New upstream version 0-1299+ds

.ci/scripts/format/script.sh

@@ -10,7 +10,7 @@ if grep -nrI '\s$' src *.yml *.txt *.md Doxyfile .gitignore .gitmodules .ci* dis
 fi
 
 # Default clang-format points to default 3.5 version one
-CLANG_FORMAT=clang-format-12
+CLANG_FORMAT=${CLANG_FORMAT:-clang-format-12}
 $CLANG_FORMAT --version
 
 if [ "$TRAVIS_EVENT_TYPE" = "pull_request" ]; then

CMakeLists.txt

@@ -3,13 +3,8 @@
 
 cmake_minimum_required(VERSION 3.22)
 
-# Dynarmic has cmake_minimum_required(3.12) and we may want to override
-# some of its variables, which is only possible in 3.13+
-set(CMAKE_POLICY_DEFAULT_CMP0077 NEW)
-
 list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/CMakeModules")
 list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/externals/cmake-modules")
-list(APPEND CMAKE_MODULE_PATH "${CMAKE_CURRENT_SOURCE_DIR}/externals/find-modules")
 include(DownloadExternals)
 include(CMakeDependentOption)
 
@@ -22,6 +17,8 @@ CMAKE_DEPENDENT_OPTION(YUZU_USE_BUNDLED_SDL2 "Download bundled SDL2 binaries" ON
 # On Linux system SDL2 is likely to be lacking HIDAPI support which have drawbacks but is needed for SDL motion
 CMAKE_DEPENDENT_OPTION(YUZU_USE_EXTERNAL_SDL2 "Compile external SDL2" ON "ENABLE_SDL2;NOT MSVC" OFF)
 
+option(ENABLE_LIBUSB "Enable the use of LibUSB" ON)
+
 option(ENABLE_OPENGL "Enable OpenGL" ON)
 mark_as_advanced(FORCE ENABLE_OPENGL)
 option(ENABLE_QT "Enable the Qt frontend" ON)
@@ -35,6 +32,8 @@ option(ENABLE_WEB_SERVICE "Enable web services (telemetry, etc.)" ON)
 
 option(YUZU_USE_BUNDLED_FFMPEG "Download/Build bundled FFmpeg" "${WIN32}")
 
+option(YUZU_USE_EXTERNAL_VULKAN_HEADERS "Use Vulkan-Headers from externals" ON)
+
 option(YUZU_USE_QT_MULTIMEDIA "Use QtMultimedia for Camera" OFF)
 
 option(YUZU_USE_QT_WEB_ENGINE "Use QtWebEngine for web applet implementation" OFF)
@@ -47,6 +46,8 @@ option(YUZU_TESTS "Compile tests" ON)
 
 option(YUZU_USE_PRECOMPILED_HEADERS "Use precompiled headers" ON)
 
+option(YUZU_ROOM "Compile LDN room server" ON)
+
 CMAKE_DEPENDENT_OPTION(YUZU_CRASH_DUMPS "Compile Windows crash dump (Minidump) support" OFF "WIN32" OFF)
 
 option(YUZU_USE_BUNDLED_VCPKG "Use vcpkg for yuzu dependencies" "${MSVC}")
@@ -201,36 +202,43 @@ set(CMAKE_RUNTIME_OUTPUT_DIRECTORY ${PROJECT_BINARY_DIR}/bin)
 # System imported libraries
 # =======================================================================
 
-find_package(enet 1.3)
+# Enforce the search mode of non-required packages for better and shorter failure messages
+find_package(enet 1.3 MODULE)
 find_package(fmt 9 REQUIRED)
-find_package(inih)
-find_package(libusb 1.0.24)
+find_package(inih MODULE)
 find_package(lz4 REQUIRED)
 find_package(nlohmann_json 3.8 REQUIRED)
-find_package(Opus 1.3)
-find_package(Vulkan 1.3.238)
+find_package(Opus 1.3 MODULE)
 find_package(ZLIB 1.2 REQUIRED)
 find_package(zstd 1.5 REQUIRED)
 
+if (NOT YUZU_USE_EXTERNAL_VULKAN_HEADERS)
+    find_package(Vulkan 1.3.238 REQUIRED)
+endif()
+
+if (ENABLE_LIBUSB)
+    find_package(libusb 1.0.24 MODULE)
+endif()
+
 if (ARCHITECTURE_x86 OR ARCHITECTURE_x86_64)
-    find_package(xbyak 6)
+    find_package(xbyak 6 CONFIG)
 endif()
 
 if (ARCHITECTURE_x86_64 OR ARCHITECTURE_arm64)
-    find_package(dynarmic 6.4.0)
+    find_package(dynarmic 6.4.0 CONFIG)
 endif()
 
 if (ENABLE_CUBEB)
-    find_package(cubeb)
+    find_package(cubeb CONFIG)
 endif()
 
 if (USE_DISCORD_PRESENCE)
-    find_package(DiscordRPC)
+    find_package(DiscordRPC MODULE)
 endif()
 
 if (ENABLE_WEB_SERVICE)
-    find_package(cpp-jwt 1.4)
-    find_package(httplib 0.11)
+    find_package(cpp-jwt 1.4 CONFIG)
+    find_package(httplib 0.11 MODULE)
 endif()
 
 if (YUZU_TESTS)

CMakeModules/Findhttplib.cmake

@@ -5,7 +5,7 @@
 include(FindPackageHandleStandardArgs)
 
 find_package(httplib QUIET CONFIG)
-if (httplib_FOUND)
+if (httplib_CONSIDERED_CONFIGS)
     find_package_handle_standard_args(httplib CONFIG_MODE)
 else()
     find_package(PkgConfig QUIET)

CMakeModules/Findlz4.cmake

@@ -4,7 +4,7 @@
 include(FindPackageHandleStandardArgs)
 
 find_package(lz4 QUIET CONFIG)
-if (lz4_FOUND)
+if (lz4_CONSIDERED_CONFIGS)
     find_package_handle_standard_args(lz4 CONFIG_MODE)
 else()
     find_package(PkgConfig QUIET)

CMakeModules/Findzstd.cmake

@@ -4,7 +4,7 @@
 include(FindPackageHandleStandardArgs)
 
 find_package(zstd QUIET CONFIG)
-if (zstd_FOUND)
+if (zstd_CONSIDERED_CONFIGS)
     find_package_handle_standard_args(zstd CONFIG_MODE)
 else()
     find_package(PkgConfig QUIET)

externals/CMakeLists.txt

@@ -1,9 +1,9 @@
 # SPDX-FileCopyrightText: 2016 Citra Emulator Project
 # SPDX-License-Identifier: GPL-2.0-or-later
 
-list(APPEND CMAKE_MODULE_PATH "${PROJECT_SOURCE_DIR}/CMakeModules")
-list(APPEND CMAKE_MODULE_PATH "${PROJECT_SOURCE_DIR}/externals/find-modules")
-include(DownloadExternals)
+# Dynarmic has cmake_minimum_required(3.12) and we may want to override
+# some of its variables, which is only possible in 3.13+
+set(CMAKE_POLICY_DEFAULT_CMP0077 NEW)
 
 # xbyak
 if ((ARCHITECTURE_x86 OR ARCHITECTURE_x86_64) AND NOT TARGET xbyak::xbyak)
@@ -12,8 +12,7 @@ endif()
 
 # Dynarmic
 if ((ARCHITECTURE_x86_64 OR ARCHITECTURE_arm64) AND NOT TARGET dynarmic::dynarmic)
-    set(DYNARMIC_NO_BUNDLED_FMT ON)
-    set(DYNARMIC_IGNORE_ASSERTS ON CACHE BOOL "" FORCE)
+    set(DYNARMIC_IGNORE_ASSERTS ON)
     add_subdirectory(dynarmic EXCLUDE_FROM_ALL)
     add_library(dynarmic::dynarmic ALIAS dynarmic)
 endif()
@@ -45,7 +44,7 @@ if (CMAKE_CXX_COMPILER_VERSION VERSION_GREATER_EQUAL "12" AND CMAKE_CXX_COMPILER
 endif()
 
 # libusb
-if (NOT TARGET libusb::usb)
+if (ENABLE_LIBUSB AND NOT TARGET libusb::usb)
     add_subdirectory(libusb EXCLUDE_FROM_ALL)
 endif()
 
@@ -60,10 +59,10 @@ if (YUZU_USE_EXTERNAL_SDL2)
             Locale Power Render)
         foreach(_SUB ${SDL_UNUSED_SUBSYSTEMS})
           string(TOUPPER ${_SUB} _OPT)
-          option(SDL_${_OPT} "" OFF)
+          set(SDL_${_OPT} OFF)
         endforeach()
 
-        option(HIDAPI "" ON)
+        set(HIDAPI ON)
     endif()
     set(SDL_STATIC ON)
     set(SDL_SHARED OFF)
@@ -83,7 +82,7 @@ endif()
 
 # Cubeb
 if (ENABLE_CUBEB AND NOT TARGET cubeb::cubeb)
-    set(BUILD_TESTS OFF CACHE BOOL "")
+    set(BUILD_TESTS OFF)
     add_subdirectory(cubeb EXCLUDE_FROM_ALL)
     add_library(cubeb::cubeb ALIAS cubeb)
 endif()
@@ -98,6 +97,7 @@ endif()
 # Sirit
 add_subdirectory(sirit EXCLUDE_FROM_ALL)
 
+# httplib
 if (ENABLE_WEB_SERVICE AND NOT TARGET httplib::httplib)
     if (NOT WIN32)
         find_package(OpenSSL 1.1)
@@ -108,7 +108,7 @@ if (ENABLE_WEB_SERVICE AND NOT TARGET httplib::httplib)
 
     if (WIN32 OR NOT OPENSSL_FOUND)
         # LibreSSL
-        set(LIBRESSL_SKIP_INSTALL ON CACHE BOOL "")
+        set(LIBRESSL_SKIP_INSTALL ON)
         set(OPENSSLDIR "/etc/ssl/")
         add_subdirectory(libressl EXCLUDE_FROM_ALL)
         target_include_directories(ssl INTERFACE ./libressl/include)
@@ -118,7 +118,6 @@ if (ENABLE_WEB_SERVICE AND NOT TARGET httplib::httplib)
             DEFINITION OPENSSL_LIBS)
     endif()
 
-    # httplib
     add_library(httplib INTERFACE)
     target_include_directories(httplib INTERFACE ./cpp-httplib)
     target_compile_definitions(httplib INTERFACE -DCPPHTTPLIB_OPENSSL_SUPPORT)
@@ -152,6 +151,6 @@ if (YUZU_USE_BUNDLED_FFMPEG)
 endif()
 
 # Vulkan-Headers
-if (NOT TARGET Vulkan::Headers)
+if (YUZU_USE_EXTERNAL_VULKAN_HEADERS)
     add_subdirectory(Vulkan-Headers EXCLUDE_FROM_ALL)
 endif()

sirit/include/sirit/sirit.h

@@ -1160,6 +1160,14 @@ public:
     /// TBD
     Id OpSubgroupAllEqualKHR(Id result_type, Id predicate);
 
+    // Result is true only in the active invocation with the lowest id in the group, otherwise
+    // result is false.
+    Id OpGroupNonUniformElect(Id result_type, Id scope);
+
+    // Result is the Value of the invocation from the active invocation with the lowest id in the
+    // group to all active invocations in the group.
+    Id OpGroupNonUniformBroadcastFirst(Id result_type, Id scope, Id value);
+
     // Result is the Value of the invocation identified by the id Id to all active invocations in
     // the group.
     Id OpGroupNonUniformBroadcast(Id result_type, Id scope, Id value, Id id);

sirit/src/instructions/group.cpp

@@ -36,6 +36,17 @@ Id Module::OpSubgroupAllEqualKHR(Id result_type, Id predicate) {
     return *code << OpId{spv::Op::OpSubgroupAllEqualKHR, result_type} << predicate << EndOp{};
 }
 
+Id Module::OpGroupNonUniformElect(Id result_type, Id scope) {
+    code->Reserve(4);
+    return *code << OpId{spv::Op::OpGroupNonUniformElect, result_type} << scope << EndOp{};
+}
+
+Id Module::OpGroupNonUniformBroadcastFirst(Id result_type, Id scope, Id value) {
+    code->Reserve(5);
+    return *code << OpId{spv::Op::OpGroupNonUniformBroadcastFirst, result_type} << scope << value
+                 << EndOp{};
+}
+
 Id Module::OpGroupNonUniformBroadcast(Id result_type, Id scope, Id value, Id id) {
     code->Reserve(6);
     return *code << OpId{spv::Op::OpGroupNonUniformBroadcast, result_type} << scope << value

src/CMakeLists.txt

@@ -161,7 +161,10 @@ add_subdirectory(video_core)
 add_subdirectory(network)
 add_subdirectory(input_common)
 add_subdirectory(shader_recompiler)
-add_subdirectory(dedicated_room)
+
+if (YUZU_ROOM)
+     add_subdirectory(dedicated_room)
+endif()
 
 if (YUZU_TESTS)
     add_subdirectory(tests)

src/audio_core/CMakeLists.txt

@@ -187,11 +187,7 @@ add_library(audio_core STATIC
     renderer/voice/voice_info.cpp
     renderer/voice/voice_info.h
     renderer/voice/voice_state.h
-    sink/cubeb_sink.cpp
-    sink/cubeb_sink.h
     sink/null_sink.h
-    sink/sdl2_sink.cpp
-    sink/sdl2_sink.h
     sink/sink.h
     sink/sink_details.cpp
     sink/sink_details.h
@@ -222,11 +218,22 @@ if (ARCHITECTURE_x86_64 OR ARCHITECTURE_arm64)
     target_link_libraries(audio_core PRIVATE dynarmic::dynarmic)
 endif()
 
-if(ENABLE_CUBEB)
+if (ENABLE_CUBEB)
+    target_sources(audio_core PRIVATE
+        sink/cubeb_sink.cpp
+        sink/cubeb_sink.h
+    )
+
     target_link_libraries(audio_core PRIVATE cubeb::cubeb)
     target_compile_definitions(audio_core PRIVATE -DHAVE_CUBEB=1)
 endif()
-if(ENABLE_SDL2)
+
+if (ENABLE_SDL2)
+    target_sources(audio_core PRIVATE
+        sink/sdl2_sink.cpp
+        sink/sdl2_sink.h
+    )
+
     target_link_libraries(audio_core PRIVATE SDL2::SDL2)
     target_compile_definitions(audio_core PRIVATE HAVE_SDL2)
 endif()

src/common/CMakeLists.txt

@@ -97,6 +97,7 @@ add_library(common STATIC
     point.h
     precompiled_headers.h
     quaternion.h
+    range_map.h
     reader_writer_queue.h
     ring_buffer.h
     ${CMAKE_CURRENT_BINARY_DIR}/scm_rev.cpp

src/common/host_memory.cpp

@@ -393,12 +393,27 @@ public:
         }
 
         // Virtual memory initialization
-        virtual_base = static_cast<u8*>(
-            mmap(nullptr, virtual_size, PROT_NONE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0));
+#if defined(__FreeBSD__)
+        virtual_base =
+            static_cast<u8*>(mmap(nullptr, virtual_size, PROT_NONE,
+                                  MAP_PRIVATE | MAP_ANONYMOUS | MAP_ALIGNED_SUPER, -1, 0));
+        if (virtual_base == MAP_FAILED) {
+            virtual_base = static_cast<u8*>(
+                mmap(nullptr, virtual_size, PROT_NONE, MAP_PRIVATE | MAP_ANONYMOUS, -1, 0));
+            if (virtual_base == MAP_FAILED) {
+                LOG_CRITICAL(HW_Memory, "mmap failed: {}", strerror(errno));
+                throw std::bad_alloc{};
+            }
+        }
+#else
+        virtual_base = static_cast<u8*>(mmap(nullptr, virtual_size, PROT_NONE,
+                                             MAP_PRIVATE | MAP_ANONYMOUS | MAP_NORESERVE, -1, 0));
         if (virtual_base == MAP_FAILED) {
             LOG_CRITICAL(HW_Memory, "mmap failed: {}", strerror(errno));
             throw std::bad_alloc{};
         }
+        madvise(virtual_base, virtual_size, MADV_HUGEPAGE);
+#endif
 
         good = true;
     }

src/common/range_map.h

@@ -0,0 +1,139 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#pragma once
+
+#include <map>
+#include <type_traits>
+
+#include "common/common_types.h"
+
+namespace Common {
+
+template <typename KeyTBase, typename ValueT>
+class RangeMap {
+private:
+    using KeyT =
+        std::conditional_t<std::is_signed_v<KeyTBase>, KeyTBase, std::make_signed_t<KeyTBase>>;
+
+public:
+    explicit RangeMap(ValueT null_value_) : null_value{null_value_} {
+        container.emplace(std::numeric_limits<KeyT>::min(), null_value);
+    };
+    ~RangeMap() = default;
+
+    void Map(KeyTBase address, KeyTBase address_end, ValueT value) {
+        KeyT new_address = static_cast<KeyT>(address);
+        KeyT new_address_end = static_cast<KeyT>(address_end);
+        if (new_address < 0) {
+            new_address = 0;
+        }
+        if (new_address_end < 0) {
+            new_address_end = 0;
+        }
+        InternalMap(new_address, new_address_end, value);
+    }
+
+    void Unmap(KeyTBase address, KeyTBase address_end) {
+        Map(address, address_end, null_value);
+    }
+
+    [[nodiscard]] size_t GetContinousSizeFrom(KeyTBase address) const {
+        const KeyT new_address = static_cast<KeyT>(address);
+        if (new_address < 0) {
+            return 0;
+        }
+        return ContinousSizeInternal(new_address);
+    }
+
+    [[nodiscard]] ValueT GetValueAt(KeyT address) const {
+        const KeyT new_address = static_cast<KeyT>(address);
+        if (new_address < 0) {
+            return null_value;
+        }
+        return GetValueInternal(new_address);
+    }
+
+private:
+    using MapType = std::map<KeyT, ValueT>;
+    using IteratorType = typename MapType::iterator;
+    using ConstIteratorType = typename MapType::const_iterator;
+
+    size_t ContinousSizeInternal(KeyT address) const {
+        const auto it = GetFirstElementBeforeOrOn(address);
+        if (it == container.end() || it->second == null_value) {
+            return 0;
+        }
+        const auto it_end = std::next(it);
+        if (it_end == container.end()) {
+            return std::numeric_limits<KeyT>::max() - address;
+        }
+        return it_end->first - address;
+    }
+
+    ValueT GetValueInternal(KeyT address) const {
+        const auto it = GetFirstElementBeforeOrOn(address);
+        if (it == container.end()) {
+            return null_value;
+        }
+        return it->second;
+    }
+
+    ConstIteratorType GetFirstElementBeforeOrOn(KeyT address) const {
+        auto it = container.lower_bound(address);
+        if (it == container.begin()) {
+            return it;
+        }
+        if (it != container.end() && (it->first == address)) {
+            return it;
+        }
+        --it;
+        return it;
+    }
+
+    ValueT GetFirstValueWithin(KeyT address) {
+        auto it = container.lower_bound(address);
+        if (it == container.begin()) {
+            return it->second;
+        }
+        if (it == container.end()) [[unlikely]] { // this would be a bug
+            return null_value;
+        }
+        --it;
+        return it->second;
+    }
+
+    ValueT GetLastValueWithin(KeyT address) {
+        auto it = container.upper_bound(address);
+        if (it == container.end()) {
+            return null_value;
+        }
+        if (it == container.begin()) [[unlikely]] { // this would be a bug
+            return it->second;
+        }
+        --it;
+        return it->second;
+    }
+
+    void InternalMap(KeyT address, KeyT address_end, ValueT value) {
+        const bool must_add_start = GetFirstValueWithin(address) != value;
+        const ValueT last_value = GetLastValueWithin(address_end);
+        const bool must_add_end = last_value != value;
+        auto it = container.lower_bound(address);
+        const auto it_end = container.upper_bound(address_end);
+        while (it != it_end) {
+            it = container.erase(it);
+        }
+        if (must_add_start) {
+            container.emplace(address, value);
+        }
+        if (must_add_end) {
+            container.emplace(address_end, last_value);
+        }
+    }
+
+    ValueT null_value;
+    MapType container;
+};
+
+} // namespace Common

src/common/settings.cpp

@@ -185,6 +185,7 @@ void RestoreGlobalState(bool is_powered_on) {
     // Renderer
     values.fsr_sharpening_slider.SetGlobal(true);
     values.renderer_backend.SetGlobal(true);
+    values.renderer_force_max_clock.SetGlobal(true);
     values.vulkan_device.SetGlobal(true);
     values.aspect_ratio.SetGlobal(true);
     values.max_anisotropy.SetGlobal(true);
@@ -200,6 +201,7 @@ void RestoreGlobalState(bool is_powered_on) {
     values.use_asynchronous_shaders.SetGlobal(true);
     values.use_fast_gpu_time.SetGlobal(true);
     values.use_pessimistic_flushes.SetGlobal(true);
+    values.use_vulkan_driver_pipeline_cache.SetGlobal(true);
     values.bg_red.SetGlobal(true);
     values.bg_green.SetGlobal(true);
     values.bg_blue.SetGlobal(true);

src/common/settings.h

@@ -415,6 +415,7 @@ struct Values {
     // Renderer
     SwitchableSetting<RendererBackend, true> renderer_backend{
         RendererBackend::Vulkan, RendererBackend::OpenGL, RendererBackend::Null, "backend"};
+    SwitchableSetting<bool> renderer_force_max_clock{true, "force_max_clock"};
     Setting<bool> renderer_debug{false, "debug"};
     Setting<bool> renderer_shader_feedback{false, "shader_feedback"};
     Setting<bool> enable_nsight_aftermath{false, "nsight_aftermath"};
@@ -451,6 +452,8 @@ struct Values {
     SwitchableSetting<bool> use_asynchronous_shaders{false, "use_asynchronous_shaders"};
     SwitchableSetting<bool> use_fast_gpu_time{true, "use_fast_gpu_time"};
     SwitchableSetting<bool> use_pessimistic_flushes{false, "use_pessimistic_flushes"};
+    SwitchableSetting<bool> use_vulkan_driver_pipeline_cache{true,
+                                                             "use_vulkan_driver_pipeline_cache"};
 
     SwitchableSetting<u8> bg_red{0, "bg_red"};
     SwitchableSetting<u8> bg_green{0, "bg_green"};
@@ -531,6 +534,7 @@ struct Values {
     Setting<bool> reporting_services{false, "reporting_services"};
     Setting<bool> quest_flag{false, "quest_flag"};
     Setting<bool> disable_macro_jit{false, "disable_macro_jit"};
+    Setting<bool> disable_macro_hle{false, "disable_macro_hle"};
     Setting<bool> extended_logging{false, "extended_logging"};
     Setting<bool> use_debug_asserts{false, "use_debug_asserts"};
     Setting<bool> use_auto_stub{false, "use_auto_stub"};

src/core/arm/dynarmic/arm_dynarmic_32.cpp

@@ -229,7 +229,11 @@ std::shared_ptr<Dynarmic::A32::Jit> ARM_Dynarmic_32::MakeJit(Common::PageTable*
     config.enable_cycle_counting = true;
 
     // Code cache size
+#ifdef ARCHITECTURE_arm64
+    config.code_cache_size = 128_MiB;
+#else
     config.code_cache_size = 512_MiB;
+#endif
 
     // Allow memory fault handling to work
     if (system.DebuggerEnabled()) {

src/core/arm/dynarmic/arm_dynarmic_64.cpp

@@ -288,7 +288,11 @@ std::shared_ptr<Dynarmic::A64::Jit> ARM_Dynarmic_64::MakeJit(Common::PageTable*
     config.enable_cycle_counting = true;
 
     // Code cache size
+#ifdef ARCHITECTURE_arm64
+    config.code_cache_size = 128_MiB;
+#else
     config.code_cache_size = 512_MiB;
+#endif
 
     // Allow memory fault handling to work
     if (system.DebuggerEnabled()) {

src/core/file_sys/vfs.cpp

@@ -194,9 +194,9 @@ std::size_t VfsFile::WriteBytes(const std::vector<u8>& data, std::size_t offset)
 
 std::string VfsFile::GetFullPath() const {
     if (GetContainingDirectory() == nullptr)
-        return "/" + GetName();
+        return '/' + GetName();
 
-    return GetContainingDirectory()->GetFullPath() + "/" + GetName();
+    return GetContainingDirectory()->GetFullPath() + '/' + GetName();
 }
 
 VirtualFile VfsDirectory::GetFileRelative(std::string_view path) const {
@@ -435,7 +435,7 @@ std::string VfsDirectory::GetFullPath() const {
     if (IsRoot())
         return GetName();
 
-    return GetParentDirectory()->GetFullPath() + "/" + GetName();
+    return GetParentDirectory()->GetFullPath() + '/' + GetName();
 }
 
 bool ReadOnlyVfsDirectory::IsWritable() const {

src/core/hid/emulated_console.cpp

@@ -40,6 +40,11 @@ void EmulatedConsole::SetTouchParams() {
         touch_params[index++] = std::move(touchscreen_param);
     }
 
+    if (Settings::values.touch_from_button_maps.empty()) {
+        LOG_WARNING(Input, "touch_from_button_maps is unset by frontend config");
+        return;
+    }
+
     const auto button_index =
         static_cast<u64>(Settings::values.touch_from_button_map_index.GetValue());
     const auto& touch_buttons = Settings::values.touch_from_button_maps[button_index].buttons;

src/core/hid/emulated_controller.cpp

@@ -11,6 +11,11 @@
 namespace Core::HID {
 constexpr s32 HID_JOYSTICK_MAX = 0x7fff;
 constexpr s32 HID_TRIGGER_MAX = 0x7fff;
+// Use a common UUID for TAS and Virtual Gamepad
+constexpr Common::UUID TAS_UUID =
+    Common::UUID{{0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x7, 0xA5, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}};
+constexpr Common::UUID VIRTUAL_UUID =
+    Common::UUID{{0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x7, 0xFF, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}};
 
 EmulatedController::EmulatedController(NpadIdType npad_id_type_) : npad_id_type(npad_id_type_) {}
 
@@ -348,10 +353,6 @@ void EmulatedController::ReloadInput() {
         }
     }
 
-    // Use a common UUID for TAS
-    static constexpr Common::UUID TAS_UUID = Common::UUID{
-        {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x7, 0xA5, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}};
-
     // Register TAS devices. No need to force update
     for (std::size_t index = 0; index < tas_button_devices.size(); ++index) {
         if (!tas_button_devices[index]) {
@@ -377,10 +378,6 @@ void EmulatedController::ReloadInput() {
         });
     }
 
-    // Use a common UUID for Virtual Gamepad
-    static constexpr Common::UUID VIRTUAL_UUID = Common::UUID{
-        {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x7, 0xFF, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}};
-
     // Register virtual devices. No need to force update
     for (std::size_t index = 0; index < virtual_button_devices.size(); ++index) {
         if (!virtual_button_devices[index]) {
@@ -780,7 +777,12 @@ void EmulatedController::SetStick(const Common::Input::CallbackStatus& callback,
 
     // Only read stick values that have the same uuid or are over the threshold to avoid flapping
     if (controller.stick_values[index].uuid != uuid) {
-        if (!stick_value.down && !stick_value.up && !stick_value.left && !stick_value.right) {
+        const bool is_tas = uuid == TAS_UUID;
+        if (is_tas && stick_value.x.value == 0 && stick_value.y.value == 0) {
+            return;
+        }
+        if (!is_tas && !stick_value.down && !stick_value.up && !stick_value.left &&
+            !stick_value.right) {
             return;
         }
     }

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

@@ -22,15 +22,19 @@ namespace {
 
 namespace Service::NIFM {
 
+// This is nn::nifm::RequestState
 enum class RequestState : u32 {
     NotSubmitted = 1,
-    Error = 1, ///< The duplicate 1 is intentional; it means both not submitted and error on HW.
-    Pending = 2,
-    Connected = 3,
+    Invalid = 1, ///< The duplicate 1 is intentional; it means both not submitted and error on HW.
+    OnHold = 2,
+    Accepted = 3,
+    Blocking = 4,
 };
 
-enum class InternetConnectionType : u8 {
-    WiFi = 1,
+// This is nn::nifm::NetworkInterfaceType
+enum class NetworkInterfaceType : u32 {
+    Invalid = 0,
+    WiFi_Ieee80211 = 1,
     Ethernet = 2,
 };
 
@@ -42,14 +46,23 @@ enum class InternetConnectionStatus : u8 {
     Connected,
 };
 
+// This is nn::nifm::NetworkProfileType
+enum class NetworkProfileType : u32 {
+    User,
+    SsidList,
+    Temporary,
+};
+
+// This is nn::nifm::IpAddressSetting
 struct IpAddressSetting {
     bool is_automatic{};
-    Network::IPv4Address current_address{};
+    Network::IPv4Address ip_address{};
     Network::IPv4Address subnet_mask{};
-    Network::IPv4Address gateway{};
+    Network::IPv4Address default_gateway{};
 };
 static_assert(sizeof(IpAddressSetting) == 0xD, "IpAddressSetting has incorrect size.");
 
+// This is nn::nifm::DnsSetting
 struct DnsSetting {
     bool is_automatic{};
     Network::IPv4Address primary_dns{};
@@ -57,18 +70,26 @@ struct DnsSetting {
 };
 static_assert(sizeof(DnsSetting) == 0x9, "DnsSetting has incorrect size.");
 
+// This is nn::nifm::AuthenticationSetting
+struct AuthenticationSetting {
+    bool is_enabled{};
+    std::array<char, 0x20> user{};
+    std::array<char, 0x20> password{};
+};
+static_assert(sizeof(AuthenticationSetting) == 0x41, "AuthenticationSetting has incorrect size.");
+
+// This is nn::nifm::ProxySetting
 struct ProxySetting {
-    bool enabled{};
+    bool is_enabled{};
     INSERT_PADDING_BYTES(1);
     u16 port{};
     std::array<char, 0x64> proxy_server{};
-    bool automatic_auth_enabled{};
-    std::array<char, 0x20> user{};
-    std::array<char, 0x20> password{};
+    AuthenticationSetting authentication{};
     INSERT_PADDING_BYTES(1);
 };
 static_assert(sizeof(ProxySetting) == 0xAA, "ProxySetting has incorrect size.");
 
+// This is nn::nifm::IpSettingData
 struct IpSettingData {
     IpAddressSetting ip_address_setting{};
     DnsSetting dns_setting{};
@@ -101,6 +122,7 @@ static_assert(sizeof(NifmWirelessSettingData) == 0x70,
               "NifmWirelessSettingData has incorrect size.");
 
 #pragma pack(push, 1)
+// This is nn::nifm::detail::sf::NetworkProfileData
 struct SfNetworkProfileData {
     IpSettingData ip_setting_data{};
     u128 uuid{};
@@ -114,13 +136,14 @@ struct SfNetworkProfileData {
 };
 static_assert(sizeof(SfNetworkProfileData) == 0x17C, "SfNetworkProfileData has incorrect size.");
 
+// This is nn::nifm::NetworkProfileData
 struct NifmNetworkProfileData {
     u128 uuid{};
     std::array<char, 0x40> network_name{};
-    u32 unknown_1{};
-    u32 unknown_2{};
-    u8 unknown_3{};
-    u8 unknown_4{};
+    NetworkProfileType network_profile_type{};
+    NetworkInterfaceType network_interface_type{};
+    bool is_auto_connect{};
+    bool is_large_capacity{};
     INSERT_PADDING_BYTES(2);
     NifmWirelessSettingData wireless_setting_data{};
     IpSettingData ip_setting_data{};
@@ -184,6 +207,7 @@ public:
 
         event1 = CreateKEvent(service_context, "IRequest:Event1");
         event2 = CreateKEvent(service_context, "IRequest:Event2");
+        state = RequestState::NotSubmitted;
     }
 
     ~IRequest() override {
@@ -196,7 +220,7 @@ private:
         LOG_WARNING(Service_NIFM, "(STUBBED) called");
 
         if (state == RequestState::NotSubmitted) {
-            UpdateState(RequestState::Pending);
+            UpdateState(RequestState::OnHold);
         }
 
         IPC::ResponseBuilder rb{ctx, 2};
@@ -219,14 +243,14 @@ private:
             switch (state) {
             case RequestState::NotSubmitted:
                 return has_connection ? ResultSuccess : ResultNetworkCommunicationDisabled;
-            case RequestState::Pending:
+            case RequestState::OnHold:
                 if (has_connection) {
-                    UpdateState(RequestState::Connected);
+                    UpdateState(RequestState::Accepted);
                 } else {
-                    UpdateState(RequestState::Error);
+                    UpdateState(RequestState::Invalid);
                 }
                 return ResultPendingConnection;
-            case RequestState::Connected:
+            case RequestState::Accepted:
             default:
                 return ResultSuccess;
             }
@@ -338,9 +362,9 @@ void IGeneralService::GetCurrentNetworkProfile(Kernel::HLERequestContext& ctx) {
             .ip_setting_data{
                 .ip_address_setting{
                     .is_automatic{true},
-                    .current_address{Network::TranslateIPv4(net_iface->ip_address)},
+                    .ip_address{Network::TranslateIPv4(net_iface->ip_address)},
                     .subnet_mask{Network::TranslateIPv4(net_iface->subnet_mask)},
-                    .gateway{Network::TranslateIPv4(net_iface->gateway)},
+                    .default_gateway{Network::TranslateIPv4(net_iface->gateway)},
                 },
                 .dns_setting{
                     .is_automatic{true},
@@ -348,12 +372,14 @@ void IGeneralService::GetCurrentNetworkProfile(Kernel::HLERequestContext& ctx) {
                     .secondary_dns{1, 0, 0, 1},
                 },
                 .proxy_setting{
-                    .enabled{false},
+                    .is_enabled{false},
                     .port{},
                     .proxy_server{},
-                    .automatic_auth_enabled{},
-                    .user{},
-                    .password{},
+                    .authentication{
+                        .is_enabled{},
+                        .user{},
+                        .password{},
+                    },
                 },
                 .mtu{1500},
             },
@@ -370,7 +396,7 @@ void IGeneralService::GetCurrentNetworkProfile(Kernel::HLERequestContext& ctx) {
     // When we're connected to a room, spoof the hosts IP address
     if (auto room_member = network.GetRoomMember().lock()) {
         if (room_member->IsConnected()) {
-            network_profile_data.ip_setting_data.ip_address_setting.current_address =
+            network_profile_data.ip_setting_data.ip_address_setting.ip_address =
                 room_member->GetFakeIpAddress();
         }
     }
@@ -444,9 +470,9 @@ void IGeneralService::GetCurrentIpConfigInfo(Kernel::HLERequestContext& ctx) {
         return IpConfigInfo{
             .ip_address_setting{
                 .is_automatic{true},
-                .current_address{Network::TranslateIPv4(net_iface->ip_address)},
+                .ip_address{Network::TranslateIPv4(net_iface->ip_address)},
                 .subnet_mask{Network::TranslateIPv4(net_iface->subnet_mask)},
-                .gateway{Network::TranslateIPv4(net_iface->gateway)},
+                .default_gateway{Network::TranslateIPv4(net_iface->gateway)},
             },
             .dns_setting{
                 .is_automatic{true},
@@ -459,7 +485,7 @@ void IGeneralService::GetCurrentIpConfigInfo(Kernel::HLERequestContext& ctx) {
     // When we're connected to a room, spoof the hosts IP address
     if (auto room_member = network.GetRoomMember().lock()) {
         if (room_member->IsConnected()) {
-            ip_config_info.ip_address_setting.current_address = room_member->GetFakeIpAddress();
+            ip_config_info.ip_address_setting.ip_address = room_member->GetFakeIpAddress();
         }
     }
 
@@ -480,7 +506,7 @@ void IGeneralService::GetInternetConnectionStatus(Kernel::HLERequestContext& ctx
     LOG_WARNING(Service_NIFM, "(STUBBED) called");
 
     struct Output {
-        InternetConnectionType type{InternetConnectionType::WiFi};
+        u8 type{static_cast<u8>(NetworkInterfaceType::WiFi_Ieee80211)};
         u8 wifi_strength{3};
         InternetConnectionStatus state{InternetConnectionStatus::Connected};
     };

src/core/internal_network/network.cpp

@@ -117,6 +117,8 @@ Errno TranslateNativeError(int e) {
         return Errno::NETUNREACH;
     case WSAEMSGSIZE:
         return Errno::MSGSIZE;
+    case WSAETIMEDOUT:
+        return Errno::TIMEDOUT;
     default:
         UNIMPLEMENTED_MSG("Unimplemented errno={}", e);
         return Errno::OTHER;
@@ -211,6 +213,8 @@ Errno TranslateNativeError(int e) {
         return Errno::NETUNREACH;
     case EMSGSIZE:
         return Errno::MSGSIZE;
+    case ETIMEDOUT:
+        return Errno::TIMEDOUT;
     default:
         UNIMPLEMENTED_MSG("Unimplemented errno={}", e);
         return Errno::OTHER;
@@ -226,7 +230,7 @@ Errno GetAndLogLastError() {
     int e = errno;
 #endif
     const Errno err = TranslateNativeError(e);
-    if (err == Errno::AGAIN) {
+    if (err == Errno::AGAIN || err == Errno::TIMEDOUT) {
         return err;
     }
     LOG_ERROR(Network, "Socket operation error: {}", Common::NativeErrorToString(e));

src/core/memory.cpp

@@ -436,7 +436,7 @@ struct Memory::Impl {
         }
 
         if (Settings::IsFastmemEnabled()) {
-            const bool is_read_enable = Settings::IsGPULevelHigh() || !cached;
+            const bool is_read_enable = !Settings::IsGPULevelExtreme() || !cached;
             system.DeviceMemory().buffer.Protect(vaddr, size, is_read_enable, !cached);
         }
 

src/dedicated_room/CMakeLists.txt

@@ -1,8 +1,6 @@
 # SPDX-FileCopyrightText: 2017 Citra Emulator Project
 # SPDX-License-Identifier: GPL-2.0-or-later
 
-set(CMAKE_MODULE_PATH ${CMAKE_MODULE_PATH} ${PROJECT_SOURCE_DIR}/CMakeModules)
-
 add_executable(yuzu-room
     precompiled_headers.h
     yuzu_room.cpp

src/input_common/CMakeLists.txt

@@ -4,14 +4,10 @@
 add_library(input_common STATIC
     drivers/camera.cpp
     drivers/camera.h
-    drivers/gc_adapter.cpp
-    drivers/gc_adapter.h
     drivers/keyboard.cpp
     drivers/keyboard.h
     drivers/mouse.cpp
     drivers/mouse.h
-    drivers/sdl_driver.cpp
-    drivers/sdl_driver.h
     drivers/tas_input.cpp
     drivers/tas_input.h
     drivers/touch_screen.cpp
@@ -62,8 +58,17 @@ if (ENABLE_SDL2)
     target_compile_definitions(input_common PRIVATE HAVE_SDL2)
 endif()
 
+if (ENABLE_LIBUSB)
+    target_sources(input_common PRIVATE
+        drivers/gc_adapter.cpp
+        drivers/gc_adapter.h
+    )
+    target_link_libraries(input_common PRIVATE libusb::usb)
+    target_compile_definitions(input_common PRIVATE HAVE_LIBUSB)
+endif()
+
 create_target_directory_groups(input_common)
-target_link_libraries(input_common PUBLIC core PRIVATE common Boost::boost libusb::usb)
+target_link_libraries(input_common PUBLIC core PRIVATE common Boost::boost)
 
 if (YUZU_USE_PRECOMPILED_HEADERS)
     target_precompile_headers(input_common PRIVATE precompiled_headers.h)

src/input_common/main.cpp

@@ -5,7 +5,6 @@
 #include "common/input.h"
 #include "common/param_package.h"
 #include "input_common/drivers/camera.h"
-#include "input_common/drivers/gc_adapter.h"
 #include "input_common/drivers/keyboard.h"
 #include "input_common/drivers/mouse.h"
 #include "input_common/drivers/tas_input.h"
@@ -19,6 +18,10 @@
 #include "input_common/input_mapping.h"
 #include "input_common/input_poller.h"
 #include "input_common/main.h"
+
+#ifdef HAVE_LIBUSB
+#include "input_common/drivers/gc_adapter.h"
+#endif
 #ifdef HAVE_SDL2
 #include "input_common/drivers/sdl_driver.h"
 #endif
@@ -45,7 +48,9 @@ struct InputSubsystem::Impl {
         RegisterEngine("keyboard", keyboard);
         RegisterEngine("mouse", mouse);
         RegisterEngine("touch", touch_screen);
+#ifdef HAVE_LIBUSB
         RegisterEngine("gcpad", gcadapter);
+#endif
         RegisterEngine("cemuhookudp", udp_client);
         RegisterEngine("tas", tas_input);
         RegisterEngine("camera", camera);
@@ -72,7 +77,9 @@ struct InputSubsystem::Impl {
         UnregisterEngine(keyboard);
         UnregisterEngine(mouse);
         UnregisterEngine(touch_screen);
+#ifdef HAVE_LIBUSB
         UnregisterEngine(gcadapter);
+#endif
         UnregisterEngine(udp_client);
         UnregisterEngine(tas_input);
         UnregisterEngine(camera);
@@ -95,8 +102,10 @@ struct InputSubsystem::Impl {
         devices.insert(devices.end(), keyboard_devices.begin(), keyboard_devices.end());
         auto mouse_devices = mouse->GetInputDevices();
         devices.insert(devices.end(), mouse_devices.begin(), mouse_devices.end());
+#ifdef HAVE_LIBUSB
         auto gcadapter_devices = gcadapter->GetInputDevices();
         devices.insert(devices.end(), gcadapter_devices.begin(), gcadapter_devices.end());
+#endif
         auto udp_devices = udp_client->GetInputDevices();
         devices.insert(devices.end(), udp_devices.begin(), udp_devices.end());
 #ifdef HAVE_SDL2
@@ -119,9 +128,11 @@ struct InputSubsystem::Impl {
         if (engine == mouse->GetEngineName()) {
             return mouse;
         }
+#ifdef HAVE_LIBUSB
         if (engine == gcadapter->GetEngineName()) {
             return gcadapter;
         }
+#endif
         if (engine == udp_client->GetEngineName()) {
             return udp_client;
         }
@@ -194,9 +205,11 @@ struct InputSubsystem::Impl {
         if (engine == mouse->GetEngineName()) {
             return true;
         }
+#ifdef HAVE_LIBUSB
         if (engine == gcadapter->GetEngineName()) {
             return true;
         }
+#endif
         if (engine == udp_client->GetEngineName()) {
             return true;
         }
@@ -217,7 +230,9 @@ struct InputSubsystem::Impl {
     void BeginConfiguration() {
         keyboard->BeginConfiguration();
         mouse->BeginConfiguration();
+#ifdef HAVE_LIBUSB
         gcadapter->BeginConfiguration();
+#endif
         udp_client->BeginConfiguration();
 #ifdef HAVE_SDL2
         sdl->BeginConfiguration();
@@ -227,7 +242,9 @@ struct InputSubsystem::Impl {
     void EndConfiguration() {
         keyboard->EndConfiguration();
         mouse->EndConfiguration();
+#ifdef HAVE_LIBUSB
         gcadapter->EndConfiguration();
+#endif
         udp_client->EndConfiguration();
 #ifdef HAVE_SDL2
         sdl->EndConfiguration();
@@ -248,7 +265,6 @@ struct InputSubsystem::Impl {
 
     std::shared_ptr<Keyboard> keyboard;
     std::shared_ptr<Mouse> mouse;
-    std::shared_ptr<GCAdapter> gcadapter;
     std::shared_ptr<TouchScreen> touch_screen;
     std::shared_ptr<TasInput::Tas> tas_input;
     std::shared_ptr<CemuhookUDP::UDPClient> udp_client;
@@ -256,6 +272,10 @@ struct InputSubsystem::Impl {
     std::shared_ptr<VirtualAmiibo> virtual_amiibo;
     std::shared_ptr<VirtualGamepad> virtual_gamepad;
 
+#ifdef HAVE_LIBUSB
+    std::shared_ptr<GCAdapter> gcadapter;
+#endif
+
 #ifdef HAVE_SDL2
     std::shared_ptr<SDLDriver> sdl;
 #endif

src/shader_recompiler/backend/glasm/emit_glasm_context_get_set.cpp

@@ -137,6 +137,15 @@ void EmitGetAttribute(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr, Scal
     case IR::Attribute::VertexId:
         ctx.Add("MOV.F {}.x,{}.id;", inst, ctx.attrib_name);
         break;
+    case IR::Attribute::BaseInstance:
+        ctx.Add("MOV.F {}.x,{}.baseInstance;", inst, ctx.attrib_name);
+        break;
+    case IR::Attribute::BaseVertex:
+        ctx.Add("MOV.F {}.x,{}.baseVertex;", inst, ctx.attrib_name);
+        break;
+    case IR::Attribute::DrawID:
+        ctx.Add("MOV.F {}.x,{}.draw.id;", inst, ctx.attrib_name);
+        break;
     case IR::Attribute::FrontFace:
         ctx.Add("CMP.F {}.x,{}.facing.x,0,-1;", inst, ctx.attrib_name);
         break;
@@ -156,6 +165,15 @@ void EmitGetAttributeU32(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr, S
     case IR::Attribute::VertexId:
         ctx.Add("MOV.S {}.x,{}.id;", inst, ctx.attrib_name);
         break;
+    case IR::Attribute::BaseInstance:
+        ctx.Add("MOV.S {}.x,{}.baseInstance;", inst, ctx.attrib_name);
+        break;
+    case IR::Attribute::BaseVertex:
+        ctx.Add("MOV.S {}.x,{}.baseVertex;", inst, ctx.attrib_name);
+        break;
+    case IR::Attribute::DrawID:
+        ctx.Add("MOV.S {}.x,{}.draw.id;", inst, ctx.attrib_name);
+        break;
     default:
         throw NotImplementedException("Get U32 attribute {}", attr);
     }

src/shader_recompiler/backend/glsl/emit_glsl.cpp

@@ -219,7 +219,7 @@ std::string EmitGLSL(const Profile& profile, const RuntimeInfo& runtime_info, IR
     EmitContext ctx{program, bindings, profile, runtime_info};
     Precolor(program);
     EmitCode(ctx, program);
-    const std::string version{fmt::format("#version 450{}\n", GlslVersionSpecifier(ctx))};
+    const std::string version{fmt::format("#version 460{}\n", GlslVersionSpecifier(ctx))};
     ctx.header.insert(0, version);
     if (program.shared_memory_size > 0) {
         const auto requested_size{program.shared_memory_size};

src/shader_recompiler/backend/glsl/emit_glsl_context_get_set.cpp

@@ -234,6 +234,15 @@ void EmitGetAttribute(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr,
     case IR::Attribute::FrontFace:
         ctx.AddF32("{}=itof(gl_FrontFacing?-1:0);", inst);
         break;
+    case IR::Attribute::BaseInstance:
+        ctx.AddF32("{}=itof(gl_BaseInstance);", inst);
+        break;
+    case IR::Attribute::BaseVertex:
+        ctx.AddF32("{}=itof(gl_BaseVertex);", inst);
+        break;
+    case IR::Attribute::DrawID:
+        ctx.AddF32("{}=itof(gl_DrawID);", inst);
+        break;
     default:
         throw NotImplementedException("Get attribute {}", attr);
     }
@@ -250,6 +259,15 @@ void EmitGetAttributeU32(EmitContext& ctx, IR::Inst& inst, IR::Attribute attr, s
     case IR::Attribute::VertexId:
         ctx.AddU32("{}=uint(gl_VertexID);", inst);
         break;
+    case IR::Attribute::BaseInstance:
+        ctx.AddU32("{}=uint(gl_BaseInstance);", inst);
+        break;
+    case IR::Attribute::BaseVertex:
+        ctx.AddU32("{}=uint(gl_BaseVertex);", inst);
+        break;
+    case IR::Attribute::DrawID:
+        ctx.AddU32("{}=uint(gl_DrawID);", inst);
+        break;
     default:
         throw NotImplementedException("Get U32 attribute {}", attr);
     }

src/shader_recompiler/backend/spirv/emit_spirv_context_get_set.cpp

@@ -321,8 +321,12 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
     case IR::Attribute::PositionY:
     case IR::Attribute::PositionZ:
     case IR::Attribute::PositionW:
-        return ctx.OpLoad(ctx.F32[1], AttrPointer(ctx, ctx.input_f32, vertex, ctx.input_position,
-                                                  ctx.Const(element)));
+        return ctx.OpLoad(
+            ctx.F32[1],
+            ctx.need_input_position_indirect
+                ? AttrPointer(ctx, ctx.input_f32, vertex, ctx.input_position, ctx.u32_zero_value,
+                              ctx.Const(element))
+                : AttrPointer(ctx, ctx.input_f32, vertex, ctx.input_position, ctx.Const(element)));
     case IR::Attribute::InstanceId:
         if (ctx.profile.support_vertex_instance_id) {
             return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.instance_id));
@@ -339,6 +343,12 @@ Id EmitGetAttribute(EmitContext& ctx, IR::Attribute attr, Id vertex) {
             const Id base{ctx.OpLoad(ctx.U32[1], ctx.base_vertex)};
             return ctx.OpBitcast(ctx.F32[1], ctx.OpISub(ctx.U32[1], index, base));
         }
+    case IR::Attribute::BaseInstance:
+        return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.base_instance));
+    case IR::Attribute::BaseVertex:
+        return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.base_vertex));
+    case IR::Attribute::DrawID:
+        return ctx.OpBitcast(ctx.F32[1], ctx.OpLoad(ctx.U32[1], ctx.draw_index));
     case IR::Attribute::FrontFace:
         return ctx.OpSelect(ctx.F32[1], ctx.OpLoad(ctx.U1, ctx.front_face),
                             ctx.OpBitcast(ctx.F32[1], ctx.Const(std::numeric_limits<u32>::max())),
@@ -380,6 +390,12 @@ Id EmitGetAttributeU32(EmitContext& ctx, IR::Attribute attr, Id) {
             const Id base{ctx.OpLoad(ctx.U32[1], ctx.base_vertex)};
             return ctx.OpISub(ctx.U32[1], index, base);
         }
+    case IR::Attribute::BaseInstance:
+        return ctx.OpLoad(ctx.U32[1], ctx.base_instance);
+    case IR::Attribute::BaseVertex:
+        return ctx.OpLoad(ctx.U32[1], ctx.base_vertex);
+    case IR::Attribute::DrawID:
+        return ctx.OpLoad(ctx.U32[1], ctx.draw_index);
     default:
         throw NotImplementedException("Read U32 attribute {}", attr);
     }

src/shader_recompiler/backend/spirv/emit_spirv_warp.cpp

@@ -58,11 +58,10 @@ Id SelectValue(EmitContext& ctx, Id in_range, Id value, Id src_thread_id) {
         ctx.OpGroupNonUniformShuffle(ctx.U32[1], SubgroupScope(ctx), value, src_thread_id), value);
 }
 
-Id GetUpperClamp(EmitContext& ctx, Id invocation_id, Id clamp) {
-    const Id thirty_two{ctx.Const(32u)};
-    const Id is_upper_partition{ctx.OpSGreaterThanEqual(ctx.U1, invocation_id, thirty_two)};
-    const Id upper_clamp{ctx.OpIAdd(ctx.U32[1], thirty_two, clamp)};
-    return ctx.OpSelect(ctx.U32[1], is_upper_partition, upper_clamp, clamp);
+Id AddPartitionBase(EmitContext& ctx, Id thread_id) {
+    const Id partition_idx{ctx.OpShiftRightLogical(ctx.U32[1], GetThreadId(ctx), ctx.Const(5u))};
+    const Id partition_base{ctx.OpShiftLeftLogical(ctx.U32[1], partition_idx, ctx.Const(5u))};
+    return ctx.OpIAdd(ctx.U32[1], thread_id, partition_base);
 }
 } // Anonymous namespace
 
@@ -145,64 +144,63 @@ Id EmitSubgroupGeMask(EmitContext& ctx) {
 Id EmitShuffleIndex(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                     Id segmentation_mask) {
     const Id not_seg_mask{ctx.OpNot(ctx.U32[1], segmentation_mask)};
-    const Id thread_id{GetThreadId(ctx)};
-    if (ctx.profile.warp_size_potentially_larger_than_guest) {
-        const Id thirty_two{ctx.Const(32u)};
-        const Id is_upper_partition{ctx.OpSGreaterThanEqual(ctx.U1, thread_id, thirty_two)};
-        const Id upper_index{ctx.OpIAdd(ctx.U32[1], thirty_two, index)};
-        const Id upper_clamp{ctx.OpIAdd(ctx.U32[1], thirty_two, clamp)};
-        index = ctx.OpSelect(ctx.U32[1], is_upper_partition, upper_index, index);
-        clamp = ctx.OpSelect(ctx.U32[1], is_upper_partition, upper_clamp, clamp);
-    }
+    const Id thread_id{EmitLaneId(ctx)};
     const Id min_thread_id{ComputeMinThreadId(ctx, thread_id, segmentation_mask)};
     const Id max_thread_id{ComputeMaxThreadId(ctx, min_thread_id, clamp, not_seg_mask)};
 
     const Id lhs{ctx.OpBitwiseAnd(ctx.U32[1], index, not_seg_mask)};
-    const Id src_thread_id{ctx.OpBitwiseOr(ctx.U32[1], lhs, min_thread_id)};
+    Id src_thread_id{ctx.OpBitwiseOr(ctx.U32[1], lhs, min_thread_id)};
     const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
 
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        src_thread_id = AddPartitionBase(ctx, src_thread_id);
+    }
+
     SetInBoundsFlag(inst, in_range);
     return SelectValue(ctx, in_range, value, src_thread_id);
 }
 
 Id EmitShuffleUp(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                  Id segmentation_mask) {
-    const Id thread_id{GetThreadId(ctx)};
-    if (ctx.profile.warp_size_potentially_larger_than_guest) {
-        clamp = GetUpperClamp(ctx, thread_id, clamp);
-    }
+    const Id thread_id{EmitLaneId(ctx)};
     const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
-    const Id src_thread_id{ctx.OpISub(ctx.U32[1], thread_id, index)};
+    Id src_thread_id{ctx.OpISub(ctx.U32[1], thread_id, index)};
     const Id in_range{ctx.OpSGreaterThanEqual(ctx.U1, src_thread_id, max_thread_id)};
 
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        src_thread_id = AddPartitionBase(ctx, src_thread_id);
+    }
+
     SetInBoundsFlag(inst, in_range);
     return SelectValue(ctx, in_range, value, src_thread_id);
 }
 
 Id EmitShuffleDown(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                    Id segmentation_mask) {
-    const Id thread_id{GetThreadId(ctx)};
-    if (ctx.profile.warp_size_potentially_larger_than_guest) {
-        clamp = GetUpperClamp(ctx, thread_id, clamp);
-    }
+    const Id thread_id{EmitLaneId(ctx)};
     const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
-    const Id src_thread_id{ctx.OpIAdd(ctx.U32[1], thread_id, index)};
+    Id src_thread_id{ctx.OpIAdd(ctx.U32[1], thread_id, index)};
     const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
 
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        src_thread_id = AddPartitionBase(ctx, src_thread_id);
+    }
+
     SetInBoundsFlag(inst, in_range);
     return SelectValue(ctx, in_range, value, src_thread_id);
 }
 
 Id EmitShuffleButterfly(EmitContext& ctx, IR::Inst* inst, Id value, Id index, Id clamp,
                         Id segmentation_mask) {
-    const Id thread_id{GetThreadId(ctx)};
-    if (ctx.profile.warp_size_potentially_larger_than_guest) {
-        clamp = GetUpperClamp(ctx, thread_id, clamp);
-    }
+    const Id thread_id{EmitLaneId(ctx)};
     const Id max_thread_id{GetMaxThreadId(ctx, thread_id, clamp, segmentation_mask)};
-    const Id src_thread_id{ctx.OpBitwiseXor(ctx.U32[1], thread_id, index)};
+    Id src_thread_id{ctx.OpBitwiseXor(ctx.U32[1], thread_id, index)};
     const Id in_range{ctx.OpSLessThanEqual(ctx.U1, src_thread_id, max_thread_id)};
 
+    if (ctx.profile.warp_size_potentially_larger_than_guest) {
+        src_thread_id = AddPartitionBase(ctx, src_thread_id);
+    }
+
     SetInBoundsFlag(inst, in_range);
     return SelectValue(ctx, in_range, value, src_thread_id);
 }

src/shader_recompiler/backend/spirv/spirv_emit_context.cpp

@@ -544,7 +544,7 @@ void EmitContext::DefineCommonTypes(const Info& info) {
         U16 = Name(TypeInt(16, false), "u16");
         S16 = Name(TypeInt(16, true), "s16");
     }
-    if (info.uses_int64) {
+    if (info.uses_int64 && profile.support_int64) {
         AddCapability(spv::Capability::Int64);
         U64 = Name(TypeInt(64, false), "u64");
     }
@@ -721,9 +721,21 @@ void EmitContext::DefineAttributeMemAccess(const Info& info) {
         size_t label_index{0};
         if (info.loads.AnyComponent(IR::Attribute::PositionX)) {
             AddLabel(labels[label_index]);
-            const Id pointer{is_array
-                                 ? OpAccessChain(input_f32, input_position, vertex, masked_index)
-                                 : OpAccessChain(input_f32, input_position, masked_index)};
+            const Id pointer{[&]() {
+                if (need_input_position_indirect) {
+                    if (is_array)
+                        return OpAccessChain(input_f32, input_position, vertex, u32_zero_value,
+                                             masked_index);
+                    else
+                        return OpAccessChain(input_f32, input_position, u32_zero_value,
+                                             masked_index);
+                } else {
+                    if (is_array)
+                        return OpAccessChain(input_f32, input_position, vertex, masked_index);
+                    else
+                        return OpAccessChain(input_f32, input_position, masked_index);
+                }
+            }()};
             const Id result{OpLoad(F32[1], pointer)};
             OpReturnValue(result);
             ++label_index;
@@ -1367,30 +1379,56 @@ void EmitContext::DefineInputs(const IR::Program& program) {
         Decorate(layer, spv::Decoration::Flat);
     }
     if (loads.AnyComponent(IR::Attribute::PositionX)) {
-        const bool is_fragment{stage != Stage::Fragment};
-        const spv::BuiltIn built_in{is_fragment ? spv::BuiltIn::Position : spv::BuiltIn::FragCoord};
-        input_position = DefineInput(*this, F32[4], true, built_in);
-        if (profile.support_geometry_shader_passthrough) {
-            if (info.passthrough.AnyComponent(IR::Attribute::PositionX)) {
-                Decorate(input_position, spv::Decoration::PassthroughNV);
+        const bool is_fragment{stage == Stage::Fragment};
+        if (!is_fragment && profile.has_broken_spirv_position_input) {
+            need_input_position_indirect = true;
+
+            const Id input_position_struct = TypeStruct(F32[4]);
+            input_position = DefineInput(*this, input_position_struct, true);
+
+            MemberDecorate(input_position_struct, 0, spv::Decoration::BuiltIn,
+                           static_cast<unsigned>(spv::BuiltIn::Position));
+            Decorate(input_position_struct, spv::Decoration::Block);
+        } else {
+            const spv::BuiltIn built_in{is_fragment ? spv::BuiltIn::FragCoord
+                                                    : spv::BuiltIn::Position};
+            input_position = DefineInput(*this, F32[4], true, built_in);
+
+            if (profile.support_geometry_shader_passthrough) {
+                if (info.passthrough.AnyComponent(IR::Attribute::PositionX)) {
+                    Decorate(input_position, spv::Decoration::PassthroughNV);
+                }
             }
         }
     }
     if (loads[IR::Attribute::InstanceId]) {
         if (profile.support_vertex_instance_id) {
             instance_id = DefineInput(*this, U32[1], true, spv::BuiltIn::InstanceId);
+            if (loads[IR::Attribute::BaseInstance]) {
+                base_instance = DefineInput(*this, U32[1], true, spv::BuiltIn::BaseVertex);
+            }
         } else {
             instance_index = DefineInput(*this, U32[1], true, spv::BuiltIn::InstanceIndex);
             base_instance = DefineInput(*this, U32[1], true, spv::BuiltIn::BaseInstance);
         }
+    } else if (loads[IR::Attribute::BaseInstance]) {
+        base_instance = DefineInput(*this, U32[1], true, spv::BuiltIn::BaseInstance);
     }
     if (loads[IR::Attribute::VertexId]) {
         if (profile.support_vertex_instance_id) {
             vertex_id = DefineInput(*this, U32[1], true, spv::BuiltIn::VertexId);
+            if (loads[IR::Attribute::BaseVertex]) {
+                base_vertex = DefineInput(*this, U32[1], true, spv::BuiltIn::BaseVertex);
+            }
         } else {
             vertex_index = DefineInput(*this, U32[1], true, spv::BuiltIn::VertexIndex);
             base_vertex = DefineInput(*this, U32[1], true, spv::BuiltIn::BaseVertex);
         }
+    } else if (loads[IR::Attribute::BaseVertex]) {
+        base_vertex = DefineInput(*this, U32[1], true, spv::BuiltIn::BaseVertex);
+    }
+    if (loads[IR::Attribute::DrawID]) {
+        draw_index = DefineInput(*this, U32[1], true, spv::BuiltIn::DrawIndex);
     }
     if (loads[IR::Attribute::FrontFace]) {
         front_face = DefineInput(*this, U1, true, spv::BuiltIn::FrontFacing);

src/shader_recompiler/backend/spirv/spirv_emit_context.h

@@ -218,6 +218,7 @@ public:
     Id base_instance{};
     Id vertex_id{};
     Id vertex_index{};
+    Id draw_index{};
     Id base_vertex{};
     Id front_face{};
     Id point_coord{};
@@ -279,6 +280,7 @@ public:
     Id write_global_func_u32x2{};
     Id write_global_func_u32x4{};
 
+    bool need_input_position_indirect{};
     Id input_position{};
     std::array<Id, 32> input_generics{};
 

src/shader_recompiler/environment.h

@@ -34,6 +34,11 @@ public:
 
     [[nodiscard]] virtual std::array<u32, 3> WorkgroupSize() const = 0;
 
+    [[nodiscard]] virtual bool HasHLEMacroState() const = 0;
+
+    [[nodiscard]] virtual std::optional<ReplaceConstant> GetReplaceConstBuffer(u32 bank,
+                                                                               u32 offset) = 0;
+
     virtual void Dump(u64 hash) = 0;
 
     [[nodiscard]] const ProgramHeader& SPH() const noexcept {
@@ -52,11 +57,16 @@ public:
         return start_address;
     }
 
+    [[nodiscard]] bool IsPropietaryDriver() const noexcept {
+        return is_propietary_driver;
+    }
+
 protected:
     ProgramHeader sph{};
     std::array<u32, 8> gp_passthrough_mask{};
     Stage stage{};
     u32 start_address{};
+    bool is_propietary_driver{};
 };
 
 } // namespace Shader

src/shader_recompiler/frontend/ir/attribute.cpp

@@ -446,6 +446,12 @@ std::string NameOf(Attribute attribute) {
         return "ViewportMask";
     case Attribute::FrontFace:
         return "FrontFace";
+    case Attribute::BaseInstance:
+        return "BaseInstance";
+    case Attribute::BaseVertex:
+        return "BaseVertex";
+    case Attribute::DrawID:
+        return "DrawID";
     }
     return fmt::format("<reserved attribute {}>", static_cast<int>(attribute));
 }

src/shader_recompiler/frontend/ir/attribute.h

@@ -219,6 +219,11 @@ enum class Attribute : u64 {
     FixedFncTexture9Q = 231,
     ViewportMask = 232,
     FrontFace = 255,
+
+    // Implementation attributes
+    BaseInstance = 256,
+    BaseVertex = 257,
+    DrawID = 258,
 };
 
 constexpr size_t NUM_GENERICS = 32;

src/shader_recompiler/frontend/ir/ir_emitter.cpp

@@ -294,6 +294,14 @@ F32 IREmitter::GetAttribute(IR::Attribute attribute, const U32& vertex) {
     return Inst<F32>(Opcode::GetAttribute, attribute, vertex);
 }
 
+U32 IREmitter::GetAttributeU32(IR::Attribute attribute) {
+    return GetAttributeU32(attribute, Imm32(0));
+}
+
+U32 IREmitter::GetAttributeU32(IR::Attribute attribute, const U32& vertex) {
+    return Inst<U32>(Opcode::GetAttributeU32, attribute, vertex);
+}
+
 void IREmitter::SetAttribute(IR::Attribute attribute, const F32& value, const U32& vertex) {
     Inst(Opcode::SetAttribute, attribute, value, vertex);
 }

src/shader_recompiler/frontend/ir/ir_emitter.h

@@ -74,6 +74,8 @@ public:
 
     [[nodiscard]] F32 GetAttribute(IR::Attribute attribute);
     [[nodiscard]] F32 GetAttribute(IR::Attribute attribute, const U32& vertex);
+    [[nodiscard]] U32 GetAttributeU32(IR::Attribute attribute);
+    [[nodiscard]] U32 GetAttributeU32(IR::Attribute attribute, const U32& vertex);
     void SetAttribute(IR::Attribute attribute, const F32& value, const U32& vertex);
 
     [[nodiscard]] F32 GetAttributeIndexed(const U32& phys_address);

src/shader_recompiler/frontend/maxwell/translate_program.cpp

@@ -171,6 +171,70 @@ std::map<IR::Attribute, IR::Attribute> GenerateLegacyToGenericMappings(
     }
     return mapping;
 }
+
+void EmitGeometryPassthrough(IR::IREmitter& ir, const IR::Program& program,
+                             const Shader::VaryingState& passthrough_mask,
+                             bool passthrough_position,
+                             std::optional<IR::Attribute> passthrough_layer_attr) {
+    for (u32 i = 0; i < program.output_vertices; i++) {
+        // Assign generics from input
+        for (u32 j = 0; j < 32; j++) {
+            if (!passthrough_mask.Generic(j)) {
+                continue;
+            }
+
+            const IR::Attribute attr = IR::Attribute::Generic0X + (j * 4);
+            ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
+        }
+
+        if (passthrough_position) {
+            // Assign position from input
+            const IR::Attribute attr = IR::Attribute::PositionX;
+            ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
+            ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
+        }
+
+        if (passthrough_layer_attr) {
+            // Assign layer
+            ir.SetAttribute(IR::Attribute::Layer, ir.GetAttribute(*passthrough_layer_attr),
+                            ir.Imm32(0));
+        }
+
+        // Emit vertex
+        ir.EmitVertex(ir.Imm32(0));
+    }
+    ir.EndPrimitive(ir.Imm32(0));
+}
+
+u32 GetOutputTopologyVertices(OutputTopology output_topology) {
+    switch (output_topology) {
+    case OutputTopology::PointList:
+        return 1;
+    case OutputTopology::LineStrip:
+        return 2;
+    default:
+        return 3;
+    }
+}
+
+void LowerGeometryPassthrough(const IR::Program& program, const HostTranslateInfo& host_info) {
+    for (IR::Block* const block : program.blocks) {
+        for (IR::Inst& inst : block->Instructions()) {
+            if (inst.GetOpcode() == IR::Opcode::Epilogue) {
+                IR::IREmitter ir{*block, IR::Block::InstructionList::s_iterator_to(inst)};
+                EmitGeometryPassthrough(
+                    ir, program, program.info.passthrough,
+                    program.info.passthrough.AnyComponent(IR::Attribute::PositionX), {});
+            }
+        }
+    }
+}
+
 } // Anonymous namespace
 
 IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Block>& block_pool,
@@ -198,6 +262,11 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
             for (size_t i = 0; i < program.info.passthrough.mask.size(); ++i) {
                 program.info.passthrough.mask[i] = ((mask[i / 32] >> (i % 32)) & 1) == 0;
             }
+
+            if (!host_info.support_geometry_shader_passthrough) {
+                program.output_vertices = GetOutputTopologyVertices(program.output_topology);
+                LowerGeometryPassthrough(program, host_info);
+            }
         }
         break;
     }
@@ -219,11 +288,11 @@ IR::Program TranslateProgram(ObjectPool<IR::Inst>& inst_pool, ObjectPool<IR::Blo
     }
     Optimization::SsaRewritePass(program);
 
-    Optimization::ConstantPropagationPass(program);
+    Optimization::ConstantPropagationPass(env, program);
 
     Optimization::PositionPass(env, program);
 
-    Optimization::GlobalMemoryToStorageBufferPass(program);
+    Optimization::GlobalMemoryToStorageBufferPass(program, host_info);
     Optimization::TexturePass(env, program, host_info);
 
     if (Settings::values.resolution_info.active) {
@@ -342,17 +411,7 @@ IR::Program GenerateGeometryPassthrough(ObjectPool<IR::Inst>& inst_pool,
     IR::Program program;
     program.stage = Stage::Geometry;
     program.output_topology = output_topology;
-    switch (output_topology) {
-    case OutputTopology::PointList:
-        program.output_vertices = 1;
-        break;
-    case OutputTopology::LineStrip:
-        program.output_vertices = 2;
-        break;
-    default:
-        program.output_vertices = 3;
-        break;
-    }
+    program.output_vertices = GetOutputTopologyVertices(output_topology);
 
     program.is_geometry_passthrough = false;
     program.info.loads.mask = source_program.info.stores.mask;
@@ -366,35 +425,8 @@ IR::Program GenerateGeometryPassthrough(ObjectPool<IR::Inst>& inst_pool,
     node.data.block = current_block;
 
     IR::IREmitter ir{*current_block};
-    for (u32 i = 0; i < program.output_vertices; i++) {
-        // Assign generics from input
-        for (u32 j = 0; j < 32; j++) {
-            if (!program.info.stores.Generic(j)) {
-                continue;
-            }
-
-            const IR::Attribute attr = IR::Attribute::Generic0X + (j * 4);
-            ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
-            ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
-            ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
-            ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
-        }
-
-        // Assign position from input
-        const IR::Attribute attr = IR::Attribute::PositionX;
-        ir.SetAttribute(attr + 0, ir.GetAttribute(attr + 0, ir.Imm32(i)), ir.Imm32(0));
-        ir.SetAttribute(attr + 1, ir.GetAttribute(attr + 1, ir.Imm32(i)), ir.Imm32(0));
-        ir.SetAttribute(attr + 2, ir.GetAttribute(attr + 2, ir.Imm32(i)), ir.Imm32(0));
-        ir.SetAttribute(attr + 3, ir.GetAttribute(attr + 3, ir.Imm32(i)), ir.Imm32(0));
-
-        // Assign layer
-        ir.SetAttribute(IR::Attribute::Layer, ir.GetAttribute(source_program.info.emulated_layer),
-                        ir.Imm32(0));
-
-        // Emit vertex
-        ir.EmitVertex(ir.Imm32(0));
-    }
-    ir.EndPrimitive(ir.Imm32(0));
+    EmitGeometryPassthrough(ir, program, program.info.stores, true,
+                            source_program.info.emulated_layer);
 
     IR::Block* return_block{block_pool.Create(inst_pool)};
     IR::IREmitter{*return_block}.Epilogue();

src/shader_recompiler/host_translate_info.h

@@ -15,6 +15,9 @@ struct HostTranslateInfo {
     bool needs_demote_reorder{}; ///< True when the device needs DemoteToHelperInvocation reordered
     bool support_snorm_render_buffer{};  ///< True when the device supports SNORM render buffers
     bool support_viewport_index_layer{}; ///< True when the device supports gl_Layer in VS
+    u32 min_ssbo_alignment{};            ///< Minimum alignment supported by the device for SSBOs
+    bool support_geometry_shader_passthrough{}; ///< True when the device supports geometry
+                                                ///< passthrough shaders
 };
 
 } // namespace Shader

src/shader_recompiler/ir_opt/constant_propagation_pass.cpp

@@ -7,6 +7,7 @@
 #include <type_traits>
 
 #include "common/bit_cast.h"
+#include "shader_recompiler/environment.h"
 #include "shader_recompiler/exception.h"
 #include "shader_recompiler/frontend/ir/ir_emitter.h"
 #include "shader_recompiler/frontend/ir/value.h"
@@ -515,6 +516,9 @@ void FoldBitCast(IR::Inst& inst, IR::Opcode reverse) {
             case IR::Attribute::PrimitiveId:
             case IR::Attribute::InstanceId:
             case IR::Attribute::VertexId:
+            case IR::Attribute::BaseVertex:
+            case IR::Attribute::BaseInstance:
+            case IR::Attribute::DrawID:
                 break;
             default:
                 return;
@@ -644,7 +648,63 @@ void FoldFSwizzleAdd(IR::Block& block, IR::Inst& inst) {
     }
 }
 
-void ConstantPropagation(IR::Block& block, IR::Inst& inst) {
+void FoldConstBuffer(Environment& env, IR::Block& block, IR::Inst& inst) {
+    const IR::Value bank{inst.Arg(0)};
+    const IR::Value offset{inst.Arg(1)};
+    if (!bank.IsImmediate() || !offset.IsImmediate()) {
+        return;
+    }
+    const auto bank_value = bank.U32();
+    const auto offset_value = offset.U32();
+    auto replacement = env.GetReplaceConstBuffer(bank_value, offset_value);
+    if (!replacement) {
+        return;
+    }
+    const auto new_attribute = [replacement]() {
+        switch (*replacement) {
+        case ReplaceConstant::BaseInstance:
+            return IR::Attribute::BaseInstance;
+        case ReplaceConstant::BaseVertex:
+            return IR::Attribute::BaseVertex;
+        case ReplaceConstant::DrawID:
+            return IR::Attribute::DrawID;
+        default:
+            throw NotImplementedException("Not implemented replacement variable {}", *replacement);
+        }
+    }();
+    IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
+    if (inst.GetOpcode() == IR::Opcode::GetCbufU32) {
+        inst.ReplaceUsesWith(ir.GetAttributeU32(new_attribute));
+    } else {
+        inst.ReplaceUsesWith(ir.GetAttribute(new_attribute));
+    }
+}
+
+void FoldDriverConstBuffer(Environment& env, IR::Block& block, IR::Inst& inst, u32 which_bank,
+                           u32 offset_start = 0, u32 offset_end = std::numeric_limits<u16>::max()) {
+    const IR::Value bank{inst.Arg(0)};
+    const IR::Value offset{inst.Arg(1)};
+    if (!bank.IsImmediate() || !offset.IsImmediate()) {
+        return;
+    }
+    const auto bank_value = bank.U32();
+    if (bank_value != which_bank) {
+        return;
+    }
+    const auto offset_value = offset.U32();
+    if (offset_value < offset_start || offset_value >= offset_end) {
+        return;
+    }
+    IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
+    if (inst.GetOpcode() == IR::Opcode::GetCbufU32) {
+        inst.ReplaceUsesWith(IR::Value{env.ReadCbufValue(bank_value, offset_value)});
+    } else {
+        inst.ReplaceUsesWith(
+            IR::Value{Common::BitCast<f32>(env.ReadCbufValue(bank_value, offset_value))});
+    }
+}
+
+void ConstantPropagation(Environment& env, IR::Block& block, IR::Inst& inst) {
     switch (inst.GetOpcode()) {
     case IR::Opcode::GetRegister:
         return FoldGetRegister(inst);
@@ -789,18 +849,28 @@ void ConstantPropagation(IR::Block& block, IR::Inst& inst) {
                                     IR::Opcode::CompositeInsertF16x4);
     case IR::Opcode::FSwizzleAdd:
         return FoldFSwizzleAdd(block, inst);
+    case IR::Opcode::GetCbufF32:
+    case IR::Opcode::GetCbufU32:
+        if (env.HasHLEMacroState()) {
+            FoldConstBuffer(env, block, inst);
+        }
+        if (env.IsPropietaryDriver()) {
+            FoldDriverConstBuffer(env, block, inst, 1);
+        }
+        break;
     default:
         break;
     }
 }
+
 } // Anonymous namespace
 
-void ConstantPropagationPass(IR::Program& program) {
+void ConstantPropagationPass(Environment& env, IR::Program& program) {
     const auto end{program.post_order_blocks.rend()};
     for (auto it = program.post_order_blocks.rbegin(); it != end; ++it) {
         IR::Block* const block{*it};
         for (IR::Inst& inst : block->Instructions()) {
-            ConstantPropagation(*block, inst);
+            ConstantPropagation(env, *block, inst);
         }
     }
 }

src/shader_recompiler/ir_opt/global_memory_to_storage_buffer_pass.cpp

@@ -11,6 +11,7 @@
 #include "shader_recompiler/frontend/ir/breadth_first_search.h"
 #include "shader_recompiler/frontend/ir/ir_emitter.h"
 #include "shader_recompiler/frontend/ir/value.h"
+#include "shader_recompiler/host_translate_info.h"
 #include "shader_recompiler/ir_opt/passes.h"
 
 namespace Shader::Optimization {
@@ -402,7 +403,7 @@ void CollectStorageBuffers(IR::Block& block, IR::Inst& inst, StorageInfo& info)
 }
 
 /// Returns the offset in indices (not bytes) for an equivalent storage instruction
-IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer) {
+IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer, u32 alignment) {
     IR::IREmitter ir{block, IR::Block::InstructionList::s_iterator_to(inst)};
     IR::U32 offset;
     if (const std::optional<LowAddrInfo> low_addr{TrackLowAddress(&inst)}) {
@@ -415,7 +416,10 @@ IR::U32 StorageOffset(IR::Block& block, IR::Inst& inst, StorageBufferAddr buffer
     }
     // Subtract the least significant 32 bits from the guest offset. The result is the storage
     // buffer offset in bytes.
-    const IR::U32 low_cbuf{ir.GetCbuf(ir.Imm32(buffer.index), ir.Imm32(buffer.offset))};
+    IR::U32 low_cbuf{ir.GetCbuf(ir.Imm32(buffer.index), ir.Imm32(buffer.offset))};
+
+    // Align the offset base to match the host alignment requirements
+    low_cbuf = ir.BitwiseAnd(low_cbuf, ir.Imm32(~(alignment - 1U)));
     return ir.ISub(offset, low_cbuf);
 }
 
@@ -510,7 +514,7 @@ void Replace(IR::Block& block, IR::Inst& inst, const IR::U32& storage_index,
 }
 } // Anonymous namespace
 
-void GlobalMemoryToStorageBufferPass(IR::Program& program) {
+void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateInfo& host_info) {
     StorageInfo info;
     for (IR::Block* const block : program.post_order_blocks) {
         for (IR::Inst& inst : block->Instructions()) {
@@ -534,7 +538,8 @@ void GlobalMemoryToStorageBufferPass(IR::Program& program) {
         const IR::U32 index{IR::Value{static_cast<u32>(info.set.index_of(it))}};
         IR::Block* const block{storage_inst.block};
         IR::Inst* const inst{storage_inst.inst};
-        const IR::U32 offset{StorageOffset(*block, *inst, storage_buffer)};
+        const IR::U32 offset{
+            StorageOffset(*block, *inst, storage_buffer, host_info.min_ssbo_alignment)};
         Replace(*block, *inst, index, offset);
     }
 }

src/shader_recompiler/ir_opt/passes.h

@@ -13,9 +13,9 @@ struct HostTranslateInfo;
 namespace Shader::Optimization {
 
 void CollectShaderInfoPass(Environment& env, IR::Program& program);
-void ConstantPropagationPass(IR::Program& program);
+void ConstantPropagationPass(Environment& env, IR::Program& program);
 void DeadCodeEliminationPass(IR::Program& program);
-void GlobalMemoryToStorageBufferPass(IR::Program& program);
+void GlobalMemoryToStorageBufferPass(IR::Program& program, const HostTranslateInfo& host_info);
 void IdentityRemovalPass(IR::Program& program);
 void LowerFp16ToFp32(IR::Program& program);
 void LowerInt64ToInt32(IR::Program& program);

src/shader_recompiler/profile.h

@@ -55,6 +55,8 @@ struct Profile {
 
     /// OpFClamp is broken and OpFMax + OpFMin should be used instead
     bool has_broken_spirv_clamp{};
+    /// The Position builtin needs to be wrapped in a struct when used as an input
+    bool has_broken_spirv_position_input{};
     /// Offset image operands with an unsigned type do not work
     bool has_broken_unsigned_image_offsets{};
     /// Signed instructions with unsigned data types are misinterpreted

src/shader_recompiler/shader_info.h

@@ -16,6 +16,12 @@
 
 namespace Shader {
 
+enum class ReplaceConstant : u32 {
+    BaseInstance,
+    BaseVertex,
+    DrawID,
+};
+
 enum class TextureType : u32 {
     Color1D,
     ColorArray1D,
@@ -59,6 +65,8 @@ enum class Interpolation {
 struct ConstantBufferDescriptor {
     u32 index;
     u32 count;
+
+    auto operator<=>(const ConstantBufferDescriptor&) const = default;
 };
 
 struct StorageBufferDescriptor {
@@ -66,6 +74,8 @@ struct StorageBufferDescriptor {
     u32 cbuf_offset;
     u32 count;
     bool is_written;
+
+    auto operator<=>(const StorageBufferDescriptor&) const = default;
 };
 
 struct TextureBufferDescriptor {
@@ -78,6 +88,8 @@ struct TextureBufferDescriptor {
     u32 secondary_shift_left;
     u32 count;
     u32 size_shift;
+
+    auto operator<=>(const TextureBufferDescriptor&) const = default;
 };
 using TextureBufferDescriptors = boost::container::small_vector<TextureBufferDescriptor, 6>;
 
@@ -89,6 +101,8 @@ struct ImageBufferDescriptor {
     u32 cbuf_offset;
     u32 count;
     u32 size_shift;
+
+    auto operator<=>(const ImageBufferDescriptor&) const = default;
 };
 using ImageBufferDescriptors = boost::container::small_vector<ImageBufferDescriptor, 2>;
 
@@ -104,6 +118,8 @@ struct TextureDescriptor {
     u32 secondary_shift_left;
     u32 count;
     u32 size_shift;
+
+    auto operator<=>(const TextureDescriptor&) const = default;
 };
 using TextureDescriptors = boost::container::small_vector<TextureDescriptor, 12>;
 
@@ -116,6 +132,8 @@ struct ImageDescriptor {
     u32 cbuf_offset;
     u32 count;
     u32 size_shift;
+
+    auto operator<=>(const ImageDescriptor&) const = default;
 };
 using ImageDescriptors = boost::container::small_vector<ImageDescriptor, 4>;
 

src/shader_recompiler/varying_state.h

@@ -11,7 +11,7 @@
 namespace Shader {
 
 struct VaryingState {
-    std::bitset<256> mask{};
+    std::bitset<512> mask{};
 
     void Set(IR::Attribute attribute, bool state = true) {
         mask[static_cast<size_t>(attribute)] = state;

src/tests/CMakeLists.txt

@@ -7,6 +7,7 @@ add_executable(tests
     common/fibers.cpp
     common/host_memory.cpp
     common/param_package.cpp
+    common/range_map.cpp
     common/ring_buffer.cpp
     common/scratch_buffer.cpp
     common/unique_function.cpp

src/tests/common/range_map.cpp

@@ -0,0 +1,70 @@
+// SPDX-FileCopyrightText: Copyright 2022 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-3.0-or-later
+
+#include <stdexcept>
+
+#include <catch2/catch.hpp>
+
+#include "common/range_map.h"
+
+enum class MappedEnum : u32 {
+    Invalid = 0,
+    Valid_1 = 1,
+    Valid_2 = 2,
+    Valid_3 = 3,
+};
+
+TEST_CASE("Range Map: Setup", "[video_core]") {
+    Common::RangeMap<u64, MappedEnum> my_map(MappedEnum::Invalid);
+    my_map.Map(3000, 3500, MappedEnum::Valid_1);
+    my_map.Unmap(3200, 3600);
+    my_map.Map(4000, 4500, MappedEnum::Valid_2);
+    my_map.Map(4200, 4400, MappedEnum::Valid_2);
+    my_map.Map(4200, 4400, MappedEnum::Valid_1);
+    REQUIRE(my_map.GetContinousSizeFrom(4200) == 200);
+    REQUIRE(my_map.GetContinousSizeFrom(3000) == 200);
+    REQUIRE(my_map.GetContinousSizeFrom(2900) == 0);
+
+    REQUIRE(my_map.GetValueAt(2900) == MappedEnum::Invalid);
+    REQUIRE(my_map.GetValueAt(3100) == MappedEnum::Valid_1);
+    REQUIRE(my_map.GetValueAt(3000) == MappedEnum::Valid_1);
+    REQUIRE(my_map.GetValueAt(3200) == MappedEnum::Invalid);
+
+    REQUIRE(my_map.GetValueAt(4199) == MappedEnum::Valid_2);
+    REQUIRE(my_map.GetValueAt(4200) == MappedEnum::Valid_1);
+    REQUIRE(my_map.GetValueAt(4400) == MappedEnum::Valid_2);
+    REQUIRE(my_map.GetValueAt(4500) == MappedEnum::Invalid);
+    REQUIRE(my_map.GetValueAt(4600) == MappedEnum::Invalid);
+
+    my_map.Unmap(0, 6000);
+    for (u64 address = 0; address < 10000; address += 1000) {
+        REQUIRE(my_map.GetContinousSizeFrom(address) == 0);
+    }
+
+    my_map.Map(1000, 3000, MappedEnum::Valid_1);
+    my_map.Map(4000, 5000, MappedEnum::Valid_1);
+    my_map.Map(2500, 4100, MappedEnum::Valid_1);
+    REQUIRE(my_map.GetContinousSizeFrom(1000) == 4000);
+
+    my_map.Map(1000, 3000, MappedEnum::Valid_1);
+    my_map.Map(4000, 5000, MappedEnum::Valid_2);
+    my_map.Map(2500, 4100, MappedEnum::Valid_3);
+    REQUIRE(my_map.GetContinousSizeFrom(1000) == 1500);
+    REQUIRE(my_map.GetContinousSizeFrom(2500) == 1600);
+    REQUIRE(my_map.GetContinousSizeFrom(4100) == 900);
+    REQUIRE(my_map.GetValueAt(900) == MappedEnum::Invalid);
+    REQUIRE(my_map.GetValueAt(1000) == MappedEnum::Valid_1);
+    REQUIRE(my_map.GetValueAt(2500) == MappedEnum::Valid_3);
+    REQUIRE(my_map.GetValueAt(4100) == MappedEnum::Valid_2);
+    REQUIRE(my_map.GetValueAt(5000) == MappedEnum::Invalid);
+
+    my_map.Map(2000, 6000, MappedEnum::Valid_3);
+    REQUIRE(my_map.GetContinousSizeFrom(1000) == 1000);
+    REQUIRE(my_map.GetContinousSizeFrom(3000) == 3000);
+    REQUIRE(my_map.GetValueAt(1000) == MappedEnum::Valid_1);
+    REQUIRE(my_map.GetValueAt(1999) == MappedEnum::Valid_1);
+    REQUIRE(my_map.GetValueAt(1500) == MappedEnum::Valid_1);
+    REQUIRE(my_map.GetValueAt(2001) == MappedEnum::Valid_3);
+    REQUIRE(my_map.GetValueAt(5999) == MappedEnum::Valid_3);
+    REQUIRE(my_map.GetValueAt(6000) == MappedEnum::Invalid);
+}

src/tests/video_core/buffer_base.cpp

@@ -538,7 +538,7 @@ TEST_CASE("BufferBase: Cached write downloads") {
     int num = 0;
     buffer.ForEachDownloadRangeAndClear(c, WORD, [&](u64 offset, u64 size) { ++num; });
     buffer.ForEachUploadRange(c, WORD, [&](u64 offset, u64 size) { ++num; });
-    REQUIRE(num == 0);
+    REQUIRE(num == 1);
     REQUIRE(!buffer.IsRegionCpuModified(c + PAGE, PAGE));
     REQUIRE(!buffer.IsRegionGpuModified(c + PAGE, PAGE));
     buffer.FlushCachedWrites();

src/video_core/CMakeLists.txt

@@ -13,6 +13,7 @@ add_library(video_core STATIC
     buffer_cache/buffer_base.h
     buffer_cache/buffer_cache.cpp
     buffer_cache/buffer_cache.h
+    cache_types.h
     cdma_pusher.cpp
     cdma_pusher.h
     compatible_formats.cpp
@@ -84,6 +85,7 @@ add_library(video_core STATIC
     gpu.h
     gpu_thread.cpp
     gpu_thread.h
+    invalidation_accumulator.h
     memory_manager.cpp
     memory_manager.h
     precompiled_headers.h
@@ -189,6 +191,8 @@ add_library(video_core STATIC
     renderer_vulkan/vk_texture_cache.cpp
     renderer_vulkan/vk_texture_cache.h
     renderer_vulkan/vk_texture_cache_base.cpp
+    renderer_vulkan/vk_turbo_mode.cpp
+    renderer_vulkan/vk_turbo_mode.h
     renderer_vulkan/vk_update_descriptor.cpp
     renderer_vulkan/vk_update_descriptor.h
     shader_cache.cpp

src/video_core/buffer_cache/buffer_base.h

@@ -430,7 +430,7 @@ private:
         if (query_begin >= SizeBytes() || size < 0) {
             return;
         }
-        u64* const untracked_words = Array<Type::Untracked>();
+        [[maybe_unused]] u64* const untracked_words = Array<Type::Untracked>();
         u64* const state_words = Array<type>();
         const u64 query_end = query_begin + std::min(static_cast<u64>(size), SizeBytes());
         u64* const words_begin = state_words + query_begin / BYTES_PER_WORD;
@@ -483,7 +483,7 @@ private:
                 NotifyRasterizer<true>(word_index, current_bits, ~u64{0});
             }
             // Exclude CPU modified pages when visiting GPU pages
-            const u64 word = current_word & ~(type == Type::GPU ? untracked_words[word_index] : 0);
+            const u64 word = current_word;
             u64 page = page_begin;
             page_begin = 0;
 
@@ -531,7 +531,7 @@ private:
     [[nodiscard]] bool IsRegionModified(u64 offset, u64 size) const noexcept {
         static_assert(type != Type::Untracked);
 
-        const u64* const untracked_words = Array<Type::Untracked>();
+        [[maybe_unused]] const u64* const untracked_words = Array<Type::Untracked>();
         const u64* const state_words = Array<type>();
         const u64 num_query_words = size / BYTES_PER_WORD + 1;
         const u64 word_begin = offset / BYTES_PER_WORD;
@@ -539,8 +539,7 @@ private:
         const u64 page_limit = Common::DivCeil(offset + size, BYTES_PER_PAGE);
         u64 page_index = (offset / BYTES_PER_PAGE) % PAGES_PER_WORD;
         for (u64 word_index = word_begin; word_index < word_end; ++word_index, page_index = 0) {
-            const u64 off_word = type == Type::GPU ? untracked_words[word_index] : 0;
-            const u64 word = state_words[word_index] & ~off_word;
+            const u64 word = state_words[word_index];
             if (word == 0) {
                 continue;
             }
@@ -564,7 +563,7 @@ private:
     [[nodiscard]] std::pair<u64, u64> ModifiedRegion(u64 offset, u64 size) const noexcept {
         static_assert(type != Type::Untracked);
 
-        const u64* const untracked_words = Array<Type::Untracked>();
+        [[maybe_unused]] const u64* const untracked_words = Array<Type::Untracked>();
         const u64* const state_words = Array<type>();
         const u64 num_query_words = size / BYTES_PER_WORD + 1;
         const u64 word_begin = offset / BYTES_PER_WORD;
@@ -574,8 +573,7 @@ private:
         u64 begin = std::numeric_limits<u64>::max();
         u64 end = 0;
         for (u64 word_index = word_begin; word_index < word_end; ++word_index) {
-            const u64 off_word = type == Type::GPU ? untracked_words[word_index] : 0;
-            const u64 word = state_words[word_index] & ~off_word;
+            const u64 word = state_words[word_index];
             if (word == 0) {
                 continue;
             }

src/video_core/buffer_cache/buffer_cache.h

@@ -200,7 +200,16 @@ public:
     /// Return true when a CPU region is modified from the CPU
     [[nodiscard]] bool IsRegionCpuModified(VAddr addr, size_t size);
 
-    std::mutex mutex;
+    void SetDrawIndirect(
+        const Tegra::Engines::DrawManager::IndirectParams* current_draw_indirect_) {
+        current_draw_indirect = current_draw_indirect_;
+    }
+
+    [[nodiscard]] std::pair<Buffer*, u32> GetDrawIndirectCount();
+
+    [[nodiscard]] std::pair<Buffer*, u32> GetDrawIndirectBuffer();
+
+    std::recursive_mutex mutex;
     Runtime& runtime;
 
 private:
@@ -272,6 +281,8 @@ private:
 
     void BindHostVertexBuffers();
 
+    void BindHostDrawIndirectBuffers();
+
     void BindHostGraphicsUniformBuffers(size_t stage);
 
     void BindHostGraphicsUniformBuffer(size_t stage, u32 index, u32 binding_index, bool needs_bind);
@@ -298,6 +309,8 @@ private:
 
     void UpdateVertexBuffer(u32 index);
 
+    void UpdateDrawIndirect();
+
     void UpdateUniformBuffers(size_t stage);
 
     void UpdateStorageBuffers(size_t stage);
@@ -372,6 +385,8 @@ private:
     SlotVector<Buffer> slot_buffers;
     DelayedDestructionRing<Buffer, 8> delayed_destruction_ring;
 
+    const Tegra::Engines::DrawManager::IndirectParams* current_draw_indirect{};
+
     u32 last_index_count = 0;
 
     Binding index_buffer;
@@ -380,6 +395,8 @@ private:
     std::array<std::array<Binding, NUM_STORAGE_BUFFERS>, NUM_STAGES> storage_buffers;
     std::array<std::array<TextureBufferBinding, NUM_TEXTURE_BUFFERS>, NUM_STAGES> texture_buffers;
     std::array<Binding, NUM_TRANSFORM_FEEDBACK_BUFFERS> transform_feedback_buffers;
+    Binding count_buffer_binding;
+    Binding indirect_buffer_binding;
 
     std::array<Binding, NUM_COMPUTE_UNIFORM_BUFFERS> compute_uniform_buffers;
     std::array<Binding, NUM_STORAGE_BUFFERS> compute_storage_buffers;
@@ -674,6 +691,9 @@ void BufferCache<P>::BindHostGeometryBuffers(bool is_indexed) {
     }
     BindHostVertexBuffers();
     BindHostTransformFeedbackBuffers();
+    if (current_draw_indirect) {
+        BindHostDrawIndirectBuffers();
+    }
 }
 
 template <class P>
@@ -823,6 +843,7 @@ bool BufferCache<P>::ShouldWaitAsyncFlushes() const noexcept {
 template <class P>
 void BufferCache<P>::CommitAsyncFlushesHigh() {
     AccumulateFlushes();
+
     if (committed_ranges.empty()) {
         return;
     }
@@ -869,7 +890,7 @@ void BufferCache<P>::CommitAsyncFlushesHigh() {
                                 buffer_id,
                             });
                             // Align up to avoid cache conflicts
-                            constexpr u64 align = 256ULL;
+                            constexpr u64 align = 8ULL;
                             constexpr u64 mask = ~(align - 1ULL);
                             total_size_bytes += (new_size + align - 1) & mask;
                             largest_copy = std::max(largest_copy, new_size);
@@ -1041,6 +1062,19 @@ void BufferCache<P>::BindHostVertexBuffers() {
     }
 }
 
+template <class P>
+void BufferCache<P>::BindHostDrawIndirectBuffers() {
+    const auto bind_buffer = [this](const Binding& binding) {
+        Buffer& buffer = slot_buffers[binding.buffer_id];
+        TouchBuffer(buffer, binding.buffer_id);
+        SynchronizeBuffer(buffer, binding.cpu_addr, binding.size);
+    };
+    if (current_draw_indirect->include_count) {
+        bind_buffer(count_buffer_binding);
+    }
+    bind_buffer(indirect_buffer_binding);
+}
+
 template <class P>
 void BufferCache<P>::BindHostGraphicsUniformBuffers(size_t stage) {
     u32 dirty = ~0U;
@@ -1272,6 +1306,9 @@ void BufferCache<P>::DoUpdateGraphicsBuffers(bool is_indexed) {
             UpdateStorageBuffers(stage);
             UpdateTextureBuffers(stage);
         }
+        if (current_draw_indirect) {
+            UpdateDrawIndirect();
+        }
     } while (has_deleted_buffers);
 }
 
@@ -1289,7 +1326,7 @@ void BufferCache<P>::UpdateIndexBuffer() {
     const auto& draw_state = maxwell3d->draw_manager->GetDrawState();
     const auto& index_array = draw_state.index_buffer;
     auto& flags = maxwell3d->dirty.flags;
-    if (!flags[Dirty::IndexBuffer] && last_index_count == index_array.count) {
+    if (!flags[Dirty::IndexBuffer]) {
         return;
     }
     flags[Dirty::IndexBuffer] = false;
@@ -1361,6 +1398,27 @@ void BufferCache<P>::UpdateVertexBuffer(u32 index) {
     };
 }
 
+template <class P>
+void BufferCache<P>::UpdateDrawIndirect() {
+    const auto update = [this](GPUVAddr gpu_addr, size_t size, Binding& binding) {
+        const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
+        if (!cpu_addr) {
+            binding = NULL_BINDING;
+            return;
+        }
+        binding = Binding{
+            .cpu_addr = *cpu_addr,
+            .size = static_cast<u32>(size),
+            .buffer_id = FindBuffer(*cpu_addr, static_cast<u32>(size)),
+        };
+    };
+    if (current_draw_indirect->include_count) {
+        update(current_draw_indirect->count_start_address, sizeof(u32), count_buffer_binding);
+    }
+    update(current_draw_indirect->indirect_start_address, current_draw_indirect->buffer_size,
+           indirect_buffer_binding);
+}
+
 template <class P>
 void BufferCache<P>::UpdateUniformBuffers(size_t stage) {
     ForEachEnabledBit(enabled_uniform_buffer_masks[stage], [&](u32 index) {
@@ -1880,14 +1938,21 @@ typename BufferCache<P>::Binding BufferCache<P>::StorageBufferBinding(GPUVAddr s
                                                                       bool is_written) const {
     const GPUVAddr gpu_addr = gpu_memory->Read<u64>(ssbo_addr);
     const u32 size = gpu_memory->Read<u32>(ssbo_addr + 8);
-    const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(gpu_addr);
+    const u32 alignment = runtime.GetStorageBufferAlignment();
+
+    const GPUVAddr aligned_gpu_addr = Common::AlignDown(gpu_addr, alignment);
+    const u32 aligned_size =
+        Common::AlignUp(static_cast<u32>(gpu_addr - aligned_gpu_addr) + size, alignment);
+
+    const std::optional<VAddr> cpu_addr = gpu_memory->GpuToCpuAddress(aligned_gpu_addr);
     if (!cpu_addr || size == 0) {
         return NULL_BINDING;
     }
-    const VAddr cpu_end = Common::AlignUp(*cpu_addr + size, Core::Memory::YUZU_PAGESIZE);
+
+    const VAddr cpu_end = Common::AlignUp(*cpu_addr + aligned_size, Core::Memory::YUZU_PAGESIZE);
     const Binding binding{
         .cpu_addr = *cpu_addr,
-        .size = is_written ? size : static_cast<u32>(cpu_end - *cpu_addr),
+        .size = is_written ? aligned_size : static_cast<u32>(cpu_end - *cpu_addr),
         .buffer_id = BufferId{},
     };
     return binding;
@@ -1941,4 +2006,16 @@ bool BufferCache<P>::HasFastUniformBufferBound(size_t stage, u32 binding_index)
     }
 }
 
+template <class P>
+std::pair<typename BufferCache<P>::Buffer*, u32> BufferCache<P>::GetDrawIndirectCount() {
+    auto& buffer = slot_buffers[count_buffer_binding.buffer_id];
+    return std::make_pair(&buffer, buffer.Offset(count_buffer_binding.cpu_addr));
+}
+
+template <class P>
+std::pair<typename BufferCache<P>::Buffer*, u32> BufferCache<P>::GetDrawIndirectBuffer() {
+    auto& buffer = slot_buffers[indirect_buffer_binding.buffer_id];
+    return std::make_pair(&buffer, buffer.Offset(indirect_buffer_binding.cpu_addr));
+}
+
 } // namespace VideoCommon

src/video_core/cache_types.h

@@ -0,0 +1,24 @@
+// SPDX-FileCopyrightText: Copyright 2020 yuzu Emulator Project
+// SPDX-License-Identifier: GPL-2.0-or-later
+
+#pragma once
+
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+
+namespace VideoCommon {
+
+enum class CacheType : u32 {
+    None = 0,
+    TextureCache = 1 << 0,
+    QueryCache = 1 << 1,
+    BufferCache = 1 << 2,
+    ShaderCache = 1 << 3,
+    NoTextureCache = QueryCache | BufferCache | ShaderCache,
+    NoBufferCache = TextureCache | QueryCache | ShaderCache,
+    NoQueryCache = TextureCache | BufferCache | ShaderCache,
+    All = TextureCache | QueryCache | BufferCache | ShaderCache,
+};
+DECLARE_ENUM_FLAG_OPERATORS(CacheType)
+
+} // namespace VideoCommon

src/video_core/dma_pusher.cpp

@@ -61,7 +61,7 @@ bool DmaPusher::Step() {
     } else {
         const CommandListHeader command_list_header{
             command_list.command_lists[dma_pushbuffer_subindex++]};
-        const GPUVAddr dma_get = command_list_header.addr;
+        dma_state.dma_get = command_list_header.addr;
 
         if (dma_pushbuffer_subindex >= command_list.command_lists.size()) {
             // We've gone through the current list, remove it from the queue
@@ -75,12 +75,22 @@ bool DmaPusher::Step() {
 
         // Push buffer non-empty, read a word
         command_headers.resize_destructive(command_list_header.size);
-        if (Settings::IsGPULevelHigh()) {
-            memory_manager.ReadBlock(dma_get, command_headers.data(),
-                                     command_list_header.size * sizeof(u32));
+        constexpr u32 MacroRegistersStart = 0xE00;
+        if (dma_state.method < MacroRegistersStart) {
+            if (Settings::IsGPULevelHigh()) {
+                memory_manager.ReadBlock(dma_state.dma_get, command_headers.data(),
+                                         command_list_header.size * sizeof(u32));
+            } else {
+                memory_manager.ReadBlockUnsafe(dma_state.dma_get, command_headers.data(),
+                                               command_list_header.size * sizeof(u32));
+            }
         } else {
-            memory_manager.ReadBlockUnsafe(dma_get, command_headers.data(),
-                                           command_list_header.size * sizeof(u32));
+            const size_t copy_size = command_list_header.size * sizeof(u32);
+            if (subchannels[dma_state.subchannel]) {
+                subchannels[dma_state.subchannel]->current_dirty =
+                    memory_manager.IsMemoryDirty(dma_state.dma_get, copy_size);
+            }
+            memory_manager.ReadBlockUnsafe(dma_state.dma_get, command_headers.data(), copy_size);
         }
         ProcessCommands(command_headers);
     }
@@ -94,6 +104,7 @@ void DmaPusher::ProcessCommands(std::span<const CommandHeader> commands) {
 
         if (dma_state.method_count) {
             // Data word of methods command
+            dma_state.dma_word_offset = static_cast<u32>(index * sizeof(u32));
             if (dma_state.non_incrementing) {
                 const u32 max_write = static_cast<u32>(
                     std::min<std::size_t>(index + dma_state.method_count, commands.size()) - index);
@@ -132,6 +143,8 @@ void DmaPusher::ProcessCommands(std::span<const CommandHeader> commands) {
             case SubmissionMode::Inline:
                 dma_state.method = command_header.method;
                 dma_state.subchannel = command_header.subchannel;
+                dma_state.dma_word_offset = static_cast<u64>(
+                    -static_cast<s64>(dma_state.dma_get)); // negate to set address as 0
                 CallMethod(command_header.arg_count);
                 dma_state.non_incrementing = true;
                 dma_increment_once = false;
@@ -164,8 +177,14 @@ void DmaPusher::CallMethod(u32 argument) const {
             dma_state.method_count,
         });
     } else {
-        subchannels[dma_state.subchannel]->CallMethod(dma_state.method, argument,
-                                                      dma_state.is_last_call);
+        auto subchannel = subchannels[dma_state.subchannel];
+        if (!subchannel->execution_mask[dma_state.method]) [[likely]] {
+            subchannel->method_sink.emplace_back(dma_state.method, argument);
+            return;
+        }
+        subchannel->ConsumeSink();
+        subchannel->current_dma_segment = dma_state.dma_get + dma_state.dma_word_offset;
+        subchannel->CallMethod(dma_state.method, argument, dma_state.is_last_call);
     }
 }
 
@@ -174,8 +193,11 @@ void DmaPusher::CallMultiMethod(const u32* base_start, u32 num_methods) const {
         puller.CallMultiMethod(dma_state.method, dma_state.subchannel, base_start, num_methods,
                                dma_state.method_count);
     } else {
-        subchannels[dma_state.subchannel]->CallMultiMethod(dma_state.method, base_start,
-                                                           num_methods, dma_state.method_count);
+        auto subchannel = subchannels[dma_state.subchannel];
+        subchannel->ConsumeSink();
+        subchannel->current_dma_segment = dma_state.dma_get + dma_state.dma_word_offset;
+        subchannel->CallMultiMethod(dma_state.method, base_start, num_methods,
+                                    dma_state.method_count);
     }
 }
 

src/video_core/dma_pusher.h

@@ -156,6 +156,8 @@ private:
         u32 subchannel;        ///< Current subchannel
         u32 method_count;      ///< Current method count
         u32 length_pending;    ///< Large NI command length pending
+        GPUVAddr dma_get;      ///< Currently read segment
+        u64 dma_word_offset;   ///< Current word ofset from address
         bool non_incrementing; ///< Current command's NI flag
         bool is_last_call;
     };

src/video_core/engines/draw_manager.cpp

@@ -91,6 +91,23 @@ void DrawManager::DrawIndex(PrimitiveTopology topology, u32 index_first, u32 ind
     ProcessDraw(true, num_instances);
 }
 
+void DrawManager::DrawArrayIndirect(PrimitiveTopology topology) {
+    draw_state.topology = topology;
+
+    ProcessDrawIndirect();
+}
+
+void DrawManager::DrawIndexedIndirect(PrimitiveTopology topology, u32 index_first,
+                                      u32 index_count) {
+    const auto& regs{maxwell3d->regs};
+    draw_state.topology = topology;
+    draw_state.index_buffer = regs.index_buffer;
+    draw_state.index_buffer.first = index_first;
+    draw_state.index_buffer.count = index_count;
+
+    ProcessDrawIndirect();
+}
+
 void DrawManager::SetInlineIndexBuffer(u32 index) {
     draw_state.inline_index_draw_indexes.push_back(static_cast<u8>(index & 0x000000ff));
     draw_state.inline_index_draw_indexes.push_back(static_cast<u8>((index & 0x0000ff00) >> 8));
@@ -198,4 +215,18 @@ void DrawManager::ProcessDraw(bool draw_indexed, u32 instance_count) {
         maxwell3d->rasterizer->Draw(draw_indexed, instance_count);
     }
 }
+
+void DrawManager::ProcessDrawIndirect() {
+    LOG_TRACE(
+        HW_GPU,
+        "called, topology={}, is_indexed={}, includes_count={}, buffer_size={}, max_draw_count={}",
+        draw_state.topology, indirect_state.is_indexed, indirect_state.include_count,
+        indirect_state.buffer_size, indirect_state.max_draw_counts);
+
+    UpdateTopology();
+
+    if (maxwell3d->ShouldExecute()) {
+        maxwell3d->rasterizer->DrawIndirect();
+    }
+}
 } // namespace Tegra::Engines

src/video_core/engines/draw_manager.h

@@ -32,6 +32,16 @@ public:
         std::vector<u8> inline_index_draw_indexes;
     };
 
+    struct IndirectParams {
+        bool is_indexed;
+        bool include_count;
+        GPUVAddr count_start_address;
+        GPUVAddr indirect_start_address;
+        size_t buffer_size;
+        size_t max_draw_counts;
+        size_t stride;
+    };
+
     explicit DrawManager(Maxwell3D* maxwell_3d);
 
     void ProcessMethodCall(u32 method, u32 argument);
@@ -46,10 +56,22 @@ public:
     void DrawIndex(PrimitiveTopology topology, u32 index_first, u32 index_count, u32 base_index,
                    u32 base_instance, u32 num_instances);
 
+    void DrawArrayIndirect(PrimitiveTopology topology);
+
+    void DrawIndexedIndirect(PrimitiveTopology topology, u32 index_first, u32 index_count);
+
     const State& GetDrawState() const {
         return draw_state;
     }
 
+    IndirectParams& GetIndirectParams() {
+        return indirect_state;
+    }
+
+    const IndirectParams& GetIndirectParams() const {
+        return indirect_state;
+    }
+
 private:
     void SetInlineIndexBuffer(u32 index);
 
@@ -63,7 +85,10 @@ private:
 
     void ProcessDraw(bool draw_indexed, u32 instance_count);
 
+    void ProcessDrawIndirect();
+
     Maxwell3D* maxwell3d{};
     State draw_state{};
+    IndirectParams indirect_state{};
 };
 } // namespace Tegra::Engines

src/video_core/engines/engine_interface.h

@@ -3,6 +3,10 @@
 
 #pragma once
 
+#include <bitset>
+#include <limits>
+#include <vector>
+
 #include "common/common_types.h"
 
 namespace Tegra::Engines {
@@ -17,6 +21,26 @@ public:
     /// Write multiple values to the register identified by method.
     virtual void CallMultiMethod(u32 method, const u32* base_start, u32 amount,
                                  u32 methods_pending) = 0;
+
+    void ConsumeSink() {
+        if (method_sink.empty()) {
+            return;
+        }
+        ConsumeSinkImpl();
+    }
+
+    std::bitset<std::numeric_limits<u16>::max()> execution_mask{};
+    std::vector<std::pair<u32, u32>> method_sink{};
+    bool current_dirty{};
+    GPUVAddr current_dma_segment;
+
+protected:
+    virtual void ConsumeSinkImpl() {
+        for (auto [method, value] : method_sink) {
+            CallMethod(method, value, true);
+        }
+        method_sink.clear();
+    }
 };
 
 } // namespace Tegra::Engines

src/video_core/engines/engine_upload.cpp

@@ -76,7 +76,7 @@ void State::ProcessData(std::span<const u8> read_buffer) {
                                        regs.dest.height, regs.dest.depth, x_offset, regs.dest.y,
                                        x_elements, regs.line_count, regs.dest.BlockHeight(),
                                        regs.dest.BlockDepth(), regs.line_length_in);
-        memory_manager.WriteBlock(address, tmp_buffer.data(), dst_size);
+        memory_manager.WriteBlockCached(address, tmp_buffer.data(), dst_size);
     }
 }
 

src/video_core/engines/fermi_2d.cpp

@@ -6,6 +6,7 @@
 #include "common/microprofile.h"
 #include "video_core/engines/fermi_2d.h"
 #include "video_core/engines/sw_blitter/blitter.h"
+#include "video_core/memory_manager.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/surface.h"
 #include "video_core/textures/decoders.h"
@@ -20,11 +21,14 @@ namespace Tegra::Engines {
 
 using namespace Texture;
 
-Fermi2D::Fermi2D(MemoryManager& memory_manager_) {
-    sw_blitter = std::make_unique<Blitter::SoftwareBlitEngine>(memory_manager_);
+Fermi2D::Fermi2D(MemoryManager& memory_manager_) : memory_manager{memory_manager_} {
+    sw_blitter = std::make_unique<Blitter::SoftwareBlitEngine>(memory_manager);
     // Nvidia's OpenGL driver seems to assume these values
     regs.src.depth = 1;
     regs.dst.depth = 1;
+
+    execution_mask.reset();
+    execution_mask[FERMI2D_REG_INDEX(pixels_from_memory.src_y0) + 1] = true;
 }
 
 Fermi2D::~Fermi2D() = default;
@@ -49,6 +53,13 @@ void Fermi2D::CallMultiMethod(u32 method, const u32* base_start, u32 amount, u32
     }
 }
 
+void Fermi2D::ConsumeSinkImpl() {
+    for (auto [method, value] : method_sink) {
+        regs.reg_array[method] = value;
+    }
+    method_sink.clear();
+}
+
 void Fermi2D::Blit() {
     MICROPROFILE_SCOPE(GPU_BlitEngine);
     LOG_DEBUG(HW_GPU, "called. source address=0x{:x}, destination address=0x{:x}",
@@ -94,6 +105,7 @@ void Fermi2D::Blit() {
         config.src_x0 = 0;
     }
 
+    memory_manager.FlushCaching();
     if (!rasterizer->AccelerateSurfaceCopy(src, regs.dst, config)) {
         sw_blitter->Blit(src, regs.dst, config);
     }

src/video_core/engines/fermi_2d.h

@@ -305,10 +305,13 @@ public:
 private:
     VideoCore::RasterizerInterface* rasterizer = nullptr;
     std::unique_ptr<Blitter::SoftwareBlitEngine> sw_blitter;
+    MemoryManager& memory_manager;
 
     /// Performs the copy from the source surface to the destination surface as configured in the
     /// registers.
     void Blit();
+
+    void ConsumeSinkImpl() override;
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \

src/video_core/engines/kepler_compute.cpp

@@ -14,7 +14,12 @@
 namespace Tegra::Engines {
 
 KeplerCompute::KeplerCompute(Core::System& system_, MemoryManager& memory_manager_)
-    : system{system_}, memory_manager{memory_manager_}, upload_state{memory_manager, regs.upload} {}
+    : system{system_}, memory_manager{memory_manager_}, upload_state{memory_manager, regs.upload} {
+    execution_mask.reset();
+    execution_mask[KEPLER_COMPUTE_REG_INDEX(exec_upload)] = true;
+    execution_mask[KEPLER_COMPUTE_REG_INDEX(data_upload)] = true;
+    execution_mask[KEPLER_COMPUTE_REG_INDEX(launch)] = true;
+}
 
 KeplerCompute::~KeplerCompute() = default;
 
@@ -23,6 +28,13 @@ void KeplerCompute::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_)
     upload_state.BindRasterizer(rasterizer);
 }
 
+void KeplerCompute::ConsumeSinkImpl() {
+    for (auto [method, value] : method_sink) {
+        regs.reg_array[method] = value;
+    }
+    method_sink.clear();
+}
+
 void KeplerCompute::CallMethod(u32 method, u32 method_argument, bool is_last_call) {
     ASSERT_MSG(method < Regs::NUM_REGS,
                "Invalid KeplerCompute register, increase the size of the Regs structure");

src/video_core/engines/kepler_compute.h

@@ -204,6 +204,8 @@ public:
 private:
     void ProcessLaunch();
 
+    void ConsumeSinkImpl() override;
+
     /// Retrieves information about a specific TIC entry from the TIC buffer.
     Texture::TICEntry GetTICEntry(u32 tic_index) const;
 

src/video_core/engines/kepler_memory.cpp

@@ -18,6 +18,17 @@ KeplerMemory::~KeplerMemory() = default;
 
 void KeplerMemory::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
     upload_state.BindRasterizer(rasterizer_);
+
+    execution_mask.reset();
+    execution_mask[KEPLERMEMORY_REG_INDEX(exec)] = true;
+    execution_mask[KEPLERMEMORY_REG_INDEX(data)] = true;
+}
+
+void KeplerMemory::ConsumeSinkImpl() {
+    for (auto [method, value] : method_sink) {
+        regs.reg_array[method] = value;
+    }
+    method_sink.clear();
 }
 
 void KeplerMemory::CallMethod(u32 method, u32 method_argument, bool is_last_call) {

src/video_core/engines/kepler_memory.h

@@ -73,6 +73,8 @@ public:
     } regs{};
 
 private:
+    void ConsumeSinkImpl() override;
+
     Core::System& system;
     Upload::State upload_state;
 };

src/video_core/engines/maxwell_3d.cpp

@@ -4,6 +4,8 @@
 #include <cstring>
 #include <optional>
 #include "common/assert.h"
+#include "common/scope_exit.h"
+#include "common/settings.h"
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "video_core/dirty_flags.h"
@@ -28,6 +30,10 @@ Maxwell3D::Maxwell3D(Core::System& system_, MemoryManager& memory_manager_)
                                                                                 regs.upload} {
     dirty.flags.flip();
     InitializeRegisterDefaults();
+    execution_mask.reset();
+    for (size_t i = 0; i < execution_mask.size(); i++) {
+        execution_mask[i] = IsMethodExecutable(static_cast<u32>(i));
+    }
 }
 
 Maxwell3D::~Maxwell3D() = default;
@@ -121,6 +127,71 @@ void Maxwell3D::InitializeRegisterDefaults() {
     shadow_state = regs;
 }
 
+bool Maxwell3D::IsMethodExecutable(u32 method) {
+    if (method >= MacroRegistersStart) {
+        return true;
+    }
+    switch (method) {
+    case MAXWELL3D_REG_INDEX(draw.end):
+    case MAXWELL3D_REG_INDEX(draw.begin):
+    case MAXWELL3D_REG_INDEX(vertex_buffer.first):
+    case MAXWELL3D_REG_INDEX(vertex_buffer.count):
+    case MAXWELL3D_REG_INDEX(index_buffer.first):
+    case MAXWELL3D_REG_INDEX(index_buffer.count):
+    case MAXWELL3D_REG_INDEX(draw_inline_index):
+    case MAXWELL3D_REG_INDEX(index_buffer32_subsequent):
+    case MAXWELL3D_REG_INDEX(index_buffer16_subsequent):
+    case MAXWELL3D_REG_INDEX(index_buffer8_subsequent):
+    case MAXWELL3D_REG_INDEX(index_buffer32_first):
+    case MAXWELL3D_REG_INDEX(index_buffer16_first):
+    case MAXWELL3D_REG_INDEX(index_buffer8_first):
+    case MAXWELL3D_REG_INDEX(inline_index_2x16.even):
+    case MAXWELL3D_REG_INDEX(inline_index_4x8.index0):
+    case MAXWELL3D_REG_INDEX(vertex_array_instance_first):
+    case MAXWELL3D_REG_INDEX(vertex_array_instance_subsequent):
+    case MAXWELL3D_REG_INDEX(wait_for_idle):
+    case MAXWELL3D_REG_INDEX(shadow_ram_control):
+    case MAXWELL3D_REG_INDEX(load_mme.instruction_ptr):
+    case MAXWELL3D_REG_INDEX(load_mme.instruction):
+    case MAXWELL3D_REG_INDEX(load_mme.start_address):
+    case MAXWELL3D_REG_INDEX(falcon[4]):
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer):
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 1:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 2:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 3:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 4:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 5:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 6:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 7:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 8:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 9:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 10:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 11:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 12:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 13:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 14:
+    case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 15:
+    case MAXWELL3D_REG_INDEX(bind_groups[0].raw_config):
+    case MAXWELL3D_REG_INDEX(bind_groups[1].raw_config):
+    case MAXWELL3D_REG_INDEX(bind_groups[2].raw_config):
+    case MAXWELL3D_REG_INDEX(bind_groups[3].raw_config):
+    case MAXWELL3D_REG_INDEX(bind_groups[4].raw_config):
+    case MAXWELL3D_REG_INDEX(topology_override):
+    case MAXWELL3D_REG_INDEX(clear_surface):
+    case MAXWELL3D_REG_INDEX(report_semaphore.query):
+    case MAXWELL3D_REG_INDEX(render_enable.mode):
+    case MAXWELL3D_REG_INDEX(clear_report_value):
+    case MAXWELL3D_REG_INDEX(sync_info):
+    case MAXWELL3D_REG_INDEX(launch_dma):
+    case MAXWELL3D_REG_INDEX(inline_data):
+    case MAXWELL3D_REG_INDEX(fragment_barrier):
+    case MAXWELL3D_REG_INDEX(tiled_cache_barrier):
+        return true;
+    default:
+        return false;
+    }
+}
+
 void Maxwell3D::ProcessMacro(u32 method, const u32* base_start, u32 amount, bool is_last_call) {
     if (executing_macro == 0) {
         // A macro call must begin by writing the macro method's register, not its argument.
@@ -130,14 +201,72 @@ void Maxwell3D::ProcessMacro(u32 method, const u32* base_start, u32 amount, bool
     }
 
     macro_params.insert(macro_params.end(), base_start, base_start + amount);
+    for (size_t i = 0; i < amount; i++) {
+        macro_addresses.push_back(current_dma_segment + i * sizeof(u32));
+    }
+    macro_segments.emplace_back(current_dma_segment, amount);
+    current_macro_dirty |= current_dirty;
+    current_dirty = false;
 
     // Call the macro when there are no more parameters in the command buffer
     if (is_last_call) {
+        ConsumeSink();
         CallMacroMethod(executing_macro, macro_params);
         macro_params.clear();
+        macro_addresses.clear();
+        macro_segments.clear();
+        current_macro_dirty = false;
     }
 }
 
+void Maxwell3D::RefreshParametersImpl() {
+    size_t current_index = 0;
+    for (auto& segment : macro_segments) {
+        if (segment.first == 0) {
+            current_index += segment.second;
+            continue;
+        }
+        memory_manager.ReadBlock(segment.first, &macro_params[current_index],
+                                 sizeof(u32) * segment.second);
+        current_index += segment.second;
+    }
+}
+
+u32 Maxwell3D::GetMaxCurrentVertices() {
+    u32 num_vertices = 0;
+    for (size_t index = 0; index < Regs::NumVertexArrays; ++index) {
+        const auto& array = regs.vertex_streams[index];
+        if (array.enable == 0) {
+            continue;
+        }
+        const auto& attribute = regs.vertex_attrib_format[index];
+        if (attribute.constant) {
+            num_vertices = std::max(num_vertices, 1U);
+            continue;
+        }
+        const auto& limit = regs.vertex_stream_limits[index];
+        const GPUVAddr gpu_addr_begin = array.Address();
+        const GPUVAddr gpu_addr_end = limit.Address() + 1;
+        const u32 address_size = static_cast<u32>(gpu_addr_end - gpu_addr_begin);
+        num_vertices = std::max(
+            num_vertices, address_size / std::max(attribute.SizeInBytes(), array.stride.Value()));
+    }
+    return num_vertices;
+}
+
+size_t Maxwell3D::EstimateIndexBufferSize() {
+    GPUVAddr start_address = regs.index_buffer.StartAddress();
+    GPUVAddr end_address = regs.index_buffer.EndAddress();
+    constexpr std::array<size_t, 4> max_sizes = {
+        std::numeric_limits<u8>::max(), std::numeric_limits<u16>::max(),
+        std::numeric_limits<u32>::max(), std::numeric_limits<u32>::max()};
+    const size_t byte_size = regs.index_buffer.FormatSizeInBytes();
+    return std::min<size_t>(
+        memory_manager.GetMemoryLayoutSize(start_address, byte_size * max_sizes[byte_size]) /
+            byte_size,
+        static_cast<size_t>(end_address - start_address));
+}
+
 u32 Maxwell3D::ProcessShadowRam(u32 method, u32 argument) {
     // Keep track of the register value in shadow_state when requested.
     const auto control = shadow_state.shadow_ram_control;
@@ -152,6 +281,29 @@ u32 Maxwell3D::ProcessShadowRam(u32 method, u32 argument) {
     return argument;
 }
 
+void Maxwell3D::ConsumeSinkImpl() {
+    SCOPE_EXIT({ method_sink.clear(); });
+    const auto control = shadow_state.shadow_ram_control;
+    if (control == Regs::ShadowRamControl::Track ||
+        control == Regs::ShadowRamControl::TrackWithFilter) {
+
+        for (auto [method, value] : method_sink) {
+            shadow_state.reg_array[method] = value;
+            ProcessDirtyRegisters(method, value);
+        }
+        return;
+    }
+    if (control == Regs::ShadowRamControl::Replay) {
+        for (auto [method, value] : method_sink) {
+            ProcessDirtyRegisters(method, shadow_state.reg_array[method]);
+        }
+        return;
+    }
+    for (auto [method, value] : method_sink) {
+        ProcessDirtyRegisters(method, value);
+    }
+}
+
 void Maxwell3D::ProcessDirtyRegisters(u32 method, u32 argument) {
     if (regs.reg_array[method] == argument) {
         return;
@@ -263,7 +415,6 @@ void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) {
 
     const u32 argument = ProcessShadowRam(method, method_argument);
     ProcessDirtyRegisters(method, argument);
-
     ProcessMethodCall(method, argument, method_argument, is_last_call);
 }
 
@@ -294,9 +445,11 @@ void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
     case MAXWELL3D_REG_INDEX(const_buffer.buffer) + 15:
         ProcessCBMultiData(base_start, amount);
         break;
-    case MAXWELL3D_REG_INDEX(inline_data):
+    case MAXWELL3D_REG_INDEX(inline_data): {
+        ASSERT(methods_pending == amount);
         upload_state.ProcessData(base_start, amount);
         return;
+    }
     default:
         for (u32 i = 0; i < amount; i++) {
             CallMethod(method, base_start[i], methods_pending - i <= 1);
@@ -332,11 +485,6 @@ void Maxwell3D::StampQueryResult(u64 payload, bool long_query) {
 }
 
 void Maxwell3D::ProcessQueryGet() {
-    // TODO(Subv): Support the other query units.
-    if (regs.report_semaphore.query.location != Regs::ReportSemaphore::Location::All) {
-        LOG_DEBUG(HW_GPU, "Locations other than ALL are unimplemented");
-    }
-
     switch (regs.report_semaphore.query.operation) {
     case Regs::ReportSemaphore::Operation::Release:
         if (regs.report_semaphore.query.short_query != 0) {
@@ -389,7 +537,11 @@ void Maxwell3D::ProcessQueryCondition() {
     case Regs::RenderEnable::Override::NeverRender:
         execute_on = false;
         break;
-    case Regs::RenderEnable::Override::UseRenderEnable:
+    case Regs::RenderEnable::Override::UseRenderEnable: {
+        if (rasterizer->AccelerateConditionalRendering()) {
+            execute_on = true;
+            return;
+        }
         switch (regs.render_enable.mode) {
         case Regs::RenderEnable::Mode::True: {
             execute_on = true;
@@ -427,6 +579,7 @@ void Maxwell3D::ProcessQueryCondition() {
         }
         break;
     }
+    }
 }
 
 void Maxwell3D::ProcessCounterReset() {
@@ -463,7 +616,8 @@ std::optional<u64> Maxwell3D::GetQueryResult() {
 }
 
 void Maxwell3D::ProcessCBBind(size_t stage_index) {
-    // Bind the buffer currently in CB_ADDRESS to the specified index in the desired shader stage.
+    // Bind the buffer currently in CB_ADDRESS to the specified index in the desired shader
+    // stage.
     const auto& bind_data = regs.bind_groups[stage_index];
     auto& buffer = state.shader_stages[stage_index].const_buffers[bind_data.shader_slot];
     buffer.enabled = bind_data.valid.Value() != 0;
@@ -490,7 +644,7 @@ void Maxwell3D::ProcessCBMultiData(const u32* start_base, u32 amount) {
 
     const GPUVAddr address{buffer_address + regs.const_buffer.offset};
     const size_t copy_size = amount * sizeof(u32);
-    memory_manager.WriteBlock(address, start_base, copy_size);
+    memory_manager.WriteBlockCached(address, start_base, copy_size);
 
     // Increment the current buffer position.
     regs.const_buffer.offset += static_cast<u32>(copy_size);
@@ -524,4 +678,10 @@ u32 Maxwell3D::GetRegisterValue(u32 method) const {
     return regs.reg_array[method];
 }
 
+void Maxwell3D::SetHLEReplacementAttributeType(u32 bank, u32 offset,
+                                               HLEReplacementAttributeType name) {
+    const u64 key = (static_cast<u64>(bank) << 32) | offset;
+    replace_table.emplace(key, name);
+}
+
 } // namespace Tegra::Engines

src/video_core/engines/maxwell_3d.h

@@ -272,6 +272,7 @@ public:
             };
 
             union {
+                u32 raw;
                 BitField<0, 1, Mode> mode;
                 BitField<4, 8, u32> pad;
             };
@@ -1217,10 +1218,12 @@ public:
 
         struct Window {
             union {
+                u32 raw_x;
                 BitField<0, 16, u32> x_min;
                 BitField<16, 16, u32> x_max;
             };
             union {
+                u32 raw_y;
                 BitField<0, 16, u32> y_min;
                 BitField<16, 16, u32> y_max;
             };
@@ -2708,7 +2711,7 @@ public:
                 u32 post_z_pixel_imask;                                                ///< 0x0F1C
                 INSERT_PADDING_BYTES_NOINIT(0x20);
                 ConstantColorRendering const_color_rendering;                          ///< 0x0F40
-                s32 stencil_back_ref;                                                  ///< 0x0F54
+                u32 stencil_back_ref;                                                  ///< 0x0F54
                 u32 stencil_back_mask;                                                 ///< 0x0F58
                 u32 stencil_back_func_mask;                                            ///< 0x0F5C
                 INSERT_PADDING_BYTES_NOINIT(0x14);
@@ -2832,9 +2835,9 @@ public:
                 Blend blend;                                                           ///< 0x133C
                 u32 stencil_enable;                                                    ///< 0x1380
                 StencilOp stencil_front_op;                                            ///< 0x1384
-                s32 stencil_front_ref;                                                 ///< 0x1394
-                s32 stencil_front_func_mask;                                           ///< 0x1398
-                s32 stencil_front_mask;                                                ///< 0x139C
+                u32 stencil_front_ref;                                                 ///< 0x1394
+                u32 stencil_front_func_mask;                                           ///< 0x1398
+                u32 stencil_front_mask;                                                ///< 0x139C
                 INSERT_PADDING_BYTES_NOINIT(0x4);
                 u32 draw_auto_start_byte_count;                                        ///< 0x13A4
                 PsSaturate frag_color_clamp;                                           ///< 0x13A8
@@ -3020,6 +3023,24 @@ public:
     /// Store temporary hw register values, used by some calls to restore state after a operation
     Regs shadow_state;
 
+    // None Engine
+    enum class EngineHint : u32 {
+        None = 0x0,
+        OnHLEMacro = 0x1,
+    };
+
+    EngineHint engine_state{EngineHint::None};
+
+    enum class HLEReplacementAttributeType : u32 {
+        BaseVertex = 0x0,
+        BaseInstance = 0x1,
+        DrawID = 0x2,
+    };
+
+    void SetHLEReplacementAttributeType(u32 bank, u32 offset, HLEReplacementAttributeType name);
+
+    std::unordered_map<u64, HLEReplacementAttributeType> replace_table;
+
     static_assert(sizeof(Regs) == Regs::NUM_REGS * sizeof(u32), "Maxwell3D Regs has wrong size");
     static_assert(std::is_trivially_copyable_v<Regs>, "Maxwell3D Regs must be trivially copyable");
 
@@ -3067,6 +3088,35 @@ public:
     std::unique_ptr<DrawManager> draw_manager;
     friend class DrawManager;
 
+    GPUVAddr GetMacroAddress(size_t index) const {
+        return macro_addresses[index];
+    }
+
+    void RefreshParameters() {
+        if (!current_macro_dirty) {
+            return;
+        }
+        RefreshParametersImpl();
+    }
+
+    bool AnyParametersDirty() const {
+        return current_macro_dirty;
+    }
+
+    u32 GetMaxCurrentVertices();
+
+    size_t EstimateIndexBufferSize();
+
+    /// Handles a write to the CLEAR_BUFFERS register.
+    void ProcessClearBuffers(u32 layer_count);
+
+    /// Handles a write to the CB_BIND register.
+    void ProcessCBBind(size_t stage_index);
+
+    /// Handles a write to the CB_DATA[i] register.
+    void ProcessCBData(u32 value);
+    void ProcessCBMultiData(const u32* start_base, u32 amount);
+
 private:
     void InitializeRegisterDefaults();
 
@@ -3076,6 +3126,8 @@ private:
 
     void ProcessDirtyRegisters(u32 method, u32 argument);
 
+    void ConsumeSinkImpl() override;
+
     void ProcessMethodCall(u32 method, u32 argument, u32 nonshadow_argument, bool is_last_call);
 
     /// Retrieves information about a specific TIC entry from the TIC buffer.
@@ -3116,16 +3168,13 @@ private:
     /// Handles writes to syncing register.
     void ProcessSyncPoint();
 
-    /// Handles a write to the CB_DATA[i] register.
-    void ProcessCBData(u32 value);
-    void ProcessCBMultiData(const u32* start_base, u32 amount);
-
-    /// Handles a write to the CB_BIND register.
-    void ProcessCBBind(size_t stage_index);
-
     /// Returns a query's value or an empty object if the value will be deferred through a cache.
     std::optional<u64> GetQueryResult();
 
+    void RefreshParametersImpl();
+
+    bool IsMethodExecutable(u32 method);
+
     Core::System& system;
     MemoryManager& memory_manager;
 
@@ -3145,6 +3194,10 @@ private:
     Upload::State upload_state;
 
     bool execute_on{true};
+
+    std::vector<std::pair<GPUVAddr, size_t>> macro_segments;
+    std::vector<GPUVAddr> macro_addresses;
+    bool current_macro_dirty{};
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \

src/video_core/engines/maxwell_dma.cpp

@@ -21,7 +21,10 @@ namespace Tegra::Engines {
 using namespace Texture;
 
 MaxwellDMA::MaxwellDMA(Core::System& system_, MemoryManager& memory_manager_)
-    : system{system_}, memory_manager{memory_manager_} {}
+    : system{system_}, memory_manager{memory_manager_} {
+    execution_mask.reset();
+    execution_mask[offsetof(Regs, launch_dma) / sizeof(u32)] = true;
+}
 
 MaxwellDMA::~MaxwellDMA() = default;
 
@@ -29,6 +32,13 @@ void MaxwellDMA::BindRasterizer(VideoCore::RasterizerInterface* rasterizer_) {
     rasterizer = rasterizer_;
 }
 
+void MaxwellDMA::ConsumeSinkImpl() {
+    for (auto [method, value] : method_sink) {
+        regs.reg_array[method] = value;
+    }
+    method_sink.clear();
+}
+
 void MaxwellDMA::CallMethod(u32 method, u32 method_argument, bool is_last_call) {
     ASSERT_MSG(method < NUM_REGS, "Invalid MaxwellDMA register");
 
@@ -59,7 +69,7 @@ void MaxwellDMA::Launch() {
     if (launch.multi_line_enable) {
         const bool is_src_pitch = launch.src_memory_layout == LaunchDMA::MemoryLayout::PITCH;
         const bool is_dst_pitch = launch.dst_memory_layout == LaunchDMA::MemoryLayout::PITCH;
-
+        memory_manager.FlushCaching();
         if (!is_src_pitch && !is_dst_pitch) {
             // If both the source and the destination are in block layout, assert.
             CopyBlockLinearToBlockLinear();
@@ -94,6 +104,7 @@ void MaxwellDMA::Launch() {
                                             reinterpret_cast<u8*>(tmp_buffer.data()),
                                             regs.line_length_in * sizeof(u32));
         } else {
+            memory_manager.FlushCaching();
             const auto convert_linear_2_blocklinear_addr = [](u64 address) {
                 return (address & ~0x1f0ULL) | ((address & 0x40) >> 2) | ((address & 0x10) << 1) |
                        ((address & 0x180) >> 1) | ((address & 0x20) << 3);
@@ -111,8 +122,8 @@ void MaxwellDMA::Launch() {
                     memory_manager.ReadBlockUnsafe(
                         convert_linear_2_blocklinear_addr(regs.offset_in + offset),
                         tmp_buffer.data(), tmp_buffer.size());
-                    memory_manager.WriteBlock(regs.offset_out + offset, tmp_buffer.data(),
-                                              tmp_buffer.size());
+                    memory_manager.WriteBlockCached(regs.offset_out + offset, tmp_buffer.data(),
+                                                    tmp_buffer.size());
                 }
             } else if (is_src_pitch && !is_dst_pitch) {
                 UNIMPLEMENTED_IF(regs.line_length_in % 16 != 0);
@@ -122,7 +133,7 @@ void MaxwellDMA::Launch() {
                 for (u32 offset = 0; offset < regs.line_length_in; offset += 16) {
                     memory_manager.ReadBlockUnsafe(regs.offset_in + offset, tmp_buffer.data(),
                                                    tmp_buffer.size());
-                    memory_manager.WriteBlock(
+                    memory_manager.WriteBlockCached(
                         convert_linear_2_blocklinear_addr(regs.offset_out + offset),
                         tmp_buffer.data(), tmp_buffer.size());
                 }
@@ -131,8 +142,8 @@ void MaxwellDMA::Launch() {
                     std::vector<u8> tmp_buffer(regs.line_length_in);
                     memory_manager.ReadBlockUnsafe(regs.offset_in, tmp_buffer.data(),
                                                    regs.line_length_in);
-                    memory_manager.WriteBlock(regs.offset_out, tmp_buffer.data(),
-                                              regs.line_length_in);
+                    memory_manager.WriteBlockCached(regs.offset_out, tmp_buffer.data(),
+                                                    regs.line_length_in);
                 }
             }
         }
@@ -194,7 +205,7 @@ void MaxwellDMA::CopyBlockLinearToPitch() {
                      src_params.origin.y, x_elements, regs.line_count, block_height, block_depth,
                      regs.pitch_out);
 
-    memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size);
+    memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
 }
 
 void MaxwellDMA::CopyPitchToBlockLinear() {
@@ -246,7 +257,7 @@ void MaxwellDMA::CopyPitchToBlockLinear() {
                    dst_params.origin.y, x_elements, regs.line_count, block_height, block_depth,
                    regs.pitch_in);
 
-    memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size);
+    memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
 }
 
 void MaxwellDMA::FastCopyBlockLinearToPitch() {
@@ -277,7 +288,7 @@ void MaxwellDMA::FastCopyBlockLinearToPitch() {
                      regs.src_params.block_size.height, regs.src_params.block_size.depth,
                      regs.pitch_out);
 
-    memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size);
+    memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
 }
 
 void MaxwellDMA::CopyBlockLinearToBlockLinear() {
@@ -337,7 +348,7 @@ void MaxwellDMA::CopyBlockLinearToBlockLinear() {
                    dst.depth, dst_x_offset, dst.origin.y, x_elements, regs.line_count,
                    dst.block_size.height, dst.block_size.depth, pitch);
 
-    memory_manager.WriteBlock(regs.offset_out, write_buffer.data(), dst_size);
+    memory_manager.WriteBlockCached(regs.offset_out, write_buffer.data(), dst_size);
 }
 
 void MaxwellDMA::ReleaseSemaphore() {

src/video_core/engines/maxwell_dma.h

@@ -231,6 +231,8 @@ private:
 
     void ReleaseSemaphore();
 
+    void ConsumeSinkImpl() override;
+
     Core::System& system;
 
     MemoryManager& memory_manager;