiz.dest_mask == NO_DEST_REG_MASK) { + // Store dest back to memory + movaps(xword[STATE + dest_offset_disp], src); + + } else { + // Not all components are enabled, so mask the result when storing to the destination + // register... + movaps(SCRATCH, xword[STATE + dest_offset_disp]); + + if (Common::GetCPUCaps().sse4_1) { + u8 mask = ((swiz.dest_mask & 1) << 3) | ((swiz.dest_mask & 8) >> 3) | + ((swiz.dest_mask & 2) << 1) | ((swiz.dest_mask & 4) >> 1); + blendps(SCRATCH, src, mask); + } else { + movaps(SCRATCH2, src); + unpckhps(SCRATCH2, SCRATCH); // Unpack X/Y components of source and destination + unpcklps(SCRATCH, src); // Unpack Z/W components of source and destination + + // Compute selector to selectively copy source components to destination for SHUFPS + // instruction + u8 sel = ((swiz.DestComponentEnabled(0) ? 1 : 0) << 0) | + ((swiz.DestComponentEnabled(1) ? 3 : 2) << 2) | + ((swiz.DestComponentEnabled(2) ? 0 : 1) << 4) | + ((swiz.DestComponentEnabled(3) ? 2 : 3) << 6); + shufps(SCRATCH, SCRATCH2, sel); + } + + // Store dest back to memory + movaps(xword[STATE + dest_offset_disp], SCRATCH); + } +} + +void JitShader::Compile_SanitizedMul(Xmm src1, Xmm src2, Xmm scratch) { + movaps(scratch, src1); + cmpordps(scratch, src2); + + mulps(src1, src2); + + movaps(src2, src1); + cmpunordps(src2, src2); + + xorps(scratch, src2); + andps(src1, scratch); +} + +void JitShader::Compile_EvaluateCondition(Instruction instr) { + // Note: NXOR is used below to check for equality + switch (instr.flow_control.op) { + case Instruction::FlowControlType::Or: + mov(eax, COND0); + mov(ebx, COND1); + xor(eax, (instr.flow_control.refx.Value() ^ 1)); + xor(ebx, (instr.flow_control.refy.Value() ^ 1)); + or (eax, ebx); + break; + + case Instruction::FlowControlType::And: + mov(eax, COND0); + mov(ebx, COND1); + xor(eax, (instr.flow_control.refx.Value() ^ 1)); + xor(ebx, (instr.flow_control.refy.Value() ^ 1)); + and(eax, ebx); + break; + + case Instruction::FlowControlType::JustX: + mov(eax, COND0); + xor(eax, (instr.flow_control.refx.Value() ^ 1)); + break; + + case Instruction::FlowControlType::JustY: + mov(eax, COND1); + xor(eax, (instr.flow_control.refy.Value() ^ 1)); + break; + } +} + +void JitShader::Compile_UniformCondition(Instruction instr) { + size_t offset = ShaderSetup::GetBoolUniformOffset(instr.flow_control.bool_uniform_id); + cmp(byte[SETUP + offset], 0); +} + +BitSet32 JitShader::PersistentCallerSavedRegs() { + return persistent_regs & ABI_ALL_CALLER_SAVED; +} + +void JitShader::Compile_ADD(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); + addps(SRC1, SRC2); + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_DP3(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); + + Compile_SanitizedMul(SRC1, SRC2, SCRATCH); + + movaps(SRC2, SRC1); + shufps(SRC2, SRC2, _MM_SHUFFLE(1, 1, 1, 1)); + + movaps(SRC3, SRC1); + shufps(SRC3, SRC3, _MM_SHUFFLE(2, 2, 2, 2)); + + shufps(SRC1, SRC1, _MM_SHUFFLE(0, 0, 0, 0)); + addps(SRC1, SRC2); + addps(SRC1, SRC3); + + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_DP4(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); + + Compile_SanitizedMul(SRC1, SRC2, SCRATCH); + + movaps(SRC2, SRC1); + shufps(SRC1, SRC1, _MM_SHUFFLE(2, 3, 0, 1)); // XYZW -> ZWXY + addps(SRC1, SRC2); + + movaps(SRC2, SRC1); + shufps(SRC1, SRC1, _MM_SHUFFLE(0, 1, 2, 3)); // XYZW -> WZYX + addps(SRC1, SRC2); + + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_DPH(Instruction instr) { + if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::DPHI) { + Compile_SwizzleSrc(instr, 1, instr.common.src1i, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2i, SRC2); + } else { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); + } + + if (Common::GetCPUCaps().sse4_1) { + // Set 4th component to 1.0 + blendps(SRC1, ONE, 0b1000); + } else { + // Set 4th component to 1.0 + movaps(SCRATCH, SRC1); + unpckhps(SCRATCH, ONE); // XYZW, 1111 -> Z1__ + unpcklpd(SRC1, SCRATCH); // XYZW, Z1__ -> XYZ1 + } + + Compile_SanitizedMul(SRC1, SRC2, SCRATCH); + + movaps(SRC2, SRC1); + shufps(SRC1, SRC1, _MM_SHUFFLE(2, 3, 0, 1)); // XYZW -> ZWXY + addps(SRC1, SRC2); + + movaps(SRC2, SRC1); + shufps(SRC1, SRC1, _MM_SHUFFLE(0, 1, 2, 3)); // XYZW -> WZYX + addps(SRC1, SRC2); + + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_EX2(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + movss(xmm0, SRC1); // ABI_PARAM1 + + ABI_PushRegistersAndAdjustStack(*this, PersistentCallerSavedRegs(), 0); + CallFarFunction(*this, exp2f); + ABI_PopRegistersAndAdjustStack(*this, PersistentCallerSavedRegs(), 0); + + shufps(xmm0, xmm0, _MM_SHUFFLE(0, 0, 0, 0)); // ABI_RETURN + movaps(SRC1, xmm0); + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_LG2(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + movss(xmm0, SRC1); // ABI_PARAM1 + + ABI_PushRegistersAndAdjustStack(*this, PersistentCallerSavedRegs(), 0); + CallFarFunction(*this, log2f); + ABI_PopRegistersAndAdjustStack(*this, PersistentCallerSavedRegs(), 0); + + shufps(xmm0, xmm0, _MM_SHUFFLE(0, 0, 0, 0)); // ABI_RETURN + movaps(SRC1, xmm0); + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_MUL(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); + Compile_SanitizedMul(SRC1, SRC2, SCRATCH); + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_SGE(Instruction instr) { + if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::SGEI) { + Compile_SwizzleSrc(instr, 1, instr.common.src1i, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2i, SRC2); + } else { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); + } + + cmpleps(SRC2, SRC1); + andps(SRC2, ONE); + + Compile_DestEnable(instr, SRC2); +} + +void JitShader::Compile_SLT(Instruction instr) { + if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::SLTI) { + Compile_SwizzleSrc(instr, 1, instr.common.src1i, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2i, SRC2); + } else { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); + } + + cmpltps(SRC1, SRC2); + andps(SRC1, ONE); + + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_FLR(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + + if (Common::GetCPUCaps().sse4_1) { + roundps(SRC1, SRC1, _MM_FROUND_FLOOR); + } else { + cvttps2dq(SRC1, SRC1); + cvtdq2ps(SRC1, SRC1); + } + + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_MAX(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); + // SSE semantics match PICA200 ones: In case of NaN, SRC2 is returned. + maxps(SRC1, SRC2); + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_MIN(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); + // SSE semantics match PICA200 ones: In case of NaN, SRC2 is returned. + minps(SRC1, SRC2); + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_MOVA(Instruction instr) { + SwizzlePattern swiz = {(*swizzle_data)[instr.common.operand_desc_id]}; + + if (!swiz.DestComponentEnabled(0) && !swiz.DestComponentEnabled(1)) { + return; // NoOp + } + + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + + // Convert floats to integers using truncation (only care about X and Y components) + cvttps2dq(SRC1, SRC1); + + // Get result + movq(rax, SRC1); + + // Handle destination enable + if (swiz.DestComponentEnabled(0) && swiz.DestComponentEnabled(1)) { + // Move and sign-extend low 32 bits + movsxd(ADDROFFS_REG_0, eax); + + // Move and sign-extend high 32 bits + shr(rax, 32); + movsxd(ADDROFFS_REG_1, eax); + + // Multiply by 16 to be used as an offset later + shl(ADDROFFS_REG_0, 4); + shl(ADDROFFS_REG_1, 4); + } else { + if (swiz.DestComponentEnabled(0)) { + // Move and sign-extend low 32 bits + movsxd(ADDROFFS_REG_0, eax); + + // Multiply by 16 to be used as an offset later + shl(ADDROFFS_REG_0, 4); + } else if (swiz.DestComponentEnabled(1)) { + // Move and sign-extend high 32 bits + shr(rax, 32); + movsxd(ADDROFFS_REG_1, eax); + + // Multiply by 16 to be used as an offset later + shl(ADDROFFS_REG_1, 4); + } + } +} + +void JitShader::Compile_MOV(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_RCP(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + + // TODO(bunnei): RCPSS is a pretty rough approximation, this might cause problems if Pica + // performs this operation more accurately. This should be checked on hardware. + rcpss(SRC1, SRC1); + shufps(SRC1, SRC1, _MM_SHUFFLE(0, 0, 0, 0)); // XYWZ -> XXXX + + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_RSQ(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + + // TODO(bunnei): RSQRTSS is a pretty rough approximation, this might cause problems if Pica + // performs this operation more accurately. This should be checked on hardware. + rsqrtss(SRC1, SRC1); + shufps(SRC1, SRC1, _MM_SHUFFLE(0, 0, 0, 0)); // XYWZ -> XXXX + + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_NOP(Instruction instr) {} + +void JitShader::Compile_END(Instruction instr) { + ABI_PopRegistersAndAdjustStack(*this, ABI_ALL_CALLEE_SAVED, 8); + ret(); +} + +void JitShader::Compile_CALL(Instruction instr) { + // Push offset of the return + push(qword, (instr.flow_control.dest_offset + instr.flow_control.num_instructions)); + + // Call the subroutine + call(instruction_labels[instr.flow_control.dest_offset]); + + // Skip over the return offset that's on the stack + add(rsp, 8); +} + +void JitShader::Compile_CALLC(Instruction instr) { + Compile_EvaluateCondition(instr); + Label b; + jz(b); + Compile_CALL(instr); + L(b); +} + +void JitShader::Compile_CALLU(Instruction instr) { + Compile_UniformCondition(instr); + Label b; + jz(b); + Compile_CALL(instr); + L(b); +} + +void JitShader::Compile_CMP(Instruction instr) { + using Op = Instruction::Common::CompareOpType::Op; + Op op_x = instr.common.compare_op.x; + Op op_y = instr.common.compare_op.y; + + Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1); + Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2); + + // SSE doesn't have greater-than (GT) or greater-equal (GE) comparison operators. You need to + // emulate them by swapping the lhs and rhs and using LT and LE. NLT and NLE can't be used here + // because they don't match when used with NaNs. + static const u8 cmp[] = {CMP_EQ, CMP_NEQ, CMP_LT, CMP_LE, CMP_LT, CMP_LE}; + + bool invert_op_x = (op_x == Op::GreaterThan || op_x == Op::GreaterEqual); + Xmm lhs_x = invert_op_x ? SRC2 : SRC1; + Xmm rhs_x = invert_op_x ? SRC1 : SRC2; + + if (op_x == op_y) { + // Compare X-component and Y-component together + cmpps(lhs_x, rhs_x, cmp[op_x]); + movq(COND0, lhs_x); + + mov(COND1, COND0); + } else { + bool invert_op_y = (op_y == Op::GreaterThan || op_y == Op::GreaterEqual); + Xmm lhs_y = invert_op_y ? SRC2 : SRC1; + Xmm rhs_y = invert_op_y ? SRC1 : SRC2; + + // Compare X-component + movaps(SCRATCH, lhs_x); + cmpss(SCRATCH, rhs_x, cmp[op_x]); + + // Compare Y-component + cmpps(lhs_y, rhs_y, cmp[op_y]); + + movq(COND0, SCRATCH); + movq(COND1, lhs_y); + } + + shr(COND0.cvt32(), 31); // ignores upper 32 bits in source + shr(COND1, 63); +} + +void JitShader::Compile_MAD(Instruction instr) { + Compile_SwizzleSrc(instr, 1, instr.mad.src1, SRC1); + + if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::MADI) { + Compile_SwizzleSrc(instr, 2, instr.mad.src2i, SRC2); + Compile_SwizzleSrc(instr, 3, instr.mad.src3i, SRC3); + } else { + Compile_SwizzleSrc(instr, 2, instr.mad.src2, SRC2); + Compile_SwizzleSrc(instr, 3, instr.mad.src3, SRC3); + } + + Compile_SanitizedMul(SRC1, SRC2, SCRATCH); + addps(SRC1, SRC3); + + Compile_DestEnable(instr, SRC1); +} + +void JitShader::Compile_IF(Instruction instr) { + Compile_Assert(instr.flow_control.dest_offset >= program_counter, + "Backwards if-statements not supported"); + Label l_else, l_endif; + + // Evaluate the "IF" condition + if (instr.opcode.Value() == OpCode::Id::IFU) { + Compile_UniformCondition(instr); + } else if (instr.opcode.Value() == OpCode::Id::IFC) { + Compile_EvaluateCondition(instr); + } + jz(l_else, T_NEAR); + + // Compile the code that corresponds to the condition evaluating as true + Compile_Block(instr.flow_control.dest_offset); + + // If there isn't an "ELSE" condition, we are done here + if (instr.flow_control.num_instructions == 0) { + L(l_else); + return; + } + + jmp(l_endif, T_NEAR); + + L(l_else); + // This code corresponds to the "ELSE" condition + // Comple the code that corresponds to the condition evaluating as false + Compile_Block(instr.flow_control.dest_offset + instr.flow_control.num_instructions); + + L(l_endif); +} + +void JitShader::Compile_LOOP(Instruction instr) { + Compile_Assert(instr.flow_control.dest_offset >= program_counter, + "Backwards loops not supported"); + Compile_Assert(!looping, "Nested loops not supported"); + + looping = true; + + // This decodes the fields from the integer uniform at index instr.flow_control.int_uniform_id. + // The Y (LOOPCOUNT_REG) and Z (LOOPINC) component are kept multiplied by 16 (Left shifted by + // 4 bits) to be used as an offset into the 16-byte vector registers later + size_t offset = ShaderSetup::GetIntUniformOffset(instr.flow_control.int_uniform_id); + mov(LOOPCOUNT, dword[SETUP + offset]); + mov(LOOPCOUNT_REG, LOOPCOUNT); + shr(LOOPCOUNT_REG, 4); + and(LOOPCOUNT_REG, 0xFF0); // Y-component is the start + mov(LOOPINC, LOOPCOUNT); + shr(LOOPINC, 12); + and(LOOPINC, 0xFF0); // Z-component is the incrementer + movzx(LOOPCOUNT, LOOPCOUNT.cvt8()); // X-component is iteration count + add(LOOPCOUNT, 1); // Iteration count is X-component + 1 + + Label l_loop_start; + L(l_loop_start); + + Compile_Block(instr.flow_control.dest_offset + 1); + + add(LOOPCOUNT_REG, LOOPINC); // Increment LOOPCOUNT_REG by Z-component + sub(LOOPCOUNT, 1); // Increment loop count by 1 + jnz(l_loop_start); // Loop if not equal + + looping = false; +} + +void JitShader::Compile_JMP(Instruction instr) { + if (instr.opcode.Value() == OpCode::Id::JMPC) + Compile_EvaluateCondition(instr); + else if (instr.opcode.Value() == OpCode::Id::JMPU) + Compile_UniformCondition(instr); + else + UNREACHABLE(); + + bool inverted_condition = + (instr.opcode.Value() == OpCode::Id::JMPU) && (instr.flow_control.num_instructions & 1); + + Label& b = instruction_labels[instr.flow_control.dest_offset]; + if (inverted_condition) { + jz(b, T_NEAR); + } else { + jnz(b, T_NEAR); + } +} + +void JitShader::Compile_Block(unsigned end) { + while (program_counter < end) { + Compile_NextInstr(); + } +} + +void JitShader::Compile_Return() { + // Peek return offset on the stack and check if we're at that offset + mov(rax, qword[rsp + 8]); + cmp(eax, (program_counter)); + + // If so, jump back to before CALL + Label b; + jnz(b); + ret(); + L(b); +} + +void JitShader::Compile_NextInstr() { + if (std::binary_search(return_offsets.begin(), return_offsets.end(), program_counter)) { + Compile_Return(); + } + + L(instruction_labels[program_counter]); + + Instruction instr = {(*program_code)[program_counter++]}; + + OpCode::Id opcode = instr.opcode.Value(); + auto instr_func = instr_table[static_cast<unsigned>(opcode)]; + + if (instr_func) { + // JIT the instruction! + ((*this).*instr_func)(instr); + } else { + // Unhandled instruction + LOG_CRITICAL(HW_GPU, "Unhandled instruction: 0x%02x (0x%08x)", + instr.opcode.Value().EffectiveOpCode(), instr.hex); + } +} + +void JitShader::FindReturnOffsets() { + return_offsets.clear(); + + for (size_t offset = 0; offset < program_code->size(); ++offset) { + Instruction instr = {(*program_code)[offset]}; + + switch (instr.opcode.Value()) { + case OpCode::Id::CALL: + case OpCode::Id::CALLC: + case OpCode::Id::CALLU: + return_offsets.push_back(instr.flow_control.dest_offset + + instr.flow_control.num_instructions); + break; + default: + break; + } + } + + // Sort for efficient binary search later + std::sort(return_offsets.begin(), return_offsets.end()); +} + +void JitShader::Compile(const std::array<u32, 1024>* program_code_, + const std::array<u32, 1024>* swizzle_data_) { + program_code = program_code_; + swizzle_data = swizzle_data_; + + // Reset flow control state + program = (CompiledShader*)getCurr(); + program_counter = 0; + looping = false; + instruction_labels.fill(Xbyak::Label()); + + // Find all `CALL` instructions and identify return locations + FindReturnOffsets(); + + // The stack pointer is 8 modulo 16 at the entry of a procedure + ABI_PushRegistersAndAdjustStack(*this, ABI_ALL_CALLEE_SAVED, 8); + + mov(SETUP, ABI_PARAM1); + mov(STATE, ABI_PARAM2); + + // Zero address/loop registers + xor(ADDROFFS_REG_0.cvt32(), ADDROFFS_REG_0.cvt32()); + xor(ADDROFFS_REG_1.cvt32(), ADDROFFS_REG_1.cvt32()); + xor(LOOPCOUNT_REG, LOOPCOUNT_REG); + + // Used to set a register to one + static const __m128 one = {1.f, 1.f, 1.f, 1.f}; + mov(rax, reinterpret_cast<size_t>(&one)); + movaps(ONE, xword[rax]); + + // Used to negate registers + static const __m128 neg = {-0.f, -0.f, -0.f, -0.f}; + mov(rax, reinterpret_cast<size_t>(&neg)); + movaps(NEGBIT, xword[rax]); + + // Jump to start of the shader program + jmp(ABI_PARAM3); + + // Compile entire program + Compile_Block(static_cast<unsigned>(program_code->size())); + + // Free memory that's no longer needed + program_code = nullptr; + swizzle_data = nullptr; + return_offsets.clear(); + return_offsets.shrink_to_fit(); + + ready(); + + ASSERT_MSG(getSize() <= MAX_SHADER_SIZE, "Compiled a shader that exceeds the allocated size!"); + LOG_DEBUG(HW_GPU, "Compiled shader size=%lu", getSize()); +} + +JitShader::JitShader() : Xbyak::CodeGenerator(MAX_SHADER_SIZE) {} + +} // namespace Shader + +} // namespace Pica diff --git a/src/video_core/shader/shader_jit_x64_compiler.h b/src/video_core/shader/shader_jit_x64_compiler.h new file mode 100644 index 000000000..29e9875ea --- /dev/null +++ b/src/video_core/shader/shader_jit_x64_compiler.h @@ -0,0 +1,125 @@ +// Copyright 2015 Citra Emulator Project +// Licensed under GPLv2 or any later version +// Refer to the license.txt file included. + +#pragma once + +#include <array> +#include <cstddef> +#include <utility> +#include <vector> +#include <nihstro/shader_bytecode.h> +#include <xbyak.h> +#include "common/bit_set.h" +#include "common/common_types.h" +#include "common/x64/emitter.h" +#include "video_core/shader/shader.h" + +using nihstro::Instruction; +using nihstro::OpCode; +using nihstro::SwizzlePattern; + +namespace Pica { + +namespace Shader { + +/// Memory allocated for each compiled shader (64Kb) +constexpr size_t MAX_SHADER_SIZE = 1024 * 64; + +/** + * This class implements the shader JIT compiler. It recompiles a Pica shader program into x86_64 + * code that can be executed on the host machine directly. + */ +class JitShader : public Xbyak::CodeGenerator { +public: + JitShader(); + + void Run(const ShaderSetup& setup, UnitState& state, unsigned offset) const { + program(&setup, &state, instruction_labels[offset].getAddress()); + } + + void Compile(const std::array<u32, 1024>* program_code, + const std::array<u32, 1024>* swizzle_data); + + void Compile_ADD(Instruction instr); + void Compile_DP3(Instruction instr); + void Compile_DP4(Instruction instr); + void Compile_DPH(Instruction instr); + void Compile_EX2(Instruction instr); + void Compile_LG2(Instruction instr); + void Compile_MUL(Instruction instr); + void Compile_SGE(Instruction instr); + void Compile_SLT(Instruction instr); + void Compile_FLR(Instruction instr); + void Compile_MAX(Instruction instr); + void Compile_MIN(Instruction instr); + void Compile_RCP(Instruction instr); + void Compile_RSQ(Instruction instr); + void Compile_MOVA(Instruction instr); + void Compile_MOV(Instruction instr); + void Compile_NOP(Instruction instr); + void Compile_END(Instruction instr); + void Compile_CALL(Instruction instr); + void Compile_CALLC(Instruction instr); + void Compile_CALLU(Instruction instr); + void Compile_IF(Instruction instr); + void Compile_LOOP(Instruction instr); + void Compile_JMP(Instruction instr); + void Compile_CMP(Instruction instr); + void Compile_MAD(Instruction instr); + +private: + void Compile_Block(unsigned end); + void Compile_NextInstr(); + + void Compile_SwizzleSrc(Instruction instr, unsigned src_num, SourceRegister src_reg, + Xbyak::Xmm dest); + void Compile_DestEnable(Instruction instr, Xbyak::Xmm dest); + + /** + * Compiles a `MUL src1, src2` operation, properly handling the PICA semantics when multiplying + * zero by inf. Clobbers `src2` and `scratch`. + */ + void Compile_SanitizedMul(Xbyak::Xmm src1, Xbyak::Xmm src2, Xbyak::Xmm scratch); + + void Compile_EvaluateCondition(Instruction instr); + void Compile_UniformCondition(Instruction instr); + + /** + * Emits the code to conditionally return from a subroutine envoked by the `CALL` instruction. + */ + void Compile_Return(); + + BitSet32 PersistentCallerSavedRegs(); + + /** + * Assertion evaluated at compile-time, but only triggered if executed at runtime. + * @param msg Message to be logged if the assertion fails. + */ + void Compile_Assert(bool condition, const char* msg); + + /** + * Analyzes the entire shader program for `CALL` instructions before emitting any code, + * identifying the locations where a return needs to be inserted. + */ + void FindReturnOffsets(); + + const std::array<u32, 1024>* program_code = nullptr; + const std::array<u32, 1024>* swizzle_data = nullptr; + + /// Mapping of Pica VS instructions to pointers in the emitted code + std::array<Xbyak::Label, 1024> instruction_labels; + + /// Offsets in code where a return needs to be inserted + std::vector<unsigned> return_offsets; + + unsigned program_counter = 0; ///< Offset of the next instruction to decode + bool looping = false; ///< True if compiling a loop, used to check for nested loops + + using CompiledShader = void(const void* setup, void* state, const u8* start_addr); + CompiledShader* program = nullptr; +}; + +} // Shader + +} // Pica diff --git a/src/video_core/video_core.cpp b/src/video_core/video_core.cpp index 8db882f59..7186a7652 100644 --- a/src/video_core/video_core.cpp +++ b/src/video_core/video_core.cpp @@ -19,7 +19,6 @@ std::unique_ptr<RendererBase> g_renderer; ///< Renderer plugin std::atomic<bool> g_hw_renderer_enabled; std::atomic<bool> g_shader_jit_enabled; -std::atomic<bool> g_scaled_resolution_enabled; std::atomic<bool> g_vsync_enabled; std::atomic<bool> g_toggle_framelimit_enabled; diff --git a/src/video_core/video_core.h b/src/video_core/video_core.h index c397c1974..4aba19ca0 100644 --- a/src/video_core/video_core.h +++ b/src/video_core/video_core.h @@ -37,7 +37,6 @@ extern EmuWindow* g_emu_window; ///< Emu window // qt ui) extern std::atomic<bool> g_hw_renderer_enabled; extern std::atomic<bool> g_shader_jit_enabled; -extern std::atomic<bool> g_scaled_resolution_enabled; extern std::atomic<bool> g_toggle_framelimit_enabled; /// Start the video core |