Revert "shader_recompiler/dead_code_elimination: Add DeadBranchElimination pass"

video_core: Fix legacy to generic location unpaired

CMakeLists.txt

@@ -541,9 +541,9 @@ add_definitions(-DBOOST_ERROR_CODE_HEADER_ONLY
 # Adjustments for MSVC + Ninja
 if (MSVC AND CMAKE_GENERATOR STREQUAL "Ninja")
     add_compile_options(
-        /wd4711 # function 'function' selected for automatic inline expansion
         /wd4464 # relative include path contains '..'
-        /wd4820 # 'identifier1': '4' bytes padding added after data member 'identifier2'
+        /wd4711 # function 'function' selected for automatic inline expansion
+        /wd4820 # 'bytes' bytes padding added after construct 'member_name'
     )
 endif()
 

src/CMakeLists.txt

@@ -58,13 +58,11 @@ if (MSVC)
 
         # Warnings
         /W3
-        /we4018 # 'expression': signed/unsigned mismatch
+        /WX
+
         /we4062 # Enumerator 'identifier' in a switch of enum 'enumeration' is not handled
-        /we4101 # 'identifier': unreferenced local variable
         /we4189 # 'identifier': local variable is initialized but not referenced
         /we4265 # 'class': class has virtual functions, but destructor is not virtual
-        /we4267 # 'var': conversion from 'size_t' to 'type', possible loss of data
-        /we4305 # 'context': truncation from 'type1' to 'type2'
         /we4388 # 'expression': signed/unsigned mismatch
         /we4389 # 'operator': signed/unsigned mismatch
         /we4456 # Declaration of 'identifier' hides previous local declaration
@@ -75,10 +73,13 @@ if (MSVC)
         /we4547 # 'operator': operator before comma has no effect; expected operator with side-effect
         /we4549 # 'operator1': operator before comma has no effect; did you intend 'operator2'?
         /we4555 # Expression has no effect; expected expression with side-effect
-        /we4715 # 'function': not all control paths return a value
-        /we4834 # Discarding return value of function with 'nodiscard' attribute
+        /we4826 # Conversion from 'type1' to 'type2' is sign-extended. This may cause unexpected runtime behavior.
         /we5038 # data member 'member1' will be initialized after data member 'member2'
+        /we5233 # explicit lambda capture 'identifier' is not used
         /we5245 # 'function': unreferenced function with internal linkage has been removed
+
+        /wd4100 # 'identifier': unreferenced formal parameter
+        /wd4324 # 'struct_name': structure was padded due to __declspec(align())
     )
 
     if (USE_CCACHE)
@@ -99,24 +100,18 @@ if (MSVC)
     set(CMAKE_EXE_LINKER_FLAGS_RELEASE "/DEBUG /MANIFEST:NO /INCREMENTAL:NO /OPT:REF,ICF" CACHE STRING "" FORCE)
 else()
     add_compile_options(
-        -Wall
-        -Werror=array-bounds
-        -Werror=implicit-fallthrough
+        -Werror=all
+        -Werror=extra
         -Werror=missing-declarations
-        -Werror=missing-field-initializers
-        -Werror=reorder
         -Werror=shadow
-        -Werror=sign-compare
-        -Werror=switch
-        -Werror=uninitialized
-        -Werror=unused-function
-        -Werror=unused-result
-        -Werror=unused-variable
-        -Wextra
-        -Wmissing-declarations
+        -Werror=unused
+
         -Wno-attributes
         -Wno-invalid-offsetof
         -Wno-unused-parameter
+
+        $<$<CXX_COMPILER_ID:Clang>:-Wno-braced-scalar-init>
+        $<$<CXX_COMPILER_ID:Clang>:-Wno-unused-private-field>
     )
 
     if (ARCHITECTURE_x86_64)

src/audio_core/CMakeLists.txt

@@ -206,20 +206,11 @@ if (MSVC)
         /we4244 # 'conversion': conversion from 'type1' to 'type2', possible loss of data
         /we4245 # 'conversion': conversion from 'type1' to 'type2', signed/unsigned mismatch
         /we4254 # 'operator': conversion from 'type1:field_bits' to 'type2:field_bits', possible loss of data
-        /we4456 # Declaration of 'identifier' hides previous local declaration
-        /we4457 # Declaration of 'identifier' hides function parameter
-        /we4458 # Declaration of 'identifier' hides class member
-        /we4459 # Declaration of 'identifier' hides global declaration
+        /we4800 # Implicit conversion from 'type' to bool. Possible information loss
     )
 else()
     target_compile_options(audio_core PRIVATE
         -Werror=conversion
-        -Werror=ignored-qualifiers
-        -Werror=shadow
-        -Werror=unused-variable
-
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-parameter>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-variable>
 
         -Wno-sign-conversion
     )

src/audio_core/renderer/behavior/info_updater.cpp

@@ -91,7 +91,7 @@ Result InfoUpdater::UpdateVoices(VoiceContext& voice_context,
             voice_info.Initialize();
 
             for (u32 channel = 0; channel < in_param.channel_count; channel++) {
-                std::memset(voice_states[channel], 0, sizeof(VoiceState));
+                *voice_states[channel] = {};
             }
         }
 

src/audio_core/renderer/command/effect/biquad_filter.cpp

@@ -94,7 +94,7 @@ void BiquadFilterCommand::Dump([[maybe_unused]] const ADSP::CommandListProcessor
 void BiquadFilterCommand::Process(const ADSP::CommandListProcessor& processor) {
     auto state_{reinterpret_cast<VoiceState::BiquadFilterState*>(state)};
     if (needs_init) {
-        std::memset(state_, 0, sizeof(VoiceState::BiquadFilterState));
+        *state_ = {};
     }
 
     auto input_buffer{

src/audio_core/renderer/command/effect/multi_tap_biquad_filter.cpp

@@ -30,7 +30,7 @@ void MultiTapBiquadFilterCommand::Process(const ADSP::CommandListProcessor& proc
     for (u32 i = 0; i < filter_tap_count; i++) {
         auto state{reinterpret_cast<VoiceState::BiquadFilterState*>(states[i])};
         if (needs_init[i]) {
-            std::memset(state, 0, sizeof(VoiceState::BiquadFilterState));
+            *state = {};
         }
 
         ApplyBiquadFilterFloat(output_buffer, input_buffer, biquads[i].b, biquads[i].a, *state,

src/common/CMakeLists.txt

@@ -156,12 +156,13 @@ if (MSVC)
   )
   target_compile_options(common PRIVATE
     /W4
-    /WX
+
+    /we4242 # 'identifier': conversion from 'type1' to 'type2', possible loss of data
+    /we4254 # 'operator': conversion from 'type1:field_bits' to 'type2:field_bits', possible loss of data
+    /we4800 # Implicit conversion from 'type' to bool. Possible information loss
   )
 else()
   target_compile_options(common PRIVATE
-    -Werror
-
     $<$<CXX_COMPILER_ID:Clang>:-fsized-deallocation>
   )
 endif()

src/common/bit_field.h

@@ -141,10 +141,6 @@ public:
     constexpr BitField(BitField&&) noexcept = default;
     constexpr BitField& operator=(BitField&&) noexcept = default;
 
-    [[nodiscard]] constexpr operator T() const {
-        return Value();
-    }
-
     constexpr void Assign(const T& value) {
 #ifdef _MSC_VER
         storage = static_cast<StorageType>((storage & ~mask) | FormatValue(value));
@@ -162,6 +158,17 @@ public:
         return ExtractValue(storage);
     }
 
+    template <typename ConvertedToType>
+    [[nodiscard]] constexpr ConvertedToType As() const {
+        static_assert(!std::is_same_v<T, ConvertedToType>,
+                      "Unnecessary cast. Use Value() instead.");
+        return static_cast<ConvertedToType>(Value());
+    }
+
+    [[nodiscard]] constexpr operator T() const {
+        return Value();
+    }
+
     [[nodiscard]] constexpr explicit operator bool() const {
         return Value() != 0;
     }

src/common/bounded_threadsafe_queue.h

@@ -21,11 +21,6 @@ constexpr size_t hardware_interference_size = std::hardware_destructive_interfer
 constexpr size_t hardware_interference_size = 64;
 #endif
 
-#ifdef _MSC_VER
-#pragma warning(push)
-#pragma warning(disable : 4324)
-#endif
-
 template <typename T, size_t capacity = 0x400>
 class MPSCQueue {
 public:
@@ -160,8 +155,4 @@ private:
     static_assert(std::is_nothrow_destructible_v<T>, "T must be nothrow destructible");
 };
 
-#ifdef _MSC_VER
-#pragma warning(pop)
-#endif
-
 } // namespace Common

src/core/CMakeLists.txt

@@ -774,19 +774,15 @@ if (MSVC)
         /we4244 # 'conversion': conversion from 'type1' to 'type2', possible loss of data
         /we4245 # 'conversion': conversion from 'type1' to 'type2', signed/unsigned mismatch
         /we4254 # 'operator': conversion from 'type1:field_bits' to 'type2:field_bits', possible loss of data
+        /we4800 # Implicit conversion from 'type' to bool. Possible information loss
     )
 else()
     target_compile_options(core PRIVATE
         -Werror=conversion
-        -Werror=ignored-qualifiers
-
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=class-memaccess>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-parameter>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-variable>
-
-        $<$<CXX_COMPILER_ID:Clang>:-fsized-deallocation>
 
         -Wno-sign-conversion
+
+        $<$<CXX_COMPILER_ID:Clang>:-fsized-deallocation>
     )
 endif()
 

src/core/file_sys/card_image.cpp

@@ -232,8 +232,8 @@ const std::vector<std::shared_ptr<NCA>>& XCI::GetNCAs() const {
 
 std::shared_ptr<NCA> XCI::GetNCAByType(NCAContentType type) const {
     const auto program_id = secure_partition->GetProgramTitleID();
-    const auto iter = std::find_if(
-        ncas.begin(), ncas.end(), [this, type, program_id](const std::shared_ptr<NCA>& nca) {
+    const auto iter =
+        std::find_if(ncas.begin(), ncas.end(), [type, program_id](const std::shared_ptr<NCA>& nca) {
             return nca->GetType() == type && nca->GetTitleId() == program_id;
         });
     return iter == ncas.end() ? nullptr : *iter;

src/core/file_sys/program_metadata.cpp

@@ -127,7 +127,7 @@ void ProgramMetadata::LoadManual(bool is_64_bit, ProgramAddressSpaceType address
 }
 
 bool ProgramMetadata::Is64BitProgram() const {
-    return npdm_header.has_64_bit_instructions;
+    return npdm_header.has_64_bit_instructions.As<bool>();
 }
 
 ProgramAddressSpaceType ProgramMetadata::GetAddressSpaceType() const {

src/core/hid/emulated_controller.cpp

@@ -1158,27 +1158,27 @@ bool EmulatedController::IsControllerSupported(bool use_temporary_value) const {
     const auto type = is_configuring && use_temporary_value ? tmp_npad_type : npad_type;
     switch (type) {
     case NpadStyleIndex::ProController:
-        return supported_style_tag.fullkey;
+        return supported_style_tag.fullkey.As<bool>();
     case NpadStyleIndex::Handheld:
-        return supported_style_tag.handheld;
+        return supported_style_tag.handheld.As<bool>();
     case NpadStyleIndex::JoyconDual:
-        return supported_style_tag.joycon_dual;
+        return supported_style_tag.joycon_dual.As<bool>();
     case NpadStyleIndex::JoyconLeft:
-        return supported_style_tag.joycon_left;
+        return supported_style_tag.joycon_left.As<bool>();
     case NpadStyleIndex::JoyconRight:
-        return supported_style_tag.joycon_right;
+        return supported_style_tag.joycon_right.As<bool>();
     case NpadStyleIndex::GameCube:
-        return supported_style_tag.gamecube;
+        return supported_style_tag.gamecube.As<bool>();
     case NpadStyleIndex::Pokeball:
-        return supported_style_tag.palma;
+        return supported_style_tag.palma.As<bool>();
     case NpadStyleIndex::NES:
-        return supported_style_tag.lark;
+        return supported_style_tag.lark.As<bool>();
     case NpadStyleIndex::SNES:
-        return supported_style_tag.lucia;
+        return supported_style_tag.lucia.As<bool>();
     case NpadStyleIndex::N64:
-        return supported_style_tag.lagoon;
+        return supported_style_tag.lagoon.As<bool>();
     case NpadStyleIndex::SegaGenesis:
-        return supported_style_tag.lager;
+        return supported_style_tag.lager.As<bool>();
     default:
         return false;
     }

src/core/hle/ipc_helpers.h

@@ -406,7 +406,7 @@ inline s32 RequestParser::Pop() {
 }
 
 // Ignore the -Wclass-memaccess warning on memcpy for non-trivially default constructible objects.
-#if defined(__GNUC__)
+#if defined(__GNUC__) && !defined(__clang__) && !defined(__INTEL_COMPILER)
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wclass-memaccess"
 #endif
@@ -417,7 +417,7 @@ void RequestParser::PopRaw(T& value) {
     std::memcpy(&value, cmdbuf + index, sizeof(T));
     index += (sizeof(T) + 3) / 4; // round up to word length
 }
-#if defined(__GNUC__)
+#if defined(__GNUC__) && !defined(__clang__) && !defined(__INTEL_COMPILER)
 #pragma GCC diagnostic pop
 #endif
 

src/core/hle/kernel/k_session_request.h

@@ -3,6 +3,8 @@
 
 #pragma once
 
+#include <array>
+
 #include "core/hle/kernel/k_auto_object.h"
 #include "core/hle/kernel/k_event.h"
 #include "core/hle/kernel/k_memory_block.h"
@@ -52,8 +54,7 @@ public:
         };
 
     public:
-        explicit SessionMappings(KernelCore& kernel_)
-            : kernel(kernel_), m_mappings(nullptr), m_num_send(), m_num_recv(), m_num_exch() {}
+        explicit SessionMappings(KernelCore& kernel_) : kernel(kernel_) {}
 
         void Initialize() {}
         void Finalize();
@@ -149,17 +150,15 @@ public:
 
     private:
         KernelCore& kernel;
-        Mapping m_static_mappings[NumStaticMappings];
-        Mapping* m_mappings;
-        u8 m_num_send;
-        u8 m_num_recv;
-        u8 m_num_exch;
+        std::array<Mapping, NumStaticMappings> m_static_mappings;
+        Mapping* m_mappings{};
+        u8 m_num_send{};
+        u8 m_num_recv{};
+        u8 m_num_exch{};
     };
 
 public:
-    explicit KSessionRequest(KernelCore& kernel_)
-        : KAutoObject(kernel_), m_mappings(kernel_), m_thread(nullptr), m_server(nullptr),
-          m_event(nullptr) {}
+    explicit KSessionRequest(KernelCore& kernel_) : KAutoObject(kernel_), m_mappings(kernel_) {}
 
     static KSessionRequest* Create(KernelCore& kernel) {
         KSessionRequest* req = KSessionRequest::Allocate(kernel);
@@ -281,7 +280,7 @@ public:
 
 private:
     // NOTE: This is public and virtual in Nintendo's kernel.
-    void Finalize() {
+    void Finalize() override {
         m_mappings.Finalize();
 
         if (m_thread) {
@@ -297,11 +296,11 @@ private:
 
 private:
     SessionMappings m_mappings;
-    KThread* m_thread;
-    KProcess* m_server;
-    KEvent* m_event;
-    uintptr_t m_address;
-    size_t m_size;
+    KThread* m_thread{};
+    KProcess* m_server{};
+    KEvent* m_event{};
+    uintptr_t m_address{};
+    size_t m_size{};
 };
 
 } // namespace Kernel

src/core/hle/kernel/svc.cpp

@@ -751,8 +751,8 @@ static void Break(Core::System& system, u32 reason, u64 info1, u64 info2) {
     }
 
     system.GetReporter().SaveSvcBreakReport(
-        static_cast<u32>(break_reason.break_type.Value()), break_reason.signal_debugger, info1,
-        info2, has_dumped_buffer ? std::make_optional(debug_buffer) : std::nullopt);
+        static_cast<u32>(break_reason.break_type.Value()), break_reason.signal_debugger.As<bool>(),
+        info1, info2, has_dumped_buffer ? std::make_optional(debug_buffer) : std::nullopt);
 
     if (!break_reason.signal_debugger) {
         LOG_CRITICAL(

src/core/hle/service/am/applets/applets.h

@@ -164,7 +164,7 @@ protected:
         u32_le size;
         u32_le library_version;
         u32_le theme_color;
-        u8 play_startup_sound;
+        bool play_startup_sound;
         u64_le system_tick;
     };
     static_assert(sizeof(CommonArguments) == 0x20, "CommonArguments has incorrect size.");

src/core/hle/service/hid/controllers/npad.cpp

@@ -745,8 +745,9 @@ void Controller_NPad::SetSupportedNpadIdTypes(u8* data, std::size_t length) {
 }
 
 void Controller_NPad::GetSupportedNpadIdTypes(u32* data, std::size_t max_length) {
-    ASSERT(max_length < supported_npad_id_types.size());
-    std::memcpy(data, supported_npad_id_types.data(), supported_npad_id_types.size());
+    const auto copy_amount = supported_npad_id_types.size() * sizeof(u32);
+    ASSERT(max_length <= copy_amount);
+    std::memcpy(data, supported_npad_id_types.data(), copy_amount);
 }
 
 std::size_t Controller_NPad::GetSupportedNpadIdTypesSize() const {
@@ -1501,25 +1502,25 @@ bool Controller_NPad::IsControllerSupported(Core::HID::NpadStyleIndex controller
         Core::HID::NpadStyleTag style = GetSupportedStyleSet();
         switch (controller) {
         case Core::HID::NpadStyleIndex::ProController:
-            return style.fullkey;
+            return style.fullkey.As<bool>();
         case Core::HID::NpadStyleIndex::JoyconDual:
-            return style.joycon_dual;
+            return style.joycon_dual.As<bool>();
         case Core::HID::NpadStyleIndex::JoyconLeft:
-            return style.joycon_left;
+            return style.joycon_left.As<bool>();
         case Core::HID::NpadStyleIndex::JoyconRight:
-            return style.joycon_right;
+            return style.joycon_right.As<bool>();
         case Core::HID::NpadStyleIndex::GameCube:
-            return style.gamecube;
+            return style.gamecube.As<bool>();
         case Core::HID::NpadStyleIndex::Pokeball:
-            return style.palma;
+            return style.palma.As<bool>();
         case Core::HID::NpadStyleIndex::NES:
-            return style.lark;
+            return style.lark.As<bool>();
         case Core::HID::NpadStyleIndex::SNES:
-            return style.lucia;
+            return style.lucia.As<bool>();
         case Core::HID::NpadStyleIndex::N64:
-            return style.lagoon;
+            return style.lagoon.As<bool>();
         case Core::HID::NpadStyleIndex::SegaGenesis:
-            return style.lager;
+            return style.lager.As<bool>();
         default:
             return false;
         }

src/core/hle/service/nvdrv/nvdrv.cpp

@@ -53,7 +53,7 @@ void InstallInterfaces(SM::ServiceManager& service_manager, NVFlinger::NVFlinger
 }
 
 Module::Module(Core::System& system)
-    : service_context{system, "nvdrv"}, events_interface{*this}, container{system.Host1x()} {
+    : container{system.Host1x()}, service_context{system, "nvdrv"}, events_interface{*this} {
     builders["/dev/nvhost-as-gpu"] = [this, &system](DeviceFD fd) {
         std::shared_ptr<Devices::nvdevice> device =
             std::make_shared<Devices::nvhost_as_gpu>(system, *this, container);

src/core/hle/service/nvdrv/nvdrv.h

@@ -97,6 +97,9 @@ private:
     friend class EventInterface;
     friend class Service::NVFlinger::NVFlinger;
 
+    /// Manages syncpoints on the host
+    NvCore::Container container;
+
     /// Id to use for the next open file descriptor.
     DeviceFD next_fd = 1;
 
@@ -108,9 +111,6 @@ private:
 
     EventInterface events_interface;
 
-    /// Manages syncpoints on the host
-    NvCore::Container container;
-
     std::unordered_map<std::string, std::function<FilesContainerType::iterator(DeviceFD)>> builders;
 };
 

src/core/memory.cpp

@@ -233,18 +233,17 @@ struct Memory::Impl {
                           current_vaddr, src_addr, size);
                 std::memset(dest_buffer, 0, copy_amount);
             },
-            [&dest_buffer](const std::size_t copy_amount, const u8* const src_ptr) {
+            [&](const std::size_t copy_amount, const u8* const src_ptr) {
                 std::memcpy(dest_buffer, src_ptr, copy_amount);
             },
-            [&system = system, &dest_buffer](const VAddr current_vaddr,
-                                             const std::size_t copy_amount,
-                                             const u8* const host_ptr) {
+            [&](const VAddr current_vaddr, const std::size_t copy_amount,
+                const u8* const host_ptr) {
                 if constexpr (!UNSAFE) {
                     system.GPU().FlushRegion(current_vaddr, copy_amount);
                 }
                 std::memcpy(dest_buffer, host_ptr, copy_amount);
             },
-            [&dest_buffer](const std::size_t copy_amount) {
+            [&](const std::size_t copy_amount) {
                 dest_buffer = static_cast<u8*>(dest_buffer) + copy_amount;
             });
     }
@@ -267,17 +266,16 @@ struct Memory::Impl {
                           "Unmapped WriteBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
                           current_vaddr, dest_addr, size);
             },
-            [&src_buffer](const std::size_t copy_amount, u8* const dest_ptr) {
+            [&](const std::size_t copy_amount, u8* const dest_ptr) {
                 std::memcpy(dest_ptr, src_buffer, copy_amount);
             },
-            [&system = system, &src_buffer](const VAddr current_vaddr,
-                                            const std::size_t copy_amount, u8* const host_ptr) {
+            [&](const VAddr current_vaddr, const std::size_t copy_amount, u8* const host_ptr) {
                 if constexpr (!UNSAFE) {
                     system.GPU().InvalidateRegion(current_vaddr, copy_amount);
                 }
                 std::memcpy(host_ptr, src_buffer, copy_amount);
             },
-            [&src_buffer](const std::size_t copy_amount) {
+            [&](const std::size_t copy_amount) {
                 src_buffer = static_cast<const u8*>(src_buffer) + copy_amount;
             });
     }
@@ -301,8 +299,7 @@ struct Memory::Impl {
             [](const std::size_t copy_amount, u8* const dest_ptr) {
                 std::memset(dest_ptr, 0, copy_amount);
             },
-            [&system = system](const VAddr current_vaddr, const std::size_t copy_amount,
-                               u8* const host_ptr) {
+            [&](const VAddr current_vaddr, const std::size_t copy_amount, u8* const host_ptr) {
                 system.GPU().InvalidateRegion(current_vaddr, copy_amount);
                 std::memset(host_ptr, 0, copy_amount);
             },
@@ -313,22 +310,20 @@ struct Memory::Impl {
                    const std::size_t size) {
         WalkBlock(
             process, dest_addr, size,
-            [this, &process, &dest_addr, &src_addr, size](const std::size_t copy_amount,
-                                                          const VAddr current_vaddr) {
+            [&](const std::size_t copy_amount, const VAddr current_vaddr) {
                 LOG_ERROR(HW_Memory,
                           "Unmapped CopyBlock @ 0x{:016X} (start address = 0x{:016X}, size = {})",
                           current_vaddr, src_addr, size);
                 ZeroBlock(process, dest_addr, copy_amount);
             },
-            [this, &process, &dest_addr](const std::size_t copy_amount, const u8* const src_ptr) {
+            [&](const std::size_t copy_amount, const u8* const src_ptr) {
                 WriteBlockImpl<false>(process, dest_addr, src_ptr, copy_amount);
             },
-            [this, &system = system, &process, &dest_addr](
-                const VAddr current_vaddr, const std::size_t copy_amount, u8* const host_ptr) {
+            [&](const VAddr current_vaddr, const std::size_t copy_amount, u8* const host_ptr) {
                 system.GPU().FlushRegion(current_vaddr, copy_amount);
                 WriteBlockImpl<false>(process, dest_addr, host_ptr, copy_amount);
             },
-            [&dest_addr, &src_addr](const std::size_t copy_amount) {
+            [&](const std::size_t copy_amount) {
                 dest_addr += static_cast<VAddr>(copy_amount);
                 src_addr += static_cast<VAddr>(copy_amount);
             });
@@ -575,7 +570,7 @@ struct Memory::Impl {
             [vaddr]() {
                 LOG_ERROR(HW_Memory, "Unmapped Read{} @ 0x{:016X}", sizeof(T) * 8, vaddr);
             },
-            [&system = system, vaddr]() { system.GPU().FlushRegion(vaddr, sizeof(T)); });
+            [&]() { system.GPU().FlushRegion(vaddr, sizeof(T)); });
         if (ptr) {
             std::memcpy(&result, ptr, sizeof(T));
         }
@@ -599,7 +594,7 @@ struct Memory::Impl {
                 LOG_ERROR(HW_Memory, "Unmapped Write{} @ 0x{:016X} = 0x{:016X}", sizeof(T) * 8,
                           vaddr, static_cast<u64>(data));
             },
-            [&system = system, vaddr]() { system.GPU().InvalidateRegion(vaddr, sizeof(T)); });
+            [&]() { system.GPU().InvalidateRegion(vaddr, sizeof(T)); });
         if (ptr) {
             std::memcpy(ptr, &data, sizeof(T));
         }
@@ -613,7 +608,7 @@ struct Memory::Impl {
                 LOG_ERROR(HW_Memory, "Unmapped WriteExclusive{} @ 0x{:016X} = 0x{:016X}",
                           sizeof(T) * 8, vaddr, static_cast<u64>(data));
             },
-            [&system = system, vaddr]() { system.GPU().InvalidateRegion(vaddr, sizeof(T)); });
+            [&]() { system.GPU().InvalidateRegion(vaddr, sizeof(T)); });
         if (ptr) {
             const auto volatile_pointer = reinterpret_cast<volatile T*>(ptr);
             return Common::AtomicCompareAndSwap(volatile_pointer, data, expected);
@@ -628,7 +623,7 @@ struct Memory::Impl {
                 LOG_ERROR(HW_Memory, "Unmapped WriteExclusive128 @ 0x{:016X} = 0x{:016X}{:016X}",
                           vaddr, static_cast<u64>(data[1]), static_cast<u64>(data[0]));
             },
-            [&system = system, vaddr]() { system.GPU().InvalidateRegion(vaddr, sizeof(u128)); });
+            [&]() { system.GPU().InvalidateRegion(vaddr, sizeof(u128)); });
         if (ptr) {
             const auto volatile_pointer = reinterpret_cast<volatile u64*>(ptr);
             return Common::AtomicCompareAndSwap(volatile_pointer, data, expected);

src/input_common/CMakeLists.txt

@@ -39,21 +39,14 @@ add_library(input_common STATIC
 if (MSVC)
     target_compile_options(input_common PRIVATE
         /W4
-        /WX
 
         /we4242 # 'identifier': conversion from 'type1' to 'type2', possible loss of data
-        /we4244 # 'conversion': conversion from 'type1' to 'type2', possible loss of data
-        /we4245 # 'conversion': conversion from 'type1' to 'type2', signed/unsigned mismatch
         /we4254 # 'operator': conversion from 'type1:field_bits' to 'type2:field_bits', possible loss of data
+        /we4800 # Implicit conversion from 'type' to bool. Possible information loss
     )
 else()
     target_compile_options(input_common PRIVATE
-        -Werror
         -Werror=conversion
-        -Werror=ignored-qualifiers
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-parameter>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-variable>
-        -Werror=unused-variable
     )
 endif()
 

src/input_common/drivers/sdl_driver.cpp

@@ -40,8 +40,8 @@ public:
     void EnableMotion() {
         if (sdl_controller) {
             SDL_GameController* controller = sdl_controller.get();
-            has_accel = SDL_GameControllerHasSensor(controller, SDL_SENSOR_ACCEL);
-            has_gyro = SDL_GameControllerHasSensor(controller, SDL_SENSOR_GYRO);
+            has_accel = SDL_GameControllerHasSensor(controller, SDL_SENSOR_ACCEL) == SDL_TRUE;
+            has_gyro = SDL_GameControllerHasSensor(controller, SDL_SENSOR_GYRO) == SDL_TRUE;
             if (has_accel) {
                 SDL_GameControllerSetSensorEnabled(controller, SDL_SENSOR_ACCEL, SDL_TRUE);
             }

src/input_common/input_poller.cpp

@@ -797,8 +797,8 @@ std::unique_ptr<Common::Input::InputDevice> InputFactory::CreateButtonDevice(
 
     const auto button_id = params.Get("button", 0);
     const auto keyboard_key = params.Get("code", 0);
-    const auto toggle = params.Get("toggle", false);
-    const auto inverted = params.Get("inverted", false);
+    const auto toggle = params.Get("toggle", false) != 0;
+    const auto inverted = params.Get("inverted", false) != 0;
     input_engine->PreSetController(identifier);
     input_engine->PreSetButton(identifier, button_id);
     input_engine->PreSetButton(identifier, keyboard_key);
@@ -820,8 +820,8 @@ std::unique_ptr<Common::Input::InputDevice> InputFactory::CreateHatButtonDevice(
 
     const auto button_id = params.Get("hat", 0);
     const auto direction = input_engine->GetHatButtonId(params.Get("direction", ""));
-    const auto toggle = params.Get("toggle", false);
-    const auto inverted = params.Get("inverted", false);
+    const auto toggle = params.Get("toggle", false) != 0;
+    const auto inverted = params.Get("inverted", false) != 0;
 
     input_engine->PreSetController(identifier);
     input_engine->PreSetHatButton(identifier, button_id);
@@ -879,7 +879,7 @@ std::unique_ptr<Common::Input::InputDevice> InputFactory::CreateAnalogDevice(
         .threshold = std::clamp(params.Get("threshold", 0.5f), 0.0f, 1.0f),
         .offset = std::clamp(params.Get("offset", 0.0f), -1.0f, 1.0f),
         .inverted = params.Get("invert", "+") == "-",
-        .toggle = static_cast<bool>(params.Get("toggle", false)),
+        .toggle = params.Get("toggle", false) != 0,
     };
     input_engine->PreSetController(identifier);
     input_engine->PreSetAxis(identifier, axis);
@@ -895,8 +895,8 @@ std::unique_ptr<Common::Input::InputDevice> InputFactory::CreateTriggerDevice(
     };
 
     const auto button = params.Get("button", 0);
-    const auto toggle = params.Get("toggle", false);
-    const auto inverted = params.Get("inverted", false);
+    const auto toggle = params.Get("toggle", false) != 0;
+    const auto inverted = params.Get("inverted", false) != 0;
 
     const auto axis = params.Get("axis", 0);
     const Common::Input::AnalogProperties properties = {
@@ -926,8 +926,8 @@ std::unique_ptr<Common::Input::InputDevice> InputFactory::CreateTouchDevice(
     };
 
     const auto button = params.Get("button", 0);
-    const auto toggle = params.Get("toggle", false);
-    const auto inverted = params.Get("inverted", false);
+    const auto toggle = params.Get("toggle", false) != 0;
+    const auto inverted = params.Get("inverted", false) != 0;
 
     const auto axis_x = params.Get("axis_x", 0);
     const Common::Input::AnalogProperties properties_x = {

src/shader_recompiler/CMakeLists.txt

@@ -241,24 +241,14 @@ target_link_libraries(shader_recompiler PUBLIC common fmt::fmt sirit)
 if (MSVC)
     target_compile_options(shader_recompiler PRIVATE
         /W4
-        /WX
-        /we4018 # 'expression' : signed/unsigned mismatch
-        /we4244 # 'argument' : conversion from 'type1' to 'type2', possible loss of data (floating-point)
-        /we4245 # 'conversion' : conversion from 'type1' to 'type2', signed/unsigned mismatch
+
+        /we4242 # 'identifier': conversion from 'type1' to 'type2', possible loss of data
         /we4254 # 'operator': conversion from 'type1:field_bits' to 'type2:field_bits', possible loss of data
-        /we4267 # 'var' : conversion from 'size_t' to 'type', possible loss of data
-        /we4305 # 'context' : truncation from 'type1' to 'type2'
         /we4800 # Implicit conversion from 'type' to bool. Possible information loss
-        /we4826 # Conversion from 'type1' to 'type2' is sign-extended. This may cause unexpected runtime behavior.
     )
 else()
     target_compile_options(shader_recompiler PRIVATE
-        -Werror
         -Werror=conversion
-        -Werror=ignored-qualifiers
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-parameter>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-variable>
-        -Werror=unused-variable
 
         # Bracket depth determines maximum size of a fold expression in Clang since 9c9974c3ccb6.
         # And this in turns limits the size of a std::array.

src/shader_recompiler/backend/glasm/emit_glasm_not_implemented.cpp

@@ -5,10 +5,6 @@
 #include "shader_recompiler/backend/glasm/glasm_emit_context.h"
 #include "shader_recompiler/frontend/ir/value.h"
 
-#ifdef _MSC_VER
-#pragma warning(disable : 4100)
-#endif
-
 namespace Shader::Backend::GLASM {
 
 #define NotImplemented() throw NotImplementedException("GLASM instruction {}", __LINE__)

src/shader_recompiler/backend/glsl/emit_glsl_not_implemented.cpp

@@ -7,10 +7,6 @@
 #include "shader_recompiler/backend/glsl/glsl_emit_context.h"
 #include "shader_recompiler/frontend/ir/value.h"
 
-#ifdef _MSC_VER
-#pragma warning(disable : 4100)
-#endif
-
 namespace Shader::Backend::GLSL {
 
 void EmitGetRegister(EmitContext& ctx) {

src/shader_recompiler/frontend/ir/microinstruction.cpp

@@ -325,11 +325,6 @@ void Inst::AddPhiOperand(Block* predecessor, const Value& value) {
     phi_args.emplace_back(predecessor, value);
 }
 
-void Inst::ErasePhiOperand(size_t index) {
-    const auto operand_it{phi_args.begin() + static_cast<ptrdiff_t>(index)};
-    phi_args.erase(operand_it);
-}
-
 void Inst::OrderPhiArgs() {
     if (op != Opcode::Phi) {
         throw LogicError("{} is not a Phi instruction", op);

src/shader_recompiler/frontend/ir/value.h

@@ -178,13 +178,9 @@ public:
 
     /// Get a pointer to the block of a phi argument.
     [[nodiscard]] Block* PhiBlock(size_t index) const;
-
     /// Add phi operand to a phi instruction.
     void AddPhiOperand(Block* predecessor, const Value& value);
 
-    // Erase the phi operand at the given index.
-    void ErasePhiOperand(size_t index);
-
     /// Orders the Phi arguments from farthest away to nearest.
     void OrderPhiArgs();
 

src/shader_recompiler/frontend/maxwell/translate_program.cpp

@@ -137,28 +137,35 @@ bool IsLegacyAttribute(IR::Attribute attribute) {
 }
 
 std::map<IR::Attribute, IR::Attribute> GenerateLegacyToGenericMappings(
-    const VaryingState& state, std::queue<IR::Attribute> ununsed_generics) {
+    const VaryingState& state, std::queue<IR::Attribute> unused_generics,
+    const std::map<IR::Attribute, IR::Attribute>& previous_stage_mapping) {
     std::map<IR::Attribute, IR::Attribute> mapping;
+    auto update_mapping = [&mapping, &unused_generics, previous_stage_mapping](IR::Attribute attr,
+                                                                               size_t count) {
+        if (previous_stage_mapping.find(attr) != previous_stage_mapping.end()) {
+            for (size_t i = 0; i < count; ++i) {
+                mapping.insert({attr + i, previous_stage_mapping.at(attr + i)});
+            }
+        } else {
+            for (size_t i = 0; i < count; ++i) {
+                mapping.insert({attr + i, unused_generics.front() + i});
+            }
+            unused_generics.pop();
+        }
+    };
     for (size_t index = 0; index < 4; ++index) {
         auto attr = IR::Attribute::ColorFrontDiffuseR + index * 4;
         if (state.AnyComponent(attr)) {
-            for (size_t i = 0; i < 4; ++i) {
-                mapping.insert({attr + i, ununsed_generics.front() + i});
-            }
-            ununsed_generics.pop();
+            update_mapping(attr, 4);
         }
     }
     if (state[IR::Attribute::FogCoordinate]) {
-        mapping.insert({IR::Attribute::FogCoordinate, ununsed_generics.front()});
-        ununsed_generics.pop();
+        update_mapping(IR::Attribute::FogCoordinate, 1);
     }
     for (size_t index = 0; index < IR::NUM_FIXEDFNCTEXTURE; ++index) {
         auto attr = IR::Attribute::FixedFncTexture0S + index * 4;
         if (state.AnyComponent(attr)) {
-            for (size_t i = 0; i < 4; ++i) {
-                mapping.insert({attr + i, ununsed_generics.front() + i});
-            }
-            ununsed_generics.pop();
+            update_mapping(attr, 4);
         }
     }
     return mapping;
@@ -265,21 +272,22 @@ IR::Program MergeDualVertexPrograms(IR::Program& vertex_a, IR::Program& vertex_b
 void ConvertLegacyToGeneric(IR::Program& program, const Shader::RuntimeInfo& runtime_info) {
     auto& stores = program.info.stores;
     if (stores.Legacy()) {
-        std::queue<IR::Attribute> ununsed_output_generics{};
+        std::queue<IR::Attribute> unused_output_generics{};
         for (size_t index = 0; index < IR::NUM_GENERICS; ++index) {
             if (!stores.Generic(index)) {
-                ununsed_output_generics.push(IR::Attribute::Generic0X + index * 4);
+                unused_output_generics.push(IR::Attribute::Generic0X + index * 4);
             }
         }
-        auto mappings = GenerateLegacyToGenericMappings(stores, ununsed_output_generics);
+        program.info.legacy_stores_mapping =
+            GenerateLegacyToGenericMappings(stores, unused_output_generics, {});
         for (IR::Block* const block : program.post_order_blocks) {
             for (IR::Inst& inst : block->Instructions()) {
                 switch (inst.GetOpcode()) {
                 case IR::Opcode::SetAttribute: {
                     const auto attr = inst.Arg(0).Attribute();
                     if (IsLegacyAttribute(attr)) {
-                        stores.Set(mappings[attr], true);
-                        inst.SetArg(0, Shader::IR::Value(mappings[attr]));
+                        stores.Set(program.info.legacy_stores_mapping[attr], true);
+                        inst.SetArg(0, Shader::IR::Value(program.info.legacy_stores_mapping[attr]));
                     }
                     break;
                 }
@@ -292,15 +300,16 @@ void ConvertLegacyToGeneric(IR::Program& program, const Shader::RuntimeInfo& run
 
     auto& loads = program.info.loads;
     if (loads.Legacy()) {
-        std::queue<IR::Attribute> ununsed_input_generics{};
+        std::queue<IR::Attribute> unused_input_generics{};
         for (size_t index = 0; index < IR::NUM_GENERICS; ++index) {
             const AttributeType input_type{runtime_info.generic_input_types[index]};
             if (!runtime_info.previous_stage_stores.Generic(index) || !loads.Generic(index) ||
                 input_type == AttributeType::Disabled) {
-                ununsed_input_generics.push(IR::Attribute::Generic0X + index * 4);
+                unused_input_generics.push(IR::Attribute::Generic0X + index * 4);
             }
         }
-        auto mappings = GenerateLegacyToGenericMappings(loads, ununsed_input_generics);
+        auto mappings = GenerateLegacyToGenericMappings(
+            loads, unused_input_generics, runtime_info.previous_stage_legacy_stores_mapping);
         for (IR::Block* const block : program.post_order_blocks) {
             for (IR::Inst& inst : block->Instructions()) {
                 switch (inst.GetOpcode()) {

src/shader_recompiler/ir_opt/dead_code_elimination_pass.cpp

@@ -1,104 +1,24 @@
 // SPDX-FileCopyrightText: Copyright 2021 yuzu Emulator Project
 // SPDX-License-Identifier: GPL-2.0-or-later
 
-#include <algorithm>
-
-#include <boost/container/small_vector.hpp>
-
 #include "shader_recompiler/frontend/ir/basic_block.h"
 #include "shader_recompiler/frontend/ir/value.h"
 #include "shader_recompiler/ir_opt/passes.h"
 
 namespace Shader::Optimization {
-namespace {
-template <bool TEST_USES>
-void DeadInstElimination(IR::Block* const block) {
-    // We iterate over the instructions in reverse order.
-    // This is because removing an instruction reduces the number of uses for earlier instructions.
-    auto it{block->end()};
-    while (it != block->begin()) {
-        --it;
-        if constexpr (TEST_USES) {
-            if (it->HasUses() || it->MayHaveSideEffects()) {
-                continue;
-            }
-        }
-        it->Invalidate();
-        it = block->Instructions().erase(it);
-    }
-}
-
-void DeletedPhiArgElimination(IR::Program& program, std::span<const IR::Block*> dead_blocks) {
-    for (IR::Block* const block : program.blocks) {
-        for (IR::Inst& phi : *block) {
-            if (!IR::IsPhi(phi)) {
-                continue;
-            }
-            for (size_t i = 0; i < phi.NumArgs(); ++i) {
-                if (std::ranges::find(dead_blocks, phi.PhiBlock(i)) == dead_blocks.end()) {
-                    continue;
-                }
-                // Phi operand at this index is an unreachable block
-                phi.ErasePhiOperand(i);
-                --i;
-            }
-        }
-    }
-}
-
-void DeadBranchElimination(IR::Program& program) {
-    boost::container::small_vector<const IR::Block*, 3> dead_blocks;
-    const auto begin_it{program.syntax_list.begin()};
-    for (auto node_it = begin_it; node_it != program.syntax_list.end(); ++node_it) {
-        if (node_it->type != IR::AbstractSyntaxNode::Type::If) {
-            continue;
-        }
-        IR::Inst* const cond_ref{node_it->data.if_node.cond.Inst()};
-        const IR::U1 cond{cond_ref->Arg(0)};
-        if (!cond.IsImmediate()) {
-            continue;
-        }
-        if (cond.U1()) {
-            continue;
-        }
-        // False immediate condition. Remove condition ref, erase the entire branch.
-        cond_ref->Invalidate();
-        // Account for nested if-statements within the if(false) branch
-        u32 nested_ifs{1u};
-        while (node_it->type != IR::AbstractSyntaxNode::Type::EndIf || nested_ifs > 0) {
-            node_it = program.syntax_list.erase(node_it);
-            switch (node_it->type) {
-            case IR::AbstractSyntaxNode::Type::If:
-                ++nested_ifs;
-                break;
-            case IR::AbstractSyntaxNode::Type::EndIf:
-                --nested_ifs;
-                break;
-            case IR::AbstractSyntaxNode::Type::Block: {
-                IR::Block* const block{node_it->data.block};
-                DeadInstElimination<false>(block);
-                dead_blocks.push_back(block);
-                break;
-            }
-            default:
-                break;
-            }
-        }
-        // Erase EndIf node of the if(false) branch
-        node_it = program.syntax_list.erase(node_it);
-        // Account for loop increment
-        --node_it;
-    }
-    if (!dead_blocks.empty()) {
-        DeletedPhiArgElimination(program, std::span(dead_blocks.data(), dead_blocks.size()));
-    }
-}
-} // namespace
 
 void DeadCodeEliminationPass(IR::Program& program) {
-    DeadBranchElimination(program);
+    // We iterate over the instructions in reverse order.
+    // This is because removing an instruction reduces the number of uses for earlier instructions.
     for (IR::Block* const block : program.post_order_blocks) {
-        DeadInstElimination<true>(block);
+        auto it{block->end()};
+        while (it != block->begin()) {
+            --it;
+            if (!it->HasUses() && !it->MayHaveSideEffects()) {
+                it->Invalidate();
+                it = block->Instructions().erase(it);
+            }
+        }
     }
 }
 

src/shader_recompiler/runtime_info.h

@@ -4,6 +4,7 @@
 #pragma once
 
 #include <array>
+#include <map>
 #include <optional>
 #include <vector>
 
@@ -60,6 +61,7 @@ struct TransformFeedbackVarying {
 struct RuntimeInfo {
     std::array<AttributeType, 32> generic_input_types{};
     VaryingState previous_stage_stores;
+    std::map<IR::Attribute, IR::Attribute> previous_stage_legacy_stores_mapping;
 
     bool convert_depth_mode{};
     bool force_early_z{};

src/shader_recompiler/shader_info.h

@@ -5,6 +5,7 @@
 
 #include <array>
 #include <bitset>
+#include <map>
 
 #include "common/common_types.h"
 #include "shader_recompiler/frontend/ir/type.h"
@@ -127,6 +128,8 @@ struct Info {
     VaryingState stores;
     VaryingState passthrough;
 
+    std::map<IR::Attribute, IR::Attribute> legacy_stores_mapping;
+
     bool loads_indexed_attributes{};
 
     std::array<bool, 8> stores_frag_color{};

src/video_core/CMakeLists.txt

@@ -279,14 +279,8 @@ if (MSVC)
 else()
     target_compile_options(video_core PRIVATE
         -Werror=conversion
-        -Wno-error=sign-conversion
-        -Werror=pessimizing-move
-        -Werror=redundant-move
-        -Werror=type-limits
 
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=class-memaccess>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-parameter>
-        $<$<CXX_COMPILER_ID:GNU>:-Werror=unused-but-set-variable>
+        -Wno-sign-conversion
     )
 endif()
 

src/video_core/engines/maxwell_3d.cpp

@@ -117,18 +117,10 @@ void Maxwell3D::InitializeRegisterDefaults() {
 
     shadow_state = regs;
 
-    draw_command[MAXWELL3D_REG_INDEX(draw.end)] = true;
-    draw_command[MAXWELL3D_REG_INDEX(draw.begin)] = true;
-    draw_command[MAXWELL3D_REG_INDEX(vertex_buffer.first)] = true;
-    draw_command[MAXWELL3D_REG_INDEX(vertex_buffer.count)] = true;
-    draw_command[MAXWELL3D_REG_INDEX(index_buffer.first)] = true;
-    draw_command[MAXWELL3D_REG_INDEX(index_buffer.count)] = true;
-    draw_command[MAXWELL3D_REG_INDEX(index_buffer32_first)] = true;
-    draw_command[MAXWELL3D_REG_INDEX(index_buffer16_first)] = true;
-    draw_command[MAXWELL3D_REG_INDEX(index_buffer8_first)] = true;
-    draw_command[MAXWELL3D_REG_INDEX(draw_inline_index)] = true;
-    draw_command[MAXWELL3D_REG_INDEX(inline_index_2x16.even)] = true;
-    draw_command[MAXWELL3D_REG_INDEX(inline_index_4x8.index0)] = true;
+    mme_inline[MAXWELL3D_REG_INDEX(draw.end)] = true;
+    mme_inline[MAXWELL3D_REG_INDEX(draw.begin)] = true;
+    mme_inline[MAXWELL3D_REG_INDEX(vertex_buffer.count)] = true;
+    mme_inline[MAXWELL3D_REG_INDEX(index_buffer.count)] = true;
 }
 
 void Maxwell3D::ProcessMacro(u32 method, const u32* base_start, u32 amount, bool is_last_call) {
@@ -216,6 +208,25 @@ void Maxwell3D::ProcessMethodCall(u32 method, u32 argument, u32 nonshadow_argume
         return ProcessCBBind(3);
     case MAXWELL3D_REG_INDEX(bind_groups[4].raw_config):
         return ProcessCBBind(4);
+    case MAXWELL3D_REG_INDEX(draw.end):
+        return DrawArrays();
+    case MAXWELL3D_REG_INDEX(index_buffer32_first):
+        regs.index_buffer.count = regs.index_buffer32_first.count;
+        regs.index_buffer.first = regs.index_buffer32_first.first;
+        dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
+        return DrawArrays();
+    case MAXWELL3D_REG_INDEX(index_buffer16_first):
+        regs.index_buffer.count = regs.index_buffer16_first.count;
+        regs.index_buffer.first = regs.index_buffer16_first.first;
+        dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
+        return DrawArrays();
+    case MAXWELL3D_REG_INDEX(index_buffer8_first):
+        regs.index_buffer.count = regs.index_buffer8_first.count;
+        regs.index_buffer.first = regs.index_buffer8_first.first;
+        dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
+        // a macro calls this one over and over, should it increase instancing?
+        // Used by Hades and likely other Vulkan games.
+        return DrawArrays();
     case MAXWELL3D_REG_INDEX(topology_override):
         use_topology_override = true;
         return;
@@ -250,13 +261,14 @@ void Maxwell3D::CallMacroMethod(u32 method, const std::vector<u32>& parameters)
 
     // Execute the current macro.
     macro_engine->Execute(macro_positions[entry], parameters);
-
-    ProcessDeferredDraw();
+    if (mme_draw.current_mode != MMEDrawMode::Undefined) {
+        FlushMMEInlineDraw();
+    }
 }
 
 void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) {
-    // It is an error to write to a register other than the current macro's ARG register before
-    // it has finished execution.
+    // It is an error to write to a register other than the current macro's ARG register before it
+    // has finished execution.
     if (executing_macro != 0) {
         ASSERT(method == executing_macro + 1);
     }
@@ -271,33 +283,9 @@ void Maxwell3D::CallMethod(u32 method, u32 method_argument, bool is_last_call) {
     ASSERT_MSG(method < Regs::NUM_REGS,
                "Invalid Maxwell3D register, increase the size of the Regs structure");
 
-    if (draw_command[method]) {
-        regs.reg_array[method] = method_argument;
-        deferred_draw_method.push_back(method);
-        auto u32_to_u8 = [&](const u32 argument) {
-            inline_index_draw_indexes.push_back(static_cast<u8>(argument & 0x000000ff));
-            inline_index_draw_indexes.push_back(static_cast<u8>((argument & 0x0000ff00) >> 8));
-            inline_index_draw_indexes.push_back(static_cast<u8>((argument & 0x00ff0000) >> 16));
-            inline_index_draw_indexes.push_back(static_cast<u8>((argument & 0xff000000) >> 24));
-        };
-        if (MAXWELL3D_REG_INDEX(draw_inline_index) == method) {
-            u32_to_u8(method_argument);
-        } else if (MAXWELL3D_REG_INDEX(inline_index_2x16.even) == method) {
-            u32_to_u8(regs.inline_index_2x16.even);
-            u32_to_u8(regs.inline_index_2x16.odd);
-        } else if (MAXWELL3D_REG_INDEX(inline_index_4x8.index0) == method) {
-            u32_to_u8(regs.inline_index_4x8.index0);
-            u32_to_u8(regs.inline_index_4x8.index1);
-            u32_to_u8(regs.inline_index_4x8.index2);
-            u32_to_u8(regs.inline_index_4x8.index3);
-        }
-    } else {
-        ProcessDeferredDraw();
-
-        const u32 argument = ProcessShadowRam(method, method_argument);
-        ProcessDirtyRegisters(method, argument);
-        ProcessMethodCall(method, argument, method_argument, is_last_call);
-    }
+    const u32 argument = ProcessShadowRam(method, method_argument);
+    ProcessDirtyRegisters(method, argument);
+    ProcessMethodCall(method, argument, method_argument, is_last_call);
 }
 
 void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
@@ -338,6 +326,55 @@ void Maxwell3D::CallMultiMethod(u32 method, const u32* base_start, u32 amount,
     }
 }
 
+void Maxwell3D::StepInstance(const MMEDrawMode expected_mode, const u32 count) {
+    if (mme_draw.current_mode == MMEDrawMode::Undefined) {
+        if (mme_draw.gl_begin_consume) {
+            mme_draw.current_mode = expected_mode;
+            mme_draw.current_count = count;
+            mme_draw.instance_count = 1;
+            mme_draw.gl_begin_consume = false;
+            mme_draw.gl_end_count = 0;
+        }
+        return;
+    } else {
+        if (mme_draw.current_mode == expected_mode && count == mme_draw.current_count &&
+            mme_draw.instance_mode && mme_draw.gl_begin_consume) {
+            mme_draw.instance_count++;
+            mme_draw.gl_begin_consume = false;
+            return;
+        } else {
+            FlushMMEInlineDraw();
+        }
+    }
+    // Tail call in case it needs to retry.
+    StepInstance(expected_mode, count);
+}
+
+void Maxwell3D::CallMethodFromMME(u32 method, u32 method_argument) {
+    if (mme_inline[method]) {
+        regs.reg_array[method] = method_argument;
+        if (method == MAXWELL3D_REG_INDEX(vertex_buffer.count) ||
+            method == MAXWELL3D_REG_INDEX(index_buffer.count)) {
+            const MMEDrawMode expected_mode = method == MAXWELL3D_REG_INDEX(vertex_buffer.count)
+                                                  ? MMEDrawMode::Array
+                                                  : MMEDrawMode::Indexed;
+            StepInstance(expected_mode, method_argument);
+        } else if (method == MAXWELL3D_REG_INDEX(draw.begin)) {
+            mme_draw.instance_mode =
+                (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Subsequent) ||
+                (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Unchanged);
+            mme_draw.gl_begin_consume = true;
+        } else {
+            mme_draw.gl_end_count++;
+        }
+    } else {
+        if (mme_draw.current_mode != MMEDrawMode::Undefined) {
+            FlushMMEInlineDraw();
+        }
+        CallMethod(method, method_argument, true);
+    }
+}
+
 void Maxwell3D::ProcessTopologyOverride() {
     using PrimitiveTopology = Maxwell3D::Regs::PrimitiveTopology;
     using PrimitiveTopologyOverride = Maxwell3D::Regs::PrimitiveTopologyOverride;
@@ -367,6 +404,41 @@ void Maxwell3D::ProcessTopologyOverride() {
     }
 }
 
+void Maxwell3D::FlushMMEInlineDraw() {
+    LOG_TRACE(HW_GPU, "called, topology={}, count={}", regs.draw.topology.Value(),
+              regs.vertex_buffer.count);
+    ASSERT_MSG(!(regs.index_buffer.count && regs.vertex_buffer.count), "Both indexed and direct?");
+    ASSERT(mme_draw.instance_count == mme_draw.gl_end_count);
+
+    // Both instance configuration registers can not be set at the same time.
+    ASSERT_MSG(regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::First ||
+                   regs.draw.instance_id != Maxwell3D::Regs::Draw::InstanceId::Unchanged,
+               "Illegal combination of instancing parameters");
+
+    ProcessTopologyOverride();
+
+    const bool is_indexed = mme_draw.current_mode == MMEDrawMode::Indexed;
+    if (ShouldExecute()) {
+        rasterizer->Draw(is_indexed, true);
+    }
+
+    // TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if
+    // the game is trying to draw indexed or direct mode. This needs to be verified on HW still -
+    // it's possible that it is incorrect and that there is some other register used to specify the
+    // drawing mode.
+    if (is_indexed) {
+        regs.index_buffer.count = 0;
+    } else {
+        regs.vertex_buffer.count = 0;
+    }
+    mme_draw.current_mode = MMEDrawMode::Undefined;
+    mme_draw.current_count = 0;
+    mme_draw.instance_count = 0;
+    mme_draw.instance_mode = false;
+    mme_draw.gl_begin_consume = false;
+    mme_draw.gl_end_count = 0;
+}
+
 void Maxwell3D::ProcessMacroUpload(u32 data) {
     macro_engine->AddCode(regs.load_mme.instruction_ptr++, data);
 }
@@ -401,9 +473,7 @@ void Maxwell3D::ProcessQueryGet() {
 
     switch (regs.report_semaphore.query.operation) {
     case Regs::ReportSemaphore::Operation::Release:
-        if (regs.report_semaphore.query.release ==
-                Regs::ReportSemaphore::Release::AfterAllPreceedingWrites ||
-            regs.report_semaphore.query.short_query != 0) {
+        if (regs.report_semaphore.query.short_query != 0) {
             const GPUVAddr sequence_address{regs.report_semaphore.Address()};
             const u32 payload = regs.report_semaphore.payload;
             std::function<void()> operation([this, sequence_address, payload] {
@@ -417,11 +487,10 @@ void Maxwell3D::ProcessQueryGet() {
             };
             const GPUVAddr sequence_address{regs.report_semaphore.Address()};
             const u32 payload = regs.report_semaphore.payload;
-            std::function<void()> operation([this, sequence_address, payload] {
+            [this, sequence_address, payload] {
                 memory_manager.Write<u64>(sequence_address + sizeof(u64), system.GPU().GetTicks());
                 memory_manager.Write<u64>(sequence_address, payload);
-            });
-            rasterizer->SyncOperation(std::move(operation));
+            }();
         }
         break;
     case Regs::ReportSemaphore::Operation::Acquire:
@@ -497,10 +566,46 @@ void Maxwell3D::ProcessCounterReset() {
 
 void Maxwell3D::ProcessSyncPoint() {
     const u32 sync_point = regs.sync_info.sync_point.Value();
-    const auto condition = regs.sync_info.condition.Value();
-    [[maybe_unused]] const u32 cache_flush = regs.sync_info.clean_l2.Value();
-    if (condition == Regs::SyncInfo::Condition::RopWritesDone) {
-        rasterizer->SignalSyncPoint(sync_point);
+    const u32 cache_flush = regs.sync_info.clean_l2.Value();
+    if (cache_flush != 0) {
+        rasterizer->InvalidateGPUCache();
+    }
+    rasterizer->SignalSyncPoint(sync_point);
+}
+
+void Maxwell3D::DrawArrays() {
+    LOG_TRACE(HW_GPU, "called, topology={}, count={}", regs.draw.topology.Value(),
+              regs.vertex_buffer.count);
+    ASSERT_MSG(!(regs.index_buffer.count && regs.vertex_buffer.count), "Both indexed and direct?");
+
+    // Both instance configuration registers can not be set at the same time.
+    ASSERT_MSG(regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::First ||
+                   regs.draw.instance_id != Maxwell3D::Regs::Draw::InstanceId::Unchanged,
+               "Illegal combination of instancing parameters");
+
+    ProcessTopologyOverride();
+
+    if (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Subsequent) {
+        // Increment the current instance *before* drawing.
+        state.current_instance++;
+    } else if (regs.draw.instance_id != Maxwell3D::Regs::Draw::InstanceId::Unchanged) {
+        // Reset the current instance to 0.
+        state.current_instance = 0;
+    }
+
+    const bool is_indexed{regs.index_buffer.count && !regs.vertex_buffer.count};
+    if (ShouldExecute()) {
+        rasterizer->Draw(is_indexed, false);
+    }
+
+    // TODO(bunnei): Below, we reset vertex count so that we can use these registers to determine if
+    // the game is trying to draw indexed or direct mode. This needs to be verified on HW still -
+    // it's possible that it is incorrect and that there is some other register used to specify the
+    // drawing mode.
+    if (is_indexed) {
+        regs.index_buffer.count = 0;
+    } else {
+        regs.vertex_buffer.count = 0;
     }
 }
 
@@ -586,95 +691,4 @@ void Maxwell3D::ProcessClearBuffers() {
     rasterizer->Clear();
 }
 
-void Maxwell3D::ProcessDeferredDraw() {
-    if (deferred_draw_method.empty()) {
-        return;
-    }
-
-    enum class DrawMode {
-        Undefined,
-        General,
-        Instance,
-    };
-    DrawMode draw_mode{DrawMode::Undefined};
-    u32 instance_count = 1;
-
-    auto first_method = deferred_draw_method[0];
-    if (MAXWELL3D_REG_INDEX(draw.begin) == first_method) {
-        // The minimum number of methods for drawing must be greater than or equal to
-        // 3[draw.begin->vertex(index)count->draw.end] to avoid errors in index mode drawing
-        if (deferred_draw_method.size() < 3) {
-            return;
-        }
-        draw_mode = (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Subsequent) ||
-                            (regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::Unchanged)
-                        ? DrawMode::Instance
-                        : DrawMode::General;
-    } else if (MAXWELL3D_REG_INDEX(index_buffer32_first) == first_method ||
-               MAXWELL3D_REG_INDEX(index_buffer16_first) == first_method ||
-               MAXWELL3D_REG_INDEX(index_buffer8_first) == first_method) {
-        draw_mode = DrawMode::General;
-    }
-
-    // Drawing will only begin with draw.begin or index_buffer method, other methods directly
-    // clear
-    if (draw_mode == DrawMode::Undefined) {
-        deferred_draw_method.clear();
-        return;
-    }
-
-    if (draw_mode == DrawMode::Instance) {
-        ASSERT_MSG(deferred_draw_method.size() % 4 == 0, "Instance mode method size error");
-        instance_count = static_cast<u32>(deferred_draw_method.size()) / 4;
-    } else {
-        if (MAXWELL3D_REG_INDEX(index_buffer32_first) == first_method) {
-            regs.index_buffer.count = regs.index_buffer32_first.count;
-            regs.index_buffer.first = regs.index_buffer32_first.first;
-            dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
-        } else if (MAXWELL3D_REG_INDEX(index_buffer32_first) == first_method) {
-            regs.index_buffer.count = regs.index_buffer16_first.count;
-            regs.index_buffer.first = regs.index_buffer16_first.first;
-            dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
-        } else if (MAXWELL3D_REG_INDEX(index_buffer32_first) == first_method) {
-            regs.index_buffer.count = regs.index_buffer8_first.count;
-            regs.index_buffer.first = regs.index_buffer8_first.first;
-            dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
-        } else {
-            auto second_method = deferred_draw_method[1];
-            if (MAXWELL3D_REG_INDEX(draw_inline_index) == second_method ||
-                MAXWELL3D_REG_INDEX(inline_index_2x16.even) == second_method ||
-                MAXWELL3D_REG_INDEX(inline_index_4x8.index0) == second_method) {
-                regs.index_buffer.count = static_cast<u32>(inline_index_draw_indexes.size() / 4);
-                regs.index_buffer.format = Regs::IndexFormat::UnsignedInt;
-            }
-        }
-    }
-
-    LOG_TRACE(HW_GPU, "called, topology={}, count={}", regs.draw.topology.Value(),
-              regs.vertex_buffer.count);
-
-    ASSERT_MSG(!(regs.index_buffer.count && regs.vertex_buffer.count), "Both indexed and direct?");
-
-    // Both instance configuration registers can not be set at the same time.
-    ASSERT_MSG(regs.draw.instance_id == Maxwell3D::Regs::Draw::InstanceId::First ||
-                   regs.draw.instance_id != Maxwell3D::Regs::Draw::InstanceId::Unchanged,
-               "Illegal combination of instancing parameters");
-
-    ProcessTopologyOverride();
-
-    const bool is_indexed = regs.index_buffer.count && !regs.vertex_buffer.count;
-    if (ShouldExecute()) {
-        rasterizer->Draw(is_indexed, instance_count);
-    }
-
-    if (is_indexed) {
-        regs.index_buffer.count = 0;
-    } else {
-        regs.vertex_buffer.count = 0;
-    }
-
-    deferred_draw_method.clear();
-    inline_index_draw_indexes.clear();
-}
-
 } // namespace Tegra::Engines

src/video_core/engines/maxwell_3d.h

@@ -1739,11 +1739,14 @@ public:
             Footprint_1x1_Virtual = 2,
         };
 
-        struct InlineIndex4x8 {
+        struct InlineIndex4x8Align {
             union {
                 BitField<0, 30, u32> count;
                 BitField<30, 2, u32> start;
             };
+        };
+
+        struct InlineIndex4x8Index {
             union {
                 BitField<0, 8, u32> index0;
                 BitField<8, 8, u32> index1;
@@ -2833,7 +2836,8 @@ public:
                 u32 depth_write_enabled;                                               ///< 0x12E8
                 u32 alpha_test_enabled;                                                ///< 0x12EC
                 INSERT_PADDING_BYTES_NOINIT(0x10);
-                InlineIndex4x8 inline_index_4x8;                                       ///< 0x1300
+                InlineIndex4x8Align inline_index_4x8_align;                            ///< 0x1300
+                InlineIndex4x8Index inline_index_4x8_index;                            ///< 0x1304
                 D3DCullMode d3d_cull_mode;                                             ///< 0x1308
                 ComparisonOp depth_test_func;                                          ///< 0x130C
                 f32 alpha_test_ref;                                                    ///< 0x1310
@@ -3044,6 +3048,8 @@ public:
         };
 
         std::array<ShaderStageInfo, Regs::MaxShaderStage> shader_stages;
+
+        u32 current_instance = 0; ///< Current instance to be used to simulate instanced rendering.
     };
 
     State state{};
@@ -3058,6 +3064,11 @@ public:
     void CallMultiMethod(u32 method, const u32* base_start, u32 amount,
                          u32 methods_pending) override;
 
+    /// Write the value to the register identified by method.
+    void CallMethodFromMME(u32 method, u32 method_argument);
+
+    void FlushMMEInlineDraw();
+
     bool ShouldExecute() const {
         return execute_on;
     }
@@ -3070,6 +3081,21 @@ public:
         return *rasterizer;
     }
 
+    enum class MMEDrawMode : u32 {
+        Undefined,
+        Array,
+        Indexed,
+    };
+
+    struct MMEDrawState {
+        MMEDrawMode current_mode{MMEDrawMode::Undefined};
+        u32 current_count{};
+        u32 instance_count{};
+        bool instance_mode{};
+        bool gl_begin_consume{};
+        u32 gl_end_count{};
+    } mme_draw;
+
     struct DirtyState {
         using Flags = std::bitset<std::numeric_limits<u8>::max()>;
         using Table = std::array<u8, Regs::NUM_REGS>;
@@ -3079,8 +3105,6 @@ public:
         Tables tables{};
     } dirty;
 
-    std::vector<u8> inline_index_draw_indexes;
-
 private:
     void InitializeRegisterDefaults();
 
@@ -3140,10 +3164,14 @@ private:
     /// Handles a write to the CB_BIND register.
     void ProcessCBBind(size_t stage_index);
 
+    /// Handles a write to the VERTEX_END_GL register, triggering a draw.
+    void DrawArrays();
+
     /// Handles use of topology overrides (e.g., to avoid using a topology assigned from a macro)
     void ProcessTopologyOverride();
 
-    void ProcessDeferredDraw();
+    // Handles a instance drawcall from MME
+    void StepInstance(MMEDrawMode expected_mode, u32 count);
 
     /// Returns a query's value or an empty object if the value will be deferred through a cache.
     std::optional<u64> GetQueryResult();
@@ -3156,6 +3184,8 @@ private:
     /// Start offsets of each macro in macro_memory
     std::array<u32, 0x80> macro_positions{};
 
+    std::array<bool, Regs::NUM_REGS> mme_inline{};
+
     /// Macro method that is currently being executed / being fed parameters.
     u32 executing_macro = 0;
     /// Parameters that have been submitted to the macro call so far.
@@ -3168,9 +3198,6 @@ private:
 
     bool execute_on{true};
     bool use_topology_override{false};
-
-    std::array<bool, Regs::NUM_REGS> draw_command{};
-    std::vector<u32> deferred_draw_method;
 };
 
 #define ASSERT_REG_POSITION(field_name, position)                                                  \
@@ -3375,7 +3402,8 @@ ASSERT_REG_POSITION(alpha_to_coverage_dither, 0x12E0);
 ASSERT_REG_POSITION(blend_per_target_enabled, 0x12E4);
 ASSERT_REG_POSITION(depth_write_enabled, 0x12E8);
 ASSERT_REG_POSITION(alpha_test_enabled, 0x12EC);
-ASSERT_REG_POSITION(inline_index_4x8, 0x1300);
+ASSERT_REG_POSITION(inline_index_4x8_align, 0x1300);
+ASSERT_REG_POSITION(inline_index_4x8_index, 0x1304);
 ASSERT_REG_POSITION(d3d_cull_mode, 0x1308);
 ASSERT_REG_POSITION(depth_test_func, 0x130C);
 ASSERT_REG_POSITION(alpha_test_ref, 0x1310);

src/video_core/engines/puller.cpp

@@ -75,11 +75,10 @@ void Puller::ProcessSemaphoreTriggerMethod() {
     if (op == GpuSemaphoreOperation::WriteLong) {
         const GPUVAddr sequence_address{regs.semaphore_address.SemaphoreAddress()};
         const u32 payload = regs.semaphore_sequence;
-        std::function<void()> operation([this, sequence_address, payload] {
+        [this, sequence_address, payload] {
             memory_manager.Write<u64>(sequence_address + sizeof(u64), gpu.GetTicks());
             memory_manager.Write<u64>(sequence_address, payload);
-        });
-        rasterizer->SignalFence(std::move(operation));
+        }();
     } else {
         do {
             const u32 word{memory_manager.Read<u32>(regs.semaphore_address.SemaphoreAddress())};

src/video_core/macro/macro_hle.cpp

@@ -22,29 +22,35 @@ void HLE_771BB18C62444DA0(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
     maxwell3d.regs.draw.topology.Assign(
         static_cast<Tegra::Engines::Maxwell3D::Regs::PrimitiveTopology>(parameters[0] & 0x3ffffff));
     maxwell3d.regs.global_base_instance_index = parameters[5];
+    maxwell3d.mme_draw.instance_count = instance_count;
     maxwell3d.regs.global_base_vertex_index = parameters[3];
     maxwell3d.regs.index_buffer.count = parameters[1];
     maxwell3d.regs.index_buffer.first = parameters[4];
 
     if (maxwell3d.ShouldExecute()) {
-        maxwell3d.Rasterizer().Draw(true, instance_count);
+        maxwell3d.Rasterizer().Draw(true, true);
     }
     maxwell3d.regs.index_buffer.count = 0;
+    maxwell3d.mme_draw.instance_count = 0;
+    maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
 }
 
 void HLE_0D61FC9FAAC9FCAD(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& parameters) {
-    const u32 instance_count = (maxwell3d.GetRegisterValue(0xD1B) & parameters[2]);
+    const u32 count = (maxwell3d.GetRegisterValue(0xD1B) & parameters[2]);
 
     maxwell3d.regs.vertex_buffer.first = parameters[3];
     maxwell3d.regs.vertex_buffer.count = parameters[1];
     maxwell3d.regs.global_base_instance_index = parameters[4];
     maxwell3d.regs.draw.topology.Assign(
         static_cast<Tegra::Engines::Maxwell3D::Regs::PrimitiveTopology>(parameters[0]));
+    maxwell3d.mme_draw.instance_count = count;
 
     if (maxwell3d.ShouldExecute()) {
-        maxwell3d.Rasterizer().Draw(false, instance_count);
+        maxwell3d.Rasterizer().Draw(false, true);
     }
     maxwell3d.regs.vertex_buffer.count = 0;
+    maxwell3d.mme_draw.instance_count = 0;
+    maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
 }
 
 void HLE_0217920100488FF7(Engines::Maxwell3D& maxwell3d, const std::vector<u32>& parameters) {
@@ -57,21 +63,24 @@ void HLE_0217920100488FF7(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
     maxwell3d.dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
     maxwell3d.regs.global_base_vertex_index = element_base;
     maxwell3d.regs.global_base_instance_index = base_instance;
-    maxwell3d.CallMethod(0x8e3, 0x640, true);
-    maxwell3d.CallMethod(0x8e4, element_base, true);
-    maxwell3d.CallMethod(0x8e5, base_instance, true);
+    maxwell3d.mme_draw.instance_count = instance_count;
+    maxwell3d.CallMethodFromMME(0x8e3, 0x640);
+    maxwell3d.CallMethodFromMME(0x8e4, element_base);
+    maxwell3d.CallMethodFromMME(0x8e5, base_instance);
     maxwell3d.regs.draw.topology.Assign(
         static_cast<Tegra::Engines::Maxwell3D::Regs::PrimitiveTopology>(parameters[0]));
     if (maxwell3d.ShouldExecute()) {
-        maxwell3d.Rasterizer().Draw(true, instance_count);
+        maxwell3d.Rasterizer().Draw(true, true);
     }
     maxwell3d.regs.vertex_id_base = 0x0;
     maxwell3d.regs.index_buffer.count = 0;
     maxwell3d.regs.global_base_vertex_index = 0x0;
     maxwell3d.regs.global_base_instance_index = 0x0;
-    maxwell3d.CallMethod(0x8e3, 0x640, true);
-    maxwell3d.CallMethod(0x8e4, 0x0, true);
-    maxwell3d.CallMethod(0x8e5, 0x0, true);
+    maxwell3d.mme_draw.instance_count = 0;
+    maxwell3d.CallMethodFromMME(0x8e3, 0x640);
+    maxwell3d.CallMethodFromMME(0x8e4, 0x0);
+    maxwell3d.CallMethodFromMME(0x8e5, 0x0);
+    maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
 }
 
 // Multidraw Indirect
@@ -82,9 +91,11 @@ void HLE_3F5E74B9C9A50164(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
         maxwell3d.regs.index_buffer.count = 0;
         maxwell3d.regs.global_base_vertex_index = 0x0;
         maxwell3d.regs.global_base_instance_index = 0x0;
-        maxwell3d.CallMethod(0x8e3, 0x640, true);
-        maxwell3d.CallMethod(0x8e4, 0x0, true);
-        maxwell3d.CallMethod(0x8e5, 0x0, true);
+        maxwell3d.mme_draw.instance_count = 0;
+        maxwell3d.CallMethodFromMME(0x8e3, 0x640);
+        maxwell3d.CallMethodFromMME(0x8e4, 0x0);
+        maxwell3d.CallMethodFromMME(0x8e5, 0x0);
+        maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
         maxwell3d.dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
     });
     const u32 start_indirect = parameters[0];
@@ -116,13 +127,15 @@ void HLE_3F5E74B9C9A50164(Engines::Maxwell3D& maxwell3d, const std::vector<u32>&
         maxwell3d.regs.index_buffer.count = num_vertices;
         maxwell3d.regs.global_base_vertex_index = base_vertex;
         maxwell3d.regs.global_base_instance_index = base_instance;
-        maxwell3d.CallMethod(0x8e3, 0x640, true);
-        maxwell3d.CallMethod(0x8e4, base_vertex, true);
-        maxwell3d.CallMethod(0x8e5, base_instance, true);
+        maxwell3d.mme_draw.instance_count = instance_count;
+        maxwell3d.CallMethodFromMME(0x8e3, 0x640);
+        maxwell3d.CallMethodFromMME(0x8e4, base_vertex);
+        maxwell3d.CallMethodFromMME(0x8e5, base_instance);
         maxwell3d.dirty.flags[VideoCommon::Dirty::IndexBuffer] = true;
         if (maxwell3d.ShouldExecute()) {
-            maxwell3d.Rasterizer().Draw(true, instance_count);
+            maxwell3d.Rasterizer().Draw(true, true);
         }
+        maxwell3d.mme_draw.current_mode = Engines::Maxwell3D::MMEDrawMode::Undefined;
     }
 }
 

src/video_core/macro/macro_interpreter.cpp

@@ -335,7 +335,7 @@ void MacroInterpreterImpl::SetMethodAddress(u32 address) {
 }
 
 void MacroInterpreterImpl::Send(u32 value) {
-    maxwell3d.CallMethod(method_address.address, value, true);
+    maxwell3d.CallMethodFromMME(method_address.address, value);
     // Increment the method address by the method increment.
     method_address.address.Assign(method_address.address.Value() +
                                   method_address.increment.Value());

src/video_core/macro/macro_jit_x64.cpp

@@ -346,7 +346,7 @@ void MacroJITx64Impl::Compile_Read(Macro::Opcode opcode) {
 }
 
 void Send(Engines::Maxwell3D* maxwell3d, Macro::MethodAddress method_address, u32 value) {
-    maxwell3d->CallMethod(method_address.address, value, true);
+    maxwell3d->CallMethodFromMME(method_address.address, value);
 }
 
 void MacroJITx64Impl::Compile_Send(Xbyak::Reg32 value) {

src/video_core/memory_manager.cpp

@@ -133,7 +133,7 @@ inline void MemoryManager::SetBigPageContinous(size_t big_page_index, bool value
 template <MemoryManager::EntryType entry_type>
 GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr, size_t size,
                                     PTEKind kind) {
-    u64 remaining_size{size};
+    [[maybe_unused]] u64 remaining_size{size};
     if constexpr (entry_type == EntryType::Mapped) {
         page_table.ReserveRange(gpu_addr, size);
     }
@@ -159,7 +159,7 @@ GPUVAddr MemoryManager::PageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cp
 template <MemoryManager::EntryType entry_type>
 GPUVAddr MemoryManager::BigPageTableOp(GPUVAddr gpu_addr, [[maybe_unused]] VAddr cpu_addr,
                                        size_t size, PTEKind kind) {
-    u64 remaining_size{size};
+    [[maybe_unused]] u64 remaining_size{size};
     for (u64 offset{}; offset < size; offset += big_page_size) {
         const GPUVAddr current_gpu_addr = gpu_addr + offset;
         [[maybe_unused]] const auto current_entry_type = GetEntry<true>(current_gpu_addr);

src/video_core/rasterizer_interface.h

@@ -40,7 +40,7 @@ public:
     virtual ~RasterizerInterface() = default;
 
     /// Dispatches a draw invocation
-    virtual void Draw(bool is_indexed, u32 instance_count) = 0;
+    virtual void Draw(bool is_indexed, bool is_instanced) = 0;
 
     /// Clear the current framebuffer
     virtual void Clear() = 0;

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -205,7 +205,7 @@ void RasterizerOpenGL::Clear() {
     ++num_queued_commands;
 }
 
-void RasterizerOpenGL::Draw(bool is_indexed, u32 instance_count) {
+void RasterizerOpenGL::Draw(bool is_indexed, bool is_instanced) {
     MICROPROFILE_SCOPE(OpenGL_Drawing);
 
     SCOPE_EXIT({ gpu.TickWork(); });
@@ -222,15 +222,14 @@ void RasterizerOpenGL::Draw(bool is_indexed, u32 instance_count) {
     pipeline->SetEngine(maxwell3d, gpu_memory);
     pipeline->Configure(is_indexed);
 
-    BindInlineIndexBuffer();
-
     SyncState();
 
     const GLenum primitive_mode = MaxwellToGL::PrimitiveTopology(maxwell3d->regs.draw.topology);
     BeginTransformFeedback(pipeline, primitive_mode);
 
     const GLuint base_instance = static_cast<GLuint>(maxwell3d->regs.global_base_instance_index);
-    const GLsizei num_instances = static_cast<GLsizei>(instance_count);
+    const GLsizei num_instances =
+        static_cast<GLsizei>(is_instanced ? maxwell3d->mme_draw.instance_count : 1);
     if (is_indexed) {
         const GLint base_vertex = static_cast<GLint>(maxwell3d->regs.global_base_vertex_index);
         const GLsizei num_vertices = static_cast<GLsizei>(maxwell3d->regs.index_buffer.count);
@@ -1130,16 +1129,6 @@ void RasterizerOpenGL::ReleaseChannel(s32 channel_id) {
     query_cache.EraseChannel(channel_id);
 }
 
-void RasterizerOpenGL::BindInlineIndexBuffer() {
-    if (maxwell3d->inline_index_draw_indexes.empty()) {
-        return;
-    }
-    const auto data_count = static_cast<u32>(maxwell3d->inline_index_draw_indexes.size());
-    auto buffer = Buffer(buffer_cache_runtime, *this, 0, data_count);
-    buffer.ImmediateUpload(0, maxwell3d->inline_index_draw_indexes);
-    buffer_cache_runtime.BindIndexBuffer(buffer, 0, data_count);
-}
-
 AccelerateDMA::AccelerateDMA(BufferCache& buffer_cache_) : buffer_cache{buffer_cache_} {}
 
 bool AccelerateDMA::BufferCopy(GPUVAddr src_address, GPUVAddr dest_address, u64 amount) {

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -68,7 +68,7 @@ public:
                               StateTracker& state_tracker_);
     ~RasterizerOpenGL() override;
 
-    void Draw(bool is_indexed, u32 instance_count) override;
+    void Draw(bool is_indexed, bool is_instanced) override;
     void Clear() override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;
@@ -199,8 +199,6 @@ private:
     /// End a transform feedback
     void EndTransformFeedback();
 
-    void BindInlineIndexBuffer();
-
     Tegra::GPU& gpu;
 
     const Device& device;

src/video_core/renderer_opengl/gl_shader_cache.cpp

@@ -63,6 +63,7 @@ Shader::RuntimeInfo MakeRuntimeInfo(const GraphicsPipelineKey& key,
     Shader::RuntimeInfo info;
     if (previous_program) {
         info.previous_stage_stores = previous_program->info.stores;
+        info.previous_stage_legacy_stores_mapping = previous_program->info.legacy_stores_mapping;
     } else {
         // Mark all stores as available for vertex shaders
         info.previous_stage_stores.mask.set();

src/video_core/renderer_vulkan/vk_pipeline_cache.cpp

@@ -134,6 +134,7 @@ Shader::RuntimeInfo MakeRuntimeInfo(std::span<const Shader::IR::Program> program
     Shader::RuntimeInfo info;
     if (previous_program) {
         info.previous_stage_stores = previous_program->info.stores;
+        info.previous_stage_legacy_stores_mapping = previous_program->info.legacy_stores_mapping;
         if (previous_program->is_geometry_passthrough) {
             info.previous_stage_stores.mask |= previous_program->info.passthrough.mask;
         }

src/video_core/renderer_vulkan/vk_rasterizer.cpp

@@ -127,10 +127,11 @@ VkRect2D GetScissorState(const Maxwell& regs, size_t index, u32 up_scale = 1, u3
     return scissor;
 }
 
-DrawParams MakeDrawParams(const Maxwell& regs, u32 num_instances, bool is_indexed) {
+DrawParams MakeDrawParams(const Maxwell& regs, u32 num_instances, bool is_instanced,
+                          bool is_indexed) {
     DrawParams params{
         .base_instance = regs.global_base_instance_index,
-        .num_instances = num_instances,
+        .num_instances = is_instanced ? num_instances : 1,
         .base_vertex = is_indexed ? regs.global_base_vertex_index : regs.vertex_buffer.first,
         .num_vertices = is_indexed ? regs.index_buffer.count : regs.vertex_buffer.count,
         .first_index = is_indexed ? regs.index_buffer.first : 0,
@@ -176,7 +177,7 @@ RasterizerVulkan::RasterizerVulkan(Core::Frontend::EmuWindow& emu_window_, Tegra
 
 RasterizerVulkan::~RasterizerVulkan() = default;
 
-void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
+void RasterizerVulkan::Draw(bool is_indexed, bool is_instanced) {
     MICROPROFILE_SCOPE(Vulkan_Drawing);
 
     SCOPE_EXIT({ gpu.TickWork(); });
@@ -193,15 +194,13 @@ void RasterizerVulkan::Draw(bool is_indexed, u32 instance_count) {
     pipeline->SetEngine(maxwell3d, gpu_memory);
     pipeline->Configure(is_indexed);
 
-    BindInlineIndexBuffer();
-
     BeginTransformFeedback();
 
     UpdateDynamicStates();
 
     const auto& regs{maxwell3d->regs};
-    const u32 num_instances{instance_count};
-    const DrawParams draw_params{MakeDrawParams(regs, num_instances, is_indexed)};
+    const u32 num_instances{maxwell3d->mme_draw.instance_count};
+    const DrawParams draw_params{MakeDrawParams(regs, num_instances, is_instanced, is_indexed)};
     scheduler.Record([draw_params](vk::CommandBuffer cmdbuf) {
         if (draw_params.is_indexed) {
             cmdbuf.DrawIndexed(draw_params.num_vertices, draw_params.num_instances,
@@ -1010,17 +1009,4 @@ void RasterizerVulkan::ReleaseChannel(s32 channel_id) {
     query_cache.EraseChannel(channel_id);
 }
 
-void RasterizerVulkan::BindInlineIndexBuffer() {
-    if (maxwell3d->inline_index_draw_indexes.empty()) {
-        return;
-    }
-    const auto data_count = static_cast<u32>(maxwell3d->inline_index_draw_indexes.size());
-    auto buffer = buffer_cache_runtime.UploadStagingBuffer(data_count);
-    std::memcpy(buffer.mapped_span.data(), maxwell3d->inline_index_draw_indexes.data(), data_count);
-    buffer_cache_runtime.BindIndexBuffer(
-        maxwell3d->regs.draw.topology, maxwell3d->regs.index_buffer.format,
-        maxwell3d->regs.index_buffer.first, maxwell3d->regs.index_buffer.count, buffer.buffer,
-        static_cast<u32>(buffer.offset), data_count);
-}
-
 } // namespace Vulkan

src/video_core/renderer_vulkan/vk_rasterizer.h

@@ -64,7 +64,7 @@ public:
                               StateTracker& state_tracker_, Scheduler& scheduler_);
     ~RasterizerVulkan() override;
 
-    void Draw(bool is_indexed, u32 instance_count) override;
+    void Draw(bool is_indexed, bool is_instanced) override;
     void Clear() override;
     void DispatchCompute() override;
     void ResetCounter(VideoCore::QueryType type) override;
@@ -141,8 +141,6 @@ private:
 
     void UpdateVertexInput(Tegra::Engines::Maxwell3D::Regs& regs);
 
-    void BindInlineIndexBuffer();
-
     Tegra::GPU& gpu;
 
     ScreenInfo& screen_info;

src/video_core/texture_cache/format_lookup_table.cpp

@@ -150,6 +150,8 @@ PixelFormat PixelFormatFromTextureInfo(TextureFormat format, ComponentType red,
         return PixelFormat::D24_UNORM_S8_UINT;
     case Hash(TextureFormat::D32S8, FLOAT, UINT, UNORM, UNORM, LINEAR):
         return PixelFormat::D32_FLOAT_S8_UINT;
+    case Hash(TextureFormat::R32_B24G8, FLOAT, UINT, UNORM, UNORM, LINEAR):
+        return PixelFormat::D32_FLOAT_S8_UINT;
     case Hash(TextureFormat::BC1_RGBA, UNORM, LINEAR):
         return PixelFormat::BC1_RGBA_UNORM;
     case Hash(TextureFormat::BC1_RGBA, UNORM, SRGB):

src/video_core/texture_cache/texture_cache.h

@@ -442,7 +442,7 @@ void TextureCache<P>::WriteMemory(VAddr cpu_addr, size_t size) {
 template <class P>
 void TextureCache<P>::DownloadMemory(VAddr cpu_addr, size_t size) {
     std::vector<ImageId> images;
-    ForEachImageInRegion(cpu_addr, size, [this, &images](ImageId image_id, ImageBase& image) {
+    ForEachImageInRegion(cpu_addr, size, [&images](ImageId image_id, ImageBase& image) {
         if (!image.IsSafeDownload()) {
             return;
         }
@@ -1502,9 +1502,9 @@ void TextureCache<P>::UnregisterImage(ImageId image_id) {
     image.flags &= ~ImageFlagBits::BadOverlap;
     lru_cache.Free(image.lru_index);
     const auto& clear_page_table =
-        [this, image_id](u64 page,
-                         std::unordered_map<u64, std::vector<ImageId>, Common::IdentityHash<u64>>&
-                             selected_page_table) {
+        [image_id](u64 page,
+                   std::unordered_map<u64, std::vector<ImageId>, Common::IdentityHash<u64>>&
+                       selected_page_table) {
             const auto page_it = selected_page_table.find(page);
             if (page_it == selected_page_table.end()) {
                 ASSERT_MSG(false, "Unregistering unregistered page=0x{:x}", page << YUZU_PAGEBITS);

src/video_core/textures/astc.cpp

@@ -1661,8 +1661,8 @@ void Decompress(std::span<const uint8_t> data, uint32_t width, uint32_t height,
     for (u32 z = 0; z < depth; ++z) {
         const u32 depth_offset = z * height * width * 4;
         for (u32 y_index = 0; y_index < rows; ++y_index) {
-            auto decompress_stride = [data, width, height, depth, block_width, block_height, output,
-                                      rows, cols, z, depth_offset, y_index] {
+            auto decompress_stride = [data, width, height, block_width, block_height, output, rows,
+                                      cols, z, depth_offset, y_index] {
                 const u32 y = y_index * block_height;
                 for (u32 x_index = 0; x_index < cols; ++x_index) {
                     const u32 block_index = (z * rows * cols) + (y_index * cols) + x_index;

src/video_core/textures/decoders.cpp

@@ -21,7 +21,7 @@ constexpr u32 pdep(u32 value) {
     u32 m = mask;
     for (u32 bit = 1; m; bit += bit) {
         if (value & bit)
-            result |= m & -m;
+            result |= m & (~m + 1);
         m &= m - 1;
     }
     return result;

src/yuzu/multiplayer/state.cpp

@@ -268,7 +268,7 @@ bool MultiplayerState::OnCloseRoom() {
             return true;
         }
         // Save ban list
-        UISettings::values.multiplayer_ban_list = std::move(room->GetBanList());
+        UISettings::values.multiplayer_ban_list = room->GetBanList();
 
         room->Destroy();
         announce_multiplayer_session->Stop();

src/yuzu/startup_checks.cpp

@@ -49,7 +49,7 @@ bool CheckEnvVars(bool* is_child) {
         *is_child = true;
         return false;
     } else if (!SetEnvironmentVariableA(IS_CHILD_ENV_VAR, ENV_VAR_ENABLED_TEXT)) {
-        std::fprintf(stderr, "SetEnvironmentVariableA failed to set %s with error %d\n",
+        std::fprintf(stderr, "SetEnvironmentVariableA failed to set %s with error %lu\n",
                      IS_CHILD_ENV_VAR, GetLastError());
         return true;
     }
@@ -62,7 +62,7 @@ bool StartupChecks(const char* arg0, bool* has_broken_vulkan, bool perform_vulka
     // Set the startup variable for child processes
     const bool env_var_set = SetEnvironmentVariableA(STARTUP_CHECK_ENV_VAR, ENV_VAR_ENABLED_TEXT);
     if (!env_var_set) {
-        std::fprintf(stderr, "SetEnvironmentVariableA failed to set %s with error %d\n",
+        std::fprintf(stderr, "SetEnvironmentVariableA failed to set %s with error %lu\n",
                      STARTUP_CHECK_ENV_VAR, GetLastError());
         return false;
     }
@@ -81,22 +81,22 @@ bool StartupChecks(const char* arg0, bool* has_broken_vulkan, bool perform_vulka
         DWORD exit_code = STILL_ACTIVE;
         const int err = GetExitCodeProcess(process_info.hProcess, &exit_code);
         if (err == 0) {
-            std::fprintf(stderr, "GetExitCodeProcess failed with error %d\n", GetLastError());
+            std::fprintf(stderr, "GetExitCodeProcess failed with error %lu\n", GetLastError());
         }
 
         // Vulkan is broken if the child crashed (return value is not zero)
         *has_broken_vulkan = (exit_code != 0);
 
         if (CloseHandle(process_info.hProcess) == 0) {
-            std::fprintf(stderr, "CloseHandle failed with error %d\n", GetLastError());
+            std::fprintf(stderr, "CloseHandle failed with error %lu\n", GetLastError());
         }
         if (CloseHandle(process_info.hThread) == 0) {
-            std::fprintf(stderr, "CloseHandle failed with error %d\n", GetLastError());
+            std::fprintf(stderr, "CloseHandle failed with error %lu\n", GetLastError());
         }
     }
 
     if (!SetEnvironmentVariableA(STARTUP_CHECK_ENV_VAR, nullptr)) {
-        std::fprintf(stderr, "SetEnvironmentVariableA failed to clear %s with error %d\n",
+        std::fprintf(stderr, "SetEnvironmentVariableA failed to clear %s with error %lu\n",
                      STARTUP_CHECK_ENV_VAR, GetLastError());
     }
 
@@ -135,7 +135,8 @@ bool SpawnChild(const char* arg0, PROCESS_INFORMATION* pi, int flags) {
     startup_info.cb = sizeof(startup_info);
 
     char p_name[255];
-    std::strncpy(p_name, arg0, 255);
+    std::strncpy(p_name, arg0, 254);
+    p_name[254] = '\0';
 
     const bool process_created = CreateProcessA(nullptr,       // lpApplicationName
                                                 p_name,        // lpCommandLine
@@ -149,7 +150,7 @@ bool SpawnChild(const char* arg0, PROCESS_INFORMATION* pi, int flags) {
                                                 pi             // lpProcessInformation
     );
     if (!process_created) {
-        std::fprintf(stderr, "CreateProcessA failed with error %d\n", GetLastError());
+        std::fprintf(stderr, "CreateProcessA failed with error %lu\n", GetLastError());
         return false;
     }