Revert "system_archive: Fix Korean and Chinese fonts"

System: Refactor CPU Core management and move ARMInterface and Schedulers to Kernel

CMakeLists.txt

@@ -350,6 +350,13 @@ function(create_target_directory_groups target_name)
     endforeach()
 endfunction()
 
+# Prevent boost from linking against libs when building
+add_definitions(-DBOOST_ERROR_CODE_HEADER_ONLY
+    -DBOOST_SYSTEM_NO_LIB
+    -DBOOST_DATE_TIME_NO_LIB
+    -DBOOST_REGEX_NO_LIB
+)
+
 enable_testing()
 add_subdirectory(externals)
 add_subdirectory(src)

CMakeModules/GenerateSCMRev.cmake

@@ -5,6 +5,10 @@ function(get_timestamp _var)
 endfunction()
 
 list(APPEND CMAKE_MODULE_PATH "${SRC_DIR}/externals/cmake-modules")
+
+# Find the package here with the known path so that the GetGit commands can find it as well
+find_package(Git QUIET PATHS "${GIT_EXECUTABLE}")
+
 # generate git/build information
 include(GetGitRevisionDescription)
 get_git_head_revision(GIT_REF_SPEC GIT_REV)

dist/qt_themes/qdarkstyle/style.qss

@@ -2,7 +2,8 @@ QToolTip {
     border: 1px solid #76797C;
     background-color: #5A7566;
     color: white;
-    padding: 0px;                /*remove padding, for fix combobox tooltip.*/
+    /*remove padding, for fix combobox tooltip.*/
+    padding: 0;
     opacity: 200;
 }
 
@@ -13,7 +14,7 @@ QWidget {
     selection-color: #eff0f1;
     background-clip: border;
     border-image: none;
-    border: 0px transparent black;
+    border: 0;
     outline: 0;
 }
 
@@ -27,10 +28,10 @@ QWidget:item:selected {
 }
 
 QCheckBox {
-    spacing: 5px;
+    spacing: 6px;
     outline: none;
     color: #eff0f1;
-    margin-bottom: 2px;
+    margin: 0 2px 1px 0;
 }
 
 QCheckBox:disabled {
@@ -163,7 +164,7 @@ QMenuBar::item:selected {
 }
 
 QMenuBar::item:pressed {
-    border: 1px solid #76797C;
+    border: 1px solid #18465d;
     background-color: #3daee9;
     color: #eff0f1;
     margin-bottom: -1px;
@@ -171,9 +172,9 @@ QMenuBar::item:pressed {
 }
 
 QMenu {
-    border: 1px solid #76797C;
+    border: 1px solid #434242;
+    padding: 2px;
     color: #eff0f1;
-    margin: 2px;
 }
 
 QMenu::icon {
@@ -181,7 +182,7 @@ QMenu::icon {
 }
 
 QMenu::item {
-    padding: 5px 30px 5px 30px;
+    padding: 5px 16px 5px 40px;
     border: 1px solid transparent;
     /* reserve space for selection border */
 }
@@ -190,22 +191,30 @@ QMenu::item:selected {
     color: #eff0f1;
 }
 
-QMenu::separator {
-    height: 2px;
-    background: lightblue;
-    margin-left: 10px;
-    margin-right: 5px;
+QMenu::item:disabled {
+    color: #54575B;
+}
+
+QMenu::item:disabled:hover,
+QMenu::item:disabled:selected {
+    background-color: #393e43;
+    color: #666;
+}
+
+QMenu::separator,
+QMenuBar::separator {
+    height: 1px;
+    background-color: #54575B;
+    margin: 2px 4px 2px 40px;
 }
 
 QMenu::indicator {
+    margin: 0 -26px 0 8px;
     width: 18px;
     height: 18px;
 }
 
-
-/* non-exclusive indicator = check box style indicator
-   (see QActionGroup::setExclusive) */
-
+/* non-exclusive indicator = check box style indicator (see QActionGroup::setExclusive) */
 QMenu::indicator:non-exclusive:unchecked {
     image: url(:/qss_icons/rc/checkbox_unchecked.png);
 }
@@ -222,9 +231,7 @@ QMenu::indicator:non-exclusive:checked:selected {
     image: url(:/qss_icons/rc/checkbox_checked_disabled.png);
 }
 
-
 /* exclusive indicator = radio button style indicator (see QActionGroup::setExclusive) */
-
 QMenu::indicator:exclusive:unchecked {
     image: url(:/qss_icons/rc/radio_unchecked.png);
 }
@@ -242,39 +249,46 @@ QMenu::indicator:exclusive:checked:selected {
 }
 
 QMenu::right-arrow {
-    margin: 5px;
+    margin-right: 10px;
     image: url(:/qss_icons/rc/right_arrow.png)
 }
 
 QWidget:disabled {
-    color: #454545;
+    color: #4f515b;
     background-color: #31363b;
 }
 
 QAbstractItemView {
-    alternate-background-color: #31363b;
+    alternate-background-color: #2c2f32;
     color: #eff0f1;
     border: 1px solid #3A3939;
     border-radius: 2px;
 }
 
-QWidget:focus,
-QMenuBar:focus {
+QAbstractItemView:disabled,
+QAbstractItemView:read-only {
+    alternate-background-color: #232629;
+}
+
+QWidget:focus {
     border: 1px solid #3daee9;
 }
 
 QTabWidget:focus,
 QCheckBox:focus,
 QRadioButton:focus,
-QSlider:focus {
+QSlider:focus,
+QTreeView:focus,
+QMenu:focus,
+QMenuBar:focus,
+QTabBar:focus {
     border: none;
 }
 
 QLineEdit {
     background-color: #232629;
     padding: 5px;
-    border-style: solid;
-    border: 1px solid #76797C;
+    border: 1px solid #54575B;
     border-radius: 2px;
     color: #eff0f1;
 }
@@ -284,9 +298,10 @@ QAbstractItemView QLineEdit {
 }
 
 QGroupBox {
-    border: 1px solid #76797C;
+    border: 1px solid #54575B;
     border-radius: 2px;
-    margin-top: 20px;
+    margin-top: 12px;
+    padding-top: 2px;
 }
 
 QGroupBox::title {
@@ -294,12 +309,12 @@ QGroupBox::title {
     subcontrol-position: top center;
     padding-left: 10px;
     padding-right: 10px;
-    padding-top: 10px;
+    padding-top: 2px;
 }
 
 QAbstractScrollArea {
     border-radius: 2px;
-    border: 1px solid #76797C;
+    border: 1px solid #54575B;
     background-color: transparent;
 }
 
@@ -318,7 +333,7 @@ QScrollBar::handle:horizontal {
 }
 
 QScrollBar::add-line:horizontal {
-    margin: 0px 3px 0px 3px;
+    margin: 0 3px;
     border-image: url(:/qss_icons/rc/right_arrow_disabled.png);
     width: 10px;
     height: 10px;
@@ -327,7 +342,7 @@ QScrollBar::add-line:horizontal {
 }
 
 QScrollBar::sub-line:horizontal {
-    margin: 0px 3px 0px 3px;
+    margin: 0 3px;
     border-image: url(:/qss_icons/rc/left_arrow_disabled.png);
     height: 10px;
     width: 10px;
@@ -378,7 +393,7 @@ QScrollBar::handle:vertical {
 }
 
 QScrollBar::sub-line:vertical {
-    margin: 3px 0px 3px 0px;
+    margin: 3px 0;
     border-image: url(:/qss_icons/rc/up_arrow_disabled.png);
     height: 10px;
     width: 10px;
@@ -387,7 +402,7 @@ QScrollBar::sub-line:vertical {
 }
 
 QScrollBar::add-line:vertical {
-    margin: 3px 0px 3px 0px;
+    margin: 3px 0;
     border-image: url(:/qss_icons/rc/down_arrow_disabled.png);
     height: 10px;
     width: 10px;
@@ -426,15 +441,14 @@ QScrollBar::sub-page:vertical {
 QTextEdit {
     background-color: #232629;
     color: #eff0f1;
-    border: 1px solid #76797C;
+    border: 1px solid #54575B;
 }
 
 QPlainTextEdit {
     background-color: #232629;
-    ;
     color: #eff0f1;
     border-radius: 2px;
-    border: 1px solid #76797C;
+    border: 1px solid #54575B;
 }
 
 QHeaderView::section {
@@ -466,15 +480,6 @@ QMainWindow::separator:hover {
     spacing: 2px;
 }
 
-QMenu::separator {
-    height: 1px;
-    background-color: #76797C;
-    color: white;
-    padding-left: 4px;
-    margin-left: 10px;
-    margin-right: 5px;
-}
-
 QFrame {
     border-radius: 2px;
     border: 1px solid #76797C;
@@ -517,25 +522,19 @@ QToolButton#qt_toolbar_ext_button {
 
 QPushButton {
     color: #eff0f1;
-    background-color: #31363b;
     border-width: 1px;
-    border-color: #76797C;
+    border-color: #54575B;
     border-style: solid;
-    padding: 5px;
+    padding: 6px 4px;
     border-radius: 2px;
     outline: none;
+    min-width: 100px;
+    background-color: #232629;
 }
 
 QPushButton:disabled {
     background-color: #31363b;
-    border-width: 1px;
     border-color: #454545;
-    border-style: solid;
-    padding-top: 5px;
-    padding-bottom: 5px;
-    padding-left: 10px;
-    padding-right: 10px;
-    border-radius: 2px;
     color: #454545;
 }
 
@@ -552,11 +551,11 @@ QPushButton:pressed {
 
 QComboBox {
     selection-background-color: #3daee9;
-    border-style: solid;
-    border: 1px solid #76797C;
+    border: 1px solid #54575B;
     border-radius: 2px;
-    padding: 5px;
+    padding: 4px 6px;
     min-width: 75px;
+    background-color: #232629;
 }
 
 QPushButton:checked {
@@ -570,15 +569,12 @@ QAbstractSpinBox:hover,
 QLineEdit:hover,
 QTextEdit:hover,
 QPlainTextEdit:hover,
-QAbstractView:hover,
-QTreeView:hover {
+QAbstractView:hover {
     border: 1px solid #3daee9;
     color: #eff0f1;
 }
 
 QComboBox:on {
-    padding-top: 3px;
-    padding-left: 4px;
     selection-background-color: #4a4a4a;
 }
 
@@ -592,6 +588,7 @@ QComboBox QAbstractItemView {
 QComboBox::drop-down {
     subcontrol-origin: padding;
     subcontrol-position: top right;
+    left: -6px;
     width: 15px;
     border-left-width: 0px;
     border-left-color: darkgray;
@@ -611,8 +608,8 @@ QComboBox::down-arrow:focus {
 }
 
 QAbstractSpinBox {
-    padding: 5px;
-    border: 1px solid #76797C;
+    padding: 4px 6px;
+    border: 1px solid #54575B;
     background-color: #232629;
     color: #eff0f1;
     border-radius: 2px;
@@ -623,12 +620,14 @@ QAbstractSpinBox:up-button {
     background-color: transparent;
     subcontrol-origin: border;
     subcontrol-position: center right;
+    left: -6px;
 }
 
 QAbstractSpinBox:down-button {
     background-color: transparent;
     subcontrol-origin: border;
     subcontrol-position: center left;
+    right: -6px;
 }
 
 QAbstractSpinBox::up-arrow,
@@ -655,22 +654,27 @@ QAbstractSpinBox::down-arrow:hover {
     image: url(:/qss_icons/rc/down_arrow.png);
 }
 
-QLabel {
-    border: 0px solid black;
+QLabel,
+QTabWidget {
+    border: 0;
 }
 
 QTabWidget {
-    border: 0px transparent black;
+    padding-top: 1px;
 }
 
 QTabWidget::pane {
     border: 1px solid #76797C;
     padding: 5px;
-    margin: 0px;
+    position: absolute;
+    top: -1px;
+    border-top-right-radius: 2px;
+    border-bottom-right-radius: 2px;
+    border-bottom-left-radius: 2px;
 }
 
 QTabWidget::tab-bar {
-    /* left: 5px; move to the right by 5px */
+    overflow: visible;
 }
 
 QTabBar {
@@ -678,10 +682,6 @@ QTabBar {
     border-radius: 3px;
 }
 
-QTabBar:focus {
-    border: 0px transparent black;
-}
-
 QTabBar::close-button {
     image: url(:/qss_icons/rc/close.png);
     background: transparent;
@@ -697,36 +697,33 @@ QTabBar::close-button:pressed {
     background: transparent;
 }
 
-
 /* TOP TABS */
-
 QTabBar::tab:top {
     color: #eff0f1;
-    border: 1px solid #76797C;
-    border-bottom: 1px transparent black;
-    background-color: #31363b;
-    padding: 5px;
-    min-width: 50px;
+    border: 1px solid #54575B;
+    background-color: #2a2f33;
+    padding: 4px 16px 5px;
+    min-width: 36px;
     border-top-left-radius: 2px;
     border-top-right-radius: 2px;
 }
 
 QTabBar::tab:top:selected {
-    color: #eff0f1;
-    background-color: #54575B;
-    border: 1px solid #76797C;
-    border-bottom: 2px solid #3daee9;
-    border-top-left-radius: 2px;
-    border-top-right-radius: 2px;
+    border-color: #76797C;
+    background-color: #31363b;
+    border-bottom-color: #31363b;
+}
+
+QTabBar::tab:top:!selected {
+    margin-top: 1px;
+    border-bottom-color: #76797C;
 }
 
 QTabBar::tab:top:!selected:hover {
     background-color: #3daee9;
 }
 
-
 /* BOTTOM TABS */
-
 QTabBar::tab:bottom {
     color: #eff0f1;
     border: 1px solid #76797C;
@@ -751,9 +748,7 @@ QTabBar::tab:bottom:!selected:hover {
     background-color: #3daee9;
 }
 
-
 /* LEFT TABS */
-
 QTabBar::tab:left {
     color: #eff0f1;
     border: 1px solid #76797C;
@@ -778,9 +773,7 @@ QTabBar::tab:left:!selected:hover {
     background-color: #3daee9;
 }
 
-
 /* RIGHT TABS */
-
 QTabBar::tab:right {
     color: #eff0f1;
     border: 1px solid #76797C;
@@ -848,7 +841,7 @@ QDockWidget::float-button:pressed {
 
 QTreeView,
 QListView {
-    border: 1px solid #76797C;
+    border: 1px solid #54575B;
     background-color: #232629;
 }
 
@@ -979,8 +972,8 @@ QSlider::handle:vertical {
 }
 
 QToolButton {
-    background-color: transparent;
-    border: 1px transparent #76797C;
+    background-color: #232629;
+    border: 1px solid #54575B;
     border-radius: 2px;
     margin: 3px;
     padding: 5px;
@@ -989,7 +982,6 @@ QToolButton {
 QToolButton[popupMode="1"] {
     /* only for MenuButtonPopup */
     padding-right: 20px;
-    /* make way for the popup button */
     border: 1px #76797C;
     border-radius: 5px;
 }
@@ -997,7 +989,6 @@ QToolButton[popupMode="1"] {
 QToolButton[popupMode="2"] {
     /* only for InstantPopup */
     padding-right: 10px;
-    /* make way for the popup button */
     border: 1px #76797C;
 }
 
@@ -1016,19 +1007,14 @@ QToolButton::menu-button:pressed {
     padding: 5px;
 }
 
-
 /* the subcontrol below is used only in the InstantPopup or DelayedPopup mode */
-
 QToolButton::menu-indicator {
     image: url(:/qss_icons/rc/down_arrow.png);
     top: -7px;
     left: -2px;
-    /* shift it a bit */
 }
 
-
 /* the subcontrols below are used only in the MenuButtonPopup mode */
-
 QToolButton::menu-button {
     border: 1px transparent #76797C;
     border-top-right-radius: 6px;
@@ -1053,14 +1039,22 @@ QPushButton::menu-indicator {
 }
 
 QTableView {
-    border: 1px solid #76797C;
+    border: 1px solid #54575B;
     gridline-color: #31363b;
     background-color: #232629;
 }
 
+QTreeView:disabled {
+    background-color: #1f2225;
+}
+
 QTableView,
 QHeaderView {
-    border-radius: 0px;
+    border-radius: 0;
+}
+
+QListView:focus {
+    border-color: #54575B;
 }
 
 QTableView::item:pressed,
@@ -1078,7 +1072,7 @@ QListView::item:selected:active {
 }
 
 QHeaderView {
-    background-color: #31363b;
+    background-color: #403F3F;
     border: 1px transparent;
     border-radius: 0px;
     margin: 0px;
@@ -1086,30 +1080,32 @@ QHeaderView {
 }
 
 QHeaderView::section {
-    background-color: #31363b;
+    background-color: #232629;
     color: #eff0f1;
-    padding: 5px;
-    border: 1px solid #76797C;
+    padding: 0 5px;
+    border: 1px solid #434242;
+    border-bottom: 0;
     border-radius: 0px;
     text-align: center;
 }
 
 QHeaderView::section::vertical::first,
 QHeaderView::section::vertical::only-one {
-    border-top: 1px solid #76797C;
+    border-top: 1px solid #31363b;
 }
 
 QHeaderView::section::vertical {
     border-top: transparent;
 }
 
+QHeaderView::section::horizontal,
 QHeaderView::section::horizontal::first,
 QHeaderView::section::horizontal::only-one {
-    border-left: 1px solid #76797C;
+    border-left: transparent;
 }
 
-QHeaderView::section::horizontal {
-    border-left: transparent;
+QHeaderView::section::horizontal::last {
+    border-right: transparent;
 }
 
 QHeaderView::section:checked {
@@ -1117,9 +1113,7 @@ QHeaderView::section:checked {
     background-color: #334e5e;
 }
 
-
-/* style the sort indicator */
-
+/* sort indicator */
 QHeaderView::down-arrow {
     image: url(:/qss_icons/rc/down_arrow.png);
 }
@@ -1149,14 +1143,13 @@ QToolBox::tab {
 }
 
 QToolBox::tab:selected {
-    /* italicize selected tabs */
     font: italic;
     background-color: #31363b;
     border-color: #3daee9;
 }
 
 QStatusBar::item {
-    border: 0px transparent dark;
+    border: 0;
 }
 
 QFrame[height="3"],
@@ -1193,7 +1186,6 @@ QProgressBar::chunk {
 
 QDateEdit {
     selection-background-color: #3daee9;
-    border-style: solid;
     border: 1px solid #3375A3;
     border-radius: 2px;
     padding: 1px;
@@ -1217,7 +1209,7 @@ QDateEdit::drop-down {
     subcontrol-origin: padding;
     subcontrol-position: top right;
     width: 15px;
-    border-left-width: 0px;
+    border-left-width: 0;
     border-left-color: darkgray;
     border-left-style: solid;
     border-top-right-radius: 3px;
@@ -1233,3 +1225,14 @@ QDateEdit::down-arrow:hover,
 QDateEdit::down-arrow:focus {
     image: url(:/qss_icons/rc/down_arrow.png);
 }
+
+QComboBox:disabled,
+QPushButton:disabled,
+QAbstractSpinBox:disabled,
+QDateEdit:disabled,
+QLineEdit:disabled,
+QTextEdit:disabled,
+QToolButton:disabled,
+QPlainTextEdit:disabled {
+    background-color: #2b2e31;
+}

src/CMakeLists.txt

@@ -77,6 +77,15 @@ else()
             add_compile_options("-static")
         endif()
     endif()
+
+    if(CMAKE_SYSTEM_NAME STREQUAL "Linux" OR MINGW)
+        # GNU ar: Create thin archive files.
+        # Requires binutils-2.19 or later.
+        set(CMAKE_C_ARCHIVE_CREATE   "<CMAKE_AR> qcTP <TARGET> <LINK_FLAGS> <OBJECTS>")
+        set(CMAKE_C_ARCHIVE_APPEND   "<CMAKE_AR> qTP  <TARGET> <LINK_FLAGS> <OBJECTS>")
+        set(CMAKE_CXX_ARCHIVE_CREATE "<CMAKE_AR> qcTP <TARGET> <LINK_FLAGS> <OBJECTS>")
+        set(CMAKE_CXX_ARCHIVE_APPEND "<CMAKE_AR> qTP  <TARGET> <LINK_FLAGS> <OBJECTS>")
+    endif()
 endif()
 
 add_subdirectory(common)

src/common/CMakeLists.txt

@@ -15,6 +15,10 @@ endif ()
 if (DEFINED ENV{DISPLAYVERSION})
   set(DISPLAY_VERSION $ENV{DISPLAYVERSION})
 endif ()
+
+# Pass the path to git to the GenerateSCMRev.cmake as well
+find_package(Git QUIET)
+
 add_custom_command(OUTPUT scm_rev.cpp
     COMMAND ${CMAKE_COMMAND}
       -DSRC_DIR="${CMAKE_SOURCE_DIR}"
@@ -23,6 +27,7 @@ add_custom_command(OUTPUT scm_rev.cpp
       -DTITLE_BAR_FORMAT_RUNNING="${TITLE_BAR_FORMAT_RUNNING}"
       -DBUILD_TAG="${BUILD_TAG}"
       -DBUILD_ID="${DISPLAY_VERSION}"
+      -DGIT_EXECUTABLE="${GIT_EXECUTABLE}"
       -P "${CMAKE_SOURCE_DIR}/CMakeModules/GenerateSCMRev.cmake"
     DEPENDS
       # WARNING! It was too much work to try and make a common location for this list,

src/common/logging/backend.cpp

@@ -120,7 +120,7 @@ private:
             duration_cast<std::chrono::microseconds>(steady_clock::now() - time_origin);
         entry.log_class = log_class;
         entry.log_level = log_level;
-        entry.filename = Common::TrimSourcePath(filename);
+        entry.filename = filename;
         entry.line_num = line_nr;
         entry.function = function;
         entry.message = std::move(message);

src/common/logging/backend.h

@@ -23,7 +23,7 @@ struct Entry {
     std::chrono::microseconds timestamp;
     Class log_class;
     Level log_level;
-    std::string filename;
+    const char* filename;
     unsigned int line_num;
     std::string function;
     std::string message;

src/common/logging/log.h

@@ -9,6 +9,15 @@
 
 namespace Log {
 
+// trims up to and including the last of ../, ..\, src/, src\ in a string
+constexpr const char* TrimSourcePath(std::string_view source) {
+    const auto rfind = [source](const std::string_view match) {
+        return source.rfind(match) == source.npos ? 0 : (source.rfind(match) + match.size());
+    };
+    auto idx = std::max({rfind("src/"), rfind("src\\"), rfind("../"), rfind("..\\")});
+    return source.data() + idx;
+}
+
 /// Specifies the severity or level of detail of the log message.
 enum class Level : u8 {
     Trace,    ///< Extremely detailed and repetitive debugging information that is likely to
@@ -141,24 +150,24 @@ void FmtLogMessage(Class log_class, Level log_level, const char* filename, unsig
 
 #ifdef _DEBUG
 #define LOG_TRACE(log_class, ...)                                                                  \
-    ::Log::FmtLogMessage(::Log::Class::log_class, ::Log::Level::Trace, __FILE__, __LINE__,         \
-                         __func__, __VA_ARGS__)
+    ::Log::FmtLogMessage(::Log::Class::log_class, ::Log::Level::Trace,                             \
+                         ::Log::TrimSourcePath(__FILE__), __LINE__, __func__, __VA_ARGS__)
 #else
 #define LOG_TRACE(log_class, fmt, ...) (void(0))
 #endif
 
 #define LOG_DEBUG(log_class, ...)                                                                  \
-    ::Log::FmtLogMessage(::Log::Class::log_class, ::Log::Level::Debug, __FILE__, __LINE__,         \
-                         __func__, __VA_ARGS__)
+    ::Log::FmtLogMessage(::Log::Class::log_class, ::Log::Level::Debug,                             \
+                         ::Log::TrimSourcePath(__FILE__), __LINE__, __func__, __VA_ARGS__)
 #define LOG_INFO(log_class, ...)                                                                   \
-    ::Log::FmtLogMessage(::Log::Class::log_class, ::Log::Level::Info, __FILE__, __LINE__,          \
-                         __func__, __VA_ARGS__)
+    ::Log::FmtLogMessage(::Log::Class::log_class, ::Log::Level::Info,                              \
+                         ::Log::TrimSourcePath(__FILE__), __LINE__, __func__, __VA_ARGS__)
 #define LOG_WARNING(log_class, ...)                                                                \
-    ::Log::FmtLogMessage(::Log::Class::log_class, ::Log::Level::Warning, __FILE__, __LINE__,       \
-                         __func__, __VA_ARGS__)
+    ::Log::FmtLogMessage(::Log::Class::log_class, ::Log::Level::Warning,                           \
+                         ::Log::TrimSourcePath(__FILE__), __LINE__, __func__, __VA_ARGS__)
 #define LOG_ERROR(log_class, ...)                                                                  \
-    ::Log::FmtLogMessage(::Log::Class::log_class, ::Log::Level::Error, __FILE__, __LINE__,         \
-                         __func__, __VA_ARGS__)
+    ::Log::FmtLogMessage(::Log::Class::log_class, ::Log::Level::Error,                             \
+                         ::Log::TrimSourcePath(__FILE__), __LINE__, __func__, __VA_ARGS__)
 #define LOG_CRITICAL(log_class, ...)                                                               \
-    ::Log::FmtLogMessage(::Log::Class::log_class, ::Log::Level::Critical, __FILE__, __LINE__,      \
-                         __func__, __VA_ARGS__)
+    ::Log::FmtLogMessage(::Log::Class::log_class, ::Log::Level::Critical,                          \
+                         ::Log::TrimSourcePath(__FILE__), __LINE__, __func__, __VA_ARGS__)

src/common/string_util.cpp

@@ -223,26 +223,4 @@ std::u16string UTF16StringFromFixedZeroTerminatedBuffer(std::u16string_view buff
     return std::u16string(buffer.begin(), buffer.begin() + len);
 }
 
-const char* TrimSourcePath(const char* path, const char* root) {
-    const char* p = path;
-
-    while (*p != '\0') {
-        const char* next_slash = p;
-        while (*next_slash != '\0' && *next_slash != '/' && *next_slash != '\\') {
-            ++next_slash;
-        }
-
-        bool is_src = Common::ComparePartialString(p, next_slash, root);
-        p = next_slash;
-
-        if (*p != '\0') {
-            ++p;
-        }
-        if (is_src) {
-            path = p;
-        }
-    }
-    return path;
-}
-
 } // namespace Common

src/common/telemetry.cpp

@@ -44,20 +44,6 @@ template class Field<std::string>;
 template class Field<const char*>;
 template class Field<std::chrono::microseconds>;
 
-#ifdef ARCHITECTURE_x86_64
-static const char* CpuVendorToStr(Common::CPUVendor vendor) {
-    switch (vendor) {
-    case Common::CPUVendor::INTEL:
-        return "Intel";
-    case Common::CPUVendor::AMD:
-        return "Amd";
-    case Common::CPUVendor::OTHER:
-        return "Other";
-    }
-    UNREACHABLE();
-}
-#endif
-
 void AppendBuildInfo(FieldCollection& fc) {
     const bool is_git_dirty{std::strstr(Common::g_scm_desc, "dirty") != nullptr};
     fc.AddField(FieldType::App, "Git_IsDirty", is_git_dirty);
@@ -71,7 +57,6 @@ void AppendCPUInfo(FieldCollection& fc) {
 #ifdef ARCHITECTURE_x86_64
     fc.AddField(FieldType::UserSystem, "CPU_Model", Common::GetCPUCaps().cpu_string);
     fc.AddField(FieldType::UserSystem, "CPU_BrandString", Common::GetCPUCaps().brand_string);
-    fc.AddField(FieldType::UserSystem, "CPU_Vendor", CpuVendorToStr(Common::GetCPUCaps().vendor));
     fc.AddField(FieldType::UserSystem, "CPU_Extension_x64_AES", Common::GetCPUCaps().aes);
     fc.AddField(FieldType::UserSystem, "CPU_Extension_x64_AVX", Common::GetCPUCaps().avx);
     fc.AddField(FieldType::UserSystem, "CPU_Extension_x64_AVX2", Common::GetCPUCaps().avx2);

src/common/thread.h

@@ -28,6 +28,15 @@ public:
         is_set = false;
     }
 
+    template <class Duration>
+    bool WaitFor(const std::chrono::duration<Duration>& time) {
+        std::unique_lock lk{mutex};
+        if (!condvar.wait_for(lk, time, [this] { return is_set; }))
+            return false;
+        is_set = false;
+        return true;
+    }
+
     template <class Clock, class Duration>
     bool WaitUntil(const std::chrono::time_point<Clock, Duration>& time) {
         std::unique_lock lk{mutex};

src/common/x64/cpu_detect.cpp

@@ -3,8 +3,6 @@
 // Refer to the license.txt file included.
 
 #include <cstring>
-#include <string>
-#include <thread>
 #include "common/common_types.h"
 #include "common/x64/cpu_detect.h"
 
@@ -51,8 +49,6 @@ namespace Common {
 static CPUCaps Detect() {
     CPUCaps caps = {};
 
-    caps.num_cores = std::thread::hardware_concurrency();
-
     // Assumes the CPU supports the CPUID instruction. Those that don't would likely not support
     // yuzu at all anyway
 
@@ -70,12 +66,6 @@ static CPUCaps Detect() {
     __cpuid(cpu_id, 0x80000000);
 
     u32 max_ex_fn = cpu_id[0];
-    if (!strcmp(caps.brand_string, "GenuineIntel"))
-        caps.vendor = CPUVendor::INTEL;
-    else if (!strcmp(caps.brand_string, "AuthenticAMD"))
-        caps.vendor = CPUVendor::AMD;
-    else
-        caps.vendor = CPUVendor::OTHER;
 
     // Set reasonable default brand string even if brand string not available
     strcpy(caps.cpu_string, caps.brand_string);
@@ -96,15 +86,9 @@ static CPUCaps Detect() {
             caps.sse4_1 = true;
         if ((cpu_id[2] >> 20) & 1)
             caps.sse4_2 = true;
-        if ((cpu_id[2] >> 22) & 1)
-            caps.movbe = true;
         if ((cpu_id[2] >> 25) & 1)
             caps.aes = true;
 
-        if ((cpu_id[3] >> 24) & 1) {
-            caps.fxsave_fxrstor = true;
-        }
-
         // AVX support requires 3 separate checks:
         //  - Is the AVX bit set in CPUID?
         //  - Is the XSAVE bit set in CPUID?
@@ -129,8 +113,6 @@ static CPUCaps Detect() {
         }
     }
 
-    caps.flush_to_zero = caps.sse;
-
     if (max_ex_fn >= 0x80000004) {
         // Extract CPU model string
         __cpuid(cpu_id, 0x80000002);
@@ -144,14 +126,8 @@ static CPUCaps Detect() {
     if (max_ex_fn >= 0x80000001) {
         // Check for more features
         __cpuid(cpu_id, 0x80000001);
-        if (cpu_id[2] & 1)
-            caps.lahf_sahf_64 = true;
-        if ((cpu_id[2] >> 5) & 1)
-            caps.lzcnt = true;
         if ((cpu_id[2] >> 16) & 1)
             caps.fma4 = true;
-        if ((cpu_id[3] >> 29) & 1)
-            caps.long_mode = true;
     }
 
     return caps;
@@ -162,48 +138,4 @@ const CPUCaps& GetCPUCaps() {
     return caps;
 }
 
-std::string GetCPUCapsString() {
-    auto caps = GetCPUCaps();
-
-    std::string sum(caps.cpu_string);
-    sum += " (";
-    sum += caps.brand_string;
-    sum += ")";
-
-    if (caps.sse)
-        sum += ", SSE";
-    if (caps.sse2) {
-        sum += ", SSE2";
-        if (!caps.flush_to_zero)
-            sum += " (without DAZ)";
-    }
-
-    if (caps.sse3)
-        sum += ", SSE3";
-    if (caps.ssse3)
-        sum += ", SSSE3";
-    if (caps.sse4_1)
-        sum += ", SSE4.1";
-    if (caps.sse4_2)
-        sum += ", SSE4.2";
-    if (caps.avx)
-        sum += ", AVX";
-    if (caps.avx2)
-        sum += ", AVX2";
-    if (caps.bmi1)
-        sum += ", BMI1";
-    if (caps.bmi2)
-        sum += ", BMI2";
-    if (caps.fma)
-        sum += ", FMA";
-    if (caps.aes)
-        sum += ", AES";
-    if (caps.movbe)
-        sum += ", MOVBE";
-    if (caps.long_mode)
-        sum += ", 64-bit support";
-
-    return sum;
-}
-
 } // namespace Common

src/common/x64/cpu_detect.h

@@ -4,23 +4,12 @@
 
 #pragma once
 
-#include <string>
-
 namespace Common {
 
-/// x86/x64 CPU vendors that may be detected by this module
-enum class CPUVendor {
-    INTEL,
-    AMD,
-    OTHER,
-};
-
 /// x86/x64 CPU capabilities that may be detected by this module
 struct CPUCaps {
-    CPUVendor vendor;
     char cpu_string[0x21];
     char brand_string[0x41];
-    int num_cores;
     bool sse;
     bool sse2;
     bool sse3;
@@ -35,20 +24,6 @@ struct CPUCaps {
     bool fma;
     bool fma4;
     bool aes;
-
-    // Support for the FXSAVE and FXRSTOR instructions
-    bool fxsave_fxrstor;
-
-    bool movbe;
-
-    // This flag indicates that the hardware supports some mode in which denormal inputs and outputs
-    // are automatically set to (signed) zero.
-    bool flush_to_zero;
-
-    // Support for LAHF and SAHF instructions in 64-bit mode
-    bool lahf_sahf_64;
-
-    bool long_mode;
 };
 
 /**
@@ -57,10 +32,4 @@ struct CPUCaps {
  */
 const CPUCaps& GetCPUCaps();
 
-/**
- * Gets a string summary of the name and supported capabilities of the host CPU
- * @return String summary
- */
-std::string GetCPUCapsString();
-
 } // namespace Common

src/core/CMakeLists.txt

@@ -15,14 +15,14 @@ add_library(core STATIC
     constants.h
     core.cpp
     core.h
-    core_cpu.cpp
-    core_cpu.h
+    core_manager.cpp
+    core_manager.h
     core_timing.cpp
     core_timing.h
     core_timing_util.cpp
     core_timing_util.h
-    cpu_core_manager.cpp
-    cpu_core_manager.h
+    cpu_manager.cpp
+    cpu_manager.h
     crypto/aes_util.cpp
     crypto/aes_util.h
     crypto/encryption_layer.cpp
@@ -158,6 +158,8 @@ add_library(core STATIC
     hle/kernel/mutex.h
     hle/kernel/object.cpp
     hle/kernel/object.h
+    hle/kernel/physical_core.cpp
+    hle/kernel/physical_core.h
     hle/kernel/process.cpp
     hle/kernel/process.h
     hle/kernel/process_capability.cpp

src/core/arm/dynarmic/arm_dynarmic.cpp

@@ -10,7 +10,7 @@
 #include "common/microprofile.h"
 #include "core/arm/dynarmic/arm_dynarmic.h"
 #include "core/core.h"
-#include "core/core_cpu.h"
+#include "core/core_manager.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/gdbstub/gdbstub.h"
@@ -141,6 +141,7 @@ std::unique_ptr<Dynarmic::A64::Jit> ARM_Dynarmic::MakeJit(Common::PageTable& pag
     config.page_table = reinterpret_cast<void**>(page_table.pointers.data());
     config.page_table_address_space_bits = address_space_bits;
     config.silently_mirror_page_table = false;
+    config.absolute_offset_page_table = true;
 
     // Multi-process state
     config.processor_id = core_index;

src/core/arm/exclusive_monitor.cpp

@@ -2,10 +2,24 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#ifdef ARCHITECTURE_x86_64
+#include "core/arm/dynarmic/arm_dynarmic.h"
+#endif
 #include "core/arm/exclusive_monitor.h"
+#include "core/memory.h"
 
 namespace Core {
 
 ExclusiveMonitor::~ExclusiveMonitor() = default;
 
+std::unique_ptr<Core::ExclusiveMonitor> MakeExclusiveMonitor(Memory::Memory& memory,
+                                                             std::size_t num_cores) {
+#ifdef ARCHITECTURE_x86_64
+    return std::make_unique<Core::DynarmicExclusiveMonitor>(memory, num_cores);
+#else
+    // TODO(merry): Passthrough exclusive monitor
+    return nullptr;
+#endif
+}
+
 } // namespace Core

src/core/arm/exclusive_monitor.h

@@ -4,8 +4,14 @@
 
 #pragma once
 
+#include <memory>
+
 #include "common/common_types.h"
 
+namespace Memory {
+class Memory;
+}
+
 namespace Core {
 
 class ExclusiveMonitor {
@@ -22,4 +28,7 @@ public:
     virtual bool ExclusiveWrite128(std::size_t core_index, VAddr vaddr, u128 value) = 0;
 };
 
+std::unique_ptr<Core::ExclusiveMonitor> MakeExclusiveMonitor(Memory::Memory& memory,
+                                                             std::size_t num_cores);
+
 } // namespace Core

src/core/core.cpp

@@ -11,9 +11,9 @@
 #include "common/string_util.h"
 #include "core/arm/exclusive_monitor.h"
 #include "core/core.h"
-#include "core/core_cpu.h"
+#include "core/core_manager.h"
 #include "core/core_timing.h"
-#include "core/cpu_core_manager.h"
+#include "core/cpu_manager.h"
 #include "core/file_sys/bis_factory.h"
 #include "core/file_sys/card_image.h"
 #include "core/file_sys/mode.h"
@@ -28,6 +28,7 @@
 #include "core/hardware_interrupt_manager.h"
 #include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/physical_core.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/scheduler.h"
 #include "core/hle/kernel/thread.h"
@@ -113,16 +114,25 @@ FileSys::VirtualFile GetGameFileFromPath(const FileSys::VirtualFilesystem& vfs,
 struct System::Impl {
     explicit Impl(System& system)
         : kernel{system}, fs_controller{system}, memory{system},
-          cpu_core_manager{system}, reporter{system}, applet_manager{system} {}
+          cpu_manager{system}, reporter{system}, applet_manager{system} {}
 
-    Cpu& CurrentCpuCore() {
-        return cpu_core_manager.GetCurrentCore();
+    CoreManager& CurrentCoreManager() {
+        return cpu_manager.GetCurrentCoreManager();
+    }
+
+    Kernel::PhysicalCore& CurrentPhysicalCore() {
+        const auto index = cpu_manager.GetActiveCoreIndex();
+        return kernel.PhysicalCore(index);
+    }
+
+    Kernel::PhysicalCore& GetPhysicalCore(std::size_t index) {
+        return kernel.PhysicalCore(index);
     }
 
     ResultStatus RunLoop(bool tight_loop) {
         status = ResultStatus::Success;
 
-        cpu_core_manager.RunLoop(tight_loop);
+        cpu_manager.RunLoop(tight_loop);
 
         return status;
     }
@@ -131,8 +141,8 @@ struct System::Impl {
         LOG_DEBUG(HW_Memory, "initialized OK");
 
         core_timing.Initialize();
-        cpu_core_manager.Initialize();
         kernel.Initialize();
+        cpu_manager.Initialize();
 
         const auto current_time = std::chrono::duration_cast<std::chrono::seconds>(
             std::chrono::system_clock::now().time_since_epoch());
@@ -205,7 +215,6 @@ struct System::Impl {
         // Main process has been loaded and been made current.
         // Begin GPU and CPU execution.
         gpu_core->Start();
-        cpu_core_manager.StartThreads();
 
         // Initialize cheat engine
         if (cheat_engine) {
@@ -272,7 +281,7 @@ struct System::Impl {
         gpu_core.reset();
 
         // Close all CPU/threading state
-        cpu_core_manager.Shutdown();
+        cpu_manager.Shutdown();
 
         // Shutdown kernel and core timing
         kernel.Shutdown();
@@ -342,7 +351,7 @@ struct System::Impl {
     std::unique_ptr<Tegra::GPU> gpu_core;
     std::unique_ptr<Hardware::InterruptManager> interrupt_manager;
     Memory::Memory memory;
-    CpuCoreManager cpu_core_manager;
+    CpuManager cpu_manager;
     bool is_powered_on = false;
     bool exit_lock = false;
 
@@ -377,12 +386,12 @@ struct System::Impl {
 System::System() : impl{std::make_unique<Impl>(*this)} {}
 System::~System() = default;
 
-Cpu& System::CurrentCpuCore() {
-    return impl->CurrentCpuCore();
+CoreManager& System::CurrentCoreManager() {
+    return impl->CurrentCoreManager();
 }
 
-const Cpu& System::CurrentCpuCore() const {
-    return impl->CurrentCpuCore();
+const CoreManager& System::CurrentCoreManager() const {
+    return impl->CurrentCoreManager();
 }
 
 System::ResultStatus System::RunLoop(bool tight_loop) {
@@ -394,7 +403,7 @@ System::ResultStatus System::SingleStep() {
 }
 
 void System::InvalidateCpuInstructionCaches() {
-    impl->cpu_core_manager.InvalidateAllInstructionCaches();
+    impl->kernel.InvalidateAllInstructionCaches();
 }
 
 System::ResultStatus System::Load(Frontend::EmuWindow& emu_window, const std::string& filepath) {
@@ -406,13 +415,11 @@ bool System::IsPoweredOn() const {
 }
 
 void System::PrepareReschedule() {
-    CurrentCpuCore().PrepareReschedule();
+    impl->CurrentPhysicalCore().Stop();
 }
 
 void System::PrepareReschedule(const u32 core_index) {
-    if (core_index < GlobalScheduler().CpuCoresCount()) {
-        CpuCore(core_index).PrepareReschedule();
-    }
+    impl->kernel.PrepareReschedule(core_index);
 }
 
 PerfStatsResults System::GetAndResetPerfStats() {
@@ -428,31 +435,31 @@ const TelemetrySession& System::TelemetrySession() const {
 }
 
 ARM_Interface& System::CurrentArmInterface() {
-    return CurrentCpuCore().ArmInterface();
+    return impl->CurrentPhysicalCore().ArmInterface();
 }
 
 const ARM_Interface& System::CurrentArmInterface() const {
-    return CurrentCpuCore().ArmInterface();
+    return impl->CurrentPhysicalCore().ArmInterface();
 }
 
 std::size_t System::CurrentCoreIndex() const {
-    return CurrentCpuCore().CoreIndex();
+    return impl->cpu_manager.GetActiveCoreIndex();
 }
 
 Kernel::Scheduler& System::CurrentScheduler() {
-    return CurrentCpuCore().Scheduler();
+    return impl->CurrentPhysicalCore().Scheduler();
 }
 
 const Kernel::Scheduler& System::CurrentScheduler() const {
-    return CurrentCpuCore().Scheduler();
+    return impl->CurrentPhysicalCore().Scheduler();
 }
 
 Kernel::Scheduler& System::Scheduler(std::size_t core_index) {
-    return CpuCore(core_index).Scheduler();
+    return impl->GetPhysicalCore(core_index).Scheduler();
 }
 
 const Kernel::Scheduler& System::Scheduler(std::size_t core_index) const {
-    return CpuCore(core_index).Scheduler();
+    return impl->GetPhysicalCore(core_index).Scheduler();
 }
 
 /// Gets the global scheduler
@@ -474,28 +481,28 @@ const Kernel::Process* System::CurrentProcess() const {
 }
 
 ARM_Interface& System::ArmInterface(std::size_t core_index) {
-    return CpuCore(core_index).ArmInterface();
+    return impl->GetPhysicalCore(core_index).ArmInterface();
 }
 
 const ARM_Interface& System::ArmInterface(std::size_t core_index) const {
-    return CpuCore(core_index).ArmInterface();
+    return impl->GetPhysicalCore(core_index).ArmInterface();
 }
 
-Cpu& System::CpuCore(std::size_t core_index) {
-    return impl->cpu_core_manager.GetCore(core_index);
+CoreManager& System::GetCoreManager(std::size_t core_index) {
+    return impl->cpu_manager.GetCoreManager(core_index);
 }
 
-const Cpu& System::CpuCore(std::size_t core_index) const {
+const CoreManager& System::GetCoreManager(std::size_t core_index) const {
     ASSERT(core_index < NUM_CPU_CORES);
-    return impl->cpu_core_manager.GetCore(core_index);
+    return impl->cpu_manager.GetCoreManager(core_index);
 }
 
 ExclusiveMonitor& System::Monitor() {
-    return impl->cpu_core_manager.GetExclusiveMonitor();
+    return impl->kernel.GetExclusiveMonitor();
 }
 
 const ExclusiveMonitor& System::Monitor() const {
-    return impl->cpu_core_manager.GetExclusiveMonitor();
+    return impl->kernel.GetExclusiveMonitor();
 }
 
 Memory::Memory& System::Memory() {

src/core/core.h

@@ -93,7 +93,7 @@ class Memory;
 namespace Core {
 
 class ARM_Interface;
-class Cpu;
+class CoreManager;
 class ExclusiveMonitor;
 class FrameLimiter;
 class PerfStats;
@@ -218,10 +218,10 @@ public:
     const ARM_Interface& ArmInterface(std::size_t core_index) const;
 
     /// Gets a CPU interface to the CPU core with the specified index
-    Cpu& CpuCore(std::size_t core_index);
+    CoreManager& GetCoreManager(std::size_t core_index);
 
     /// Gets a CPU interface to the CPU core with the specified index
-    const Cpu& CpuCore(std::size_t core_index) const;
+    const CoreManager& GetCoreManager(std::size_t core_index) const;
 
     /// Gets a reference to the exclusive monitor
     ExclusiveMonitor& Monitor();
@@ -364,10 +364,10 @@ private:
     System();
 
     /// Returns the currently running CPU core
-    Cpu& CurrentCpuCore();
+    CoreManager& CurrentCoreManager();
 
     /// Returns the currently running CPU core
-    const Cpu& CurrentCpuCore() const;
+    const CoreManager& CurrentCoreManager() const;
 
     /**
      * Initialize the emulated system.

src/core/core_cpu.cpp

@@ -1,127 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include <condition_variable>
-#include <mutex>
-
-#include "common/logging/log.h"
-#ifdef ARCHITECTURE_x86_64
-#include "core/arm/dynarmic/arm_dynarmic.h"
-#endif
-#include "core/arm/exclusive_monitor.h"
-#include "core/arm/unicorn/arm_unicorn.h"
-#include "core/core.h"
-#include "core/core_cpu.h"
-#include "core/core_timing.h"
-#include "core/hle/kernel/scheduler.h"
-#include "core/hle/kernel/thread.h"
-#include "core/hle/lock.h"
-#include "core/settings.h"
-
-namespace Core {
-
-void CpuBarrier::NotifyEnd() {
-    std::unique_lock lock{mutex};
-    end = true;
-    condition.notify_all();
-}
-
-bool CpuBarrier::Rendezvous() {
-    if (!Settings::values.use_multi_core) {
-        // Meaningless when running in single-core mode
-        return true;
-    }
-
-    if (!end) {
-        std::unique_lock lock{mutex};
-
-        --cores_waiting;
-        if (!cores_waiting) {
-            cores_waiting = NUM_CPU_CORES;
-            condition.notify_all();
-            return true;
-        }
-
-        condition.wait(lock);
-        return true;
-    }
-
-    return false;
-}
-
-Cpu::Cpu(System& system, ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier,
-         std::size_t core_index)
-    : cpu_barrier{cpu_barrier}, global_scheduler{system.GlobalScheduler()},
-      core_timing{system.CoreTiming()}, core_index{core_index} {
-#ifdef ARCHITECTURE_x86_64
-    arm_interface = std::make_unique<ARM_Dynarmic>(system, exclusive_monitor, core_index);
-#else
-    arm_interface = std::make_unique<ARM_Unicorn>(system);
-    LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
-#endif
-
-    scheduler = std::make_unique<Kernel::Scheduler>(system, *arm_interface, core_index);
-}
-
-Cpu::~Cpu() = default;
-
-std::unique_ptr<ExclusiveMonitor> Cpu::MakeExclusiveMonitor(
-    [[maybe_unused]] Memory::Memory& memory, [[maybe_unused]] std::size_t num_cores) {
-#ifdef ARCHITECTURE_x86_64
-    return std::make_unique<DynarmicExclusiveMonitor>(memory, num_cores);
-#else
-    // TODO(merry): Passthrough exclusive monitor
-    return nullptr;
-#endif
-}
-
-void Cpu::RunLoop(bool tight_loop) {
-    // Wait for all other CPU cores to complete the previous slice, such that they run in lock-step
-    if (!cpu_barrier.Rendezvous()) {
-        // If rendezvous failed, session has been killed
-        return;
-    }
-
-    Reschedule();
-
-    // If we don't have a currently active thread then don't execute instructions,
-    // instead advance to the next event and try to yield to the next thread
-    if (Kernel::GetCurrentThread() == nullptr) {
-        LOG_TRACE(Core, "Core-{} idling", core_index);
-        core_timing.Idle();
-    } else {
-        if (tight_loop) {
-            arm_interface->Run();
-        } else {
-            arm_interface->Step();
-        }
-        // We are stopping a run, exclusive state must be cleared
-        arm_interface->ClearExclusiveState();
-    }
-    core_timing.Advance();
-
-    Reschedule();
-}
-
-void Cpu::SingleStep() {
-    return RunLoop(false);
-}
-
-void Cpu::PrepareReschedule() {
-    arm_interface->PrepareReschedule();
-}
-
-void Cpu::Reschedule() {
-    // Lock the global kernel mutex when we manipulate the HLE state
-    std::lock_guard lock(HLE::g_hle_lock);
-
-    global_scheduler.SelectThread(core_index);
-    scheduler->TryDoContextSwitch();
-}
-
-void Cpu::Shutdown() {
-    scheduler->Shutdown();
-}
-
-} // namespace Core

src/core/core_cpu.h

@@ -1,120 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <atomic>
-#include <condition_variable>
-#include <cstddef>
-#include <memory>
-#include <mutex>
-#include "common/common_types.h"
-
-namespace Kernel {
-class GlobalScheduler;
-class Scheduler;
-} // namespace Kernel
-
-namespace Core {
-class System;
-}
-
-namespace Core::Timing {
-class CoreTiming;
-}
-
-namespace Memory {
-class Memory;
-}
-
-namespace Core {
-
-class ARM_Interface;
-class ExclusiveMonitor;
-
-constexpr unsigned NUM_CPU_CORES{4};
-
-class CpuBarrier {
-public:
-    bool IsAlive() const {
-        return !end;
-    }
-
-    void NotifyEnd();
-
-    bool Rendezvous();
-
-private:
-    unsigned cores_waiting{NUM_CPU_CORES};
-    std::mutex mutex;
-    std::condition_variable condition;
-    std::atomic<bool> end{};
-};
-
-class Cpu {
-public:
-    Cpu(System& system, ExclusiveMonitor& exclusive_monitor, CpuBarrier& cpu_barrier,
-        std::size_t core_index);
-    ~Cpu();
-
-    void RunLoop(bool tight_loop = true);
-
-    void SingleStep();
-
-    void PrepareReschedule();
-
-    ARM_Interface& ArmInterface() {
-        return *arm_interface;
-    }
-
-    const ARM_Interface& ArmInterface() const {
-        return *arm_interface;
-    }
-
-    Kernel::Scheduler& Scheduler() {
-        return *scheduler;
-    }
-
-    const Kernel::Scheduler& Scheduler() const {
-        return *scheduler;
-    }
-
-    bool IsMainCore() const {
-        return core_index == 0;
-    }
-
-    std::size_t CoreIndex() const {
-        return core_index;
-    }
-
-    void Shutdown();
-
-    /**
-     * Creates an exclusive monitor to handle exclusive reads/writes.
-     *
-     * @param memory The current memory subsystem that the monitor may wish
-     *               to keep track of.
-     *
-     * @param num_cores The number of cores to assume about the CPU.
-     *
-     * @returns The constructed exclusive monitor instance, or nullptr if the current
-     *          CPU backend is unable to use an exclusive monitor.
-     */
-    static std::unique_ptr<ExclusiveMonitor> MakeExclusiveMonitor(Memory::Memory& memory,
-                                                                  std::size_t num_cores);
-
-private:
-    void Reschedule();
-
-    std::unique_ptr<ARM_Interface> arm_interface;
-    CpuBarrier& cpu_barrier;
-    Kernel::GlobalScheduler& global_scheduler;
-    std::unique_ptr<Kernel::Scheduler> scheduler;
-    Timing::CoreTiming& core_timing;
-
-    std::atomic<bool> reschedule_pending = false;
-    std::size_t core_index;
-};
-
-} // namespace Core

src/core/core_manager.cpp

@@ -0,0 +1,70 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <condition_variable>
+#include <mutex>
+
+#include "common/logging/log.h"
+#ifdef ARCHITECTURE_x86_64
+#include "core/arm/dynarmic/arm_dynarmic.h"
+#endif
+#include "core/arm/exclusive_monitor.h"
+#include "core/arm/unicorn/arm_unicorn.h"
+#include "core/core.h"
+#include "core/core_manager.h"
+#include "core/core_timing.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/physical_core.h"
+#include "core/hle/kernel/scheduler.h"
+#include "core/hle/kernel/thread.h"
+#include "core/hle/lock.h"
+#include "core/settings.h"
+
+namespace Core {
+
+CoreManager::CoreManager(System& system, std::size_t core_index)
+    : global_scheduler{system.GlobalScheduler()}, physical_core{system.Kernel().PhysicalCore(
+                                                      core_index)},
+      core_timing{system.CoreTiming()}, core_index{core_index} {}
+
+CoreManager::~CoreManager() = default;
+
+void CoreManager::RunLoop(bool tight_loop) {
+    Reschedule();
+
+    // If we don't have a currently active thread then don't execute instructions,
+    // instead advance to the next event and try to yield to the next thread
+    if (Kernel::GetCurrentThread() == nullptr) {
+        LOG_TRACE(Core, "Core-{} idling", core_index);
+        core_timing.Idle();
+    } else {
+        if (tight_loop) {
+            physical_core.Run();
+        } else {
+            physical_core.Step();
+        }
+    }
+    core_timing.Advance();
+
+    Reschedule();
+}
+
+void CoreManager::SingleStep() {
+    return RunLoop(false);
+}
+
+void CoreManager::PrepareReschedule() {
+    physical_core.Stop();
+}
+
+void CoreManager::Reschedule() {
+    // Lock the global kernel mutex when we manipulate the HLE state
+    std::lock_guard lock(HLE::g_hle_lock);
+
+    global_scheduler.SelectThread(core_index);
+
+    physical_core.Scheduler().TryDoContextSwitch();
+}
+
+} // namespace Core

src/core/core_manager.h

@@ -0,0 +1,63 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <atomic>
+#include <cstddef>
+#include <memory>
+#include "common/common_types.h"
+
+namespace Kernel {
+class GlobalScheduler;
+class PhysicalCore;
+} // namespace Kernel
+
+namespace Core {
+class System;
+}
+
+namespace Core::Timing {
+class CoreTiming;
+}
+
+namespace Memory {
+class Memory;
+}
+
+namespace Core {
+
+constexpr unsigned NUM_CPU_CORES{4};
+
+class CoreManager {
+public:
+    CoreManager(System& system, std::size_t core_index);
+    ~CoreManager();
+
+    void RunLoop(bool tight_loop = true);
+
+    void SingleStep();
+
+    void PrepareReschedule();
+
+    bool IsMainCore() const {
+        return core_index == 0;
+    }
+
+    std::size_t CoreIndex() const {
+        return core_index;
+    }
+
+private:
+    void Reschedule();
+
+    Kernel::GlobalScheduler& global_scheduler;
+    Kernel::PhysicalCore& physical_core;
+    Timing::CoreTiming& core_timing;
+
+    std::atomic<bool> reschedule_pending = false;
+    std::size_t core_index;
+};
+
+} // namespace Core

src/core/cpu_core_manager.cpp

@@ -1,152 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#include "common/assert.h"
-#include "core/arm/exclusive_monitor.h"
-#include "core/core.h"
-#include "core/core_cpu.h"
-#include "core/core_timing.h"
-#include "core/cpu_core_manager.h"
-#include "core/gdbstub/gdbstub.h"
-#include "core/settings.h"
-
-namespace Core {
-namespace {
-void RunCpuCore(const System& system, Cpu& cpu_state) {
-    while (system.IsPoweredOn()) {
-        cpu_state.RunLoop(true);
-    }
-}
-} // Anonymous namespace
-
-CpuCoreManager::CpuCoreManager(System& system) : system{system} {}
-CpuCoreManager::~CpuCoreManager() = default;
-
-void CpuCoreManager::Initialize() {
-    barrier = std::make_unique<CpuBarrier>();
-    exclusive_monitor = Cpu::MakeExclusiveMonitor(system.Memory(), cores.size());
-
-    for (std::size_t index = 0; index < cores.size(); ++index) {
-        cores[index] = std::make_unique<Cpu>(system, *exclusive_monitor, *barrier, index);
-    }
-}
-
-void CpuCoreManager::StartThreads() {
-    // Create threads for CPU cores 1-3, and build thread_to_cpu map
-    // CPU core 0 is run on the main thread
-    thread_to_cpu[std::this_thread::get_id()] = cores[0].get();
-    if (!Settings::values.use_multi_core) {
-        return;
-    }
-
-    for (std::size_t index = 0; index < core_threads.size(); ++index) {
-        core_threads[index] = std::make_unique<std::thread>(RunCpuCore, std::cref(system),
-                                                            std::ref(*cores[index + 1]));
-        thread_to_cpu[core_threads[index]->get_id()] = cores[index + 1].get();
-    }
-}
-
-void CpuCoreManager::Shutdown() {
-    barrier->NotifyEnd();
-    if (Settings::values.use_multi_core) {
-        for (auto& thread : core_threads) {
-            thread->join();
-            thread.reset();
-        }
-    }
-
-    thread_to_cpu.clear();
-    for (auto& cpu_core : cores) {
-        cpu_core->Shutdown();
-        cpu_core.reset();
-    }
-
-    exclusive_monitor.reset();
-    barrier.reset();
-}
-
-Cpu& CpuCoreManager::GetCore(std::size_t index) {
-    return *cores.at(index);
-}
-
-const Cpu& CpuCoreManager::GetCore(std::size_t index) const {
-    return *cores.at(index);
-}
-
-ExclusiveMonitor& CpuCoreManager::GetExclusiveMonitor() {
-    return *exclusive_monitor;
-}
-
-const ExclusiveMonitor& CpuCoreManager::GetExclusiveMonitor() const {
-    return *exclusive_monitor;
-}
-
-Cpu& CpuCoreManager::GetCurrentCore() {
-    if (Settings::values.use_multi_core) {
-        const auto& search = thread_to_cpu.find(std::this_thread::get_id());
-        ASSERT(search != thread_to_cpu.end());
-        ASSERT(search->second);
-        return *search->second;
-    }
-
-    // Otherwise, use single-threaded mode active_core variable
-    return *cores[active_core];
-}
-
-const Cpu& CpuCoreManager::GetCurrentCore() const {
-    if (Settings::values.use_multi_core) {
-        const auto& search = thread_to_cpu.find(std::this_thread::get_id());
-        ASSERT(search != thread_to_cpu.end());
-        ASSERT(search->second);
-        return *search->second;
-    }
-
-    // Otherwise, use single-threaded mode active_core variable
-    return *cores[active_core];
-}
-
-void CpuCoreManager::RunLoop(bool tight_loop) {
-    // Update thread_to_cpu in case Core 0 is run from a different host thread
-    thread_to_cpu[std::this_thread::get_id()] = cores[0].get();
-
-    if (GDBStub::IsServerEnabled()) {
-        GDBStub::HandlePacket();
-
-        // If the loop is halted and we want to step, use a tiny (1) number of instructions to
-        // execute. Otherwise, get out of the loop function.
-        if (GDBStub::GetCpuHaltFlag()) {
-            if (GDBStub::GetCpuStepFlag()) {
-                tight_loop = false;
-            } else {
-                return;
-            }
-        }
-    }
-
-    auto& core_timing = system.CoreTiming();
-    core_timing.ResetRun();
-    bool keep_running{};
-    do {
-        keep_running = false;
-        for (active_core = 0; active_core < NUM_CPU_CORES; ++active_core) {
-            core_timing.SwitchContext(active_core);
-            if (core_timing.CanCurrentContextRun()) {
-                cores[active_core]->RunLoop(tight_loop);
-            }
-            keep_running |= core_timing.CanCurrentContextRun();
-        }
-    } while (keep_running);
-
-    if (GDBStub::IsServerEnabled()) {
-        GDBStub::SetCpuStepFlag(false);
-    }
-}
-
-void CpuCoreManager::InvalidateAllInstructionCaches() {
-    for (auto& cpu : cores) {
-        cpu->ArmInterface().ClearInstructionCache();
-    }
-}
-
-} // namespace Core

src/core/cpu_core_manager.h

@@ -1,62 +0,0 @@
-// Copyright 2018 yuzu emulator team
-// Licensed under GPLv2 or any later version
-// Refer to the license.txt file included.
-
-#pragma once
-
-#include <array>
-#include <map>
-#include <memory>
-#include <thread>
-
-namespace Core {
-
-class Cpu;
-class CpuBarrier;
-class ExclusiveMonitor;
-class System;
-
-class CpuCoreManager {
-public:
-    explicit CpuCoreManager(System& system);
-    CpuCoreManager(const CpuCoreManager&) = delete;
-    CpuCoreManager(CpuCoreManager&&) = delete;
-
-    ~CpuCoreManager();
-
-    CpuCoreManager& operator=(const CpuCoreManager&) = delete;
-    CpuCoreManager& operator=(CpuCoreManager&&) = delete;
-
-    void Initialize();
-    void StartThreads();
-    void Shutdown();
-
-    Cpu& GetCore(std::size_t index);
-    const Cpu& GetCore(std::size_t index) const;
-
-    Cpu& GetCurrentCore();
-    const Cpu& GetCurrentCore() const;
-
-    ExclusiveMonitor& GetExclusiveMonitor();
-    const ExclusiveMonitor& GetExclusiveMonitor() const;
-
-    void RunLoop(bool tight_loop);
-
-    void InvalidateAllInstructionCaches();
-
-private:
-    static constexpr std::size_t NUM_CPU_CORES = 4;
-
-    std::unique_ptr<ExclusiveMonitor> exclusive_monitor;
-    std::unique_ptr<CpuBarrier> barrier;
-    std::array<std::unique_ptr<Cpu>, NUM_CPU_CORES> cores;
-    std::array<std::unique_ptr<std::thread>, NUM_CPU_CORES - 1> core_threads;
-    std::size_t active_core{}; ///< Active core, only used in single thread mode
-
-    /// Map of guest threads to CPU cores
-    std::map<std::thread::id, Cpu*> thread_to_cpu;
-
-    System& system;
-};
-
-} // namespace Core

src/core/cpu_manager.cpp

@@ -0,0 +1,83 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/assert.h"
+#include "core/arm/exclusive_monitor.h"
+#include "core/core.h"
+#include "core/core_manager.h"
+#include "core/core_timing.h"
+#include "core/cpu_manager.h"
+#include "core/gdbstub/gdbstub.h"
+#include "core/settings.h"
+
+namespace Core {
+
+CpuManager::CpuManager(System& system) : system{system} {}
+CpuManager::~CpuManager() = default;
+
+void CpuManager::Initialize() {
+    for (std::size_t index = 0; index < core_managers.size(); ++index) {
+        core_managers[index] = std::make_unique<CoreManager>(system, index);
+    }
+}
+
+void CpuManager::Shutdown() {
+    for (auto& cpu_core : core_managers) {
+        cpu_core.reset();
+    }
+}
+
+CoreManager& CpuManager::GetCoreManager(std::size_t index) {
+    return *core_managers.at(index);
+}
+
+const CoreManager& CpuManager::GetCoreManager(std::size_t index) const {
+    return *core_managers.at(index);
+}
+
+CoreManager& CpuManager::GetCurrentCoreManager() {
+    // Otherwise, use single-threaded mode active_core variable
+    return *core_managers[active_core];
+}
+
+const CoreManager& CpuManager::GetCurrentCoreManager() const {
+    // Otherwise, use single-threaded mode active_core variable
+    return *core_managers[active_core];
+}
+
+void CpuManager::RunLoop(bool tight_loop) {
+    if (GDBStub::IsServerEnabled()) {
+        GDBStub::HandlePacket();
+
+        // If the loop is halted and we want to step, use a tiny (1) number of instructions to
+        // execute. Otherwise, get out of the loop function.
+        if (GDBStub::GetCpuHaltFlag()) {
+            if (GDBStub::GetCpuStepFlag()) {
+                tight_loop = false;
+            } else {
+                return;
+            }
+        }
+    }
+
+    auto& core_timing = system.CoreTiming();
+    core_timing.ResetRun();
+    bool keep_running{};
+    do {
+        keep_running = false;
+        for (active_core = 0; active_core < NUM_CPU_CORES; ++active_core) {
+            core_timing.SwitchContext(active_core);
+            if (core_timing.CanCurrentContextRun()) {
+                core_managers[active_core]->RunLoop(tight_loop);
+            }
+            keep_running |= core_timing.CanCurrentContextRun();
+        }
+    } while (keep_running);
+
+    if (GDBStub::IsServerEnabled()) {
+        GDBStub::SetCpuStepFlag(false);
+    }
+}
+
+} // namespace Core

src/core/cpu_manager.h

@@ -0,0 +1,50 @@
+// Copyright 2018 yuzu emulator team
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <memory>
+
+namespace Core {
+
+class CoreManager;
+class System;
+
+class CpuManager {
+public:
+    explicit CpuManager(System& system);
+    CpuManager(const CpuManager&) = delete;
+    CpuManager(CpuManager&&) = delete;
+
+    ~CpuManager();
+
+    CpuManager& operator=(const CpuManager&) = delete;
+    CpuManager& operator=(CpuManager&&) = delete;
+
+    void Initialize();
+    void Shutdown();
+
+    CoreManager& GetCoreManager(std::size_t index);
+    const CoreManager& GetCoreManager(std::size_t index) const;
+
+    CoreManager& GetCurrentCoreManager();
+    const CoreManager& GetCurrentCoreManager() const;
+
+    std::size_t GetActiveCoreIndex() const {
+        return active_core;
+    }
+
+    void RunLoop(bool tight_loop);
+
+private:
+    static constexpr std::size_t NUM_CPU_CORES = 4;
+
+    std::array<std::unique_ptr<CoreManager>, NUM_CPU_CORES> core_managers;
+    std::size_t active_core{}; ///< Active core, only used in single thread mode
+
+    System& system;
+};
+
+} // namespace Core

src/core/gdbstub/gdbstub.cpp

@@ -35,7 +35,7 @@
 #include "common/swap.h"
 #include "core/arm/arm_interface.h"
 #include "core/core.h"
-#include "core/core_cpu.h"
+#include "core/core_manager.h"
 #include "core/gdbstub/gdbstub.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/scheduler.h"

src/core/hle/kernel/address_arbiter.cpp

@@ -8,7 +8,6 @@
 #include "common/assert.h"
 #include "common/common_types.h"
 #include "core/core.h"
-#include "core/core_cpu.h"
 #include "core/hle/kernel/address_arbiter.h"
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/scheduler.h"

src/core/hle/kernel/kernel.cpp

@@ -3,13 +3,15 @@
 // Refer to the license.txt file included.
 
 #include <atomic>
+#include <functional>
 #include <memory>
 #include <mutex>
 #include <utility>
 
 #include "common/assert.h"
 #include "common/logging/log.h"
-
+#include "core/arm/arm_interface.h"
+#include "core/arm/exclusive_monitor.h"
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
@@ -17,6 +19,7 @@
 #include "core/hle/kernel/errors.h"
 #include "core/hle/kernel/handle_table.h"
 #include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/physical_core.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/resource_limit.h"
 #include "core/hle/kernel/scheduler.h"
@@ -98,6 +101,7 @@ struct KernelCore::Impl {
     void Initialize(KernelCore& kernel) {
         Shutdown();
 
+        InitializePhysicalCores(kernel);
         InitializeSystemResourceLimit(kernel);
         InitializeThreads();
         InitializePreemption();
@@ -121,6 +125,21 @@ struct KernelCore::Impl {
         global_scheduler.Shutdown();
 
         named_ports.clear();
+
+        for (auto& core : cores) {
+            core.Shutdown();
+        }
+        cores.clear();
+
+        exclusive_monitor.reset(nullptr);
+    }
+
+    void InitializePhysicalCores(KernelCore& kernel) {
+        exclusive_monitor =
+            Core::MakeExclusiveMonitor(system.Memory(), global_scheduler.CpuCoresCount());
+        for (std::size_t i = 0; i < global_scheduler.CpuCoresCount(); i++) {
+            cores.emplace_back(system, kernel, i, *exclusive_monitor);
+        }
     }
 
     // Creates the default system resource limit
@@ -186,6 +205,9 @@ struct KernelCore::Impl {
     /// the ConnectToPort SVC.
     NamedPortTable named_ports;
 
+    std::unique_ptr<Core::ExclusiveMonitor> exclusive_monitor;
+    std::vector<Kernel::PhysicalCore> cores;
+
     // System context
     Core::System& system;
 };
@@ -240,6 +262,34 @@ const Kernel::GlobalScheduler& KernelCore::GlobalScheduler() const {
     return impl->global_scheduler;
 }
 
+Kernel::PhysicalCore& KernelCore::PhysicalCore(std::size_t id) {
+    return impl->cores[id];
+}
+
+const Kernel::PhysicalCore& KernelCore::PhysicalCore(std::size_t id) const {
+    return impl->cores[id];
+}
+
+Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() {
+    return *impl->exclusive_monitor;
+}
+
+const Core::ExclusiveMonitor& KernelCore::GetExclusiveMonitor() const {
+    return *impl->exclusive_monitor;
+}
+
+void KernelCore::InvalidateAllInstructionCaches() {
+    for (std::size_t i = 0; i < impl->global_scheduler.CpuCoresCount(); i++) {
+        PhysicalCore(i).ArmInterface().ClearInstructionCache();
+    }
+}
+
+void KernelCore::PrepareReschedule(std::size_t id) {
+    if (id < impl->global_scheduler.CpuCoresCount()) {
+        impl->cores[id].Stop();
+    }
+}
+
 void KernelCore::AddNamedPort(std::string name, std::shared_ptr<ClientPort> port) {
     impl->named_ports.emplace(std::move(name), std::move(port));
 }

src/core/hle/kernel/kernel.h

@@ -11,8 +11,9 @@
 #include "core/hle/kernel/object.h"
 
 namespace Core {
+class ExclusiveMonitor;
 class System;
-}
+} // namespace Core
 
 namespace Core::Timing {
 class CoreTiming;
@@ -25,6 +26,7 @@ class AddressArbiter;
 class ClientPort;
 class GlobalScheduler;
 class HandleTable;
+class PhysicalCore;
 class Process;
 class ResourceLimit;
 class Thread;
@@ -84,6 +86,21 @@ public:
     /// Gets the sole instance of the global scheduler
     const Kernel::GlobalScheduler& GlobalScheduler() const;
 
+    /// Gets the an instance of the respective physical CPU core.
+    Kernel::PhysicalCore& PhysicalCore(std::size_t id);
+
+    /// Gets the an instance of the respective physical CPU core.
+    const Kernel::PhysicalCore& PhysicalCore(std::size_t id) const;
+
+    /// Stops execution of 'id' core, in order to reschedule a new thread.
+    void PrepareReschedule(std::size_t id);
+
+    Core::ExclusiveMonitor& GetExclusiveMonitor();
+
+    const Core::ExclusiveMonitor& GetExclusiveMonitor() const;
+
+    void InvalidateAllInstructionCaches();
+
     /// Adds a port to the named port table
     void AddNamedPort(std::string name, std::shared_ptr<ClientPort> port);
 

src/core/hle/kernel/physical_core.cpp

@@ -0,0 +1,52 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/logging/log.h"
+#include "core/arm/arm_interface.h"
+#ifdef ARCHITECTURE_x86_64
+#include "core/arm/dynarmic/arm_dynarmic.h"
+#endif
+#include "core/arm/exclusive_monitor.h"
+#include "core/arm/unicorn/arm_unicorn.h"
+#include "core/core.h"
+#include "core/hle/kernel/kernel.h"
+#include "core/hle/kernel/physical_core.h"
+#include "core/hle/kernel/scheduler.h"
+#include "core/hle/kernel/thread.h"
+
+namespace Kernel {
+
+PhysicalCore::PhysicalCore(Core::System& system, KernelCore& kernel, std::size_t id,
+                           Core::ExclusiveMonitor& exclusive_monitor)
+    : core_index{id}, kernel{kernel} {
+#ifdef ARCHITECTURE_x86_64
+    arm_interface = std::make_shared<Core::ARM_Dynarmic>(system, exclusive_monitor, core_index);
+#else
+    arm_interface = std::make_shared<Core::ARM_Unicorn>(system);
+    LOG_WARNING(Core, "CPU JIT requested, but Dynarmic not available");
+#endif
+
+    scheduler = std::make_shared<Kernel::Scheduler>(system, *arm_interface, core_index);
+}
+
+PhysicalCore::~PhysicalCore() = default;
+
+void PhysicalCore::Run() {
+    arm_interface->Run();
+    arm_interface->ClearExclusiveState();
+}
+
+void PhysicalCore::Step() {
+    arm_interface->Step();
+}
+
+void PhysicalCore::Stop() {
+    arm_interface->PrepareReschedule();
+}
+
+void PhysicalCore::Shutdown() {
+    scheduler->Shutdown();
+}
+
+} // namespace Kernel

src/core/hle/kernel/physical_core.h

@@ -0,0 +1,74 @@
+// Copyright 2020 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <cstddef>
+#include <memory>
+
+namespace Kernel {
+class Scheduler;
+} // namespace Kernel
+
+namespace Core {
+class ARM_Interface;
+class ExclusiveMonitor;
+class System;
+} // namespace Core
+
+namespace Kernel {
+
+class PhysicalCore {
+public:
+    PhysicalCore(Core::System& system, KernelCore& kernel, std::size_t id,
+                 Core::ExclusiveMonitor& exclusive_monitor);
+
+    ~PhysicalCore();
+
+    /// Execute current jit state
+    void Run();
+    /// Execute a single instruction in current jit.
+    void Step();
+    /// Stop JIT execution/exit
+    void Stop();
+
+    // Shutdown this physical core.
+    void Shutdown();
+
+    Core::ARM_Interface& ArmInterface() {
+        return *arm_interface;
+    }
+
+    const Core::ARM_Interface& ArmInterface() const {
+        return *arm_interface;
+    }
+
+    bool IsMainCore() const {
+        return core_index == 0;
+    }
+
+    bool IsSystemCore() const {
+        return core_index == 3;
+    }
+
+    std::size_t CoreIndex() const {
+        return core_index;
+    }
+
+    Kernel::Scheduler& Scheduler() {
+        return *scheduler;
+    }
+
+    const Kernel::Scheduler& Scheduler() const {
+        return *scheduler;
+    }
+
+private:
+    std::size_t core_index;
+    KernelCore& kernel;
+    std::shared_ptr<Core::ARM_Interface> arm_interface;
+    std::shared_ptr<Kernel::Scheduler> scheduler;
+};
+
+} // namespace Kernel

src/core/hle/kernel/physical_memory.h

@@ -14,6 +14,9 @@ namespace Kernel {
 // - Second to ensure all host backing memory used is aligned to 256 bytes due
 // to strict alignment restrictions on GPU memory.
 
-using PhysicalMemory = std::vector<u8, Common::AlignmentAllocator<u8, 256>>;
+using PhysicalMemoryVector = std::vector<u8, Common::AlignmentAllocator<u8, 256>>;
+class PhysicalMemory final : public PhysicalMemoryVector {
+    using PhysicalMemoryVector::PhysicalMemoryVector;
+};
 
 } // namespace Kernel

src/core/hle/kernel/process.cpp

@@ -317,6 +317,8 @@ void Process::FreeTLSRegion(VAddr tls_address) {
 }
 
 void Process::LoadModule(CodeSet module_, VAddr base_addr) {
+    code_memory_size += module_.memory.size();
+
     const auto memory = std::make_shared<PhysicalMemory>(std::move(module_.memory));
 
     const auto MapSegment = [&](const CodeSet::Segment& segment, VMAPermission permissions,
@@ -332,8 +334,6 @@ void Process::LoadModule(CodeSet module_, VAddr base_addr) {
     MapSegment(module_.CodeSegment(), VMAPermission::ReadExecute, MemoryState::Code);
     MapSegment(module_.RODataSegment(), VMAPermission::Read, MemoryState::CodeData);
     MapSegment(module_.DataSegment(), VMAPermission::ReadWrite, MemoryState::CodeData);
-
-    code_memory_size += module_.memory.size();
 }
 
 Process::Process(Core::System& system)

src/core/hle/kernel/scheduler.cpp

@@ -14,7 +14,6 @@
 #include "common/logging/log.h"
 #include "core/arm/arm_interface.h"
 #include "core/core.h"
-#include "core/core_cpu.h"
 #include "core/core_timing.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"

src/core/hle/kernel/svc.cpp

@@ -15,7 +15,7 @@
 #include "common/string_util.h"
 #include "core/arm/exclusive_monitor.h"
 #include "core/core.h"
-#include "core/core_cpu.h"
+#include "core/core_manager.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/hle/kernel/address_arbiter.h"

src/core/hle/kernel/thread.cpp

@@ -13,7 +13,6 @@
 #include "common/thread_queue_list.h"
 #include "core/arm/arm_interface.h"
 #include "core/core.h"
-#include "core/core_cpu.h"
 #include "core/core_timing.h"
 #include "core/core_timing_util.h"
 #include "core/hle/kernel/errors.h"
@@ -356,7 +355,7 @@ void Thread::SetActivity(ThreadActivity value) {
         // Set status if not waiting
         if (status == ThreadStatus::Ready || status == ThreadStatus::Running) {
             SetStatus(ThreadStatus::Paused);
-            Core::System::GetInstance().CpuCore(processor_id).PrepareReschedule();
+            kernel.PrepareReschedule(processor_id);
         }
     } else if (status == ThreadStatus::Paused) {
         // Ready to reschedule

src/core/hle/kernel/vm_manager.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <algorithm>
+#include <cstring>
 #include <iterator>
 #include <utility>
 #include "common/alignment.h"
@@ -269,18 +270,9 @@ ResultVal<VAddr> VMManager::SetHeapSize(u64 size) {
     // If necessary, expand backing vector to cover new heap extents in
     // the case of allocating. Otherwise, shrink the backing memory,
     // if a smaller heap has been requested.
-    const u64 old_heap_size = GetCurrentHeapSize();
-    if (size > old_heap_size) {
-        const u64 alloc_size = size - old_heap_size;
-
-        heap_memory->insert(heap_memory->end(), alloc_size, 0);
-        RefreshMemoryBlockMappings(heap_memory.get());
-    } else if (size < old_heap_size) {
-        heap_memory->resize(size);
-        heap_memory->shrink_to_fit();
-
-        RefreshMemoryBlockMappings(heap_memory.get());
-    }
+    heap_memory->resize(size);
+    heap_memory->shrink_to_fit();
+    RefreshMemoryBlockMappings(heap_memory.get());
 
     heap_end = heap_region_base + size;
     ASSERT(GetCurrentHeapSize() == heap_memory->size());
@@ -752,24 +744,20 @@ void VMManager::MergeAdjacentVMA(VirtualMemoryArea& left, const VirtualMemoryAre
     // Always merge allocated memory blocks, even when they don't share the same backing block.
     if (left.type == VMAType::AllocatedMemoryBlock &&
         (left.backing_block != right.backing_block || left.offset + left.size != right.offset)) {
-        const auto right_begin = right.backing_block->begin() + right.offset;
-        const auto right_end = right_begin + right.size;
 
         // Check if we can save work.
         if (left.offset == 0 && left.size == left.backing_block->size()) {
             // Fast case: left is an entire backing block.
-            left.backing_block->insert(left.backing_block->end(), right_begin, right_end);
+            left.backing_block->resize(left.size + right.size);
+            std::memcpy(left.backing_block->data() + left.size,
+                        right.backing_block->data() + right.offset, right.size);
         } else {
             // Slow case: make a new memory block for left and right.
-            const auto left_begin = left.backing_block->begin() + left.offset;
-            const auto left_end = left_begin + left.size;
-            const auto left_size = static_cast<std::size_t>(std::distance(left_begin, left_end));
-            const auto right_size = static_cast<std::size_t>(std::distance(right_begin, right_end));
-
             auto new_memory = std::make_shared<PhysicalMemory>();
-            new_memory->reserve(left_size + right_size);
-            new_memory->insert(new_memory->end(), left_begin, left_end);
-            new_memory->insert(new_memory->end(), right_begin, right_end);
+            new_memory->resize(left.size + right.size);
+            std::memcpy(new_memory->data(), left.backing_block->data() + left.offset, left.size);
+            std::memcpy(new_memory->data() + left.size, right.backing_block->data() + right.offset,
+                        right.size);
 
             left.backing_block = std::move(new_memory);
             left.offset = 0;
@@ -792,8 +780,7 @@ void VMManager::UpdatePageTableForVMA(const VirtualMemoryArea& vma) {
         memory.UnmapRegion(page_table, vma.base, vma.size);
         break;
     case VMAType::AllocatedMemoryBlock:
-        memory.MapMemoryRegion(page_table, vma.base, vma.size,
-                               vma.backing_block->data() + vma.offset);
+        memory.MapMemoryRegion(page_table, vma.base, vma.size, *vma.backing_block, vma.offset);
         break;
     case VMAType::BackingMemory:
         memory.MapMemoryRegion(page_table, vma.base, vma.size, vma.backing_memory);

src/core/hle/kernel/wait_object.cpp

@@ -7,7 +7,6 @@
 #include "common/common_types.h"
 #include "common/logging/log.h"
 #include "core/core.h"
-#include "core/core_cpu.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/object.h"
 #include "core/hle/kernel/process.h"
@@ -96,7 +95,7 @@ void WaitObject::WakeupWaitingThread(std::shared_ptr<Thread> thread) {
     }
     if (resume) {
         thread->ResumeFromWait();
-        Core::System::GetInstance().PrepareReschedule(thread->GetProcessorID());
+        kernel.PrepareReschedule(thread->GetProcessorID());
     }
 }
 

src/core/hle/service/nifm/nifm.cpp

@@ -9,6 +9,7 @@
 #include "core/hle/kernel/writable_event.h"
 #include "core/hle/service/nifm/nifm.h"
 #include "core/hle/service/service.h"
+#include "core/settings.h"
 
 namespace Service::NIFM {
 
@@ -86,7 +87,12 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(RESULT_SUCCESS);
-        rb.PushEnum(RequestState::Connected);
+
+        if (Settings::values.bcat_backend == "none") {
+            rb.PushEnum(RequestState::NotSubmitted);
+        } else {
+            rb.PushEnum(RequestState::Connected);
+        }
     }
 
     void GetResult(Kernel::HLERequestContext& ctx) {
@@ -194,14 +200,22 @@ private:
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(RESULT_SUCCESS);
-        rb.Push<u8>(1);
+        if (Settings::values.bcat_backend == "none") {
+            rb.Push<u8>(0);
+        } else {
+            rb.Push<u8>(1);
+        }
     }
     void IsAnyInternetRequestAccepted(Kernel::HLERequestContext& ctx) {
         LOG_WARNING(Service_NIFM, "(STUBBED) called");
 
         IPC::ResponseBuilder rb{ctx, 3};
         rb.Push(RESULT_SUCCESS);
-        rb.Push<u8>(1);
+        if (Settings::values.bcat_backend == "none") {
+            rb.Push<u8>(0);
+        } else {
+            rb.Push<u8>(1);
+        }
     }
     Core::System& system;
 };

src/core/hle/service/nvflinger/nvflinger.cpp

@@ -88,6 +88,12 @@ std::optional<u64> NVFlinger::CreateLayer(u64 display_id) {
     return layer_id;
 }
 
+void NVFlinger::CloseLayer(u64 layer_id) {
+    for (auto& display : displays) {
+        display.CloseLayer(layer_id);
+    }
+}
+
 std::optional<u32> NVFlinger::FindBufferQueueId(u64 display_id, u64 layer_id) const {
     const auto* const layer = FindLayer(display_id, layer_id);
 
@@ -192,7 +198,7 @@ void NVFlinger::Compose() {
 
         const auto& igbp_buffer = buffer->get().igbp_buffer;
 
-        const auto& gpu = system.GPU();
+        auto& gpu = system.GPU();
         const auto& multi_fence = buffer->get().multi_fence;
         for (u32 fence_id = 0; fence_id < multi_fence.num_fences; fence_id++) {
             const auto& fence = multi_fence.fences[fence_id];

src/core/hle/service/nvflinger/nvflinger.h

@@ -54,6 +54,9 @@ public:
     /// If an invalid display ID is specified, then an empty optional is returned.
     std::optional<u64> CreateLayer(u64 display_id);
 
+    /// Closes a layer on all displays for the given layer ID.
+    void CloseLayer(u64 layer_id);
+
     /// Finds the buffer queue ID of the specified layer in the specified display.
     ///
     /// If an invalid display ID or layer ID is provided, then an empty optional is returned.

src/core/hle/service/sockets/bsd.cpp

@@ -42,6 +42,26 @@ void BSD::Socket(Kernel::HLERequestContext& ctx) {
     rb.Push<u32>(0); // bsd errno
 }
 
+void BSD::Select(Kernel::HLERequestContext& ctx) {
+    LOG_WARNING(Service, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 4};
+
+    rb.Push(RESULT_SUCCESS);
+    rb.Push<u32>(0); // ret
+    rb.Push<u32>(0); // bsd errno
+}
+
+void BSD::Bind(Kernel::HLERequestContext& ctx) {
+    LOG_WARNING(Service, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 4};
+
+    rb.Push(RESULT_SUCCESS);
+    rb.Push<u32>(0); // ret
+    rb.Push<u32>(0); // bsd errno
+}
+
 void BSD::Connect(Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service, "(STUBBED) called");
 
@@ -52,6 +72,26 @@ void BSD::Connect(Kernel::HLERequestContext& ctx) {
     rb.Push<u32>(0); // bsd errno
 }
 
+void BSD::Listen(Kernel::HLERequestContext& ctx) {
+    LOG_WARNING(Service, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 4};
+
+    rb.Push(RESULT_SUCCESS);
+    rb.Push<u32>(0); // ret
+    rb.Push<u32>(0); // bsd errno
+}
+
+void BSD::SetSockOpt(Kernel::HLERequestContext& ctx) {
+    LOG_WARNING(Service, "(STUBBED) called");
+
+    IPC::ResponseBuilder rb{ctx, 4};
+
+    rb.Push(RESULT_SUCCESS);
+    rb.Push<u32>(0); // ret
+    rb.Push<u32>(0); // bsd errno
+}
+
 void BSD::SendTo(Kernel::HLERequestContext& ctx) {
     LOG_WARNING(Service, "(STUBBED) called");
 
@@ -80,7 +120,7 @@ BSD::BSD(const char* name) : ServiceFramework(name) {
         {2, &BSD::Socket, "Socket"},
         {3, nullptr, "SocketExempt"},
         {4, nullptr, "Open"},
-        {5, nullptr, "Select"},
+        {5, &BSD::Select, "Select"},
         {6, nullptr, "Poll"},
         {7, nullptr, "Sysctl"},
         {8, nullptr, "Recv"},
@@ -88,15 +128,15 @@ BSD::BSD(const char* name) : ServiceFramework(name) {
         {10, nullptr, "Send"},
         {11, &BSD::SendTo, "SendTo"},
         {12, nullptr, "Accept"},
-        {13, nullptr, "Bind"},
+        {13, &BSD::Bind, "Bind"},
         {14, &BSD::Connect, "Connect"},
         {15, nullptr, "GetPeerName"},
         {16, nullptr, "GetSockName"},
         {17, nullptr, "GetSockOpt"},
-        {18, nullptr, "Listen"},
+        {18, &BSD::Listen, "Listen"},
         {19, nullptr, "Ioctl"},
         {20, nullptr, "Fcntl"},
-        {21, nullptr, "SetSockOpt"},
+        {21, &BSD::SetSockOpt, "SetSockOpt"},
         {22, nullptr, "Shutdown"},
         {23, nullptr, "ShutdownAllSockets"},
         {24, nullptr, "Write"},

src/core/hle/service/sockets/bsd.h

@@ -18,7 +18,11 @@ private:
     void RegisterClient(Kernel::HLERequestContext& ctx);
     void StartMonitoring(Kernel::HLERequestContext& ctx);
     void Socket(Kernel::HLERequestContext& ctx);
+    void Select(Kernel::HLERequestContext& ctx);
+    void Bind(Kernel::HLERequestContext& ctx);
     void Connect(Kernel::HLERequestContext& ctx);
+    void Listen(Kernel::HLERequestContext& ctx);
+    void SetSockOpt(Kernel::HLERequestContext& ctx);
     void SendTo(Kernel::HLERequestContext& ctx);
     void Close(Kernel::HLERequestContext& ctx);
 

src/core/hle/service/time/time_zone_manager.cpp

@@ -820,7 +820,7 @@ static ResultCode ToCalendarTimeImpl(const TimeZoneRule& rules, s64 time, Calend
     const ResultCode result{
         ToCalendarTimeInternal(rules, time, calendar_time, calendar.additiona_info)};
     calendar.time.year = static_cast<s16>(calendar_time.year);
-    calendar.time.month = calendar_time.month;
+    calendar.time.month = calendar_time.month + 1; // Internal impl. uses 0-indexed month
     calendar.time.day = calendar_time.day;
     calendar.time.hour = calendar_time.hour;
     calendar.time.minute = calendar_time.minute;
@@ -874,7 +874,7 @@ ResultCode TimeZoneManager::ToPosixTime(const TimeZoneRule& rules,
 
     CalendarTimeInternal internal_time{};
     internal_time.year = calendar_time.year;
-    internal_time.month = calendar_time.month;
+    internal_time.month = calendar_time.month - 1; // Internal impl. uses 0-indexed month
     internal_time.day = calendar_time.day;
     internal_time.hour = calendar_time.hour;
     internal_time.minute = calendar_time.minute;
@@ -1019,6 +1019,15 @@ ResultCode TimeZoneManager::ToPosixTime(const TimeZoneRule& rules,
     return RESULT_SUCCESS;
 }
 
+ResultCode TimeZoneManager::ToPosixTimeWithMyRule(const CalendarTime& calendar_time,
+                                                  s64& posix_time) const {
+    if (is_initialized) {
+        return ToPosixTime(time_zone_rule, calendar_time, posix_time);
+    }
+    posix_time = 0;
+    return ERROR_UNINITIALIZED_CLOCK;
+}
+
 ResultCode TimeZoneManager::GetDeviceLocationName(LocationName& value) const {
     if (!is_initialized) {
         return ERROR_UNINITIALIZED_CLOCK;

src/core/hle/service/time/time_zone_manager.h

@@ -39,6 +39,7 @@ public:
     ResultCode ParseTimeZoneRuleBinary(TimeZoneRule& rules, FileSys::VirtualFile& vfs_file) const;
     ResultCode ToPosixTime(const TimeZoneRule& rules, const CalendarTime& calendar_time,
                            s64& posix_time) const;
+    ResultCode ToPosixTimeWithMyRule(const CalendarTime& calendar_time, s64& posix_time) const;
 
 private:
     bool is_initialized{};

src/core/hle/service/time/time_zone_service.cpp

@@ -22,7 +22,7 @@ ITimeZoneService ::ITimeZoneService(TimeZone::TimeZoneContentManager& time_zone_
         {100, &ITimeZoneService::ToCalendarTime, "ToCalendarTime"},
         {101, &ITimeZoneService::ToCalendarTimeWithMyRule, "ToCalendarTimeWithMyRule"},
         {201, &ITimeZoneService::ToPosixTime, "ToPosixTime"},
-        {202, nullptr, "ToPosixTimeWithMyRule"},
+        {202, &ITimeZoneService::ToPosixTimeWithMyRule, "ToPosixTimeWithMyRule"},
     };
     RegisterHandlers(functions);
 }
@@ -145,4 +145,26 @@ void ITimeZoneService::ToPosixTime(Kernel::HLERequestContext& ctx) {
     ctx.WriteBuffer(&posix_time, sizeof(s64));
 }
 
+void ITimeZoneService::ToPosixTimeWithMyRule(Kernel::HLERequestContext& ctx) {
+    LOG_DEBUG(Service_Time, "called");
+
+    IPC::RequestParser rp{ctx};
+    const auto calendar_time{rp.PopRaw<TimeZone::CalendarTime>()};
+
+    s64 posix_time{};
+    if (const ResultCode result{
+            time_zone_content_manager.GetTimeZoneManager().ToPosixTimeWithMyRule(calendar_time,
+                                                                                 posix_time)};
+        result != RESULT_SUCCESS) {
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(result);
+        return;
+    }
+
+    IPC::ResponseBuilder rb{ctx, 3};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushRaw<u32>(1); // Number of times we're returning
+    ctx.WriteBuffer(&posix_time, sizeof(s64));
+}
+
 } // namespace Service::Time

src/core/hle/service/time/time_zone_service.h

@@ -22,6 +22,7 @@ private:
     void ToCalendarTime(Kernel::HLERequestContext& ctx);
     void ToCalendarTimeWithMyRule(Kernel::HLERequestContext& ctx);
     void ToPosixTime(Kernel::HLERequestContext& ctx);
+    void ToPosixTimeWithMyRule(Kernel::HLERequestContext& ctx);
 
 private:
     TimeZone::TimeZoneContentManager& time_zone_content_manager;

src/core/hle/service/vi/display/vi_display.cpp

@@ -24,11 +24,11 @@ Display::Display(u64 id, std::string name, Core::System& system) : id{id}, name{
 Display::~Display() = default;
 
 Layer& Display::GetLayer(std::size_t index) {
-    return layers.at(index);
+    return *layers.at(index);
 }
 
 const Layer& Display::GetLayer(std::size_t index) const {
-    return layers.at(index);
+    return *layers.at(index);
 }
 
 std::shared_ptr<Kernel::ReadableEvent> Display::GetVSyncEvent() const {
@@ -43,29 +43,38 @@ void Display::CreateLayer(u64 id, NVFlinger::BufferQueue& buffer_queue) {
     // TODO(Subv): Support more than 1 layer.
     ASSERT_MSG(layers.empty(), "Only one layer is supported per display at the moment");
 
-    layers.emplace_back(id, buffer_queue);
+    layers.emplace_back(std::make_shared<Layer>(id, buffer_queue));
+}
+
+void Display::CloseLayer(u64 id) {
+    layers.erase(
+        std::remove_if(layers.begin(), layers.end(),
+                       [id](const std::shared_ptr<Layer>& layer) { return layer->GetID() == id; }),
+        layers.end());
 }
 
 Layer* Display::FindLayer(u64 id) {
-    const auto itr = std::find_if(layers.begin(), layers.end(),
-                                  [id](const VI::Layer& layer) { return layer.GetID() == id; });
+    const auto itr =
+        std::find_if(layers.begin(), layers.end(),
+                     [id](const std::shared_ptr<Layer>& layer) { return layer->GetID() == id; });
 
     if (itr == layers.end()) {
         return nullptr;
     }
 
-    return &*itr;
+    return itr->get();
 }
 
 const Layer* Display::FindLayer(u64 id) const {
-    const auto itr = std::find_if(layers.begin(), layers.end(),
-                                  [id](const VI::Layer& layer) { return layer.GetID() == id; });
+    const auto itr =
+        std::find_if(layers.begin(), layers.end(),
+                     [id](const std::shared_ptr<Layer>& layer) { return layer->GetID() == id; });
 
     if (itr == layers.end()) {
         return nullptr;
     }
 
-    return &*itr;
+    return itr->get();
 }
 
 } // namespace Service::VI

src/core/hle/service/vi/display/vi_display.h

@@ -4,6 +4,7 @@
 
 #pragma once
 
+#include <memory>
 #include <string>
 #include <vector>
 
@@ -69,6 +70,12 @@ public:
     ///
     void CreateLayer(u64 id, NVFlinger::BufferQueue& buffer_queue);
 
+    /// Closes and removes a layer from this display with the given ID.
+    ///
+    /// @param id           The ID assigned to the layer to close.
+    ///
+    void CloseLayer(u64 id);
+
     /// Attempts to find a layer with the given ID.
     ///
     /// @param id The layer ID.
@@ -91,7 +98,7 @@ private:
     u64 id;
     std::string name;
 
-    std::vector<Layer> layers;
+    std::vector<std::shared_ptr<Layer>> layers;
     Kernel::EventPair vsync_event;
 };
 

src/core/hle/service/vi/vi.cpp

@@ -1066,6 +1066,18 @@ private:
         rb.Push<u64>(ctx.WriteBuffer(native_window.Serialize()));
     }
 
+    void CloseLayer(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        const auto layer_id{rp.Pop<u64>()};
+
+        LOG_DEBUG(Service_VI, "called. layer_id=0x{:016X}", layer_id);
+
+        nv_flinger->CloseLayer(layer_id);
+
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+    }
+
     void CreateStrayLayer(Kernel::HLERequestContext& ctx) {
         IPC::RequestParser rp{ctx};
         const u32 flags = rp.Pop<u32>();
@@ -1178,7 +1190,7 @@ IApplicationDisplayService::IApplicationDisplayService(
         {1101, &IApplicationDisplayService::SetDisplayEnabled, "SetDisplayEnabled"},
         {1102, &IApplicationDisplayService::GetDisplayResolution, "GetDisplayResolution"},
         {2020, &IApplicationDisplayService::OpenLayer, "OpenLayer"},
-        {2021, nullptr, "CloseLayer"},
+        {2021, &IApplicationDisplayService::CloseLayer, "CloseLayer"},
         {2030, &IApplicationDisplayService::CreateStrayLayer, "CreateStrayLayer"},
         {2031, &IApplicationDisplayService::DestroyStrayLayer, "DestroyStrayLayer"},
         {2101, &IApplicationDisplayService::SetLayerScalingMode, "SetLayerScalingMode"},

src/core/loader/elf.cpp

@@ -335,7 +335,8 @@ Kernel::CodeSet ElfReader::LoadInto(VAddr vaddr) {
             codeset_segment->addr = segment_addr;
             codeset_segment->size = aligned_size;
 
-            memcpy(&program_image[current_image_position], GetSegmentPtr(i), p->p_filesz);
+            std::memcpy(program_image.data() + current_image_position, GetSegmentPtr(i),
+                        p->p_filesz);
             current_image_position += aligned_size;
         }
     }

src/core/loader/kip.cpp

@@ -2,6 +2,7 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <cstring>
 #include "core/file_sys/kernel_executable.h"
 #include "core/file_sys/program_metadata.h"
 #include "core/gdbstub/gdbstub.h"
@@ -76,8 +77,8 @@ AppLoader::LoadResult AppLoader_KIP::Load(Kernel::Process& process) {
         segment.addr = offset;
         segment.offset = offset;
         segment.size = PageAlignSize(static_cast<u32>(data.size()));
-        program_image.resize(offset);
-        program_image.insert(program_image.end(), data.begin(), data.end());
+        program_image.resize(offset + data.size());
+        std::memcpy(program_image.data() + offset, data.data(), data.size());
     };
 
     load_segment(codeset.CodeSegment(), kip->GetTextSection(), kip->GetTextOffset());

src/core/loader/nso.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <cinttypes>
+#include <cstring>
 #include <vector>
 
 #include "common/common_funcs.h"
@@ -96,15 +97,21 @@ std::optional<VAddr> AppLoader_NSO::LoadModule(Kernel::Process& process,
         if (nso_header.IsSegmentCompressed(i)) {
             data = DecompressSegment(data, nso_header.segments[i]);
         }
-        program_image.resize(nso_header.segments[i].location);
-        program_image.insert(program_image.end(), data.begin(), data.end());
+        program_image.resize(nso_header.segments[i].location +
+                             PageAlignSize(static_cast<u32>(data.size())));
+        std::memcpy(program_image.data() + nso_header.segments[i].location, data.data(),
+                    data.size());
         codeset.segments[i].addr = nso_header.segments[i].location;
         codeset.segments[i].offset = nso_header.segments[i].location;
         codeset.segments[i].size = PageAlignSize(static_cast<u32>(data.size()));
     }
 
-    if (should_pass_arguments && !Settings::values.program_args.empty()) {
-        const auto arg_data = Settings::values.program_args;
+    if (should_pass_arguments) {
+        std::vector<u8> arg_data{Settings::values.program_args.begin(),
+                                 Settings::values.program_args.end()};
+        if (arg_data.empty()) {
+            arg_data.resize(NSO_ARGUMENT_DEFAULT_SIZE);
+        }
         codeset.DataSegment().size += NSO_ARGUMENT_DATA_ALLOCATION_SIZE;
         NSOArgumentHeader args_header{
             NSO_ARGUMENT_DATA_ALLOCATION_SIZE, static_cast<u32_le>(arg_data.size()), {}};
@@ -139,12 +146,12 @@ std::optional<VAddr> AppLoader_NSO::LoadModule(Kernel::Process& process,
         std::vector<u8> pi_header;
         pi_header.insert(pi_header.begin(), reinterpret_cast<u8*>(&nso_header),
                          reinterpret_cast<u8*>(&nso_header) + sizeof(NSOHeader));
-        pi_header.insert(pi_header.begin() + sizeof(NSOHeader), program_image.begin(),
-                         program_image.end());
+        pi_header.insert(pi_header.begin() + sizeof(NSOHeader), program_image.data(),
+                         program_image.data() + program_image.size());
 
         pi_header = pm->PatchNSO(pi_header, file.GetName());
 
-        std::copy(pi_header.begin() + sizeof(NSOHeader), pi_header.end(), program_image.begin());
+        std::copy(pi_header.begin() + sizeof(NSOHeader), pi_header.end(), program_image.data());
     }
 
     // Apply cheats if they exist and the program has a valid title ID

src/core/loader/nso.h

@@ -56,6 +56,8 @@ static_assert(sizeof(NSOHeader) == 0x100, "NSOHeader has incorrect size.");
 static_assert(std::is_trivially_copyable_v<NSOHeader>, "NSOHeader must be trivially copyable.");
 
 constexpr u64 NSO_ARGUMENT_DATA_ALLOCATION_SIZE = 0x9000;
+// NOTE: Official software default argument state is unverified.
+constexpr u64 NSO_ARGUMENT_DEFAULT_SIZE = 1;
 
 struct NSOArgumentHeader {
     u32_le allocated_size;

src/core/memory.cpp

@@ -14,6 +14,7 @@
 #include "common/swap.h"
 #include "core/arm/arm_interface.h"
 #include "core/core.h"
+#include "core/hle/kernel/physical_memory.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/vm_manager.h"
 #include "core/memory.h"
@@ -38,6 +39,11 @@ struct Memory::Impl {
         system.ArmInterface(3).PageTableChanged(*current_page_table, address_space_width);
     }
 
+    void MapMemoryRegion(Common::PageTable& page_table, VAddr base, u64 size,
+                         Kernel::PhysicalMemory& memory, VAddr offset) {
+        MapMemoryRegion(page_table, base, size, memory.data() + offset);
+    }
+
     void MapMemoryRegion(Common::PageTable& page_table, VAddr base, u64 size, u8* target) {
         ASSERT_MSG((size & PAGE_MASK) == 0, "non-page aligned size: {:016X}", size);
         ASSERT_MSG((base & PAGE_MASK) == 0, "non-page aligned base: {:016X}", base);
@@ -146,7 +152,7 @@ struct Memory::Impl {
     u8* GetPointer(const VAddr vaddr) {
         u8* const page_pointer = current_page_table->pointers[vaddr >> PAGE_BITS];
         if (page_pointer != nullptr) {
-            return page_pointer + (vaddr & PAGE_MASK);
+            return page_pointer + vaddr;
         }
 
         if (current_page_table->attributes[vaddr >> PAGE_BITS] ==
@@ -229,7 +235,8 @@ struct Memory::Impl {
             case Common::PageType::Memory: {
                 DEBUG_ASSERT(page_table.pointers[page_index]);
 
-                const u8* const src_ptr = page_table.pointers[page_index] + page_offset;
+                const u8* const src_ptr =
+                    page_table.pointers[page_index] + page_offset + (page_index << PAGE_BITS);
                 std::memcpy(dest_buffer, src_ptr, copy_amount);
                 break;
             }
@@ -276,7 +283,8 @@ struct Memory::Impl {
             case Common::PageType::Memory: {
                 DEBUG_ASSERT(page_table.pointers[page_index]);
 
-                u8* const dest_ptr = page_table.pointers[page_index] + page_offset;
+                u8* const dest_ptr =
+                    page_table.pointers[page_index] + page_offset + (page_index << PAGE_BITS);
                 std::memcpy(dest_ptr, src_buffer, copy_amount);
                 break;
             }
@@ -322,7 +330,8 @@ struct Memory::Impl {
             case Common::PageType::Memory: {
                 DEBUG_ASSERT(page_table.pointers[page_index]);
 
-                u8* dest_ptr = page_table.pointers[page_index] + page_offset;
+                u8* dest_ptr =
+                    page_table.pointers[page_index] + page_offset + (page_index << PAGE_BITS);
                 std::memset(dest_ptr, 0, copy_amount);
                 break;
             }
@@ -368,7 +377,8 @@ struct Memory::Impl {
             }
             case Common::PageType::Memory: {
                 DEBUG_ASSERT(page_table.pointers[page_index]);
-                const u8* src_ptr = page_table.pointers[page_index] + page_offset;
+                const u8* src_ptr =
+                    page_table.pointers[page_index] + page_offset + (page_index << PAGE_BITS);
                 WriteBlock(process, dest_addr, src_ptr, copy_amount);
                 break;
             }
@@ -446,7 +456,8 @@ struct Memory::Impl {
                         page_type = Common::PageType::Unmapped;
                     } else {
                         page_type = Common::PageType::Memory;
-                        current_page_table->pointers[vaddr >> PAGE_BITS] = pointer;
+                        current_page_table->pointers[vaddr >> PAGE_BITS] =
+                            pointer - (vaddr & ~PAGE_MASK);
                     }
                     break;
                 }
@@ -493,7 +504,9 @@ struct Memory::Impl {
                       memory);
         } else {
             while (base != end) {
-                page_table.pointers[base] = memory;
+                page_table.pointers[base] = memory - (base << PAGE_BITS);
+                ASSERT_MSG(page_table.pointers[base],
+                           "memory mapping base yield a nullptr within the table");
 
                 base += 1;
                 memory += PAGE_SIZE;
@@ -518,7 +531,7 @@ struct Memory::Impl {
         if (page_pointer != nullptr) {
             // NOTE: Avoid adding any extra logic to this fast-path block
             T value;
-            std::memcpy(&value, &page_pointer[vaddr & PAGE_MASK], sizeof(T));
+            std::memcpy(&value, &page_pointer[vaddr], sizeof(T));
             return value;
         }
 
@@ -559,7 +572,7 @@ struct Memory::Impl {
         u8* const page_pointer = current_page_table->pointers[vaddr >> PAGE_BITS];
         if (page_pointer != nullptr) {
             // NOTE: Avoid adding any extra logic to this fast-path block
-            std::memcpy(&page_pointer[vaddr & PAGE_MASK], &data, sizeof(T));
+            std::memcpy(&page_pointer[vaddr], &data, sizeof(T));
             return;
         }
 
@@ -594,6 +607,11 @@ void Memory::SetCurrentPageTable(Kernel::Process& process) {
     impl->SetCurrentPageTable(process);
 }
 
+void Memory::MapMemoryRegion(Common::PageTable& page_table, VAddr base, u64 size,
+                             Kernel::PhysicalMemory& memory, VAddr offset) {
+    impl->MapMemoryRegion(page_table, base, size, memory, offset);
+}
+
 void Memory::MapMemoryRegion(Common::PageTable& page_table, VAddr base, u64 size, u8* target) {
     impl->MapMemoryRegion(page_table, base, size, target);
 }

src/core/memory.h

@@ -19,8 +19,9 @@ class System;
 }
 
 namespace Kernel {
+class PhysicalMemory;
 class Process;
-}
+} // namespace Kernel
 
 namespace Memory {
 
@@ -65,6 +66,19 @@ public:
      */
     void SetCurrentPageTable(Kernel::Process& process);
 
+    /**
+     * Maps an physical buffer onto a region of the emulated process address space.
+     *
+     * @param page_table The page table of the emulated process.
+     * @param base       The address to start mapping at. Must be page-aligned.
+     * @param size       The amount of bytes to map. Must be page-aligned.
+     * @param memory     Physical buffer with the memory backing the mapping. Must be of length
+     *                   at least `size + offset`.
+     * @param offset     The offset within the physical memory. Must be page-aligned.
+     */
+    void MapMemoryRegion(Common::PageTable& page_table, VAddr base, u64 size,
+                         Kernel::PhysicalMemory& memory, VAddr offset);
+
     /**
      * Maps an allocated buffer onto a region of the emulated process address space.
      *

src/core/settings.h

@@ -401,6 +401,9 @@ struct Values {
     std::string motion_device;
     TouchscreenInput touchscreen;
     std::atomic_bool is_device_reload_pending{true};
+    std::string udp_input_address;
+    u16 udp_input_port;
+    u8 udp_pad_index;
 
     // Core
     bool use_multi_core;

src/input_common/CMakeLists.txt

@@ -9,6 +9,12 @@ add_library(input_common STATIC
     motion_emu.h
     sdl/sdl.cpp
     sdl/sdl.h
+    udp/client.cpp
+    udp/client.h
+    udp/protocol.cpp
+    udp/protocol.h
+    udp/udp.cpp
+    udp/udp.h
 )
 
 if(SDL2_FOUND)
@@ -21,4 +27,4 @@ if(SDL2_FOUND)
 endif()
 
 create_target_directory_groups(input_common)
-target_link_libraries(input_common PUBLIC core PRIVATE common)
+target_link_libraries(input_common PUBLIC core PRIVATE common ${Boost_LIBRARIES})

src/input_common/main.cpp

@@ -9,6 +9,7 @@
 #include "input_common/keyboard.h"
 #include "input_common/main.h"
 #include "input_common/motion_emu.h"
+#include "input_common/udp/udp.h"
 #ifdef HAVE_SDL2
 #include "input_common/sdl/sdl.h"
 #endif
@@ -18,6 +19,7 @@ namespace InputCommon {
 static std::shared_ptr<Keyboard> keyboard;
 static std::shared_ptr<MotionEmu> motion_emu;
 static std::unique_ptr<SDL::State> sdl;
+static std::unique_ptr<CemuhookUDP::State> udp;
 
 void Init() {
     keyboard = std::make_shared<Keyboard>();
@@ -28,6 +30,8 @@ void Init() {
     Input::RegisterFactory<Input::MotionDevice>("motion_emu", motion_emu);
 
     sdl = SDL::Init();
+
+    udp = CemuhookUDP::Init();
 }
 
 void Shutdown() {
@@ -72,11 +76,13 @@ std::string GenerateAnalogParamFromKeys(int key_up, int key_down, int key_left,
 namespace Polling {
 
 std::vector<std::unique_ptr<DevicePoller>> GetPollers(DeviceType type) {
+    std::vector<std::unique_ptr<DevicePoller>> pollers;
+
 #ifdef HAVE_SDL2
-    return sdl->GetPollers(type);
-#else
-    return {};
+    pollers = sdl->GetPollers(type);
 #endif
+
+    return pollers;
 }
 
 } // namespace Polling

src/input_common/udp/client.cpp

@@ -0,0 +1,287 @@
+// Copyright 2018 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <array>
+#include <chrono>
+#include <cstring>
+#include <functional>
+#include <thread>
+#include <boost/asio.hpp>
+#include <boost/bind.hpp>
+#include "common/logging/log.h"
+#include "input_common/udp/client.h"
+#include "input_common/udp/protocol.h"
+
+using boost::asio::ip::address_v4;
+using boost::asio::ip::udp;
+
+namespace InputCommon::CemuhookUDP {
+
+struct SocketCallback {
+    std::function<void(Response::Version)> version;
+    std::function<void(Response::PortInfo)> port_info;
+    std::function<void(Response::PadData)> pad_data;
+};
+
+class Socket {
+public:
+    using clock = std::chrono::system_clock;
+
+    explicit Socket(const std::string& host, u16 port, u8 pad_index, u32 client_id,
+                    SocketCallback callback)
+        : client_id(client_id), timer(io_service),
+          send_endpoint(udp::endpoint(address_v4::from_string(host), port)),
+          socket(io_service, udp::endpoint(udp::v4(), 0)), pad_index(pad_index),
+          callback(std::move(callback)) {}
+
+    void Stop() {
+        io_service.stop();
+    }
+
+    void Loop() {
+        io_service.run();
+    }
+
+    void StartSend(const clock::time_point& from) {
+        timer.expires_at(from + std::chrono::seconds(3));
+        timer.async_wait([this](const boost::system::error_code& error) { HandleSend(error); });
+    }
+
+    void StartReceive() {
+        socket.async_receive_from(
+            boost::asio::buffer(receive_buffer), receive_endpoint,
+            [this](const boost::system::error_code& error, std::size_t bytes_transferred) {
+                HandleReceive(error, bytes_transferred);
+            });
+    }
+
+private:
+    void HandleReceive(const boost::system::error_code& error, std::size_t bytes_transferred) {
+        if (auto type = Response::Validate(receive_buffer.data(), bytes_transferred)) {
+            switch (*type) {
+            case Type::Version: {
+                Response::Version version;
+                std::memcpy(&version, &receive_buffer[sizeof(Header)], sizeof(Response::Version));
+                callback.version(std::move(version));
+                break;
+            }
+            case Type::PortInfo: {
+                Response::PortInfo port_info;
+                std::memcpy(&port_info, &receive_buffer[sizeof(Header)],
+                            sizeof(Response::PortInfo));
+                callback.port_info(std::move(port_info));
+                break;
+            }
+            case Type::PadData: {
+                Response::PadData pad_data;
+                std::memcpy(&pad_data, &receive_buffer[sizeof(Header)], sizeof(Response::PadData));
+                callback.pad_data(std::move(pad_data));
+                break;
+            }
+            }
+        }
+        StartReceive();
+    }
+
+    void HandleSend(const boost::system::error_code& error) {
+        // Send a request for getting port info for the pad
+        Request::PortInfo port_info{1, {pad_index, 0, 0, 0}};
+        const auto port_message = Request::Create(port_info, client_id);
+        std::memcpy(&send_buffer1, &port_message, PORT_INFO_SIZE);
+        socket.send_to(boost::asio::buffer(send_buffer1), send_endpoint);
+
+        // Send a request for getting pad data for the pad
+        Request::PadData pad_data{Request::PadData::Flags::Id, pad_index, EMPTY_MAC_ADDRESS};
+        const auto pad_message = Request::Create(pad_data, client_id);
+        std::memcpy(send_buffer2.data(), &pad_message, PAD_DATA_SIZE);
+        socket.send_to(boost::asio::buffer(send_buffer2), send_endpoint);
+        StartSend(timer.expiry());
+    }
+
+    SocketCallback callback;
+    boost::asio::io_service io_service;
+    boost::asio::basic_waitable_timer<clock> timer;
+    udp::socket socket;
+
+    u32 client_id{};
+    u8 pad_index{};
+
+    static constexpr std::size_t PORT_INFO_SIZE = sizeof(Message<Request::PortInfo>);
+    static constexpr std::size_t PAD_DATA_SIZE = sizeof(Message<Request::PadData>);
+    std::array<u8, PORT_INFO_SIZE> send_buffer1;
+    std::array<u8, PAD_DATA_SIZE> send_buffer2;
+    udp::endpoint send_endpoint;
+
+    std::array<u8, MAX_PACKET_SIZE> receive_buffer;
+    udp::endpoint receive_endpoint;
+};
+
+static void SocketLoop(Socket* socket) {
+    socket->StartReceive();
+    socket->StartSend(Socket::clock::now());
+    socket->Loop();
+}
+
+Client::Client(std::shared_ptr<DeviceStatus> status, const std::string& host, u16 port,
+               u8 pad_index, u32 client_id)
+    : status(status) {
+    StartCommunication(host, port, pad_index, client_id);
+}
+
+Client::~Client() {
+    socket->Stop();
+    thread.join();
+}
+
+void Client::ReloadSocket(const std::string& host, u16 port, u8 pad_index, u32 client_id) {
+    socket->Stop();
+    thread.join();
+    StartCommunication(host, port, pad_index, client_id);
+}
+
+void Client::OnVersion(Response::Version data) {
+    LOG_TRACE(Input, "Version packet received: {}", data.version);
+}
+
+void Client::OnPortInfo(Response::PortInfo data) {
+    LOG_TRACE(Input, "PortInfo packet received: {}", data.model);
+}
+
+void Client::OnPadData(Response::PadData data) {
+    LOG_TRACE(Input, "PadData packet received");
+    if (data.packet_counter <= packet_sequence) {
+        LOG_WARNING(
+            Input,
+            "PadData packet dropped because its stale info. Current count: {} Packet count: {}",
+            packet_sequence, data.packet_counter);
+        return;
+    }
+    packet_sequence = data.packet_counter;
+    // TODO: Check how the Switch handles motions and how the CemuhookUDP motion
+    // directions correspond to the ones of the Switch
+    Common::Vec3f accel = Common::MakeVec<float>(data.accel.x, data.accel.y, data.accel.z);
+    Common::Vec3f gyro = Common::MakeVec<float>(data.gyro.pitch, data.gyro.yaw, data.gyro.roll);
+    {
+        std::lock_guard guard(status->update_mutex);
+
+        status->motion_status = {accel, gyro};
+
+        // TODO: add a setting for "click" touch. Click touch refers to a device that differentiates
+        // between a simple "tap" and a hard press that causes the touch screen to click.
+        const bool is_active = data.touch_1.is_active != 0;
+
+        float x = 0;
+        float y = 0;
+
+        if (is_active && status->touch_calibration) {
+            const u16 min_x = status->touch_calibration->min_x;
+            const u16 max_x = status->touch_calibration->max_x;
+            const u16 min_y = status->touch_calibration->min_y;
+            const u16 max_y = status->touch_calibration->max_y;
+
+            x = (std::clamp(static_cast<u16>(data.touch_1.x), min_x, max_x) - min_x) /
+                static_cast<float>(max_x - min_x);
+            y = (std::clamp(static_cast<u16>(data.touch_1.y), min_y, max_y) - min_y) /
+                static_cast<float>(max_y - min_y);
+        }
+
+        status->touch_status = {x, y, is_active};
+    }
+}
+
+void Client::StartCommunication(const std::string& host, u16 port, u8 pad_index, u32 client_id) {
+    SocketCallback callback{[this](Response::Version version) { OnVersion(version); },
+                            [this](Response::PortInfo info) { OnPortInfo(info); },
+                            [this](Response::PadData data) { OnPadData(data); }};
+    LOG_INFO(Input, "Starting communication with UDP input server on {}:{}", host, port);
+    socket = std::make_unique<Socket>(host, port, pad_index, client_id, callback);
+    thread = std::thread{SocketLoop, this->socket.get()};
+}
+
+void TestCommunication(const std::string& host, u16 port, u8 pad_index, u32 client_id,
+                       std::function<void()> success_callback,
+                       std::function<void()> failure_callback) {
+    std::thread([=] {
+        Common::Event success_event;
+        SocketCallback callback{[](Response::Version version) {}, [](Response::PortInfo info) {},
+                                [&](Response::PadData data) { success_event.Set(); }};
+        Socket socket{host, port, pad_index, client_id, callback};
+        std::thread worker_thread{SocketLoop, &socket};
+        bool result = success_event.WaitFor(std::chrono::seconds(8));
+        socket.Stop();
+        worker_thread.join();
+        if (result) {
+            success_callback();
+        } else {
+            failure_callback();
+        }
+    })
+        .detach();
+}
+
+CalibrationConfigurationJob::CalibrationConfigurationJob(
+    const std::string& host, u16 port, u8 pad_index, u32 client_id,
+    std::function<void(Status)> status_callback,
+    std::function<void(u16, u16, u16, u16)> data_callback) {
+
+    std::thread([=] {
+        constexpr u16 CALIBRATION_THRESHOLD = 100;
+
+        u16 min_x{UINT16_MAX};
+        u16 min_y{UINT16_MAX};
+        u16 max_x{};
+        u16 max_y{};
+
+        Status current_status{Status::Initialized};
+        SocketCallback callback{[](Response::Version version) {}, [](Response::PortInfo info) {},
+                                [&](Response::PadData data) {
+                                    if (current_status == Status::Initialized) {
+                                        // Receiving data means the communication is ready now
+                                        current_status = Status::Ready;
+                                        status_callback(current_status);
+                                    }
+                                    if (!data.touch_1.is_active) {
+                                        return;
+                                    }
+                                    LOG_DEBUG(Input, "Current touch: {} {}", data.touch_1.x,
+                                              data.touch_1.y);
+                                    min_x = std::min(min_x, static_cast<u16>(data.touch_1.x));
+                                    min_y = std::min(min_y, static_cast<u16>(data.touch_1.y));
+                                    if (current_status == Status::Ready) {
+                                        // First touch - min data (min_x/min_y)
+                                        current_status = Status::Stage1Completed;
+                                        status_callback(current_status);
+                                    }
+                                    if (data.touch_1.x - min_x > CALIBRATION_THRESHOLD &&
+                                        data.touch_1.y - min_y > CALIBRATION_THRESHOLD) {
+                                        // Set the current position as max value and finishes
+                                        // configuration
+                                        max_x = data.touch_1.x;
+                                        max_y = data.touch_1.y;
+                                        current_status = Status::Completed;
+                                        data_callback(min_x, min_y, max_x, max_y);
+                                        status_callback(current_status);
+
+                                        complete_event.Set();
+                                    }
+                                }};
+        Socket socket{host, port, pad_index, client_id, callback};
+        std::thread worker_thread{SocketLoop, &socket};
+        complete_event.Wait();
+        socket.Stop();
+        worker_thread.join();
+    })
+        .detach();
+}
+
+CalibrationConfigurationJob::~CalibrationConfigurationJob() {
+    Stop();
+}
+
+void CalibrationConfigurationJob::Stop() {
+    complete_event.Set();
+}
+
+} // namespace InputCommon::CemuhookUDP

src/input_common/udp/client.h

@@ -0,0 +1,96 @@
+// Copyright 2018 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <functional>
+#include <memory>
+#include <mutex>
+#include <optional>
+#include <string>
+#include <thread>
+#include <tuple>
+#include <vector>
+#include "common/common_types.h"
+#include "common/thread.h"
+#include "common/vector_math.h"
+
+namespace InputCommon::CemuhookUDP {
+
+constexpr u16 DEFAULT_PORT = 26760;
+constexpr char DEFAULT_ADDR[] = "127.0.0.1";
+
+class Socket;
+
+namespace Response {
+struct PadData;
+struct PortInfo;
+struct Version;
+} // namespace Response
+
+struct DeviceStatus {
+    std::mutex update_mutex;
+    std::tuple<Common::Vec3<float>, Common::Vec3<float>> motion_status;
+    std::tuple<float, float, bool> touch_status;
+
+    // calibration data for scaling the device's touch area to 3ds
+    struct CalibrationData {
+        u16 min_x{};
+        u16 min_y{};
+        u16 max_x{};
+        u16 max_y{};
+    };
+    std::optional<CalibrationData> touch_calibration;
+};
+
+class Client {
+public:
+    explicit Client(std::shared_ptr<DeviceStatus> status, const std::string& host = DEFAULT_ADDR,
+                    u16 port = DEFAULT_PORT, u8 pad_index = 0, u32 client_id = 24872);
+    ~Client();
+    void ReloadSocket(const std::string& host = "127.0.0.1", u16 port = 26760, u8 pad_index = 0,
+                      u32 client_id = 24872);
+
+private:
+    void OnVersion(Response::Version);
+    void OnPortInfo(Response::PortInfo);
+    void OnPadData(Response::PadData);
+    void StartCommunication(const std::string& host, u16 port, u8 pad_index, u32 client_id);
+
+    std::unique_ptr<Socket> socket;
+    std::shared_ptr<DeviceStatus> status;
+    std::thread thread;
+    u64 packet_sequence = 0;
+};
+
+/// An async job allowing configuration of the touchpad calibration.
+class CalibrationConfigurationJob {
+public:
+    enum class Status {
+        Initialized,
+        Ready,
+        Stage1Completed,
+        Completed,
+    };
+    /**
+     * Constructs and starts the job with the specified parameter.
+     *
+     * @param status_callback Callback for job status updates
+     * @param data_callback Called when calibration data is ready
+     */
+    explicit CalibrationConfigurationJob(const std::string& host, u16 port, u8 pad_index,
+                                         u32 client_id, std::function<void(Status)> status_callback,
+                                         std::function<void(u16, u16, u16, u16)> data_callback);
+    ~CalibrationConfigurationJob();
+    void Stop();
+
+private:
+    Common::Event complete_event;
+};
+
+void TestCommunication(const std::string& host, u16 port, u8 pad_index, u32 client_id,
+                       std::function<void()> success_callback,
+                       std::function<void()> failure_callback);
+
+} // namespace InputCommon::CemuhookUDP

src/input_common/udp/protocol.cpp

@@ -0,0 +1,79 @@
+// Copyright 2018 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cstddef>
+#include <cstring>
+#include "common/logging/log.h"
+#include "input_common/udp/protocol.h"
+
+namespace InputCommon::CemuhookUDP {
+
+static constexpr std::size_t GetSizeOfResponseType(Type t) {
+    switch (t) {
+    case Type::Version:
+        return sizeof(Response::Version);
+    case Type::PortInfo:
+        return sizeof(Response::PortInfo);
+    case Type::PadData:
+        return sizeof(Response::PadData);
+    }
+    return 0;
+}
+
+namespace Response {
+
+/**
+ * Returns Type if the packet is valid, else none
+ *
+ * Note: Modifies the buffer to zero out the crc (since thats the easiest way to check without
+ * copying the buffer)
+ */
+std::optional<Type> Validate(u8* data, std::size_t size) {
+    if (size < sizeof(Header)) {
+        LOG_DEBUG(Input, "Invalid UDP packet received");
+        return std::nullopt;
+    }
+    Header header{};
+    std::memcpy(&header, data, sizeof(Header));
+    if (header.magic != SERVER_MAGIC) {
+        LOG_ERROR(Input, "UDP Packet has an unexpected magic value");
+        return std::nullopt;
+    }
+    if (header.protocol_version != PROTOCOL_VERSION) {
+        LOG_ERROR(Input, "UDP Packet protocol mismatch");
+        return std::nullopt;
+    }
+    if (header.type < Type::Version || header.type > Type::PadData) {
+        LOG_ERROR(Input, "UDP Packet is an unknown type");
+        return std::nullopt;
+    }
+
+    // Packet size must equal sizeof(Header) + sizeof(Data)
+    // and also verify that the packet info mentions the correct size. Since the spec includes the
+    // type of the packet as part of the data, we need to include it in size calculations here
+    // ie: payload_length == sizeof(T) + sizeof(Type)
+    const std::size_t data_len = GetSizeOfResponseType(header.type);
+    if (header.payload_length != data_len + sizeof(Type) || size < data_len + sizeof(Header)) {
+        LOG_ERROR(
+            Input,
+            "UDP Packet payload length doesn't match. Received: {} PayloadLength: {} Expected: {}",
+            size, header.payload_length, data_len + sizeof(Type));
+        return std::nullopt;
+    }
+
+    const u32 crc32 = header.crc;
+    boost::crc_32_type result;
+    // zero out the crc in the buffer and then run the crc against it
+    std::memset(&data[offsetof(Header, crc)], 0, sizeof(u32_le));
+
+    result.process_bytes(data, data_len + sizeof(Header));
+    if (crc32 != result.checksum()) {
+        LOG_ERROR(Input, "UDP Packet CRC check failed. Offset: {}", offsetof(Header, crc));
+        return std::nullopt;
+    }
+    return header.type;
+}
+} // namespace Response
+
+} // namespace InputCommon::CemuhookUDP

src/input_common/udp/protocol.h

@@ -0,0 +1,256 @@
+// Copyright 2018 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <optional>
+#include <type_traits>
+#include <vector>
+#include <boost/crc.hpp>
+#include "common/bit_field.h"
+#include "common/swap.h"
+
+namespace InputCommon::CemuhookUDP {
+
+constexpr std::size_t MAX_PACKET_SIZE = 100;
+constexpr u16 PROTOCOL_VERSION = 1001;
+constexpr u32 CLIENT_MAGIC = 0x43555344; // DSUC (but flipped for LE)
+constexpr u32 SERVER_MAGIC = 0x53555344; // DSUS (but flipped for LE)
+
+enum class Type : u32 {
+    Version = 0x00100000,
+    PortInfo = 0x00100001,
+    PadData = 0x00100002,
+};
+
+struct Header {
+    u32_le magic{};
+    u16_le protocol_version{};
+    u16_le payload_length{};
+    u32_le crc{};
+    u32_le id{};
+    ///> In the protocol, the type of the packet is not part of the header, but its convenient to
+    ///> include in the header so the callee doesn't have to duplicate the type twice when building
+    ///> the data
+    Type type{};
+};
+static_assert(sizeof(Header) == 20, "UDP Message Header struct has wrong size");
+static_assert(std::is_trivially_copyable_v<Header>, "UDP Message Header is not trivially copyable");
+
+using MacAddress = std::array<u8, 6>;
+constexpr MacAddress EMPTY_MAC_ADDRESS = {0, 0, 0, 0, 0, 0};
+
+#pragma pack(push, 1)
+template <typename T>
+struct Message {
+    Header header{};
+    T data;
+};
+#pragma pack(pop)
+
+template <typename T>
+constexpr Type GetMessageType();
+
+namespace Request {
+
+struct Version {};
+/**
+ * Requests the server to send information about what controllers are plugged into the ports
+ * In citra's case, we only have one controller, so for simplicity's sake, we can just send a
+ * request explicitly for the first controller port and leave it at that. In the future it would be
+ * nice to make this configurable
+ */
+constexpr u32 MAX_PORTS = 4;
+struct PortInfo {
+    u32_le pad_count{}; ///> Number of ports to request data for
+    std::array<u8, MAX_PORTS> port;
+};
+static_assert(std::is_trivially_copyable_v<PortInfo>,
+              "UDP Request PortInfo is not trivially copyable");
+
+/**
+ * Request the latest pad information from the server. If the server hasn't received this message
+ * from the client in a reasonable time frame, the server will stop sending updates. The default
+ * timeout seems to be 5 seconds.
+ */
+struct PadData {
+    enum class Flags : u8 {
+        AllPorts,
+        Id,
+        Mac,
+    };
+    /// Determines which method will be used as a look up for the controller
+    Flags flags{};
+    /// Index of the port of the controller to retrieve data about
+    u8 port_id{};
+    /// Mac address of the controller to retrieve data about
+    MacAddress mac;
+};
+static_assert(sizeof(PadData) == 8, "UDP Request PadData struct has wrong size");
+static_assert(std::is_trivially_copyable_v<PadData>,
+              "UDP Request PadData is not trivially copyable");
+
+/**
+ * Creates a message with the proper header data that can be sent to the server.
+ * @param T data Request body to send
+ * @param client_id ID of the udp client (usually not checked on the server)
+ */
+template <typename T>
+Message<T> Create(const T data, const u32 client_id = 0) {
+    boost::crc_32_type crc;
+    Header header{
+        CLIENT_MAGIC, PROTOCOL_VERSION, sizeof(T) + sizeof(Type), 0, client_id, GetMessageType<T>(),
+    };
+    Message<T> message{header, data};
+    crc.process_bytes(&message, sizeof(Message<T>));
+    message.header.crc = crc.checksum();
+    return message;
+}
+} // namespace Request
+
+namespace Response {
+
+struct Version {
+    u16_le version{};
+};
+static_assert(sizeof(Version) == 2, "UDP Response Version struct has wrong size");
+static_assert(std::is_trivially_copyable_v<Version>,
+              "UDP Response Version is not trivially copyable");
+
+struct PortInfo {
+    u8 id{};
+    u8 state{};
+    u8 model{};
+    u8 connection_type{};
+    MacAddress mac;
+    u8 battery{};
+    u8 is_pad_active{};
+};
+static_assert(sizeof(PortInfo) == 12, "UDP Response PortInfo struct has wrong size");
+static_assert(std::is_trivially_copyable_v<PortInfo>,
+              "UDP Response PortInfo is not trivially copyable");
+
+#pragma pack(push, 1)
+struct PadData {
+    PortInfo info{};
+    u32_le packet_counter{};
+
+    u16_le digital_button{};
+    // The following union isn't trivially copyable but we don't use this input anyway.
+    // union DigitalButton {
+    //     u16_le button;
+    //     BitField<0, 1, u16> button_1;   // Share
+    //     BitField<1, 1, u16> button_2;   // L3
+    //     BitField<2, 1, u16> button_3;   // R3
+    //     BitField<3, 1, u16> button_4;   // Options
+    //     BitField<4, 1, u16> button_5;   // Up
+    //     BitField<5, 1, u16> button_6;   // Right
+    //     BitField<6, 1, u16> button_7;   // Down
+    //     BitField<7, 1, u16> button_8;   // Left
+    //     BitField<8, 1, u16> button_9;   // L2
+    //     BitField<9, 1, u16> button_10;  // R2
+    //     BitField<10, 1, u16> button_11; // L1
+    //     BitField<11, 1, u16> button_12; // R1
+    //     BitField<12, 1, u16> button_13; // Triangle
+    //     BitField<13, 1, u16> button_14; // Circle
+    //     BitField<14, 1, u16> button_15; // Cross
+    //     BitField<15, 1, u16> button_16; // Square
+    // } digital_button;
+
+    u8 home;
+    /// If the device supports a "click" on the touchpad, this will change to 1 when a click happens
+    u8 touch_hard_press{};
+    u8 left_stick_x{};
+    u8 left_stick_y{};
+    u8 right_stick_x{};
+    u8 right_stick_y{};
+
+    struct AnalogButton {
+        u8 button_8{};
+        u8 button_7{};
+        u8 button_6{};
+        u8 button_5{};
+        u8 button_12{};
+        u8 button_11{};
+        u8 button_10{};
+        u8 button_9{};
+        u8 button_16{};
+        u8 button_15{};
+        u8 button_14{};
+        u8 button_13{};
+    } analog_button;
+
+    struct TouchPad {
+        u8 is_active{};
+        u8 id{};
+        u16_le x{};
+        u16_le y{};
+    } touch_1, touch_2;
+
+    u64_le motion_timestamp;
+
+    struct Accelerometer {
+        float x{};
+        float y{};
+        float z{};
+    } accel;
+
+    struct Gyroscope {
+        float pitch{};
+        float yaw{};
+        float roll{};
+    } gyro;
+};
+#pragma pack(pop)
+
+static_assert(sizeof(PadData) == 80, "UDP Response PadData struct has wrong size ");
+static_assert(std::is_trivially_copyable_v<PadData>,
+              "UDP Response PadData is not trivially copyable");
+
+static_assert(sizeof(Message<PadData>) == MAX_PACKET_SIZE,
+              "UDP MAX_PACKET_SIZE is no longer larger than Message<PadData>");
+
+static_assert(sizeof(PadData::AnalogButton) == 12,
+              "UDP Response AnalogButton struct has wrong size ");
+static_assert(sizeof(PadData::TouchPad) == 6, "UDP Response TouchPad struct has wrong size ");
+static_assert(sizeof(PadData::Accelerometer) == 12,
+              "UDP Response Accelerometer struct has wrong size ");
+static_assert(sizeof(PadData::Gyroscope) == 12, "UDP Response Gyroscope struct has wrong size ");
+
+/**
+ * Create a Response Message from the data
+ * @param data array of bytes sent from the server
+ * @return boost::none if it failed to parse or Type if it succeeded. The client can then safely
+ * copy the data into the appropriate struct for that Type
+ */
+std::optional<Type> Validate(u8* data, std::size_t size);
+
+} // namespace Response
+
+template <>
+constexpr Type GetMessageType<Request::Version>() {
+    return Type::Version;
+}
+template <>
+constexpr Type GetMessageType<Request::PortInfo>() {
+    return Type::PortInfo;
+}
+template <>
+constexpr Type GetMessageType<Request::PadData>() {
+    return Type::PadData;
+}
+template <>
+constexpr Type GetMessageType<Response::Version>() {
+    return Type::Version;
+}
+template <>
+constexpr Type GetMessageType<Response::PortInfo>() {
+    return Type::PortInfo;
+}
+template <>
+constexpr Type GetMessageType<Response::PadData>() {
+    return Type::PadData;
+}
+} // namespace InputCommon::CemuhookUDP

src/input_common/udp/udp.cpp

@@ -0,0 +1,96 @@
+// Copyright 2018 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include "common/logging/log.h"
+#include "common/param_package.h"
+#include "core/frontend/input.h"
+#include "core/settings.h"
+#include "input_common/udp/client.h"
+#include "input_common/udp/udp.h"
+
+namespace InputCommon::CemuhookUDP {
+
+class UDPTouchDevice final : public Input::TouchDevice {
+public:
+    explicit UDPTouchDevice(std::shared_ptr<DeviceStatus> status_) : status(std::move(status_)) {}
+    std::tuple<float, float, bool> GetStatus() const {
+        std::lock_guard guard(status->update_mutex);
+        return status->touch_status;
+    }
+
+private:
+    std::shared_ptr<DeviceStatus> status;
+};
+
+class UDPMotionDevice final : public Input::MotionDevice {
+public:
+    explicit UDPMotionDevice(std::shared_ptr<DeviceStatus> status_) : status(std::move(status_)) {}
+    std::tuple<Common::Vec3<float>, Common::Vec3<float>> GetStatus() const {
+        std::lock_guard guard(status->update_mutex);
+        return status->motion_status;
+    }
+
+private:
+    std::shared_ptr<DeviceStatus> status;
+};
+
+class UDPTouchFactory final : public Input::Factory<Input::TouchDevice> {
+public:
+    explicit UDPTouchFactory(std::shared_ptr<DeviceStatus> status_) : status(std::move(status_)) {}
+
+    std::unique_ptr<Input::TouchDevice> Create(const Common::ParamPackage& params) override {
+        {
+            std::lock_guard guard(status->update_mutex);
+            status->touch_calibration.emplace();
+            // These default values work well for DS4 but probably not other touch inputs
+            status->touch_calibration->min_x = params.Get("min_x", 100);
+            status->touch_calibration->min_y = params.Get("min_y", 50);
+            status->touch_calibration->max_x = params.Get("max_x", 1800);
+            status->touch_calibration->max_y = params.Get("max_y", 850);
+        }
+        return std::make_unique<UDPTouchDevice>(status);
+    }
+
+private:
+    std::shared_ptr<DeviceStatus> status;
+};
+
+class UDPMotionFactory final : public Input::Factory<Input::MotionDevice> {
+public:
+    explicit UDPMotionFactory(std::shared_ptr<DeviceStatus> status_) : status(std::move(status_)) {}
+
+    std::unique_ptr<Input::MotionDevice> Create(const Common::ParamPackage& params) override {
+        return std::make_unique<UDPMotionDevice>(status);
+    }
+
+private:
+    std::shared_ptr<DeviceStatus> status;
+};
+
+State::State() {
+    auto status = std::make_shared<DeviceStatus>();
+    client =
+        std::make_unique<Client>(status, Settings::values.udp_input_address,
+                                 Settings::values.udp_input_port, Settings::values.udp_pad_index);
+
+    Input::RegisterFactory<Input::TouchDevice>("cemuhookudp",
+                                               std::make_shared<UDPTouchFactory>(status));
+    Input::RegisterFactory<Input::MotionDevice>("cemuhookudp",
+                                                std::make_shared<UDPMotionFactory>(status));
+}
+
+State::~State() {
+    Input::UnregisterFactory<Input::TouchDevice>("cemuhookudp");
+    Input::UnregisterFactory<Input::MotionDevice>("cemuhookudp");
+}
+
+void State::ReloadUDPClient() {
+    client->ReloadSocket(Settings::values.udp_input_address, Settings::values.udp_input_port,
+                         Settings::values.udp_pad_index);
+}
+
+std::unique_ptr<State> Init() {
+    return std::make_unique<State>();
+}
+} // namespace InputCommon::CemuhookUDP

src/input_common/udp/udp.h

@@ -0,0 +1,27 @@
+// Copyright 2018 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <memory>
+#include <unordered_map>
+#include "input_common/main.h"
+#include "input_common/udp/client.h"
+
+namespace InputCommon::CemuhookUDP {
+
+class UDPTouchDevice;
+class UDPMotionDevice;
+
+class State {
+public:
+    State();
+    ~State();
+    void ReloadUDPClient();
+
+private:
+    std::unique_ptr<Client> client;
+};
+
+std::unique_ptr<State> Init();
+
+} // namespace InputCommon::CemuhookUDP

src/video_core/CMakeLists.txt

@@ -153,14 +153,31 @@ if (ENABLE_VULKAN)
         renderer_vulkan/fixed_pipeline_state.h
         renderer_vulkan/maxwell_to_vk.cpp
         renderer_vulkan/maxwell_to_vk.h
+        renderer_vulkan/renderer_vulkan.h
+        renderer_vulkan/vk_blit_screen.cpp
+        renderer_vulkan/vk_blit_screen.h
         renderer_vulkan/vk_buffer_cache.cpp
         renderer_vulkan/vk_buffer_cache.h
+        renderer_vulkan/vk_compute_pass.cpp
+        renderer_vulkan/vk_compute_pass.h
+        renderer_vulkan/vk_compute_pipeline.cpp
+        renderer_vulkan/vk_compute_pipeline.h
+        renderer_vulkan/vk_descriptor_pool.cpp
+        renderer_vulkan/vk_descriptor_pool.h
         renderer_vulkan/vk_device.cpp
         renderer_vulkan/vk_device.h
+        renderer_vulkan/vk_graphics_pipeline.cpp
+        renderer_vulkan/vk_graphics_pipeline.h
         renderer_vulkan/vk_image.cpp
         renderer_vulkan/vk_image.h
         renderer_vulkan/vk_memory_manager.cpp
         renderer_vulkan/vk_memory_manager.h
+        renderer_vulkan/vk_pipeline_cache.cpp
+        renderer_vulkan/vk_pipeline_cache.h
+        renderer_vulkan/vk_rasterizer.cpp
+        renderer_vulkan/vk_rasterizer.h
+        renderer_vulkan/vk_renderpass_cache.cpp
+        renderer_vulkan/vk_renderpass_cache.h
         renderer_vulkan/vk_resource_manager.cpp
         renderer_vulkan/vk_resource_manager.h
         renderer_vulkan/vk_sampler_cache.cpp
@@ -169,12 +186,19 @@ if (ENABLE_VULKAN)
         renderer_vulkan/vk_scheduler.h
         renderer_vulkan/vk_shader_decompiler.cpp
         renderer_vulkan/vk_shader_decompiler.h
+        renderer_vulkan/vk_shader_util.cpp
+        renderer_vulkan/vk_shader_util.h
         renderer_vulkan/vk_staging_buffer_pool.cpp
         renderer_vulkan/vk_staging_buffer_pool.h
         renderer_vulkan/vk_stream_buffer.cpp
         renderer_vulkan/vk_stream_buffer.h
         renderer_vulkan/vk_swapchain.cpp
-        renderer_vulkan/vk_swapchain.h)
+        renderer_vulkan/vk_swapchain.h
+        renderer_vulkan/vk_texture_cache.cpp
+        renderer_vulkan/vk_texture_cache.h
+        renderer_vulkan/vk_update_descriptor.cpp
+        renderer_vulkan/vk_update_descriptor.h
+    )
 
     target_include_directories(video_core PRIVATE sirit ../../externals/Vulkan-Headers/include)
     target_compile_definitions(video_core PRIVATE HAS_VULKAN)

src/video_core/engines/maxwell_3d.cpp

@@ -91,6 +91,7 @@ void Maxwell3D::InitializeRegisterDefaults() {
     regs.rasterize_enable = 1;
     regs.rt_separate_frag_data = 1;
     regs.framebuffer_srgb = 1;
+    regs.cull.front_face = Maxwell3D::Regs::Cull::FrontFace::ClockWise;
 
     mme_inline[MAXWELL3D_REG_INDEX(draw.vertex_end_gl)] = true;
     mme_inline[MAXWELL3D_REG_INDEX(draw.vertex_begin_gl)] = true;

src/video_core/engines/maxwell_3d.h

@@ -1018,7 +1018,14 @@ public:
                     }
                 } instanced_arrays;
 
-                INSERT_UNION_PADDING_WORDS(0x6);
+                INSERT_UNION_PADDING_WORDS(0x4);
+
+                union {
+                    BitField<0, 1, u32> enable;
+                    BitField<4, 8, u32> unk4;
+                } vp_point_size;
+
+                INSERT_UNION_PADDING_WORDS(1);
 
                 Cull cull;
 
@@ -1271,8 +1278,6 @@ public:
 
     } dirty{};
 
-    std::array<u8, Regs::NUM_REGS> dirty_pointers{};
-
     /// Reads a register value located at the input method address
     u32 GetRegisterValue(u32 method) const;
 
@@ -1367,6 +1372,8 @@ private:
 
     bool execute_on{true};
 
+    std::array<u8, Regs::NUM_REGS> dirty_pointers{};
+
     /// Retrieves information about a specific TIC entry from the TIC buffer.
     Texture::TICEntry GetTICEntry(u32 tic_index) const;
 
@@ -1503,6 +1510,7 @@ ASSERT_REG_POSITION(primitive_restart, 0x591);
 ASSERT_REG_POSITION(index_array, 0x5F2);
 ASSERT_REG_POSITION(polygon_offset_clamp, 0x61F);
 ASSERT_REG_POSITION(instanced_arrays, 0x620);
+ASSERT_REG_POSITION(vp_point_size, 0x644);
 ASSERT_REG_POSITION(cull, 0x646);
 ASSERT_REG_POSITION(pixel_center_integer, 0x649);
 ASSERT_REG_POSITION(viewport_transform_enabled, 0x64B);

src/video_core/engines/shader_bytecode.h

@@ -215,6 +215,40 @@ enum class F2fRoundingOp : u64 {
     Trunc = 11,
 };
 
+enum class AtomicOp : u64 {
+    Add = 0,
+    Min = 1,
+    Max = 2,
+    Inc = 3,
+    Dec = 4,
+    And = 5,
+    Or = 6,
+    Xor = 7,
+    Exch = 8,
+};
+
+enum class GlobalAtomicOp : u64 {
+    Add = 0,
+    Min = 1,
+    Max = 2,
+    Inc = 3,
+    Dec = 4,
+    And = 5,
+    Or = 6,
+    Xor = 7,
+    Exch = 8,
+    SafeAdd = 10,
+};
+
+enum class GlobalAtomicType : u64 {
+    U32 = 0,
+    S32 = 1,
+    U64 = 2,
+    F32_FTZ_RN = 3,
+    F16x2_FTZ_RN = 4,
+    S64 = 5,
+};
+
 enum class UniformType : u64 {
     UnsignedByte = 0,
     SignedByte = 1,
@@ -236,6 +270,13 @@ enum class StoreType : u64 {
     Bits128 = 6,
 };
 
+enum class AtomicType : u64 {
+    U32 = 0,
+    S32 = 1,
+    U64 = 2,
+    S64 = 3,
+};
+
 enum class IMinMaxExchange : u64 {
     None = 0,
     XLo = 1,
@@ -938,6 +979,22 @@ union Instruction {
         BitField<46, 2, u64> cache_mode;
     } stg;
 
+    union {
+        BitField<52, 4, GlobalAtomicOp> operation;
+        BitField<49, 3, GlobalAtomicType> type;
+        BitField<28, 20, s64> offset;
+    } atom;
+
+    union {
+        BitField<52, 4, AtomicOp> operation;
+        BitField<28, 2, AtomicType> type;
+        BitField<30, 22, s64> offset;
+
+        s32 GetImmediateOffset() const {
+            return static_cast<s32>(offset << 2);
+        }
+    } atoms;
+
     union {
         BitField<32, 1, PhysicalAttributeDirection> direction;
         BitField<47, 3, AttributeSize> size;
@@ -1659,9 +1716,11 @@ public:
         ST_A,
         ST_L,
         ST_S,
-        ST,   // Store in generic memory
-        STG,  // Store in global memory
-        AL2P, // Transforms attribute memory into physical memory
+        ST,    // Store in generic memory
+        STG,   // Store in global memory
+        ATOM,  // Atomic operation on global memory
+        ATOMS, // Atomic operation on shared memory
+        AL2P,  // Transforms attribute memory into physical memory
         TEX,
         TEX_B,  // Texture Load Bindless
         TXQ,    // Texture Query
@@ -1964,6 +2023,8 @@ private:
             INST("1110111101010---", Id::ST_L, Type::Memory, "ST_L"),
             INST("101-------------", Id::ST, Type::Memory, "ST"),
             INST("1110111011011---", Id::STG, Type::Memory, "STG"),
+            INST("11101101--------", Id::ATOM, Type::Memory, "ATOM"),
+            INST("11101100--------", Id::ATOMS, Type::Memory, "ATOMS"),
             INST("1110111110100---", Id::AL2P, Type::Memory, "AL2P"),
             INST("110000----111---", Id::TEX, Type::Texture, "TEX"),
             INST("1101111010111---", Id::TEX_B, Type::Texture, "TEX_B"),

src/video_core/gpu.cpp

@@ -66,19 +66,20 @@ const DmaPusher& GPU::DmaPusher() const {
     return *dma_pusher;
 }
 
-void GPU::WaitFence(u32 syncpoint_id, u32 value) const {
+void GPU::WaitFence(u32 syncpoint_id, u32 value) {
     // Synced GPU, is always in sync
     if (!is_async) {
         return;
     }
     MICROPROFILE_SCOPE(GPU_wait);
-    while (syncpoints[syncpoint_id].load(std::memory_order_relaxed) < value) {
-    }
+    std::unique_lock lock{sync_mutex};
+    sync_cv.wait(lock, [=]() { return syncpoints[syncpoint_id].load() >= value; });
 }
 
 void GPU::IncrementSyncPoint(const u32 syncpoint_id) {
     syncpoints[syncpoint_id]++;
     std::lock_guard lock{sync_mutex};
+    sync_cv.notify_all();
     if (!syncpt_interrupts[syncpoint_id].empty()) {
         u32 value = syncpoints[syncpoint_id].load();
         auto it = syncpt_interrupts[syncpoint_id].begin();

src/video_core/gpu.h

@@ -6,6 +6,7 @@
 
 #include <array>
 #include <atomic>
+#include <condition_variable>
 #include <list>
 #include <memory>
 #include <mutex>
@@ -181,7 +182,7 @@ public:
     virtual void WaitIdle() const = 0;
 
     /// Allows the CPU/NvFlinger to wait on the GPU before presenting a frame.
-    void WaitFence(u32 syncpoint_id, u32 value) const;
+    void WaitFence(u32 syncpoint_id, u32 value);
 
     void IncrementSyncPoint(u32 syncpoint_id);
 
@@ -312,6 +313,8 @@ private:
 
     std::mutex sync_mutex;
 
+    std::condition_variable sync_cv;
+
     const bool is_async;
 };
 

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -1272,6 +1272,7 @@ void RasterizerOpenGL::SyncPointState() {
     const auto& regs = system.GPU().Maxwell3D().regs;
     // Limit the point size to 1 since nouveau sometimes sets a point size of 0 (and that's invalid
     // in OpenGL).
+    state.point.program_control = regs.vp_point_size.enable != 0;
     state.point.size = std::max(1.0f, regs.point_size);
 }
 

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -34,9 +34,6 @@ using VideoCommon::Shader::ShaderIR;
 
 namespace {
 
-// One UBO is always reserved for emulation values on staged shaders
-constexpr u32 STAGE_RESERVED_UBOS = 1;
-
 constexpr u32 STAGE_MAIN_OFFSET = 10;
 constexpr u32 KERNEL_MAIN_OFFSET = 0;
 
@@ -243,7 +240,6 @@ CachedProgram BuildShader(const Device& device, u64 unique_identifier, ShaderTyp
     if (!code_b.empty()) {
         ir_b.emplace(code_b, main_offset, COMPILER_SETTINGS, locker);
     }
-    const auto entries = GLShader::GetEntries(ir);
 
     std::string source = fmt::format(R"(// {}
 #version 430 core
@@ -264,6 +260,10 @@ CachedProgram BuildShader(const Device& device, u64 unique_identifier, ShaderTyp
                   "#extension GL_NV_shader_thread_group : require\n"
                   "#extension GL_NV_shader_thread_shuffle : require\n";
     }
+    // This pragma stops Nvidia's driver from over optimizing math (probably using fp16 operations)
+    // on places where we don't want to.
+    // Thanks to Ryujinx for finding this workaround.
+    source += "#pragma optionNV(fastmath off)\n";
 
     if (shader_type == ShaderType::Geometry) {
         const auto [glsl_topology, max_vertices] = GetPrimitiveDescription(variant.primitive_mode);
@@ -314,9 +314,10 @@ std::unordered_set<GLenum> GetSupportedFormats() {
 
 CachedShader::CachedShader(const ShaderParameters& params, ShaderType shader_type,
                            GLShader::ShaderEntries entries, ProgramCode code, ProgramCode code_b)
-    : RasterizerCacheObject{params.host_ptr}, system{params.system}, disk_cache{params.disk_cache},
-      device{params.device}, cpu_addr{params.cpu_addr}, unique_identifier{params.unique_identifier},
-      shader_type{shader_type}, entries{entries}, code{std::move(code)}, code_b{std::move(code_b)} {
+    : RasterizerCacheObject{params.host_ptr}, system{params.system},
+      disk_cache{params.disk_cache}, device{params.device}, cpu_addr{params.cpu_addr},
+      unique_identifier{params.unique_identifier}, shader_type{shader_type},
+      entries{std::move(entries)}, code{std::move(code)}, code_b{std::move(code_b)} {
     if (!params.precompiled_variants) {
         return;
     }

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -751,6 +751,9 @@ private:
 
     Expression Visit(const Node& node) {
         if (const auto operation = std::get_if<OperationNode>(&*node)) {
+            if (const auto amend_index = operation->GetAmendIndex()) {
+                Visit(ir.GetAmendNode(*amend_index)).CheckVoid();
+            }
             const auto operation_index = static_cast<std::size_t>(operation->GetCode());
             if (operation_index >= operation_decompilers.size()) {
                 UNREACHABLE_MSG("Out of bounds operation: {}", operation_index);
@@ -872,6 +875,9 @@ private:
         }
 
         if (const auto conditional = std::get_if<ConditionalNode>(&*node)) {
+            if (const auto amend_index = conditional->GetAmendIndex()) {
+                Visit(ir.GetAmendNode(*amend_index)).CheckVoid();
+            }
             // It's invalid to call conditional on nested nodes, use an operation instead
             code.AddLine("if ({}) {{", Visit(conditional->GetCondition()).AsBool());
             ++code.scope;
@@ -1013,7 +1019,6 @@ private:
                 }
                 return {{"gl_ViewportIndex", Type::Int}};
             case 3:
-                UNIMPLEMENTED_MSG("Requires some state changes for gl_PointSize to work in shader");
                 return {{"gl_PointSize", Type::Float}};
             }
             return {};
@@ -1850,6 +1855,13 @@ private:
                 Type::Uint};
     }
 
+    template <const std::string_view& opname, Type type>
+    Expression Atomic(Operation operation) {
+        return {fmt::format("atomic{}({}, {})", opname, Visit(operation[0]).GetCode(),
+                            Visit(operation[1]).As(type)),
+                type};
+    }
+
     Expression Branch(Operation operation) {
         const auto target = std::get_if<ImmediateNode>(&*operation[0]);
         UNIMPLEMENTED_IF(!target);
@@ -2188,6 +2200,8 @@ private:
         &GLSLDecompiler::AtomicImage<Func::Xor>,
         &GLSLDecompiler::AtomicImage<Func::Exchange>,
 
+        &GLSLDecompiler::Atomic<Func::Add, Type::Uint>,
+
         &GLSLDecompiler::Branch,
         &GLSLDecompiler::BranchIndirect,
         &GLSLDecompiler::PushFlowStack,
@@ -2307,7 +2321,7 @@ public:
     explicit ExprDecompiler(GLSLDecompiler& decomp) : decomp{decomp} {}
 
     void operator()(const ExprAnd& expr) {
-        inner += "( ";
+        inner += '(';
         std::visit(*this, *expr.operand1);
         inner += " && ";
         std::visit(*this, *expr.operand2);
@@ -2315,7 +2329,7 @@ public:
     }
 
     void operator()(const ExprOr& expr) {
-        inner += "( ";
+        inner += '(';
         std::visit(*this, *expr.operand1);
         inner += " || ";
         std::visit(*this, *expr.operand2);
@@ -2333,28 +2347,7 @@ public:
     }
 
     void operator()(const ExprCondCode& expr) {
-        const Node cc = decomp.ir.GetConditionCode(expr.cc);
-        std::string target;
-
-        if (const auto pred = std::get_if<PredicateNode>(&*cc)) {
-            const auto index = pred->GetIndex();
-            switch (index) {
-            case Tegra::Shader::Pred::NeverExecute:
-                target = "false";
-                break;
-            case Tegra::Shader::Pred::UnusedIndex:
-                target = "true";
-                break;
-            default:
-                target = decomp.GetPredicate(index);
-                break;
-            }
-        } else if (const auto flag = std::get_if<InternalFlagNode>(&*cc)) {
-            target = decomp.GetInternalFlag(flag->GetFlag());
-        } else {
-            UNREACHABLE();
-        }
-        inner += target;
+        inner += decomp.Visit(decomp.ir.GetConditionCode(expr.cc)).AsBool();
     }
 
     void operator()(const ExprVar& expr) {
@@ -2366,8 +2359,7 @@ public:
     }
 
     void operator()(VideoCommon::Shader::ExprGprEqual& expr) {
-        inner +=
-            "( ftou(" + decomp.GetRegister(expr.gpr) + ") == " + std::to_string(expr.value) + ')';
+        inner += fmt::format("(ftou({}) == {})", decomp.GetRegister(expr.gpr), expr.value);
     }
 
     const std::string& GetResult() const {
@@ -2375,8 +2367,8 @@ public:
     }
 
 private:
-    std::string inner;
     GLSLDecompiler& decomp;
+    std::string inner;
 };
 
 class ASTDecompiler {

src/video_core/renderer_opengl/gl_state.cpp

@@ -127,6 +127,7 @@ void OpenGLState::ApplyClipDistances() {
 }
 
 void OpenGLState::ApplyPointSize() {
+    Enable(GL_PROGRAM_POINT_SIZE, cur_state.point.program_control, point.program_control);
     if (UpdateValue(cur_state.point.size, point.size)) {
         glPointSize(point.size);
     }

src/video_core/renderer_opengl/gl_state.h

@@ -131,7 +131,8 @@ public:
     std::array<Viewport, Tegra::Engines::Maxwell3D::Regs::NumViewports> viewports;
 
     struct {
-        float size = 1.0f; // GL_POINT_SIZE
+        bool program_control = false; // GL_PROGRAM_POINT_SIZE
+        GLfloat size = 1.0f;          // GL_POINT_SIZE
     } point;
 
     struct {

src/video_core/renderer_opengl/gl_texture_cache.cpp

@@ -44,7 +44,7 @@ struct FormatTuple {
 
 constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format_tuples = {{
     {GL_RGBA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, false},                        // ABGR8U
-    {GL_RGBA8, GL_RGBA, GL_BYTE, false},                                            // ABGR8S
+    {GL_RGBA8_SNORM, GL_RGBA, GL_BYTE, false},                                      // ABGR8S
     {GL_RGBA8UI, GL_RGBA_INTEGER, GL_UNSIGNED_BYTE, false},                         // ABGR8UI
     {GL_RGB565, GL_RGB, GL_UNSIGNED_SHORT_5_6_5_REV, false},                        // B5G6R5U
     {GL_RGB10_A2, GL_RGBA, GL_UNSIGNED_INT_2_10_10_10_REV, false},                  // A2B10G10R10U
@@ -83,9 +83,9 @@ constexpr std::array<FormatTuple, VideoCore::Surface::MaxPixelFormat> tex_format
     {GL_RGB32F, GL_RGB, GL_FLOAT, false},                                           // RGB32F
     {GL_SRGB8_ALPHA8, GL_RGBA, GL_UNSIGNED_INT_8_8_8_8_REV, false},                 // RGBA8_SRGB
     {GL_RG8, GL_RG, GL_UNSIGNED_BYTE, false},                                       // RG8U
-    {GL_RG8, GL_RG, GL_BYTE, false},                                                // RG8S
+    {GL_RG8_SNORM, GL_RG, GL_BYTE, false},                                          // RG8S
     {GL_RG32UI, GL_RG_INTEGER, GL_UNSIGNED_INT, false},                             // RG32UI
-    {GL_RGB16F, GL_RGBA16, GL_HALF_FLOAT, false},                                   // RGBX16F
+    {GL_RGB16F, GL_RGBA, GL_HALF_FLOAT, false},                                     // RGBX16F
     {GL_R32UI, GL_RED_INTEGER, GL_UNSIGNED_INT, false},                             // R32UI
     {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},                                   // ASTC_2D_8X8
     {GL_RGBA8, GL_RGBA, GL_UNSIGNED_BYTE, false},                                   // ASTC_2D_8X5
@@ -176,6 +176,19 @@ GLint GetSwizzleSource(SwizzleSource source) {
     return GL_NONE;
 }
 
+GLenum GetComponent(PixelFormat format, bool is_first) {
+    switch (format) {
+    case PixelFormat::Z24S8:
+    case PixelFormat::Z32FS8:
+        return is_first ? GL_DEPTH_COMPONENT : GL_STENCIL_INDEX;
+    case PixelFormat::S8Z24:
+        return is_first ? GL_STENCIL_INDEX : GL_DEPTH_COMPONENT;
+    default:
+        UNREACHABLE();
+        return GL_DEPTH_COMPONENT;
+    }
+}
+
 void ApplyTextureDefaults(const SurfaceParams& params, GLuint texture) {
     if (params.IsBuffer()) {
         return;
@@ -184,7 +197,7 @@ void ApplyTextureDefaults(const SurfaceParams& params, GLuint texture) {
     glTextureParameteri(texture, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
     glTextureParameteri(texture, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
     glTextureParameteri(texture, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
-    glTextureParameteri(texture, GL_TEXTURE_MAX_LEVEL, params.num_levels - 1);
+    glTextureParameteri(texture, GL_TEXTURE_MAX_LEVEL, static_cast<GLint>(params.num_levels - 1));
     if (params.num_levels == 1) {
         glTextureParameterf(texture, GL_TEXTURE_LOD_BIAS, 1000.0f);
     }
@@ -253,14 +266,12 @@ void CachedSurface::DownloadTexture(std::vector<u8>& staging_buffer) {
         glPixelStorei(GL_PACK_ALIGNMENT, std::min(8U, params.GetRowAlignment(level)));
         glPixelStorei(GL_PACK_ROW_LENGTH, static_cast<GLint>(params.GetMipWidth(level)));
         const std::size_t mip_offset = params.GetHostMipmapLevelOffset(level);
+        u8* const mip_data = staging_buffer.data() + mip_offset;
+        const GLsizei size = static_cast<GLsizei>(params.GetHostMipmapSize(level));
         if (is_compressed) {
-            glGetCompressedTextureImage(texture.handle, level,
-                                        static_cast<GLsizei>(params.GetHostMipmapSize(level)),
-                                        staging_buffer.data() + mip_offset);
+            glGetCompressedTextureImage(texture.handle, level, size, mip_data);
         } else {
-            glGetTextureImage(texture.handle, level, format, type,
-                              static_cast<GLsizei>(params.GetHostMipmapSize(level)),
-                              staging_buffer.data() + mip_offset);
+            glGetTextureImage(texture.handle, level, format, type, size, mip_data);
         }
     }
 }
@@ -418,11 +429,21 @@ void CachedSurfaceView::ApplySwizzle(SwizzleSource x_source, SwizzleSource y_sou
     if (new_swizzle == swizzle)
         return;
     swizzle = new_swizzle;
-    const std::array<GLint, 4> gl_swizzle = {GetSwizzleSource(x_source), GetSwizzleSource(y_source),
-                                             GetSwizzleSource(z_source),
-                                             GetSwizzleSource(w_source)};
+    const std::array gl_swizzle = {GetSwizzleSource(x_source), GetSwizzleSource(y_source),
+                                   GetSwizzleSource(z_source), GetSwizzleSource(w_source)};
     const GLuint handle = GetTexture();
-    glTextureParameteriv(handle, GL_TEXTURE_SWIZZLE_RGBA, gl_swizzle.data());
+    const PixelFormat format = surface.GetSurfaceParams().pixel_format;
+    switch (format) {
+    case PixelFormat::Z24S8:
+    case PixelFormat::Z32FS8:
+    case PixelFormat::S8Z24:
+        glTextureParameteri(handle, GL_DEPTH_STENCIL_TEXTURE_MODE,
+                            GetComponent(format, x_source == SwizzleSource::R));
+        break;
+    default:
+        glTextureParameteriv(handle, GL_TEXTURE_SWIZZLE_RGBA, gl_swizzle.data());
+        break;
+    }
 }
 
 OGLTextureView CachedSurfaceView::CreateTextureView() const {
@@ -531,8 +552,11 @@ void TextureCacheOpenGL::ImageBlit(View& src_view, View& dst_view,
     const Common::Rectangle<u32>& dst_rect = copy_config.dst_rect;
     const bool is_linear = copy_config.filter == Tegra::Engines::Fermi2D::Filter::Linear;
 
-    glBlitFramebuffer(src_rect.left, src_rect.top, src_rect.right, src_rect.bottom, dst_rect.left,
-                      dst_rect.top, dst_rect.right, dst_rect.bottom, buffers,
+    glBlitFramebuffer(static_cast<GLint>(src_rect.left), static_cast<GLint>(src_rect.top),
+                      static_cast<GLint>(src_rect.right), static_cast<GLint>(src_rect.bottom),
+                      static_cast<GLint>(dst_rect.left), static_cast<GLint>(dst_rect.top),
+                      static_cast<GLint>(dst_rect.right), static_cast<GLint>(dst_rect.bottom),
+                      buffers,
                       is_linear && (buffers == GL_COLOR_BUFFER_BIT) ? GL_LINEAR : GL_NEAREST);
 }
 

src/video_core/renderer_opengl/utils.cpp

@@ -6,16 +6,20 @@
 #include <vector>
 
 #include <fmt/format.h>
-
 #include <glad/glad.h>
 
-#include "common/assert.h"
 #include "common/common_types.h"
-#include "common/scope_exit.h"
 #include "video_core/renderer_opengl/utils.h"
 
 namespace OpenGL {
 
+struct VertexArrayPushBuffer::Entry {
+    GLuint binding_index{};
+    const GLuint* buffer{};
+    GLintptr offset{};
+    GLsizei stride{};
+};
+
 VertexArrayPushBuffer::VertexArrayPushBuffer() = default;
 
 VertexArrayPushBuffer::~VertexArrayPushBuffer() = default;
@@ -47,6 +51,13 @@ void VertexArrayPushBuffer::Bind() {
     }
 }
 
+struct BindBuffersRangePushBuffer::Entry {
+    GLuint binding;
+    const GLuint* buffer;
+    GLintptr offset;
+    GLsizeiptr size;
+};
+
 BindBuffersRangePushBuffer::BindBuffersRangePushBuffer(GLenum target) : target{target} {}
 
 BindBuffersRangePushBuffer::~BindBuffersRangePushBuffer() = default;

src/video_core/renderer_opengl/utils.h

@@ -26,12 +26,7 @@ public:
     void Bind();
 
 private:
-    struct Entry {
-        GLuint binding_index{};
-        const GLuint* buffer{};
-        GLintptr offset{};
-        GLsizei stride{};
-    };
+    struct Entry;
 
     GLuint vao{};
     const GLuint* index_buffer{};
@@ -50,12 +45,7 @@ public:
     void Bind();
 
 private:
-    struct Entry {
-        GLuint binding;
-        const GLuint* buffer;
-        GLintptr offset;
-        GLsizeiptr size;
-    };
+    struct Entry;
 
     GLenum target;
     std::vector<Entry> entries;

src/video_core/renderer_vulkan/fixed_pipeline_state.cpp

@@ -109,6 +109,9 @@ constexpr FixedPipelineState::Rasterizer GetRasterizerState(const Maxwell& regs)
     const auto topology = static_cast<std::size_t>(regs.draw.topology.Value());
     const bool depth_bias_enabled = enabled_lut[PolygonOffsetEnableLUT[topology]];
 
+    const auto& clip = regs.view_volume_clip_control;
+    const bool depth_clamp_enabled = clip.depth_clamp_near == 1 || clip.depth_clamp_far == 1;
+
     Maxwell::Cull::FrontFace front_face = regs.cull.front_face;
     if (regs.screen_y_control.triangle_rast_flip != 0 &&
         regs.viewport_transform[0].scale_y > 0.0f) {
@@ -119,8 +122,9 @@ constexpr FixedPipelineState::Rasterizer GetRasterizerState(const Maxwell& regs)
     }
 
     const bool gl_ndc = regs.depth_mode == Maxwell::DepthMode::MinusOneToOne;
-    return FixedPipelineState::Rasterizer(regs.cull.enabled, depth_bias_enabled, gl_ndc,
-                                          regs.cull.cull_face, front_face);
+    return FixedPipelineState::Rasterizer(regs.cull.enabled, depth_bias_enabled,
+                                          depth_clamp_enabled, gl_ndc, regs.cull.cull_face,
+                                          front_face);
 }
 
 } // Anonymous namespace
@@ -222,15 +226,17 @@ bool FixedPipelineState::Tessellation::operator==(const Tessellation& rhs) const
 std::size_t FixedPipelineState::Rasterizer::Hash() const noexcept {
     return static_cast<std::size_t>(cull_enable) ^
            (static_cast<std::size_t>(depth_bias_enable) << 1) ^
-           (static_cast<std::size_t>(ndc_minus_one_to_one) << 2) ^
+           (static_cast<std::size_t>(depth_clamp_enable) << 2) ^
+           (static_cast<std::size_t>(ndc_minus_one_to_one) << 3) ^
            (static_cast<std::size_t>(cull_face) << 24) ^
            (static_cast<std::size_t>(front_face) << 48);
 }
 
 bool FixedPipelineState::Rasterizer::operator==(const Rasterizer& rhs) const noexcept {
-    return std::tie(cull_enable, depth_bias_enable, ndc_minus_one_to_one, cull_face, front_face) ==
-           std::tie(rhs.cull_enable, rhs.depth_bias_enable, rhs.ndc_minus_one_to_one, rhs.cull_face,
-                    rhs.front_face);
+    return std::tie(cull_enable, depth_bias_enable, depth_clamp_enable, ndc_minus_one_to_one,
+                    cull_face, front_face) ==
+           std::tie(rhs.cull_enable, rhs.depth_bias_enable, rhs.depth_clamp_enable,
+                    rhs.ndc_minus_one_to_one, rhs.cull_face, rhs.front_face);
 }
 
 std::size_t FixedPipelineState::DepthStencil::Hash() const noexcept {

src/video_core/renderer_vulkan/fixed_pipeline_state.h

@@ -170,15 +170,17 @@ struct FixedPipelineState {
     };
 
     struct Rasterizer {
-        constexpr Rasterizer(bool cull_enable, bool depth_bias_enable, bool ndc_minus_one_to_one,
-                             Maxwell::Cull::CullFace cull_face, Maxwell::Cull::FrontFace front_face)
+        constexpr Rasterizer(bool cull_enable, bool depth_bias_enable, bool depth_clamp_enable,
+                             bool ndc_minus_one_to_one, Maxwell::Cull::CullFace cull_face,
+                             Maxwell::Cull::FrontFace front_face)
             : cull_enable{cull_enable}, depth_bias_enable{depth_bias_enable},
-              ndc_minus_one_to_one{ndc_minus_one_to_one}, cull_face{cull_face}, front_face{
-                                                                                    front_face} {}
+              depth_clamp_enable{depth_clamp_enable}, ndc_minus_one_to_one{ndc_minus_one_to_one},
+              cull_face{cull_face}, front_face{front_face} {}
         Rasterizer() = default;
 
         bool cull_enable;
         bool depth_bias_enable;
+        bool depth_clamp_enable;
         bool ndc_minus_one_to_one;
         Maxwell::Cull::CullFace cull_face;
         Maxwell::Cull::FrontFace front_face;

src/video_core/renderer_vulkan/maxwell_to_vk.cpp

@@ -44,7 +44,7 @@ vk::SamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filt
     return {};
 }
 
-vk::SamplerAddressMode WrapMode(Tegra::Texture::WrapMode wrap_mode,
+vk::SamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode,
                                 Tegra::Texture::TextureFilter filter) {
     switch (wrap_mode) {
     case Tegra::Texture::WrapMode::Wrap:
@@ -56,7 +56,12 @@ vk::SamplerAddressMode WrapMode(Tegra::Texture::WrapMode wrap_mode,
     case Tegra::Texture::WrapMode::Border:
         return vk::SamplerAddressMode::eClampToBorder;
     case Tegra::Texture::WrapMode::Clamp:
-        // TODO(Rodrigo): Emulate GL_CLAMP properly
+        if (device.GetDriverID() == vk::DriverIdKHR::eNvidiaProprietary) {
+            // Nvidia's Vulkan driver defaults to GL_CLAMP on invalid enumerations, we can hack this
+            // by sending an invalid enumeration.
+            return static_cast<vk::SamplerAddressMode>(0xcafe);
+        }
+        // TODO(Rodrigo): Emulate GL_CLAMP properly on other vendors
         switch (filter) {
         case Tegra::Texture::TextureFilter::Nearest:
             return vk::SamplerAddressMode::eClampToEdge;

src/video_core/renderer_vulkan/maxwell_to_vk.h

@@ -22,7 +22,7 @@ vk::Filter Filter(Tegra::Texture::TextureFilter filter);
 
 vk::SamplerMipmapMode MipmapMode(Tegra::Texture::TextureMipmapFilter mipmap_filter);
 
-vk::SamplerAddressMode WrapMode(Tegra::Texture::WrapMode wrap_mode,
+vk::SamplerAddressMode WrapMode(const VKDevice& device, Tegra::Texture::WrapMode wrap_mode,
                                 Tegra::Texture::TextureFilter filter);
 
 vk::CompareOp DepthCompareFunction(Tegra::Texture::DepthCompareFunc depth_compare_func);

src/video_core/renderer_vulkan/renderer_vulkan.h

@@ -0,0 +1,72 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <optional>
+#include <vector>
+#include "video_core/renderer_base.h"
+#include "video_core/renderer_vulkan/declarations.h"
+
+namespace Core {
+class System;
+}
+
+namespace Vulkan {
+
+class VKBlitScreen;
+class VKDevice;
+class VKFence;
+class VKMemoryManager;
+class VKResourceManager;
+class VKSwapchain;
+class VKScheduler;
+class VKImage;
+
+struct VKScreenInfo {
+    VKImage* image{};
+    u32 width{};
+    u32 height{};
+    bool is_srgb{};
+};
+
+class RendererVulkan final : public VideoCore::RendererBase {
+public:
+    explicit RendererVulkan(Core::Frontend::EmuWindow& window, Core::System& system);
+    ~RendererVulkan() override;
+
+    /// Swap buffers (render frame)
+    void SwapBuffers(const Tegra::FramebufferConfig* framebuffer) override;
+
+    /// Initialize the renderer
+    bool Init() override;
+
+    /// Shutdown the renderer
+    void ShutDown() override;
+
+private:
+    std::optional<vk::DebugUtilsMessengerEXT> CreateDebugCallback(
+        const vk::DispatchLoaderDynamic& dldi);
+
+    bool PickDevices(const vk::DispatchLoaderDynamic& dldi);
+
+    void Report() const;
+
+    Core::System& system;
+
+    vk::Instance instance;
+    vk::SurfaceKHR surface;
+
+    VKScreenInfo screen_info;
+
+    UniqueDebugUtilsMessengerEXT debug_callback;
+    std::unique_ptr<VKDevice> device;
+    std::unique_ptr<VKSwapchain> swapchain;
+    std::unique_ptr<VKMemoryManager> memory_manager;
+    std::unique_ptr<VKResourceManager> resource_manager;
+    std::unique_ptr<VKScheduler> scheduler;
+    std::unique_ptr<VKBlitScreen> blit_screen;
+};
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_blit_screen.cpp

@@ -0,0 +1,627 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <algorithm>
+#include <array>
+#include <cstring>
+#include <memory>
+#include <tuple>
+#include <vector>
+
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "common/math_util.h"
+
+#include "core/core.h"
+#include "core/frontend/emu_window.h"
+#include "core/memory.h"
+
+#include "video_core/gpu.h"
+#include "video_core/morton.h"
+#include "video_core/rasterizer_interface.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/renderer_vulkan.h"
+#include "video_core/renderer_vulkan/vk_blit_screen.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_image.h"
+#include "video_core/renderer_vulkan/vk_memory_manager.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_shader_util.h"
+#include "video_core/renderer_vulkan/vk_swapchain.h"
+#include "video_core/surface.h"
+
+namespace Vulkan {
+
+namespace {
+
+// Generated from the "shaders/" directory, read the instructions there.
+constexpr u8 blit_vertex_code[] = {
+    0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x27, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x06, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x47, 0x4c, 0x53, 0x4c, 0x2e, 0x73, 0x74, 0x64, 0x2e, 0x34, 0x35, 0x30,
+    0x00, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
+    0x0f, 0x00, 0x09, 0x00, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x6d, 0x61, 0x69, 0x6e,
+    0x00, 0x00, 0x00, 0x00, 0x0d, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x24, 0x00, 0x00, 0x00,
+    0x25, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x0b, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x0b, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00,
+    0x0b, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
+    0x48, 0x00, 0x05, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00,
+    0x04, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x48, 0x00, 0x04, 0x00, 0x11, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x05, 0x00, 0x00, 0x00,
+    0x48, 0x00, 0x05, 0x00, 0x11, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x11, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x07, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x11, 0x00, 0x00, 0x00,
+    0x02, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x13, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x13, 0x00, 0x00, 0x00, 0x21, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x19, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x24, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x25, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x13, 0x00, 0x02, 0x00, 0x02, 0x00, 0x00, 0x00, 0x21, 0x00, 0x03, 0x00,
+    0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x16, 0x00, 0x03, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x20, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x04, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00, 0x08, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x08, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x04, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x06, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
+    0x0c, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00,
+    0x0c, 0x00, 0x00, 0x00, 0x0d, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00,
+    0x0e, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00,
+    0x0e, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x18, 0x00, 0x04, 0x00,
+    0x10, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x03, 0x00,
+    0x11, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x12, 0x00, 0x00, 0x00,
+    0x02, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x12, 0x00, 0x00, 0x00,
+    0x13, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x14, 0x00, 0x00, 0x00,
+    0x02, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x17, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00,
+    0x19, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x1b, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x1c, 0x00, 0x00, 0x00, 0x00, 0x00, 0x80, 0x3f, 0x20, 0x00, 0x04, 0x00, 0x21, 0x00, 0x00, 0x00,
+    0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x23, 0x00, 0x00, 0x00,
+    0x03, 0x00, 0x00, 0x00, 0x17, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x23, 0x00, 0x00, 0x00,
+    0x24, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00,
+    0x25, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x36, 0x00, 0x05, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00,
+    0x05, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00, 0x14, 0x00, 0x00, 0x00, 0x15, 0x00, 0x00, 0x00,
+    0x13, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x10, 0x00, 0x00, 0x00,
+    0x16, 0x00, 0x00, 0x00, 0x15, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00,
+    0x1a, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x51, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x1d, 0x00, 0x00, 0x00, 0x1a, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x51, 0x00, 0x05, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00, 0x1a, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
+    0x50, 0x00, 0x07, 0x00, 0x07, 0x00, 0x00, 0x00, 0x1f, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x00, 0x00,
+    0x1e, 0x00, 0x00, 0x00, 0x1b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x91, 0x00, 0x05, 0x00,
+    0x07, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x16, 0x00, 0x00, 0x00, 0x1f, 0x00, 0x00, 0x00,
+    0x41, 0x00, 0x05, 0x00, 0x21, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x0d, 0x00, 0x00, 0x00,
+    0x0f, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x03, 0x00, 0x22, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00,
+    0x3d, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00, 0x26, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00,
+    0x3e, 0x00, 0x03, 0x00, 0x24, 0x00, 0x00, 0x00, 0x26, 0x00, 0x00, 0x00, 0xfd, 0x00, 0x01, 0x00,
+    0x38, 0x00, 0x01, 0x00};
+
+constexpr u8 blit_fragment_code[] = {
+    0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x14, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x06, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x47, 0x4c, 0x53, 0x4c, 0x2e, 0x73, 0x74, 0x64, 0x2e, 0x34, 0x35, 0x30,
+    0x00, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
+    0x0f, 0x00, 0x07, 0x00, 0x04, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x6d, 0x61, 0x69, 0x6e,
+    0x00, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x10, 0x00, 0x03, 0x00,
+    0x04, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00,
+    0x1e, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x0d, 0x00, 0x00, 0x00,
+    0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x0d, 0x00, 0x00, 0x00,
+    0x21, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x11, 0x00, 0x00, 0x00,
+    0x1e, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x13, 0x00, 0x02, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x21, 0x00, 0x03, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x16, 0x00, 0x03, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x08, 0x00, 0x00, 0x00,
+    0x03, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x08, 0x00, 0x00, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x19, 0x00, 0x09, 0x00, 0x0a, 0x00, 0x00, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1b, 0x00, 0x03, 0x00,
+    0x0b, 0x00, 0x00, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x0c, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x0c, 0x00, 0x00, 0x00,
+    0x0d, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00, 0x0f, 0x00, 0x00, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x10, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x10, 0x00, 0x00, 0x00,
+    0x11, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x36, 0x00, 0x05, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00,
+    0x05, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x00, 0x00,
+    0x0d, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x12, 0x00, 0x00, 0x00,
+    0x11, 0x00, 0x00, 0x00, 0x57, 0x00, 0x05, 0x00, 0x07, 0x00, 0x00, 0x00, 0x13, 0x00, 0x00, 0x00,
+    0x0e, 0x00, 0x00, 0x00, 0x12, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x03, 0x00, 0x09, 0x00, 0x00, 0x00,
+    0x13, 0x00, 0x00, 0x00, 0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00};
+
+struct ScreenRectVertex {
+    ScreenRectVertex() = default;
+    explicit ScreenRectVertex(f32 x, f32 y, f32 u, f32 v) : position{{x, y}}, tex_coord{{u, v}} {}
+
+    std::array<f32, 2> position;
+    std::array<f32, 2> tex_coord;
+
+    static vk::VertexInputBindingDescription GetDescription() {
+        return vk::VertexInputBindingDescription(0, sizeof(ScreenRectVertex),
+                                                 vk::VertexInputRate::eVertex);
+    }
+
+    static std::array<vk::VertexInputAttributeDescription, 2> GetAttributes() {
+        return {vk::VertexInputAttributeDescription(0, 0, vk::Format::eR32G32Sfloat,
+                                                    offsetof(ScreenRectVertex, position)),
+                vk::VertexInputAttributeDescription(1, 0, vk::Format::eR32G32Sfloat,
+                                                    offsetof(ScreenRectVertex, tex_coord))};
+    }
+};
+
+constexpr std::array<f32, 4 * 4> MakeOrthographicMatrix(f32 width, f32 height) {
+    // clang-format off
+    return { 2.f / width, 0.f,          0.f, 0.f,
+             0.f,         2.f / height, 0.f, 0.f,
+             0.f,         0.f,          1.f, 0.f,
+            -1.f,        -1.f,          0.f, 1.f};
+    // clang-format on
+}
+
+std::size_t GetBytesPerPixel(const Tegra::FramebufferConfig& framebuffer) {
+    using namespace VideoCore::Surface;
+    return GetBytesPerPixel(PixelFormatFromGPUPixelFormat(framebuffer.pixel_format));
+}
+
+std::size_t GetSizeInBytes(const Tegra::FramebufferConfig& framebuffer) {
+    return static_cast<std::size_t>(framebuffer.stride) *
+           static_cast<std::size_t>(framebuffer.height) * GetBytesPerPixel(framebuffer);
+}
+
+vk::Format GetFormat(const Tegra::FramebufferConfig& framebuffer) {
+    switch (framebuffer.pixel_format) {
+    case Tegra::FramebufferConfig::PixelFormat::ABGR8:
+        return vk::Format::eA8B8G8R8UnormPack32;
+    case Tegra::FramebufferConfig::PixelFormat::RGB565:
+        return vk::Format::eR5G6B5UnormPack16;
+    default:
+        UNIMPLEMENTED_MSG("Unknown framebuffer pixel format: {}",
+                          static_cast<u32>(framebuffer.pixel_format));
+        return vk::Format::eA8B8G8R8UnormPack32;
+    }
+}
+
+} // Anonymous namespace
+
+struct VKBlitScreen::BufferData {
+    struct {
+        std::array<f32, 4 * 4> modelview_matrix;
+    } uniform;
+
+    std::array<ScreenRectVertex, 4> vertices;
+
+    // Unaligned image data goes here
+};
+
+VKBlitScreen::VKBlitScreen(Core::System& system, Core::Frontend::EmuWindow& render_window,
+                           VideoCore::RasterizerInterface& rasterizer, const VKDevice& device,
+                           VKResourceManager& resource_manager, VKMemoryManager& memory_manager,
+                           VKSwapchain& swapchain, VKScheduler& scheduler,
+                           const VKScreenInfo& screen_info)
+    : system{system}, render_window{render_window}, rasterizer{rasterizer}, device{device},
+      resource_manager{resource_manager}, memory_manager{memory_manager}, swapchain{swapchain},
+      scheduler{scheduler}, image_count{swapchain.GetImageCount()}, screen_info{screen_info} {
+    watches.resize(image_count);
+    std::generate(watches.begin(), watches.end(),
+                  []() { return std::make_unique<VKFenceWatch>(); });
+
+    CreateStaticResources();
+    CreateDynamicResources();
+}
+
+VKBlitScreen::~VKBlitScreen() = default;
+
+void VKBlitScreen::Recreate() {
+    CreateDynamicResources();
+}
+
+std::tuple<VKFence&, vk::Semaphore> VKBlitScreen::Draw(const Tegra::FramebufferConfig& framebuffer,
+                                                       bool use_accelerated) {
+    RefreshResources(framebuffer);
+
+    // Finish any pending renderpass
+    scheduler.RequestOutsideRenderPassOperationContext();
+
+    const std::size_t image_index = swapchain.GetImageIndex();
+    watches[image_index]->Watch(scheduler.GetFence());
+
+    VKImage* blit_image = use_accelerated ? screen_info.image : raw_images[image_index].get();
+
+    UpdateDescriptorSet(image_index, blit_image->GetPresentView());
+
+    BufferData data;
+    SetUniformData(data, framebuffer);
+    SetVertexData(data, framebuffer);
+
+    auto map = buffer_commit->Map();
+    std::memcpy(map.GetAddress(), &data, sizeof(data));
+
+    if (!use_accelerated) {
+        const u64 image_offset = GetRawImageOffset(framebuffer, image_index);
+
+        const auto pixel_format =
+            VideoCore::Surface::PixelFormatFromGPUPixelFormat(framebuffer.pixel_format);
+        const VAddr framebuffer_addr = framebuffer.address + framebuffer.offset;
+        const auto host_ptr = system.Memory().GetPointer(framebuffer_addr);
+        rasterizer.FlushRegion(ToCacheAddr(host_ptr), GetSizeInBytes(framebuffer));
+
+        // TODO(Rodrigo): Read this from HLE
+        constexpr u32 block_height_log2 = 4;
+        VideoCore::MortonSwizzle(VideoCore::MortonSwizzleMode::MortonToLinear, pixel_format,
+                                 framebuffer.stride, block_height_log2, framebuffer.height, 0, 1, 1,
+                                 map.GetAddress() + image_offset, host_ptr);
+
+        blit_image->Transition(0, 1, 0, 1, vk::PipelineStageFlagBits::eTransfer,
+                               vk::AccessFlagBits::eTransferWrite,
+                               vk::ImageLayout::eTransferDstOptimal);
+
+        const vk::BufferImageCopy copy(image_offset, 0, 0,
+                                       {vk::ImageAspectFlagBits::eColor, 0, 0, 1}, {0, 0, 0},
+                                       {framebuffer.width, framebuffer.height, 1});
+        scheduler.Record([buffer_handle = *buffer, image = blit_image->GetHandle(),
+                          copy](auto cmdbuf, auto& dld) {
+            cmdbuf.copyBufferToImage(buffer_handle, image, vk::ImageLayout::eTransferDstOptimal,
+                                     {copy}, dld);
+        });
+    }
+    map.Release();
+
+    blit_image->Transition(0, 1, 0, 1, vk::PipelineStageFlagBits::eFragmentShader,
+                           vk::AccessFlagBits::eShaderRead,
+                           vk::ImageLayout::eShaderReadOnlyOptimal);
+
+    scheduler.Record([renderpass = *renderpass, framebuffer = *framebuffers[image_index],
+                      descriptor_set = descriptor_sets[image_index], buffer = *buffer,
+                      size = swapchain.GetSize(), pipeline = *pipeline,
+                      layout = *pipeline_layout](auto cmdbuf, auto& dld) {
+        const vk::ClearValue clear_color{std::array{0.0f, 0.0f, 0.0f, 1.0f}};
+        const vk::RenderPassBeginInfo renderpass_bi(renderpass, framebuffer, {{0, 0}, size}, 1,
+                                                    &clear_color);
+
+        cmdbuf.beginRenderPass(renderpass_bi, vk::SubpassContents::eInline, dld);
+        cmdbuf.bindPipeline(vk::PipelineBindPoint::eGraphics, pipeline, dld);
+        cmdbuf.setViewport(
+            0,
+            {{0.0f, 0.0f, static_cast<f32>(size.width), static_cast<f32>(size.height), 0.0f, 1.0f}},
+            dld);
+        cmdbuf.setScissor(0, {{{0, 0}, size}}, dld);
+
+        cmdbuf.bindVertexBuffers(0, {buffer}, {offsetof(BufferData, vertices)}, dld);
+        cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eGraphics, layout, 0, {descriptor_set}, {},
+                                  dld);
+        cmdbuf.draw(4, 1, 0, 0, dld);
+        cmdbuf.endRenderPass(dld);
+    });
+
+    return {scheduler.GetFence(), *semaphores[image_index]};
+}
+
+void VKBlitScreen::CreateStaticResources() {
+    CreateShaders();
+    CreateSemaphores();
+    CreateDescriptorPool();
+    CreateDescriptorSetLayout();
+    CreateDescriptorSets();
+    CreatePipelineLayout();
+    CreateSampler();
+}
+
+void VKBlitScreen::CreateDynamicResources() {
+    CreateRenderPass();
+    CreateFramebuffers();
+    CreateGraphicsPipeline();
+}
+
+void VKBlitScreen::RefreshResources(const Tegra::FramebufferConfig& framebuffer) {
+    if (framebuffer.width == raw_width && framebuffer.height == raw_height && !raw_images.empty()) {
+        return;
+    }
+    raw_width = framebuffer.width;
+    raw_height = framebuffer.height;
+    ReleaseRawImages();
+
+    CreateStagingBuffer(framebuffer);
+    CreateRawImages(framebuffer);
+}
+
+void VKBlitScreen::CreateShaders() {
+    vertex_shader = BuildShader(device, sizeof(blit_vertex_code), blit_vertex_code);
+    fragment_shader = BuildShader(device, sizeof(blit_fragment_code), blit_fragment_code);
+}
+
+void VKBlitScreen::CreateSemaphores() {
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+
+    semaphores.resize(image_count);
+    for (std::size_t i = 0; i < image_count; ++i) {
+        semaphores[i] = dev.createSemaphoreUnique({}, nullptr, dld);
+    }
+}
+
+void VKBlitScreen::CreateDescriptorPool() {
+    const std::array<vk::DescriptorPoolSize, 2> pool_sizes{
+        vk::DescriptorPoolSize{vk::DescriptorType::eUniformBuffer, static_cast<u32>(image_count)},
+        vk::DescriptorPoolSize{vk::DescriptorType::eCombinedImageSampler,
+                               static_cast<u32>(image_count)}};
+    const vk::DescriptorPoolCreateInfo pool_ci(
+        {}, static_cast<u32>(image_count), static_cast<u32>(pool_sizes.size()), pool_sizes.data());
+    const auto dev = device.GetLogical();
+    descriptor_pool = dev.createDescriptorPoolUnique(pool_ci, nullptr, device.GetDispatchLoader());
+}
+
+void VKBlitScreen::CreateRenderPass() {
+    const vk::AttachmentDescription color_attachment(
+        {}, swapchain.GetImageFormat(), vk::SampleCountFlagBits::e1, vk::AttachmentLoadOp::eClear,
+        vk::AttachmentStoreOp::eStore, vk::AttachmentLoadOp::eDontCare,
+        vk::AttachmentStoreOp::eDontCare, vk::ImageLayout::eUndefined,
+        vk::ImageLayout::ePresentSrcKHR);
+
+    const vk::AttachmentReference color_attachment_ref(0, vk::ImageLayout::eColorAttachmentOptimal);
+
+    const vk::SubpassDescription subpass_description({}, vk::PipelineBindPoint::eGraphics, 0,
+                                                     nullptr, 1, &color_attachment_ref, nullptr,
+                                                     nullptr, 0, nullptr);
+
+    const vk::SubpassDependency dependency(
+        VK_SUBPASS_EXTERNAL, 0, vk::PipelineStageFlagBits::eColorAttachmentOutput,
+        vk::PipelineStageFlagBits::eColorAttachmentOutput, {},
+        vk::AccessFlagBits::eColorAttachmentRead | vk::AccessFlagBits::eColorAttachmentWrite, {});
+
+    const vk::RenderPassCreateInfo renderpass_ci({}, 1, &color_attachment, 1, &subpass_description,
+                                                 1, &dependency);
+
+    const auto dev = device.GetLogical();
+    renderpass = dev.createRenderPassUnique(renderpass_ci, nullptr, device.GetDispatchLoader());
+}
+
+void VKBlitScreen::CreateDescriptorSetLayout() {
+    const std::array<vk::DescriptorSetLayoutBinding, 2> layout_bindings{
+        vk::DescriptorSetLayoutBinding(0, vk::DescriptorType::eUniformBuffer, 1,
+                                       vk::ShaderStageFlagBits::eVertex, nullptr),
+        vk::DescriptorSetLayoutBinding(1, vk::DescriptorType::eCombinedImageSampler, 1,
+                                       vk::ShaderStageFlagBits::eFragment, nullptr)};
+    const vk::DescriptorSetLayoutCreateInfo descriptor_layout_ci(
+        {}, static_cast<u32>(layout_bindings.size()), layout_bindings.data());
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    descriptor_set_layout = dev.createDescriptorSetLayoutUnique(descriptor_layout_ci, nullptr, dld);
+}
+
+void VKBlitScreen::CreateDescriptorSets() {
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+
+    descriptor_sets.resize(image_count);
+    for (std::size_t i = 0; i < image_count; ++i) {
+        const vk::DescriptorSetLayout layout = *descriptor_set_layout;
+        const vk::DescriptorSetAllocateInfo descriptor_set_ai(*descriptor_pool, 1, &layout);
+        const vk::Result result =
+            dev.allocateDescriptorSets(&descriptor_set_ai, &descriptor_sets[i], dld);
+        ASSERT(result == vk::Result::eSuccess);
+    }
+}
+
+void VKBlitScreen::CreatePipelineLayout() {
+    const vk::PipelineLayoutCreateInfo pipeline_layout_ci({}, 1, &descriptor_set_layout.get(), 0,
+                                                          nullptr);
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    pipeline_layout = dev.createPipelineLayoutUnique(pipeline_layout_ci, nullptr, dld);
+}
+
+void VKBlitScreen::CreateGraphicsPipeline() {
+    const std::array shader_stages = {
+        vk::PipelineShaderStageCreateInfo({}, vk::ShaderStageFlagBits::eVertex, *vertex_shader,
+                                          "main", nullptr),
+        vk::PipelineShaderStageCreateInfo({}, vk::ShaderStageFlagBits::eFragment, *fragment_shader,
+                                          "main", nullptr)};
+
+    const auto vertex_binding_description = ScreenRectVertex::GetDescription();
+    const auto vertex_attrs_description = ScreenRectVertex::GetAttributes();
+    const vk::PipelineVertexInputStateCreateInfo vertex_input(
+        {}, 1, &vertex_binding_description, static_cast<u32>(vertex_attrs_description.size()),
+        vertex_attrs_description.data());
+
+    const vk::PipelineInputAssemblyStateCreateInfo input_assembly(
+        {}, vk::PrimitiveTopology::eTriangleStrip, false);
+
+    // Set a dummy viewport, it's going to be replaced by dynamic states.
+    const vk::Viewport viewport(0.0f, 0.0f, 1.0f, 1.0f, 0.0f, 1.0f);
+    const vk::Rect2D scissor({0, 0}, {1, 1});
+
+    const vk::PipelineViewportStateCreateInfo viewport_state({}, 1, &viewport, 1, &scissor);
+
+    const vk::PipelineRasterizationStateCreateInfo rasterizer(
+        {}, false, false, vk::PolygonMode::eFill, vk::CullModeFlagBits::eNone,
+        vk::FrontFace::eClockwise, false, 0.0f, 0.0f, 0.0f, 1.0f);
+
+    const vk::PipelineMultisampleStateCreateInfo multisampling({}, vk::SampleCountFlagBits::e1,
+                                                               false, 0.0f, nullptr, false, false);
+
+    const vk::PipelineColorBlendAttachmentState color_blend_attachment(
+        false, vk::BlendFactor::eZero, vk::BlendFactor::eZero, vk::BlendOp::eAdd,
+        vk::BlendFactor::eZero, vk::BlendFactor::eZero, vk::BlendOp::eAdd,
+        vk::ColorComponentFlagBits::eR | vk::ColorComponentFlagBits::eG |
+            vk::ColorComponentFlagBits::eB | vk::ColorComponentFlagBits::eA);
+
+    const vk::PipelineColorBlendStateCreateInfo color_blending(
+        {}, false, vk::LogicOp::eCopy, 1, &color_blend_attachment, {0.0f, 0.0f, 0.0f, 0.0f});
+
+    const std::array<vk::DynamicState, 2> dynamic_states = {vk::DynamicState::eViewport,
+                                                            vk::DynamicState::eScissor};
+
+    const vk::PipelineDynamicStateCreateInfo dynamic_state(
+        {}, static_cast<u32>(dynamic_states.size()), dynamic_states.data());
+
+    const vk::GraphicsPipelineCreateInfo pipeline_ci(
+        {}, static_cast<u32>(shader_stages.size()), shader_stages.data(), &vertex_input,
+        &input_assembly, nullptr, &viewport_state, &rasterizer, &multisampling, nullptr,
+        &color_blending, &dynamic_state, *pipeline_layout, *renderpass, 0, nullptr, 0);
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    pipeline = dev.createGraphicsPipelineUnique({}, pipeline_ci, nullptr, dld);
+}
+
+void VKBlitScreen::CreateSampler() {
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    const vk::SamplerCreateInfo sampler_ci(
+        {}, vk::Filter::eLinear, vk::Filter::eLinear, vk::SamplerMipmapMode::eLinear,
+        vk::SamplerAddressMode::eClampToBorder, vk::SamplerAddressMode::eClampToBorder,
+        vk::SamplerAddressMode::eClampToBorder, 0.0f, false, 0.0f, false, vk::CompareOp::eNever,
+        0.0f, 0.0f, vk::BorderColor::eFloatOpaqueBlack, false);
+    sampler = dev.createSamplerUnique(sampler_ci, nullptr, dld);
+}
+
+void VKBlitScreen::CreateFramebuffers() {
+    const vk::Extent2D size{swapchain.GetSize()};
+    framebuffers.clear();
+    framebuffers.resize(image_count);
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+
+    for (std::size_t i = 0; i < image_count; ++i) {
+        const vk::ImageView image_view{swapchain.GetImageViewIndex(i)};
+        const vk::FramebufferCreateInfo framebuffer_ci({}, *renderpass, 1, &image_view, size.width,
+                                                       size.height, 1);
+        framebuffers[i] = dev.createFramebufferUnique(framebuffer_ci, nullptr, dld);
+    }
+}
+
+void VKBlitScreen::ReleaseRawImages() {
+    for (std::size_t i = 0; i < raw_images.size(); ++i) {
+        watches[i]->Wait();
+    }
+    raw_images.clear();
+    raw_buffer_commits.clear();
+    buffer.reset();
+    buffer_commit.reset();
+}
+
+void VKBlitScreen::CreateStagingBuffer(const Tegra::FramebufferConfig& framebuffer) {
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+
+    const vk::BufferCreateInfo buffer_ci({}, CalculateBufferSize(framebuffer),
+                                         vk::BufferUsageFlagBits::eTransferSrc |
+                                             vk::BufferUsageFlagBits::eVertexBuffer |
+                                             vk::BufferUsageFlagBits::eUniformBuffer,
+                                         vk::SharingMode::eExclusive, 0, nullptr);
+    buffer = dev.createBufferUnique(buffer_ci, nullptr, dld);
+    buffer_commit = memory_manager.Commit(*buffer, true);
+}
+
+void VKBlitScreen::CreateRawImages(const Tegra::FramebufferConfig& framebuffer) {
+    raw_images.resize(image_count);
+    raw_buffer_commits.resize(image_count);
+
+    const auto format = GetFormat(framebuffer);
+    for (std::size_t i = 0; i < image_count; ++i) {
+        const vk::ImageCreateInfo image_ci(
+            {}, vk::ImageType::e2D, format, {framebuffer.width, framebuffer.height, 1}, 1, 1,
+            vk::SampleCountFlagBits::e1, vk::ImageTiling::eOptimal,
+            vk::ImageUsageFlagBits::eTransferDst | vk::ImageUsageFlagBits::eSampled,
+            vk::SharingMode::eExclusive, 0, nullptr, vk::ImageLayout::eUndefined);
+
+        raw_images[i] =
+            std::make_unique<VKImage>(device, scheduler, image_ci, vk::ImageAspectFlagBits::eColor);
+        raw_buffer_commits[i] = memory_manager.Commit(raw_images[i]->GetHandle(), false);
+    }
+}
+
+void VKBlitScreen::UpdateDescriptorSet(std::size_t image_index, vk::ImageView image_view) const {
+    const vk::DescriptorSet descriptor_set = descriptor_sets[image_index];
+
+    const vk::DescriptorBufferInfo buffer_info(*buffer, offsetof(BufferData, uniform),
+                                               sizeof(BufferData::uniform));
+    const vk::WriteDescriptorSet ubo_write(descriptor_set, 0, 0, 1,
+                                           vk::DescriptorType::eUniformBuffer, nullptr,
+                                           &buffer_info, nullptr);
+
+    const vk::DescriptorImageInfo image_info(*sampler, image_view,
+                                             vk::ImageLayout::eShaderReadOnlyOptimal);
+    const vk::WriteDescriptorSet sampler_write(descriptor_set, 1, 0, 1,
+                                               vk::DescriptorType::eCombinedImageSampler,
+                                               &image_info, nullptr, nullptr);
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    dev.updateDescriptorSets({ubo_write, sampler_write}, {}, dld);
+}
+
+void VKBlitScreen::SetUniformData(BufferData& data,
+                                  const Tegra::FramebufferConfig& framebuffer) const {
+    const auto& layout = render_window.GetFramebufferLayout();
+    data.uniform.modelview_matrix =
+        MakeOrthographicMatrix(static_cast<f32>(layout.width), static_cast<f32>(layout.height));
+}
+
+void VKBlitScreen::SetVertexData(BufferData& data,
+                                 const Tegra::FramebufferConfig& framebuffer) const {
+    const auto& framebuffer_transform_flags = framebuffer.transform_flags;
+    const auto& framebuffer_crop_rect = framebuffer.crop_rect;
+
+    static constexpr Common::Rectangle<f32> texcoords{0.f, 0.f, 1.f, 1.f};
+    auto left = texcoords.left;
+    auto right = texcoords.right;
+
+    switch (framebuffer_transform_flags) {
+    case Tegra::FramebufferConfig::TransformFlags::Unset:
+        break;
+    case Tegra::FramebufferConfig::TransformFlags::FlipV:
+        // Flip the framebuffer vertically
+        left = texcoords.right;
+        right = texcoords.left;
+        break;
+    default:
+        UNIMPLEMENTED_MSG("Unsupported framebuffer_transform_flags={}",
+                          static_cast<u32>(framebuffer_transform_flags));
+        break;
+    }
+
+    UNIMPLEMENTED_IF(framebuffer_crop_rect.top != 0);
+    UNIMPLEMENTED_IF(framebuffer_crop_rect.left != 0);
+
+    // Scale the output by the crop width/height. This is commonly used with 1280x720 rendering
+    // (e.g. handheld mode) on a 1920x1080 framebuffer.
+    f32 scale_u = 1.0f;
+    f32 scale_v = 1.0f;
+    if (framebuffer_crop_rect.GetWidth() > 0) {
+        scale_u = static_cast<f32>(framebuffer_crop_rect.GetWidth()) /
+                  static_cast<f32>(screen_info.width);
+    }
+    if (framebuffer_crop_rect.GetHeight() > 0) {
+        scale_v = static_cast<f32>(framebuffer_crop_rect.GetHeight()) /
+                  static_cast<f32>(screen_info.height);
+    }
+
+    const auto& screen = render_window.GetFramebufferLayout().screen;
+    const auto x = static_cast<f32>(screen.left);
+    const auto y = static_cast<f32>(screen.top);
+    const auto w = static_cast<f32>(screen.GetWidth());
+    const auto h = static_cast<f32>(screen.GetHeight());
+    data.vertices[0] = ScreenRectVertex(x, y, texcoords.top * scale_u, left * scale_v);
+    data.vertices[1] = ScreenRectVertex(x + w, y, texcoords.bottom * scale_u, left * scale_v);
+    data.vertices[2] = ScreenRectVertex(x, y + h, texcoords.top * scale_u, right * scale_v);
+    data.vertices[3] = ScreenRectVertex(x + w, y + h, texcoords.bottom * scale_u, right * scale_v);
+}
+
+u64 VKBlitScreen::CalculateBufferSize(const Tegra::FramebufferConfig& framebuffer) const {
+    return sizeof(BufferData) + GetSizeInBytes(framebuffer) * image_count;
+}
+
+u64 VKBlitScreen::GetRawImageOffset(const Tegra::FramebufferConfig& framebuffer,
+                                    std::size_t image_index) const {
+    constexpr auto first_image_offset = static_cast<u64>(sizeof(BufferData));
+    return first_image_offset + GetSizeInBytes(framebuffer) * image_index;
+}
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_blit_screen.h

@@ -0,0 +1,119 @@
+// Copyright 2018 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <memory>
+#include <tuple>
+
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_memory_manager.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+
+namespace Core {
+class System;
+}
+
+namespace Core::Frontend {
+class EmuWindow;
+}
+
+namespace Tegra {
+struct FramebufferConfig;
+}
+
+namespace VideoCore {
+class RasterizerInterface;
+}
+
+namespace Vulkan {
+
+struct ScreenInfo;
+class RasterizerVulkan;
+class VKDevice;
+class VKFence;
+class VKImage;
+class VKScheduler;
+class VKSwapchain;
+
+class VKBlitScreen final {
+public:
+    explicit VKBlitScreen(Core::System& system, Core::Frontend::EmuWindow& render_window,
+                          VideoCore::RasterizerInterface& rasterizer, const VKDevice& device,
+                          VKResourceManager& resource_manager, VKMemoryManager& memory_manager,
+                          VKSwapchain& swapchain, VKScheduler& scheduler,
+                          const VKScreenInfo& screen_info);
+    ~VKBlitScreen();
+
+    void Recreate();
+
+    std::tuple<VKFence&, vk::Semaphore> Draw(const Tegra::FramebufferConfig& framebuffer,
+                                             bool use_accelerated);
+
+private:
+    struct BufferData;
+
+    void CreateStaticResources();
+    void CreateShaders();
+    void CreateSemaphores();
+    void CreateDescriptorPool();
+    void CreateRenderPass();
+    void CreateDescriptorSetLayout();
+    void CreateDescriptorSets();
+    void CreatePipelineLayout();
+    void CreateGraphicsPipeline();
+    void CreateSampler();
+
+    void CreateDynamicResources();
+    void CreateFramebuffers();
+
+    void RefreshResources(const Tegra::FramebufferConfig& framebuffer);
+    void ReleaseRawImages();
+    void CreateStagingBuffer(const Tegra::FramebufferConfig& framebuffer);
+    void CreateRawImages(const Tegra::FramebufferConfig& framebuffer);
+
+    void UpdateDescriptorSet(std::size_t image_index, vk::ImageView image_view) const;
+    void SetUniformData(BufferData& data, const Tegra::FramebufferConfig& framebuffer) const;
+    void SetVertexData(BufferData& data, const Tegra::FramebufferConfig& framebuffer) const;
+
+    u64 CalculateBufferSize(const Tegra::FramebufferConfig& framebuffer) const;
+    u64 GetRawImageOffset(const Tegra::FramebufferConfig& framebuffer,
+                          std::size_t image_index) const;
+
+    Core::System& system;
+    Core::Frontend::EmuWindow& render_window;
+    VideoCore::RasterizerInterface& rasterizer;
+    const VKDevice& device;
+    VKResourceManager& resource_manager;
+    VKMemoryManager& memory_manager;
+    VKSwapchain& swapchain;
+    VKScheduler& scheduler;
+    const std::size_t image_count;
+    const VKScreenInfo& screen_info;
+
+    UniqueShaderModule vertex_shader;
+    UniqueShaderModule fragment_shader;
+    UniqueDescriptorPool descriptor_pool;
+    UniqueDescriptorSetLayout descriptor_set_layout;
+    UniquePipelineLayout pipeline_layout;
+    UniquePipeline pipeline;
+    UniqueRenderPass renderpass;
+    std::vector<UniqueFramebuffer> framebuffers;
+    std::vector<vk::DescriptorSet> descriptor_sets;
+    UniqueSampler sampler;
+
+    UniqueBuffer buffer;
+    VKMemoryCommit buffer_commit;
+
+    std::vector<std::unique_ptr<VKFenceWatch>> watches;
+
+    std::vector<UniqueSemaphore> semaphores;
+    std::vector<std::unique_ptr<VKImage>> raw_images;
+    std::vector<VKMemoryCommit> raw_buffer_commits;
+    u32 raw_width = 0;
+    u32 raw_height = 0;
+};
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_buffer_cache.cpp

@@ -2,124 +2,145 @@
 // Licensed under GPLv2 or any later version
 // Refer to the license.txt file included.
 
+#include <algorithm>
 #include <cstring>
 #include <memory>
 #include <optional>
 #include <tuple>
 
-#include "common/alignment.h"
 #include "common/assert.h"
-#include "core/memory.h"
-#include "video_core/memory_manager.h"
+#include "common/bit_util.h"
+#include "core/core.h"
 #include "video_core/renderer_vulkan/declarations.h"
 #include "video_core/renderer_vulkan/vk_buffer_cache.h"
+#include "video_core/renderer_vulkan/vk_device.h"
 #include "video_core/renderer_vulkan/vk_scheduler.h"
 #include "video_core/renderer_vulkan/vk_stream_buffer.h"
 
 namespace Vulkan {
 
-CachedBufferEntry::CachedBufferEntry(VAddr cpu_addr, std::size_t size, u64 offset,
-                                     std::size_t alignment, u8* host_ptr)
-    : RasterizerCacheObject{host_ptr}, cpu_addr{cpu_addr}, size{size}, offset{offset},
-      alignment{alignment} {}
+namespace {
 
-VKBufferCache::VKBufferCache(Tegra::MemoryManager& tegra_memory_manager,
-                             Memory::Memory& cpu_memory_,
-                             VideoCore::RasterizerInterface& rasterizer, const VKDevice& device,
-                             VKMemoryManager& memory_manager, VKScheduler& scheduler, u64 size)
-    : RasterizerCache{rasterizer}, tegra_memory_manager{tegra_memory_manager}, cpu_memory{
-                                                                                   cpu_memory_} {
-    const auto usage = vk::BufferUsageFlagBits::eVertexBuffer |
-                       vk::BufferUsageFlagBits::eIndexBuffer |
-                       vk::BufferUsageFlagBits::eUniformBuffer;
-    const auto access = vk::AccessFlagBits::eVertexAttributeRead | vk::AccessFlagBits::eIndexRead |
-                        vk::AccessFlagBits::eUniformRead;
-    stream_buffer =
-        std::make_unique<VKStreamBuffer>(device, memory_manager, scheduler, size, usage, access,
-                                         vk::PipelineStageFlagBits::eAllCommands);
-    buffer_handle = stream_buffer->GetBuffer();
+const auto BufferUsage =
+    vk::BufferUsageFlagBits::eVertexBuffer | vk::BufferUsageFlagBits::eIndexBuffer |
+    vk::BufferUsageFlagBits::eUniformBuffer | vk::BufferUsageFlagBits::eStorageBuffer;
+
+const auto UploadPipelineStage =
+    vk::PipelineStageFlagBits::eTransfer | vk::PipelineStageFlagBits::eVertexInput |
+    vk::PipelineStageFlagBits::eVertexShader | vk::PipelineStageFlagBits::eFragmentShader |
+    vk::PipelineStageFlagBits::eComputeShader;
+
+const auto UploadAccessBarriers =
+    vk::AccessFlagBits::eTransferRead | vk::AccessFlagBits::eShaderRead |
+    vk::AccessFlagBits::eUniformRead | vk::AccessFlagBits::eVertexAttributeRead |
+    vk::AccessFlagBits::eIndexRead;
+
+auto CreateStreamBuffer(const VKDevice& device, VKScheduler& scheduler) {
+    return std::make_unique<VKStreamBuffer>(device, scheduler, BufferUsage);
 }
 
+} // Anonymous namespace
+
+CachedBufferBlock::CachedBufferBlock(const VKDevice& device, VKMemoryManager& memory_manager,
+                                     CacheAddr cache_addr, std::size_t size)
+    : VideoCommon::BufferBlock{cache_addr, size} {
+    const vk::BufferCreateInfo buffer_ci({}, static_cast<vk::DeviceSize>(size),
+                                         BufferUsage | vk::BufferUsageFlagBits::eTransferSrc |
+                                             vk::BufferUsageFlagBits::eTransferDst,
+                                         vk::SharingMode::eExclusive, 0, nullptr);
+
+    const auto& dld{device.GetDispatchLoader()};
+    const auto dev{device.GetLogical()};
+    buffer.handle = dev.createBufferUnique(buffer_ci, nullptr, dld);
+    buffer.commit = memory_manager.Commit(*buffer.handle, false);
+}
+
+CachedBufferBlock::~CachedBufferBlock() = default;
+
+VKBufferCache::VKBufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system,
+                             const VKDevice& device, VKMemoryManager& memory_manager,
+                             VKScheduler& scheduler, VKStagingBufferPool& staging_pool)
+    : VideoCommon::BufferCache<Buffer, vk::Buffer, VKStreamBuffer>{rasterizer, system,
+                                                                   CreateStreamBuffer(device,
+                                                                                      scheduler)},
+      device{device}, memory_manager{memory_manager}, scheduler{scheduler}, staging_pool{
+                                                                                staging_pool} {}
+
 VKBufferCache::~VKBufferCache() = default;
 
-u64 VKBufferCache::UploadMemory(GPUVAddr gpu_addr, std::size_t size, u64 alignment, bool cache) {
-    const auto cpu_addr{tegra_memory_manager.GpuToCpuAddress(gpu_addr)};
-    ASSERT_MSG(cpu_addr, "Invalid GPU address");
-
-    // Cache management is a big overhead, so only cache entries with a given size.
-    // TODO: Figure out which size is the best for given games.
-    cache &= size >= 2048;
-
-    u8* const host_ptr{cpu_memory.GetPointer(*cpu_addr)};
-    if (cache) {
-        const auto entry = TryGet(host_ptr);
-        if (entry) {
-            if (entry->GetSize() >= size && entry->GetAlignment() == alignment) {
-                return entry->GetOffset();
-            }
-            Unregister(entry);
-        }
-    }
-
-    AlignBuffer(alignment);
-    const u64 uploaded_offset = buffer_offset;
-
-    if (host_ptr == nullptr) {
-        return uploaded_offset;
-    }
-
-    std::memcpy(buffer_ptr, host_ptr, size);
-    buffer_ptr += size;
-    buffer_offset += size;
-
-    if (cache) {
-        auto entry = std::make_shared<CachedBufferEntry>(*cpu_addr, size, uploaded_offset,
-                                                         alignment, host_ptr);
-        Register(entry);
-    }
-
-    return uploaded_offset;
+Buffer VKBufferCache::CreateBlock(CacheAddr cache_addr, std::size_t size) {
+    return std::make_shared<CachedBufferBlock>(device, memory_manager, cache_addr, size);
 }
 
-u64 VKBufferCache::UploadHostMemory(const u8* raw_pointer, std::size_t size, u64 alignment) {
-    AlignBuffer(alignment);
-    std::memcpy(buffer_ptr, raw_pointer, size);
-    const u64 uploaded_offset = buffer_offset;
-
-    buffer_ptr += size;
-    buffer_offset += size;
-    return uploaded_offset;
+const vk::Buffer* VKBufferCache::ToHandle(const Buffer& buffer) {
+    return buffer->GetHandle();
 }
 
-std::tuple<u8*, u64> VKBufferCache::ReserveMemory(std::size_t size, u64 alignment) {
-    AlignBuffer(alignment);
-    u8* const uploaded_ptr = buffer_ptr;
-    const u64 uploaded_offset = buffer_offset;
-
-    buffer_ptr += size;
-    buffer_offset += size;
-    return {uploaded_ptr, uploaded_offset};
+const vk::Buffer* VKBufferCache::GetEmptyBuffer(std::size_t size) {
+    size = std::max(size, std::size_t(4));
+    const auto& empty = staging_pool.GetUnusedBuffer(size, false);
+    scheduler.RequestOutsideRenderPassOperationContext();
+    scheduler.Record([size, buffer = *empty.handle](vk::CommandBuffer cmdbuf, auto& dld) {
+        cmdbuf.fillBuffer(buffer, 0, size, 0, dld);
+    });
+    return &*empty.handle;
 }
 
-void VKBufferCache::Reserve(std::size_t max_size) {
-    bool invalidate;
-    std::tie(buffer_ptr, buffer_offset_base, invalidate) = stream_buffer->Reserve(max_size);
-    buffer_offset = buffer_offset_base;
+void VKBufferCache::UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
+                                    const u8* data) {
+    const auto& staging = staging_pool.GetUnusedBuffer(size, true);
+    std::memcpy(staging.commit->Map(size), data, size);
 
-    if (invalidate) {
-        InvalidateAll();
-    }
+    scheduler.RequestOutsideRenderPassOperationContext();
+    scheduler.Record([staging = *staging.handle, buffer = *buffer->GetHandle(), offset,
+                      size](auto cmdbuf, auto& dld) {
+        cmdbuf.copyBuffer(staging, buffer, {{0, offset, size}}, dld);
+        cmdbuf.pipelineBarrier(
+            vk::PipelineStageFlagBits::eTransfer, UploadPipelineStage, {}, {},
+            {vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferWrite, UploadAccessBarriers,
+                                     VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, buffer,
+                                     offset, size)},
+            {}, dld);
+    });
 }
 
-void VKBufferCache::Send() {
-    stream_buffer->Send(buffer_offset - buffer_offset_base);
+void VKBufferCache::DownloadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
+                                      u8* data) {
+    const auto& staging = staging_pool.GetUnusedBuffer(size, true);
+    scheduler.RequestOutsideRenderPassOperationContext();
+    scheduler.Record([staging = *staging.handle, buffer = *buffer->GetHandle(), offset,
+                      size](auto cmdbuf, auto& dld) {
+        cmdbuf.pipelineBarrier(
+            vk::PipelineStageFlagBits::eVertexShader | vk::PipelineStageFlagBits::eFragmentShader |
+                vk::PipelineStageFlagBits::eComputeShader,
+            vk::PipelineStageFlagBits::eTransfer, {}, {},
+            {vk::BufferMemoryBarrier(vk::AccessFlagBits::eShaderWrite,
+                                     vk::AccessFlagBits::eTransferRead, VK_QUEUE_FAMILY_IGNORED,
+                                     VK_QUEUE_FAMILY_IGNORED, buffer, offset, size)},
+            {}, dld);
+        cmdbuf.copyBuffer(buffer, staging, {{offset, 0, size}}, dld);
+    });
+    scheduler.Finish();
+
+    std::memcpy(data, staging.commit->Map(size), size);
 }
 
-void VKBufferCache::AlignBuffer(std::size_t alignment) {
-    // Align the offset, not the mapped pointer
-    const u64 offset_aligned = Common::AlignUp(buffer_offset, alignment);
-    buffer_ptr += offset_aligned - buffer_offset;
-    buffer_offset = offset_aligned;
+void VKBufferCache::CopyBlock(const Buffer& src, const Buffer& dst, std::size_t src_offset,
+                              std::size_t dst_offset, std::size_t size) {
+    scheduler.RequestOutsideRenderPassOperationContext();
+    scheduler.Record([src_buffer = *src->GetHandle(), dst_buffer = *dst->GetHandle(), src_offset,
+                      dst_offset, size](auto cmdbuf, auto& dld) {
+        cmdbuf.copyBuffer(src_buffer, dst_buffer, {{src_offset, dst_offset, size}}, dld);
+        cmdbuf.pipelineBarrier(
+            vk::PipelineStageFlagBits::eTransfer, UploadPipelineStage, {}, {},
+            {vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferRead,
+                                     vk::AccessFlagBits::eShaderWrite, VK_QUEUE_FAMILY_IGNORED,
+                                     VK_QUEUE_FAMILY_IGNORED, src_buffer, src_offset, size),
+             vk::BufferMemoryBarrier(vk::AccessFlagBits::eTransferWrite, UploadAccessBarriers,
+                                     VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, dst_buffer,
+                                     dst_offset, size)},
+            {}, dld);
+    });
 }
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_buffer_cache.h

@@ -5,105 +5,74 @@
 #pragma once
 
 #include <memory>
-#include <tuple>
+#include <unordered_map>
+#include <vector>
 
 #include "common/common_types.h"
-#include "video_core/gpu.h"
+#include "video_core/buffer_cache/buffer_cache.h"
 #include "video_core/rasterizer_cache.h"
 #include "video_core/renderer_vulkan/declarations.h"
-#include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_memory_manager.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+#include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
+#include "video_core/renderer_vulkan/vk_stream_buffer.h"
 
-namespace Memory {
-class Memory;
-}
-
-namespace Tegra {
-class MemoryManager;
+namespace Core {
+class System;
 }
 
 namespace Vulkan {
 
 class VKDevice;
-class VKFence;
 class VKMemoryManager;
-class VKStreamBuffer;
+class VKScheduler;
 
-class CachedBufferEntry final : public RasterizerCacheObject {
+class CachedBufferBlock final : public VideoCommon::BufferBlock {
 public:
-    explicit CachedBufferEntry(VAddr cpu_addr, std::size_t size, u64 offset, std::size_t alignment,
-                               u8* host_ptr);
+    explicit CachedBufferBlock(const VKDevice& device, VKMemoryManager& memory_manager,
+                               CacheAddr cache_addr, std::size_t size);
+    ~CachedBufferBlock();
 
-    VAddr GetCpuAddr() const override {
-        return cpu_addr;
-    }
-
-    std::size_t GetSizeInBytes() const override {
-        return size;
-    }
-
-    std::size_t GetSize() const {
-        return size;
-    }
-
-    u64 GetOffset() const {
-        return offset;
-    }
-
-    std::size_t GetAlignment() const {
-        return alignment;
+    const vk::Buffer* GetHandle() const {
+        return &*buffer.handle;
     }
 
 private:
-    VAddr cpu_addr{};
-    std::size_t size{};
-    u64 offset{};
-    std::size_t alignment{};
+    VKBuffer buffer;
 };
 
-class VKBufferCache final : public RasterizerCache<std::shared_ptr<CachedBufferEntry>> {
+using Buffer = std::shared_ptr<CachedBufferBlock>;
+
+class VKBufferCache final : public VideoCommon::BufferCache<Buffer, vk::Buffer, VKStreamBuffer> {
 public:
-    explicit VKBufferCache(Tegra::MemoryManager& tegra_memory_manager, Memory::Memory& cpu_memory_,
-                           VideoCore::RasterizerInterface& rasterizer, const VKDevice& device,
-                           VKMemoryManager& memory_manager, VKScheduler& scheduler, u64 size);
+    explicit VKBufferCache(VideoCore::RasterizerInterface& rasterizer, Core::System& system,
+                           const VKDevice& device, VKMemoryManager& memory_manager,
+                           VKScheduler& scheduler, VKStagingBufferPool& staging_pool);
     ~VKBufferCache();
 
-    /// Uploads data from a guest GPU address. Returns host's buffer offset where it's been
-    /// allocated.
-    u64 UploadMemory(GPUVAddr gpu_addr, std::size_t size, u64 alignment = 4, bool cache = true);
-
-    /// Uploads from a host memory. Returns host's buffer offset where it's been allocated.
-    u64 UploadHostMemory(const u8* raw_pointer, std::size_t size, u64 alignment = 4);
-
-    /// Reserves memory to be used by host's CPU. Returns mapped address and offset.
-    std::tuple<u8*, u64> ReserveMemory(std::size_t size, u64 alignment = 4);
-
-    /// Reserves a region of memory to be used in subsequent upload/reserve operations.
-    void Reserve(std::size_t max_size);
-
-    /// Ensures that the set data is sent to the device.
-    void Send();
-
-    /// Returns the buffer cache handle.
-    vk::Buffer GetBuffer() const {
-        return buffer_handle;
-    }
+    const vk::Buffer* GetEmptyBuffer(std::size_t size) override;
 
 protected:
-    // We do not have to flush this cache as things in it are never modified by us.
-    void FlushObjectInner(const std::shared_ptr<CachedBufferEntry>& object) override {}
+    void WriteBarrier() override {}
+
+    Buffer CreateBlock(CacheAddr cache_addr, std::size_t size) override;
+
+    const vk::Buffer* ToHandle(const Buffer& buffer) override;
+
+    void UploadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
+                         const u8* data) override;
+
+    void DownloadBlockData(const Buffer& buffer, std::size_t offset, std::size_t size,
+                           u8* data) override;
+
+    void CopyBlock(const Buffer& src, const Buffer& dst, std::size_t src_offset,
+                   std::size_t dst_offset, std::size_t size) override;
 
 private:
-    void AlignBuffer(std::size_t alignment);
-
-    Tegra::MemoryManager& tegra_memory_manager;
-    Memory::Memory& cpu_memory;
-
-    std::unique_ptr<VKStreamBuffer> stream_buffer;
-    vk::Buffer buffer_handle;
-
-    u8* buffer_ptr = nullptr;
-    u64 buffer_offset = 0;
-    u64 buffer_offset_base = 0;
+    const VKDevice& device;
+    VKMemoryManager& memory_manager;
+    VKScheduler& scheduler;
+    VKStagingBufferPool& staging_pool;
 };
 
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_compute_pass.cpp

@@ -0,0 +1,339 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <cstring>
+#include <memory>
+#include <optional>
+#include <utility>
+#include <vector>
+#include "common/alignment.h"
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_compute_pass.h"
+#include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_staging_buffer_pool.h"
+#include "video_core/renderer_vulkan/vk_update_descriptor.h"
+
+namespace Vulkan {
+
+namespace {
+
+// Quad array SPIR-V module. Generated from the "shaders/" directory, read the instructions there.
+constexpr u8 quad_array[] = {
+    0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x54, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x06, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x47, 0x4c, 0x53, 0x4c, 0x2e, 0x73, 0x74, 0x64, 0x2e, 0x34, 0x35, 0x30,
+    0x00, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x03, 0x00, 0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
+    0x0f, 0x00, 0x06, 0x00, 0x05, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x6d, 0x61, 0x69, 0x6e,
+    0x00, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x10, 0x00, 0x06, 0x00, 0x04, 0x00, 0x00, 0x00,
+    0x11, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
+    0x47, 0x00, 0x04, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00,
+    0x47, 0x00, 0x04, 0x00, 0x13, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00,
+    0x48, 0x00, 0x05, 0x00, 0x14, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x14, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
+    0x47, 0x00, 0x04, 0x00, 0x16, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x47, 0x00, 0x04, 0x00, 0x16, 0x00, 0x00, 0x00, 0x21, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x48, 0x00, 0x05, 0x00, 0x29, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00, 0x29, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x47, 0x00, 0x04, 0x00, 0x4a, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00,
+    0x13, 0x00, 0x02, 0x00, 0x02, 0x00, 0x00, 0x00, 0x21, 0x00, 0x03, 0x00, 0x03, 0x00, 0x00, 0x00,
+    0x02, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00, 0x09, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x03, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x0a, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x0d, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x03, 0x00, 0x13, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x1e, 0x00, 0x03, 0x00, 0x14, 0x00, 0x00, 0x00, 0x13, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
+    0x15, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x14, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00,
+    0x15, 0x00, 0x00, 0x00, 0x16, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00,
+    0x18, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x14, 0x00, 0x02, 0x00,
+    0x1b, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x03, 0x00, 0x29, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x20, 0x00, 0x04, 0x00, 0x2a, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x29, 0x00, 0x00, 0x00,
+    0x3b, 0x00, 0x04, 0x00, 0x2a, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00,
+    0x2b, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00,
+    0x20, 0x00, 0x04, 0x00, 0x2d, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x31, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00,
+    0x1c, 0x00, 0x04, 0x00, 0x34, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00,
+    0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x35, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00,
+    0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x36, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x37, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
+    0x2c, 0x00, 0x09, 0x00, 0x34, 0x00, 0x00, 0x00, 0x38, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00,
+    0x35, 0x00, 0x00, 0x00, 0x36, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x36, 0x00, 0x00, 0x00,
+    0x37, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x3a, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00,
+    0x34, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x44, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00, 0x47, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x49, 0x00, 0x00, 0x00,
+    0x00, 0x04, 0x00, 0x00, 0x2c, 0x00, 0x06, 0x00, 0x09, 0x00, 0x00, 0x00, 0x4a, 0x00, 0x00, 0x00,
+    0x49, 0x00, 0x00, 0x00, 0x35, 0x00, 0x00, 0x00, 0x35, 0x00, 0x00, 0x00, 0x36, 0x00, 0x05, 0x00,
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
+    0xf8, 0x00, 0x02, 0x00, 0x05, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x3a, 0x00, 0x00, 0x00,
+    0x3b, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00, 0x4c, 0x00, 0x00, 0x00,
+    0xf8, 0x00, 0x02, 0x00, 0x4c, 0x00, 0x00, 0x00, 0xf6, 0x00, 0x04, 0x00, 0x4b, 0x00, 0x00, 0x00,
+    0x4e, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00, 0x4d, 0x00, 0x00, 0x00,
+    0xf8, 0x00, 0x02, 0x00, 0x4d, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00, 0x0d, 0x00, 0x00, 0x00,
+    0x0e, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x00, 0x00, 0x84, 0x00, 0x05, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x12, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00,
+    0x44, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x17, 0x00, 0x00, 0x00, 0x16, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00, 0x18, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00,
+    0x17, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x1a, 0x00, 0x00, 0x00,
+    0x19, 0x00, 0x00, 0x00, 0xae, 0x00, 0x05, 0x00, 0x1b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00,
+    0x12, 0x00, 0x00, 0x00, 0x1a, 0x00, 0x00, 0x00, 0xf7, 0x00, 0x03, 0x00, 0x1e, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0xfa, 0x00, 0x04, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x00, 0x00,
+    0x1e, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00,
+    0x4b, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1e, 0x00, 0x00, 0x00, 0xf9, 0x00, 0x02, 0x00,
+    0x21, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x21, 0x00, 0x00, 0x00, 0xf5, 0x00, 0x07, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x53, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00,
+    0x48, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0xb0, 0x00, 0x05, 0x00, 0x1b, 0x00, 0x00, 0x00,
+    0x27, 0x00, 0x00, 0x00, 0x53, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0xf6, 0x00, 0x04, 0x00,
+    0x23, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xfa, 0x00, 0x04, 0x00,
+    0x27, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00,
+    0x22, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00, 0x2d, 0x00, 0x00, 0x00, 0x2e, 0x00, 0x00, 0x00,
+    0x2b, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x2f, 0x00, 0x00, 0x00, 0x2e, 0x00, 0x00, 0x00, 0x84, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x32, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x31, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x33, 0x00, 0x00, 0x00, 0x2f, 0x00, 0x00, 0x00, 0x32, 0x00, 0x00, 0x00,
+    0x3e, 0x00, 0x03, 0x00, 0x3b, 0x00, 0x00, 0x00, 0x38, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00,
+    0x07, 0x00, 0x00, 0x00, 0x3c, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x00, 0x00, 0x53, 0x00, 0x00, 0x00,
+    0x3d, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x00, 0x00, 0x3c, 0x00, 0x00, 0x00,
+    0x80, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x00, 0x00, 0x33, 0x00, 0x00, 0x00,
+    0x3d, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00, 0x42, 0x00, 0x00, 0x00,
+    0x12, 0x00, 0x00, 0x00, 0x53, 0x00, 0x00, 0x00, 0x41, 0x00, 0x06, 0x00, 0x44, 0x00, 0x00, 0x00,
+    0x45, 0x00, 0x00, 0x00, 0x16, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x00, 0x00, 0x42, 0x00, 0x00, 0x00,
+    0x3e, 0x00, 0x03, 0x00, 0x45, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x00, 0x00, 0x80, 0x00, 0x05, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x48, 0x00, 0x00, 0x00, 0x53, 0x00, 0x00, 0x00, 0x47, 0x00, 0x00, 0x00,
+    0xf9, 0x00, 0x02, 0x00, 0x21, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x23, 0x00, 0x00, 0x00,
+    0xf9, 0x00, 0x02, 0x00, 0x4b, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x4e, 0x00, 0x00, 0x00,
+    0xf9, 0x00, 0x02, 0x00, 0x4c, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x4b, 0x00, 0x00, 0x00,
+    0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00};
+
+// Uint8 SPIR-V module. Generated from the "shaders/" directory.
+constexpr u8 uint8_pass[] = {
+    0x03, 0x02, 0x23, 0x07, 0x00, 0x00, 0x01, 0x00, 0x07, 0x00, 0x08, 0x00, 0x2f, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x01, 0x00, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00,
+    0x51, 0x11, 0x00, 0x00, 0x11, 0x00, 0x02, 0x00, 0x61, 0x11, 0x00, 0x00, 0x0a, 0x00, 0x07, 0x00,
+    0x53, 0x50, 0x56, 0x5f, 0x4b, 0x48, 0x52, 0x5f, 0x31, 0x36, 0x62, 0x69, 0x74, 0x5f, 0x73, 0x74,
+    0x6f, 0x72, 0x61, 0x67, 0x65, 0x00, 0x00, 0x00, 0x0a, 0x00, 0x07, 0x00, 0x53, 0x50, 0x56, 0x5f,
+    0x4b, 0x48, 0x52, 0x5f, 0x38, 0x62, 0x69, 0x74, 0x5f, 0x73, 0x74, 0x6f, 0x72, 0x61, 0x67, 0x65,
+    0x00, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x06, 0x00, 0x01, 0x00, 0x00, 0x00, 0x47, 0x4c, 0x53, 0x4c,
+    0x2e, 0x73, 0x74, 0x64, 0x2e, 0x34, 0x35, 0x30, 0x00, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x03, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x06, 0x00, 0x05, 0x00, 0x00, 0x00,
+    0x04, 0x00, 0x00, 0x00, 0x6d, 0x61, 0x69, 0x6e, 0x00, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00,
+    0x10, 0x00, 0x06, 0x00, 0x04, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x0b, 0x00, 0x00, 0x00,
+    0x0b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x12, 0x00, 0x00, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x48, 0x00, 0x04, 0x00, 0x13, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x13, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00,
+    0x13, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x15, 0x00, 0x00, 0x00,
+    0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x15, 0x00, 0x00, 0x00,
+    0x21, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x1f, 0x00, 0x00, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x48, 0x00, 0x04, 0x00, 0x20, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x48, 0x00, 0x05, 0x00, 0x20, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x03, 0x00,
+    0x20, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x22, 0x00, 0x00, 0x00,
+    0x22, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x22, 0x00, 0x00, 0x00,
+    0x21, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x47, 0x00, 0x04, 0x00, 0x2e, 0x00, 0x00, 0x00,
+    0x0b, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x13, 0x00, 0x02, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x21, 0x00, 0x03, 0x00, 0x03, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
+    0x07, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x17, 0x00, 0x04, 0x00,
+    0x09, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
+    0x0a, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x09, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00,
+    0x0a, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00,
+    0x0d, 0x00, 0x00, 0x00, 0x01, 0x00, 0x00, 0x00, 0x06, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00,
+    0x11, 0x00, 0x00, 0x00, 0x08, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x03, 0x00,
+    0x12, 0x00, 0x00, 0x00, 0x11, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x03, 0x00, 0x13, 0x00, 0x00, 0x00,
+    0x12, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x14, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x13, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x14, 0x00, 0x00, 0x00, 0x15, 0x00, 0x00, 0x00,
+    0x02, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00, 0x20, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x14, 0x00, 0x02, 0x00, 0x1a, 0x00, 0x00, 0x00, 0x15, 0x00, 0x04, 0x00,
+    0x1e, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x03, 0x00,
+    0x1f, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x00, 0x00, 0x1e, 0x00, 0x03, 0x00, 0x20, 0x00, 0x00, 0x00,
+    0x1f, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x21, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x20, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x21, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00,
+    0x02, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x17, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
+    0x00, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x26, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x11, 0x00, 0x00, 0x00, 0x20, 0x00, 0x04, 0x00, 0x2a, 0x00, 0x00, 0x00, 0x02, 0x00, 0x00, 0x00,
+    0x1e, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x00, 0x00,
+    0x00, 0x04, 0x00, 0x00, 0x2b, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x2d, 0x00, 0x00, 0x00,
+    0x01, 0x00, 0x00, 0x00, 0x2c, 0x00, 0x06, 0x00, 0x09, 0x00, 0x00, 0x00, 0x2e, 0x00, 0x00, 0x00,
+    0x2c, 0x00, 0x00, 0x00, 0x2d, 0x00, 0x00, 0x00, 0x2d, 0x00, 0x00, 0x00, 0x36, 0x00, 0x05, 0x00,
+    0x02, 0x00, 0x00, 0x00, 0x04, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x03, 0x00, 0x00, 0x00,
+    0xf8, 0x00, 0x02, 0x00, 0x05, 0x00, 0x00, 0x00, 0x3b, 0x00, 0x04, 0x00, 0x07, 0x00, 0x00, 0x00,
+    0x08, 0x00, 0x00, 0x00, 0x07, 0x00, 0x00, 0x00, 0x41, 0x00, 0x05, 0x00, 0x0d, 0x00, 0x00, 0x00,
+    0x0e, 0x00, 0x00, 0x00, 0x0b, 0x00, 0x00, 0x00, 0x0c, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x0e, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x03, 0x00,
+    0x08, 0x00, 0x00, 0x00, 0x0f, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x10, 0x00, 0x00, 0x00, 0x08, 0x00, 0x00, 0x00, 0x44, 0x00, 0x05, 0x00, 0x06, 0x00, 0x00, 0x00,
+    0x16, 0x00, 0x00, 0x00, 0x15, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00,
+    0x17, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00, 0x16, 0x00, 0x00, 0x00, 0x7c, 0x00, 0x04, 0x00,
+    0x06, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00, 0x18, 0x00, 0x00, 0x00, 0xb0, 0x00, 0x05, 0x00,
+    0x1a, 0x00, 0x00, 0x00, 0x1b, 0x00, 0x00, 0x00, 0x10, 0x00, 0x00, 0x00, 0x19, 0x00, 0x00, 0x00,
+    0xf7, 0x00, 0x03, 0x00, 0x1d, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0x00, 0xfa, 0x00, 0x04, 0x00,
+    0x1b, 0x00, 0x00, 0x00, 0x1c, 0x00, 0x00, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00,
+    0x1c, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x24, 0x00, 0x00, 0x00,
+    0x08, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00, 0x06, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00,
+    0x08, 0x00, 0x00, 0x00, 0x41, 0x00, 0x06, 0x00, 0x26, 0x00, 0x00, 0x00, 0x27, 0x00, 0x00, 0x00,
+    0x15, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00, 0x25, 0x00, 0x00, 0x00, 0x3d, 0x00, 0x04, 0x00,
+    0x11, 0x00, 0x00, 0x00, 0x28, 0x00, 0x00, 0x00, 0x27, 0x00, 0x00, 0x00, 0x71, 0x00, 0x04, 0x00,
+    0x1e, 0x00, 0x00, 0x00, 0x29, 0x00, 0x00, 0x00, 0x28, 0x00, 0x00, 0x00, 0x41, 0x00, 0x06, 0x00,
+    0x2a, 0x00, 0x00, 0x00, 0x2b, 0x00, 0x00, 0x00, 0x22, 0x00, 0x00, 0x00, 0x23, 0x00, 0x00, 0x00,
+    0x24, 0x00, 0x00, 0x00, 0x3e, 0x00, 0x03, 0x00, 0x2b, 0x00, 0x00, 0x00, 0x29, 0x00, 0x00, 0x00,
+    0xf9, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00, 0xf8, 0x00, 0x02, 0x00, 0x1d, 0x00, 0x00, 0x00,
+    0xfd, 0x00, 0x01, 0x00, 0x38, 0x00, 0x01, 0x00};
+
+} // Anonymous namespace
+
+VKComputePass::VKComputePass(const VKDevice& device, VKDescriptorPool& descriptor_pool,
+                             const std::vector<vk::DescriptorSetLayoutBinding>& bindings,
+                             const std::vector<vk::DescriptorUpdateTemplateEntry>& templates,
+                             const std::vector<vk::PushConstantRange> push_constants,
+                             std::size_t code_size, const u8* code) {
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+
+    const vk::DescriptorSetLayoutCreateInfo descriptor_layout_ci(
+        {}, static_cast<u32>(bindings.size()), bindings.data());
+    descriptor_set_layout = dev.createDescriptorSetLayoutUnique(descriptor_layout_ci, nullptr, dld);
+
+    const vk::PipelineLayoutCreateInfo pipeline_layout_ci({}, 1, &*descriptor_set_layout,
+                                                          static_cast<u32>(push_constants.size()),
+                                                          push_constants.data());
+    layout = dev.createPipelineLayoutUnique(pipeline_layout_ci, nullptr, dld);
+
+    if (!templates.empty()) {
+        const vk::DescriptorUpdateTemplateCreateInfo template_ci(
+            {}, static_cast<u32>(templates.size()), templates.data(),
+            vk::DescriptorUpdateTemplateType::eDescriptorSet, *descriptor_set_layout,
+            vk::PipelineBindPoint::eGraphics, *layout, 0);
+        descriptor_template = dev.createDescriptorUpdateTemplateUnique(template_ci, nullptr, dld);
+
+        descriptor_allocator.emplace(descriptor_pool, *descriptor_set_layout);
+    }
+
+    auto code_copy = std::make_unique<u32[]>(code_size / sizeof(u32) + 1);
+    std::memcpy(code_copy.get(), code, code_size);
+    const vk::ShaderModuleCreateInfo module_ci({}, code_size, code_copy.get());
+    module = dev.createShaderModuleUnique(module_ci, nullptr, dld);
+
+    const vk::PipelineShaderStageCreateInfo stage_ci({}, vk::ShaderStageFlagBits::eCompute, *module,
+                                                     "main", nullptr);
+
+    const vk::ComputePipelineCreateInfo pipeline_ci({}, stage_ci, *layout, nullptr, 0);
+    pipeline = dev.createComputePipelineUnique(nullptr, pipeline_ci, nullptr, dld);
+}
+
+VKComputePass::~VKComputePass() = default;
+
+vk::DescriptorSet VKComputePass::CommitDescriptorSet(
+    VKUpdateDescriptorQueue& update_descriptor_queue, VKFence& fence) {
+    if (!descriptor_template) {
+        return {};
+    }
+    const auto set = descriptor_allocator->Commit(fence);
+    update_descriptor_queue.Send(*descriptor_template, set);
+    return set;
+}
+
+QuadArrayPass::QuadArrayPass(const VKDevice& device, VKScheduler& scheduler,
+                             VKDescriptorPool& descriptor_pool,
+                             VKStagingBufferPool& staging_buffer_pool,
+                             VKUpdateDescriptorQueue& update_descriptor_queue)
+    : VKComputePass(device, descriptor_pool,
+                    {vk::DescriptorSetLayoutBinding(0, vk::DescriptorType::eStorageBuffer, 1,
+                                                    vk::ShaderStageFlagBits::eCompute, nullptr)},
+                    {vk::DescriptorUpdateTemplateEntry(0, 0, 1, vk::DescriptorType::eStorageBuffer,
+                                                       0, sizeof(DescriptorUpdateEntry))},
+                    {vk::PushConstantRange(vk::ShaderStageFlagBits::eCompute, 0, sizeof(u32))},
+                    std::size(quad_array), quad_array),
+      scheduler{scheduler}, staging_buffer_pool{staging_buffer_pool},
+      update_descriptor_queue{update_descriptor_queue} {}
+
+QuadArrayPass::~QuadArrayPass() = default;
+
+std::pair<const vk::Buffer&, vk::DeviceSize> QuadArrayPass::Assemble(u32 num_vertices, u32 first) {
+    const u32 num_triangle_vertices = num_vertices * 6 / 4;
+    const std::size_t staging_size = num_triangle_vertices * sizeof(u32);
+    auto& buffer = staging_buffer_pool.GetUnusedBuffer(staging_size, false);
+
+    update_descriptor_queue.Acquire();
+    update_descriptor_queue.AddBuffer(&*buffer.handle, 0, staging_size);
+    const auto set = CommitDescriptorSet(update_descriptor_queue, scheduler.GetFence());
+
+    scheduler.RequestOutsideRenderPassOperationContext();
+
+    ASSERT(num_vertices % 4 == 0);
+    const u32 num_quads = num_vertices / 4;
+    scheduler.Record([layout = *layout, pipeline = *pipeline, buffer = *buffer.handle, num_quads,
+                      first, set](auto cmdbuf, auto& dld) {
+        constexpr u32 dispatch_size = 1024;
+        cmdbuf.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline, dld);
+        cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eCompute, layout, 0, {set}, {}, dld);
+        cmdbuf.pushConstants(layout, vk::ShaderStageFlagBits::eCompute, 0, sizeof(first), &first,
+                             dld);
+        cmdbuf.dispatch(Common::AlignUp(num_quads, dispatch_size) / dispatch_size, 1, 1, dld);
+
+        const vk::BufferMemoryBarrier barrier(
+            vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eVertexAttributeRead,
+            VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, buffer, 0,
+            static_cast<vk::DeviceSize>(num_quads) * 6 * sizeof(u32));
+        cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eComputeShader,
+                               vk::PipelineStageFlagBits::eVertexInput, {}, {}, {barrier}, {}, dld);
+    });
+    return {*buffer.handle, 0};
+}
+
+Uint8Pass::Uint8Pass(const VKDevice& device, VKScheduler& scheduler,
+                     VKDescriptorPool& descriptor_pool, VKStagingBufferPool& staging_buffer_pool,
+                     VKUpdateDescriptorQueue& update_descriptor_queue)
+    : VKComputePass(device, descriptor_pool,
+                    {vk::DescriptorSetLayoutBinding(0, vk::DescriptorType::eStorageBuffer, 1,
+                                                    vk::ShaderStageFlagBits::eCompute, nullptr),
+                     vk::DescriptorSetLayoutBinding(1, vk::DescriptorType::eStorageBuffer, 1,
+                                                    vk::ShaderStageFlagBits::eCompute, nullptr)},
+                    {vk::DescriptorUpdateTemplateEntry(0, 0, 2, vk::DescriptorType::eStorageBuffer,
+                                                       0, sizeof(DescriptorUpdateEntry))},
+                    {}, std::size(uint8_pass), uint8_pass),
+      scheduler{scheduler}, staging_buffer_pool{staging_buffer_pool},
+      update_descriptor_queue{update_descriptor_queue} {}
+
+Uint8Pass::~Uint8Pass() = default;
+
+std::pair<const vk::Buffer*, u64> Uint8Pass::Assemble(u32 num_vertices, vk::Buffer src_buffer,
+                                                      u64 src_offset) {
+    const auto staging_size = static_cast<u32>(num_vertices * sizeof(u16));
+    auto& buffer = staging_buffer_pool.GetUnusedBuffer(staging_size, false);
+
+    update_descriptor_queue.Acquire();
+    update_descriptor_queue.AddBuffer(&src_buffer, src_offset, num_vertices);
+    update_descriptor_queue.AddBuffer(&*buffer.handle, 0, staging_size);
+    const auto set = CommitDescriptorSet(update_descriptor_queue, scheduler.GetFence());
+
+    scheduler.RequestOutsideRenderPassOperationContext();
+    scheduler.Record([layout = *layout, pipeline = *pipeline, buffer = *buffer.handle, set,
+                      num_vertices](auto cmdbuf, auto& dld) {
+        constexpr u32 dispatch_size = 1024;
+        cmdbuf.bindPipeline(vk::PipelineBindPoint::eCompute, pipeline, dld);
+        cmdbuf.bindDescriptorSets(vk::PipelineBindPoint::eCompute, layout, 0, {set}, {}, dld);
+        cmdbuf.dispatch(Common::AlignUp(num_vertices, dispatch_size) / dispatch_size, 1, 1, dld);
+
+        const vk::BufferMemoryBarrier barrier(
+            vk::AccessFlagBits::eShaderWrite, vk::AccessFlagBits::eVertexAttributeRead,
+            VK_QUEUE_FAMILY_IGNORED, VK_QUEUE_FAMILY_IGNORED, buffer, 0,
+            static_cast<vk::DeviceSize>(num_vertices) * sizeof(u16));
+        cmdbuf.pipelineBarrier(vk::PipelineStageFlagBits::eComputeShader,
+                               vk::PipelineStageFlagBits::eVertexInput, {}, {}, {barrier}, {}, dld);
+    });
+    return {&*buffer.handle, 0};
+}
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_compute_pass.h

@@ -0,0 +1,77 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <optional>
+#include <utility>
+#include <vector>
+#include "common/common_types.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+
+namespace Vulkan {
+
+class VKDevice;
+class VKFence;
+class VKScheduler;
+class VKStagingBufferPool;
+class VKUpdateDescriptorQueue;
+
+class VKComputePass {
+public:
+    explicit VKComputePass(const VKDevice& device, VKDescriptorPool& descriptor_pool,
+                           const std::vector<vk::DescriptorSetLayoutBinding>& bindings,
+                           const std::vector<vk::DescriptorUpdateTemplateEntry>& templates,
+                           const std::vector<vk::PushConstantRange> push_constants,
+                           std::size_t code_size, const u8* code);
+    ~VKComputePass();
+
+protected:
+    vk::DescriptorSet CommitDescriptorSet(VKUpdateDescriptorQueue& update_descriptor_queue,
+                                          VKFence& fence);
+
+    UniqueDescriptorUpdateTemplate descriptor_template;
+    UniquePipelineLayout layout;
+    UniquePipeline pipeline;
+
+private:
+    UniqueDescriptorSetLayout descriptor_set_layout;
+    std::optional<DescriptorAllocator> descriptor_allocator;
+    UniqueShaderModule module;
+};
+
+class QuadArrayPass final : public VKComputePass {
+public:
+    explicit QuadArrayPass(const VKDevice& device, VKScheduler& scheduler,
+                           VKDescriptorPool& descriptor_pool,
+                           VKStagingBufferPool& staging_buffer_pool,
+                           VKUpdateDescriptorQueue& update_descriptor_queue);
+    ~QuadArrayPass();
+
+    std::pair<const vk::Buffer&, vk::DeviceSize> Assemble(u32 num_vertices, u32 first);
+
+private:
+    VKScheduler& scheduler;
+    VKStagingBufferPool& staging_buffer_pool;
+    VKUpdateDescriptorQueue& update_descriptor_queue;
+};
+
+class Uint8Pass final : public VKComputePass {
+public:
+    explicit Uint8Pass(const VKDevice& device, VKScheduler& scheduler,
+                       VKDescriptorPool& descriptor_pool, VKStagingBufferPool& staging_buffer_pool,
+                       VKUpdateDescriptorQueue& update_descriptor_queue);
+    ~Uint8Pass();
+
+    std::pair<const vk::Buffer*, u64> Assemble(u32 num_vertices, vk::Buffer src_buffer,
+                                               u64 src_offset);
+
+private:
+    VKScheduler& scheduler;
+    VKStagingBufferPool& staging_buffer_pool;
+    VKUpdateDescriptorQueue& update_descriptor_queue;
+};
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_compute_pipeline.cpp

@@ -0,0 +1,112 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <memory>
+#include <vector>
+
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_compute_pipeline.h"
+#include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_pipeline_cache.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_shader_decompiler.h"
+#include "video_core/renderer_vulkan/vk_update_descriptor.h"
+
+namespace Vulkan {
+
+VKComputePipeline::VKComputePipeline(const VKDevice& device, VKScheduler& scheduler,
+                                     VKDescriptorPool& descriptor_pool,
+                                     VKUpdateDescriptorQueue& update_descriptor_queue,
+                                     const SPIRVShader& shader)
+    : device{device}, scheduler{scheduler}, entries{shader.entries},
+      descriptor_set_layout{CreateDescriptorSetLayout()},
+      descriptor_allocator{descriptor_pool, *descriptor_set_layout},
+      update_descriptor_queue{update_descriptor_queue}, layout{CreatePipelineLayout()},
+      descriptor_template{CreateDescriptorUpdateTemplate()},
+      shader_module{CreateShaderModule(shader.code)}, pipeline{CreatePipeline()} {}
+
+VKComputePipeline::~VKComputePipeline() = default;
+
+vk::DescriptorSet VKComputePipeline::CommitDescriptorSet() {
+    if (!descriptor_template) {
+        return {};
+    }
+    const auto set = descriptor_allocator.Commit(scheduler.GetFence());
+    update_descriptor_queue.Send(*descriptor_template, set);
+    return set;
+}
+
+UniqueDescriptorSetLayout VKComputePipeline::CreateDescriptorSetLayout() const {
+    std::vector<vk::DescriptorSetLayoutBinding> bindings;
+    u32 binding = 0;
+    const auto AddBindings = [&](vk::DescriptorType descriptor_type, std::size_t num_entries) {
+        // TODO(Rodrigo): Maybe make individual bindings here?
+        for (u32 bindpoint = 0; bindpoint < static_cast<u32>(num_entries); ++bindpoint) {
+            bindings.emplace_back(binding++, descriptor_type, 1, vk::ShaderStageFlagBits::eCompute,
+                                  nullptr);
+        }
+    };
+    AddBindings(vk::DescriptorType::eUniformBuffer, entries.const_buffers.size());
+    AddBindings(vk::DescriptorType::eStorageBuffer, entries.global_buffers.size());
+    AddBindings(vk::DescriptorType::eUniformTexelBuffer, entries.texel_buffers.size());
+    AddBindings(vk::DescriptorType::eCombinedImageSampler, entries.samplers.size());
+    AddBindings(vk::DescriptorType::eStorageImage, entries.images.size());
+
+    const vk::DescriptorSetLayoutCreateInfo descriptor_set_layout_ci(
+        {}, static_cast<u32>(bindings.size()), bindings.data());
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    return dev.createDescriptorSetLayoutUnique(descriptor_set_layout_ci, nullptr, dld);
+}
+
+UniquePipelineLayout VKComputePipeline::CreatePipelineLayout() const {
+    const vk::PipelineLayoutCreateInfo layout_ci({}, 1, &*descriptor_set_layout, 0, nullptr);
+    const auto dev = device.GetLogical();
+    return dev.createPipelineLayoutUnique(layout_ci, nullptr, device.GetDispatchLoader());
+}
+
+UniqueDescriptorUpdateTemplate VKComputePipeline::CreateDescriptorUpdateTemplate() const {
+    std::vector<vk::DescriptorUpdateTemplateEntry> template_entries;
+    u32 binding = 0;
+    u32 offset = 0;
+    FillDescriptorUpdateTemplateEntries(device, entries, binding, offset, template_entries);
+    if (template_entries.empty()) {
+        // If the shader doesn't use descriptor sets, skip template creation.
+        return UniqueDescriptorUpdateTemplate{};
+    }
+
+    const vk::DescriptorUpdateTemplateCreateInfo template_ci(
+        {}, static_cast<u32>(template_entries.size()), template_entries.data(),
+        vk::DescriptorUpdateTemplateType::eDescriptorSet, *descriptor_set_layout,
+        vk::PipelineBindPoint::eGraphics, *layout, DESCRIPTOR_SET);
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    return dev.createDescriptorUpdateTemplateUnique(template_ci, nullptr, dld);
+}
+
+UniqueShaderModule VKComputePipeline::CreateShaderModule(const std::vector<u32>& code) const {
+    const vk::ShaderModuleCreateInfo module_ci({}, code.size() * sizeof(u32), code.data());
+    const auto dev = device.GetLogical();
+    return dev.createShaderModuleUnique(module_ci, nullptr, device.GetDispatchLoader());
+}
+
+UniquePipeline VKComputePipeline::CreatePipeline() const {
+    vk::PipelineShaderStageCreateInfo shader_stage_ci({}, vk::ShaderStageFlagBits::eCompute,
+                                                      *shader_module, "main", nullptr);
+    vk::PipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci;
+    subgroup_size_ci.requiredSubgroupSize = GuestWarpSize;
+    if (entries.uses_warps && device.IsGuestWarpSizeSupported(vk::ShaderStageFlagBits::eCompute)) {
+        shader_stage_ci.pNext = &subgroup_size_ci;
+    }
+
+    const vk::ComputePipelineCreateInfo create_info({}, shader_stage_ci, *layout, {}, 0);
+    const auto dev = device.GetLogical();
+    return dev.createComputePipelineUnique({}, create_info, nullptr, device.GetDispatchLoader());
+}
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_compute_pipeline.h

@@ -0,0 +1,66 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+
+#include "common/common_types.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+#include "video_core/renderer_vulkan/vk_shader_decompiler.h"
+
+namespace Vulkan {
+
+class VKDevice;
+class VKScheduler;
+class VKUpdateDescriptorQueue;
+
+class VKComputePipeline final {
+public:
+    explicit VKComputePipeline(const VKDevice& device, VKScheduler& scheduler,
+                               VKDescriptorPool& descriptor_pool,
+                               VKUpdateDescriptorQueue& update_descriptor_queue,
+                               const SPIRVShader& shader);
+    ~VKComputePipeline();
+
+    vk::DescriptorSet CommitDescriptorSet();
+
+    vk::Pipeline GetHandle() const {
+        return *pipeline;
+    }
+
+    vk::PipelineLayout GetLayout() const {
+        return *layout;
+    }
+
+    const ShaderEntries& GetEntries() {
+        return entries;
+    }
+
+private:
+    UniqueDescriptorSetLayout CreateDescriptorSetLayout() const;
+
+    UniquePipelineLayout CreatePipelineLayout() const;
+
+    UniqueDescriptorUpdateTemplate CreateDescriptorUpdateTemplate() const;
+
+    UniqueShaderModule CreateShaderModule(const std::vector<u32>& code) const;
+
+    UniquePipeline CreatePipeline() const;
+
+    const VKDevice& device;
+    VKScheduler& scheduler;
+    ShaderEntries entries;
+
+    UniqueDescriptorSetLayout descriptor_set_layout;
+    DescriptorAllocator descriptor_allocator;
+    VKUpdateDescriptorQueue& update_descriptor_queue;
+    UniquePipelineLayout layout;
+    UniqueDescriptorUpdateTemplate descriptor_template;
+    UniqueShaderModule shader_module;
+    UniquePipeline pipeline;
+};
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_descriptor_pool.cpp

@@ -0,0 +1,89 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <memory>
+#include <vector>
+
+#include "common/common_types.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+
+namespace Vulkan {
+
+// Prefer small grow rates to avoid saturating the descriptor pool with barely used pipelines.
+constexpr std::size_t SETS_GROW_RATE = 0x20;
+
+DescriptorAllocator::DescriptorAllocator(VKDescriptorPool& descriptor_pool,
+                                         vk::DescriptorSetLayout layout)
+    : VKFencedPool{SETS_GROW_RATE}, descriptor_pool{descriptor_pool}, layout{layout} {}
+
+DescriptorAllocator::~DescriptorAllocator() = default;
+
+vk::DescriptorSet DescriptorAllocator::Commit(VKFence& fence) {
+    return *descriptors[CommitResource(fence)];
+}
+
+void DescriptorAllocator::Allocate(std::size_t begin, std::size_t end) {
+    auto new_sets = descriptor_pool.AllocateDescriptors(layout, end - begin);
+    descriptors.insert(descriptors.end(), std::make_move_iterator(new_sets.begin()),
+                       std::make_move_iterator(new_sets.end()));
+}
+
+VKDescriptorPool::VKDescriptorPool(const VKDevice& device)
+    : device{device}, active_pool{AllocateNewPool()} {}
+
+VKDescriptorPool::~VKDescriptorPool() = default;
+
+vk::DescriptorPool VKDescriptorPool::AllocateNewPool() {
+    static constexpr u32 num_sets = 0x20000;
+    static constexpr vk::DescriptorPoolSize pool_sizes[] = {
+        {vk::DescriptorType::eUniformBuffer, num_sets * 90},
+        {vk::DescriptorType::eStorageBuffer, num_sets * 60},
+        {vk::DescriptorType::eUniformTexelBuffer, num_sets * 64},
+        {vk::DescriptorType::eCombinedImageSampler, num_sets * 64},
+        {vk::DescriptorType::eStorageImage, num_sets * 40}};
+
+    const vk::DescriptorPoolCreateInfo create_info(
+        vk::DescriptorPoolCreateFlagBits::eFreeDescriptorSet, num_sets,
+        static_cast<u32>(std::size(pool_sizes)), std::data(pool_sizes));
+    const auto dev = device.GetLogical();
+    return *pools.emplace_back(
+        dev.createDescriptorPoolUnique(create_info, nullptr, device.GetDispatchLoader()));
+}
+
+std::vector<UniqueDescriptorSet> VKDescriptorPool::AllocateDescriptors(
+    vk::DescriptorSetLayout layout, std::size_t count) {
+    std::vector layout_copies(count, layout);
+    vk::DescriptorSetAllocateInfo allocate_info(active_pool, static_cast<u32>(count),
+                                                layout_copies.data());
+
+    std::vector<vk::DescriptorSet> sets(count);
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    switch (const auto result = dev.allocateDescriptorSets(&allocate_info, sets.data(), dld)) {
+    case vk::Result::eSuccess:
+        break;
+    case vk::Result::eErrorOutOfPoolMemory:
+        active_pool = AllocateNewPool();
+        allocate_info.descriptorPool = active_pool;
+        if (dev.allocateDescriptorSets(&allocate_info, sets.data(), dld) == vk::Result::eSuccess) {
+            break;
+        }
+        [[fallthrough]];
+    default:
+        vk::throwResultException(result, "vk::Device::allocateDescriptorSetsUnique");
+    }
+
+    vk::PoolFree deleter(dev, active_pool, dld);
+    std::vector<UniqueDescriptorSet> unique_sets;
+    unique_sets.reserve(count);
+    for (const auto set : sets) {
+        unique_sets.push_back(UniqueDescriptorSet{set, deleter});
+    }
+    return unique_sets;
+}
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_descriptor_pool.h

@@ -0,0 +1,56 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <memory>
+#include <vector>
+
+#include "common/common_types.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+
+namespace Vulkan {
+
+class VKDescriptorPool;
+
+class DescriptorAllocator final : public VKFencedPool {
+public:
+    explicit DescriptorAllocator(VKDescriptorPool& descriptor_pool, vk::DescriptorSetLayout layout);
+    ~DescriptorAllocator() override;
+
+    DescriptorAllocator(const DescriptorAllocator&) = delete;
+
+    vk::DescriptorSet Commit(VKFence& fence);
+
+protected:
+    void Allocate(std::size_t begin, std::size_t end) override;
+
+private:
+    VKDescriptorPool& descriptor_pool;
+    const vk::DescriptorSetLayout layout;
+
+    std::vector<UniqueDescriptorSet> descriptors;
+};
+
+class VKDescriptorPool final {
+    friend DescriptorAllocator;
+
+public:
+    explicit VKDescriptorPool(const VKDevice& device);
+    ~VKDescriptorPool();
+
+private:
+    vk::DescriptorPool AllocateNewPool();
+
+    std::vector<UniqueDescriptorSet> AllocateDescriptors(vk::DescriptorSetLayout layout,
+                                                         std::size_t count);
+
+    const VKDevice& device;
+
+    std::vector<UniqueDescriptorPool> pools;
+    vk::DescriptorPool active_pool;
+};
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_graphics_pipeline.cpp

@@ -0,0 +1,270 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#include <vector>
+#include "common/assert.h"
+#include "common/common_types.h"
+#include "common/microprofile.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/fixed_pipeline_state.h"
+#include "video_core/renderer_vulkan/maxwell_to_vk.h"
+#include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+#include "video_core/renderer_vulkan/vk_device.h"
+#include "video_core/renderer_vulkan/vk_graphics_pipeline.h"
+#include "video_core/renderer_vulkan/vk_pipeline_cache.h"
+#include "video_core/renderer_vulkan/vk_renderpass_cache.h"
+#include "video_core/renderer_vulkan/vk_scheduler.h"
+#include "video_core/renderer_vulkan/vk_update_descriptor.h"
+
+namespace Vulkan {
+
+MICROPROFILE_DECLARE(Vulkan_PipelineCache);
+
+namespace {
+
+vk::StencilOpState GetStencilFaceState(const FixedPipelineState::StencilFace& face) {
+    return vk::StencilOpState(MaxwellToVK::StencilOp(face.action_stencil_fail),
+                              MaxwellToVK::StencilOp(face.action_depth_pass),
+                              MaxwellToVK::StencilOp(face.action_depth_fail),
+                              MaxwellToVK::ComparisonOp(face.test_func), 0, 0, 0);
+}
+
+bool SupportsPrimitiveRestart(vk::PrimitiveTopology topology) {
+    static constexpr std::array unsupported_topologies = {
+        vk::PrimitiveTopology::ePointList,
+        vk::PrimitiveTopology::eLineList,
+        vk::PrimitiveTopology::eTriangleList,
+        vk::PrimitiveTopology::eLineListWithAdjacency,
+        vk::PrimitiveTopology::eTriangleListWithAdjacency,
+        vk::PrimitiveTopology::ePatchList};
+    return std::find(std::begin(unsupported_topologies), std::end(unsupported_topologies),
+                     topology) == std::end(unsupported_topologies);
+}
+
+} // Anonymous namespace
+
+VKGraphicsPipeline::VKGraphicsPipeline(const VKDevice& device, VKScheduler& scheduler,
+                                       VKDescriptorPool& descriptor_pool,
+                                       VKUpdateDescriptorQueue& update_descriptor_queue,
+                                       VKRenderPassCache& renderpass_cache,
+                                       const GraphicsPipelineCacheKey& key,
+                                       const std::vector<vk::DescriptorSetLayoutBinding>& bindings,
+                                       const SPIRVProgram& program)
+    : device{device}, scheduler{scheduler}, fixed_state{key.fixed_state}, hash{key.Hash()},
+      descriptor_set_layout{CreateDescriptorSetLayout(bindings)},
+      descriptor_allocator{descriptor_pool, *descriptor_set_layout},
+      update_descriptor_queue{update_descriptor_queue}, layout{CreatePipelineLayout()},
+      descriptor_template{CreateDescriptorUpdateTemplate(program)}, modules{CreateShaderModules(
+                                                                        program)},
+      renderpass{renderpass_cache.GetRenderPass(key.renderpass_params)}, pipeline{CreatePipeline(
+                                                                             key.renderpass_params,
+                                                                             program)} {}
+
+VKGraphicsPipeline::~VKGraphicsPipeline() = default;
+
+vk::DescriptorSet VKGraphicsPipeline::CommitDescriptorSet() {
+    if (!descriptor_template) {
+        return {};
+    }
+    const auto set = descriptor_allocator.Commit(scheduler.GetFence());
+    update_descriptor_queue.Send(*descriptor_template, set);
+    return set;
+}
+
+UniqueDescriptorSetLayout VKGraphicsPipeline::CreateDescriptorSetLayout(
+    const std::vector<vk::DescriptorSetLayoutBinding>& bindings) const {
+    const vk::DescriptorSetLayoutCreateInfo descriptor_set_layout_ci(
+        {}, static_cast<u32>(bindings.size()), bindings.data());
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    return dev.createDescriptorSetLayoutUnique(descriptor_set_layout_ci, nullptr, dld);
+}
+
+UniquePipelineLayout VKGraphicsPipeline::CreatePipelineLayout() const {
+    const vk::PipelineLayoutCreateInfo pipeline_layout_ci({}, 1, &*descriptor_set_layout, 0,
+                                                          nullptr);
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    return dev.createPipelineLayoutUnique(pipeline_layout_ci, nullptr, dld);
+}
+
+UniqueDescriptorUpdateTemplate VKGraphicsPipeline::CreateDescriptorUpdateTemplate(
+    const SPIRVProgram& program) const {
+    std::vector<vk::DescriptorUpdateTemplateEntry> template_entries;
+    u32 binding = 0;
+    u32 offset = 0;
+    for (const auto& stage : program) {
+        if (stage) {
+            FillDescriptorUpdateTemplateEntries(device, stage->entries, binding, offset,
+                                                template_entries);
+        }
+    }
+    if (template_entries.empty()) {
+        // If the shader doesn't use descriptor sets, skip template creation.
+        return UniqueDescriptorUpdateTemplate{};
+    }
+
+    const vk::DescriptorUpdateTemplateCreateInfo template_ci(
+        {}, static_cast<u32>(template_entries.size()), template_entries.data(),
+        vk::DescriptorUpdateTemplateType::eDescriptorSet, *descriptor_set_layout,
+        vk::PipelineBindPoint::eGraphics, *layout, DESCRIPTOR_SET);
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    return dev.createDescriptorUpdateTemplateUnique(template_ci, nullptr, dld);
+}
+
+std::vector<UniqueShaderModule> VKGraphicsPipeline::CreateShaderModules(
+    const SPIRVProgram& program) const {
+    std::vector<UniqueShaderModule> modules;
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    for (std::size_t i = 0; i < Maxwell::MaxShaderStage; ++i) {
+        const auto& stage = program[i];
+        if (!stage) {
+            continue;
+        }
+        const vk::ShaderModuleCreateInfo module_ci({}, stage->code.size() * sizeof(u32),
+                                                   stage->code.data());
+        modules.emplace_back(dev.createShaderModuleUnique(module_ci, nullptr, dld));
+    }
+    return modules;
+}
+
+UniquePipeline VKGraphicsPipeline::CreatePipeline(const RenderPassParams& renderpass_params,
+                                                  const SPIRVProgram& program) const {
+    const auto& vi = fixed_state.vertex_input;
+    const auto& ia = fixed_state.input_assembly;
+    const auto& ds = fixed_state.depth_stencil;
+    const auto& cd = fixed_state.color_blending;
+    const auto& ts = fixed_state.tessellation;
+    const auto& rs = fixed_state.rasterizer;
+
+    std::vector<vk::VertexInputBindingDescription> vertex_bindings;
+    std::vector<vk::VertexInputBindingDivisorDescriptionEXT> vertex_binding_divisors;
+    for (std::size_t i = 0; i < vi.num_bindings; ++i) {
+        const auto& binding = vi.bindings[i];
+        const bool instanced = binding.divisor != 0;
+        const auto rate = instanced ? vk::VertexInputRate::eInstance : vk::VertexInputRate::eVertex;
+        vertex_bindings.emplace_back(binding.index, binding.stride, rate);
+        if (instanced) {
+            vertex_binding_divisors.emplace_back(binding.index, binding.divisor);
+        }
+    }
+
+    std::vector<vk::VertexInputAttributeDescription> vertex_attributes;
+    const auto& input_attributes = program[0]->entries.attributes;
+    for (std::size_t i = 0; i < vi.num_attributes; ++i) {
+        const auto& attribute = vi.attributes[i];
+        if (input_attributes.find(attribute.index) == input_attributes.end()) {
+            // Skip attributes not used by the vertex shaders.
+            continue;
+        }
+        vertex_attributes.emplace_back(attribute.index, attribute.buffer,
+                                       MaxwellToVK::VertexFormat(attribute.type, attribute.size),
+                                       attribute.offset);
+    }
+
+    vk::PipelineVertexInputStateCreateInfo vertex_input_ci(
+        {}, static_cast<u32>(vertex_bindings.size()), vertex_bindings.data(),
+        static_cast<u32>(vertex_attributes.size()), vertex_attributes.data());
+
+    const vk::PipelineVertexInputDivisorStateCreateInfoEXT vertex_input_divisor_ci(
+        static_cast<u32>(vertex_binding_divisors.size()), vertex_binding_divisors.data());
+    if (!vertex_binding_divisors.empty()) {
+        vertex_input_ci.pNext = &vertex_input_divisor_ci;
+    }
+
+    const auto primitive_topology = MaxwellToVK::PrimitiveTopology(device, ia.topology);
+    const vk::PipelineInputAssemblyStateCreateInfo input_assembly_ci(
+        {}, primitive_topology,
+        ia.primitive_restart_enable && SupportsPrimitiveRestart(primitive_topology));
+
+    const vk::PipelineTessellationStateCreateInfo tessellation_ci({}, ts.patch_control_points);
+
+    const vk::PipelineViewportStateCreateInfo viewport_ci({}, Maxwell::NumViewports, nullptr,
+                                                          Maxwell::NumViewports, nullptr);
+
+    // TODO(Rodrigo): Find out what's the default register value for front face
+    const vk::PipelineRasterizationStateCreateInfo rasterizer_ci(
+        {}, rs.depth_clamp_enable, false, vk::PolygonMode::eFill,
+        rs.cull_enable ? MaxwellToVK::CullFace(rs.cull_face) : vk::CullModeFlagBits::eNone,
+        MaxwellToVK::FrontFace(rs.front_face), rs.depth_bias_enable, 0.0f, 0.0f, 0.0f, 1.0f);
+
+    const vk::PipelineMultisampleStateCreateInfo multisampling_ci(
+        {}, vk::SampleCountFlagBits::e1, false, 0.0f, nullptr, false, false);
+
+    const vk::CompareOp depth_test_compare = ds.depth_test_enable
+                                                 ? MaxwellToVK::ComparisonOp(ds.depth_test_function)
+                                                 : vk::CompareOp::eAlways;
+
+    const vk::PipelineDepthStencilStateCreateInfo depth_stencil_ci(
+        {}, ds.depth_test_enable, ds.depth_write_enable, depth_test_compare, ds.depth_bounds_enable,
+        ds.stencil_enable, GetStencilFaceState(ds.front_stencil),
+        GetStencilFaceState(ds.back_stencil), 0.0f, 0.0f);
+
+    std::array<vk::PipelineColorBlendAttachmentState, Maxwell::NumRenderTargets> cb_attachments;
+    const std::size_t num_attachments =
+        std::min(cd.attachments_count, renderpass_params.color_attachments.size());
+    for (std::size_t i = 0; i < num_attachments; ++i) {
+        constexpr std::array component_table{
+            vk::ColorComponentFlagBits::eR, vk::ColorComponentFlagBits::eG,
+            vk::ColorComponentFlagBits::eB, vk::ColorComponentFlagBits::eA};
+        const auto& blend = cd.attachments[i];
+
+        vk::ColorComponentFlags color_components{};
+        for (std::size_t j = 0; j < component_table.size(); ++j) {
+            if (blend.components[j])
+                color_components |= component_table[j];
+        }
+
+        cb_attachments[i] = vk::PipelineColorBlendAttachmentState(
+            blend.enable, MaxwellToVK::BlendFactor(blend.src_rgb_func),
+            MaxwellToVK::BlendFactor(blend.dst_rgb_func),
+            MaxwellToVK::BlendEquation(blend.rgb_equation),
+            MaxwellToVK::BlendFactor(blend.src_a_func), MaxwellToVK::BlendFactor(blend.dst_a_func),
+            MaxwellToVK::BlendEquation(blend.a_equation), color_components);
+    }
+    const vk::PipelineColorBlendStateCreateInfo color_blending_ci({}, false, vk::LogicOp::eCopy,
+                                                                  static_cast<u32>(num_attachments),
+                                                                  cb_attachments.data(), {});
+
+    constexpr std::array dynamic_states = {
+        vk::DynamicState::eViewport,         vk::DynamicState::eScissor,
+        vk::DynamicState::eDepthBias,        vk::DynamicState::eBlendConstants,
+        vk::DynamicState::eDepthBounds,      vk::DynamicState::eStencilCompareMask,
+        vk::DynamicState::eStencilWriteMask, vk::DynamicState::eStencilReference};
+    const vk::PipelineDynamicStateCreateInfo dynamic_state_ci(
+        {}, static_cast<u32>(dynamic_states.size()), dynamic_states.data());
+
+    vk::PipelineShaderStageRequiredSubgroupSizeCreateInfoEXT subgroup_size_ci;
+    subgroup_size_ci.requiredSubgroupSize = GuestWarpSize;
+
+    std::vector<vk::PipelineShaderStageCreateInfo> shader_stages;
+    std::size_t module_index = 0;
+    for (std::size_t stage = 0; stage < Maxwell::MaxShaderStage; ++stage) {
+        if (!program[stage]) {
+            continue;
+        }
+        const auto stage_enum = static_cast<Tegra::Engines::ShaderType>(stage);
+        const auto vk_stage = MaxwellToVK::ShaderStage(stage_enum);
+        auto& stage_ci = shader_stages.emplace_back(vk::PipelineShaderStageCreateFlags{}, vk_stage,
+                                                    *modules[module_index++], "main", nullptr);
+        if (program[stage]->entries.uses_warps && device.IsGuestWarpSizeSupported(vk_stage)) {
+            stage_ci.pNext = &subgroup_size_ci;
+        }
+    }
+
+    const vk::GraphicsPipelineCreateInfo create_info(
+        {}, static_cast<u32>(shader_stages.size()), shader_stages.data(), &vertex_input_ci,
+        &input_assembly_ci, &tessellation_ci, &viewport_ci, &rasterizer_ci, &multisampling_ci,
+        &depth_stencil_ci, &color_blending_ci, &dynamic_state_ci, *layout, renderpass, 0, {}, 0);
+
+    const auto dev = device.GetLogical();
+    const auto& dld = device.GetDispatchLoader();
+    return dev.createGraphicsPipelineUnique(nullptr, create_info, nullptr, dld);
+}
+
+} // namespace Vulkan

src/video_core/renderer_vulkan/vk_graphics_pipeline.h

@@ -0,0 +1,90 @@
+// Copyright 2019 yuzu Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+#include <memory>
+#include <optional>
+#include <unordered_map>
+#include <vector>
+
+#include "video_core/engines/maxwell_3d.h"
+#include "video_core/renderer_vulkan/declarations.h"
+#include "video_core/renderer_vulkan/fixed_pipeline_state.h"
+#include "video_core/renderer_vulkan/vk_descriptor_pool.h"
+#include "video_core/renderer_vulkan/vk_renderpass_cache.h"
+#include "video_core/renderer_vulkan/vk_resource_manager.h"
+#include "video_core/renderer_vulkan/vk_shader_decompiler.h"
+
+namespace Vulkan {
+
+using Maxwell = Tegra::Engines::Maxwell3D::Regs;
+
+struct GraphicsPipelineCacheKey;
+
+class VKDescriptorPool;
+class VKDevice;
+class VKRenderPassCache;
+class VKScheduler;
+class VKUpdateDescriptorQueue;
+
+using SPIRVProgram = std::array<std::optional<SPIRVShader>, Maxwell::MaxShaderStage>;
+
+class VKGraphicsPipeline final {
+public:
+    explicit VKGraphicsPipeline(const VKDevice& device, VKScheduler& scheduler,
+                                VKDescriptorPool& descriptor_pool,
+                                VKUpdateDescriptorQueue& update_descriptor_queue,
+                                VKRenderPassCache& renderpass_cache,
+                                const GraphicsPipelineCacheKey& key,
+                                const std::vector<vk::DescriptorSetLayoutBinding>& bindings,
+                                const SPIRVProgram& program);
+    ~VKGraphicsPipeline();
+
+    vk::DescriptorSet CommitDescriptorSet();
+
+    vk::Pipeline GetHandle() const {
+        return *pipeline;
+    }
+
+    vk::PipelineLayout GetLayout() const {
+        return *layout;
+    }
+
+    vk::RenderPass GetRenderPass() const {
+        return renderpass;
+    }
+
+private:
+    UniqueDescriptorSetLayout CreateDescriptorSetLayout(
+        const std::vector<vk::DescriptorSetLayoutBinding>& bindings) const;
+
+    UniquePipelineLayout CreatePipelineLayout() const;
+
+    UniqueDescriptorUpdateTemplate CreateDescriptorUpdateTemplate(
+        const SPIRVProgram& program) const;
+
+    std::vector<UniqueShaderModule> CreateShaderModules(const SPIRVProgram& program) const;
+
+    UniquePipeline CreatePipeline(const RenderPassParams& renderpass_params,
+                                  const SPIRVProgram& program) const;
+
+    const VKDevice& device;
+    VKScheduler& scheduler;
+    const FixedPipelineState fixed_state;
+    const u64 hash;
+
+    UniqueDescriptorSetLayout descriptor_set_layout;
+    DescriptorAllocator descriptor_allocator;
+    VKUpdateDescriptorQueue& update_descriptor_queue;
+    UniquePipelineLayout layout;
+    UniqueDescriptorUpdateTemplate descriptor_template;
+    std::vector<UniqueShaderModule> modules;
+
+    vk::RenderPass renderpass;
+    UniquePipeline pipeline;
+};
+
+} // namespace Vulkan