gl_shader_decompiler: Skip RRO instruction.

ShaderGen: Implemented fsetp/kil and predicated instruction execution.

src/core/core.cpp

@@ -12,10 +12,13 @@
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/gdbstub/gdbstub.h"
+#include "core/hle/kernel/client_port.h"
 #include "core/hle/kernel/kernel.h"
 #include "core/hle/kernel/process.h"
 #include "core/hle/kernel/thread.h"
 #include "core/hle/service/service.h"
+#include "core/hle/service/sm/controller.h"
+#include "core/hle/service/sm/sm.h"
 #include "core/hw/hw.h"
 #include "core/loader/loader.h"
 #include "core/memory_setup.h"
@@ -26,6 +29,8 @@ namespace Core {
 
 /*static*/ System System::s_instance;
 
+System::~System() = default;
+
 System::ResultStatus System::RunLoop(bool tight_loop) {
     status = ResultStatus::Success;
     if (!cpu_core) {
@@ -167,10 +172,12 @@ System::ResultStatus System::Init(EmuWindow* emu_window, u32 system_mode) {
 
     telemetry_session = std::make_unique<Core::TelemetrySession>();
 
+    service_manager = std::make_shared<Service::SM::ServiceManager>();
+
     HW::Init();
     Kernel::Init(system_mode);
     scheduler = std::make_unique<Kernel::Scheduler>(cpu_core.get());
-    Service::Init();
+    Service::Init(service_manager);
     GDBStub::Init();
 
     if (!VideoCore::Init(emu_window)) {
@@ -200,17 +207,26 @@ void System::Shutdown() {
     VideoCore::Shutdown();
     GDBStub::Shutdown();
     Service::Shutdown();
-    scheduler = nullptr;
+    scheduler.reset();
     Kernel::Shutdown();
     HW::Shutdown();
-    telemetry_session = nullptr;
-    gpu_core = nullptr;
-    cpu_core = nullptr;
+    service_manager.reset();
+    telemetry_session.reset();
+    gpu_core.reset();
+    cpu_core.reset();
     CoreTiming::Shutdown();
 
-    app_loader = nullptr;
+    app_loader.reset();
 
     LOG_DEBUG(Core, "Shutdown OK");
 }
 
+Service::SM::ServiceManager& System::ServiceManager() {
+    return *service_manager;
+}
+
+const Service::SM::ServiceManager& System::ServiceManager() const {
+    return *service_manager;
+}
+
 } // namespace Core

src/core/core.h

@@ -19,10 +19,16 @@
 class EmuWindow;
 class ARM_Interface;
 
+namespace Service::SM {
+class ServiceManager;
+}
+
 namespace Core {
 
 class System {
 public:
+    ~System();
+
     /**
      * Gets the instance of the System singleton class.
      * @returns Reference to the instance of the System singleton class.
@@ -137,6 +143,9 @@ public:
         return *app_loader;
     }
 
+    Service::SM::ServiceManager& ServiceManager();
+    const Service::SM::ServiceManager& ServiceManager() const;
+
     void SetGPUDebugContext(std::shared_ptr<Tegra::DebugContext> context) {
         debug_context = std::move(context);
     }
@@ -171,6 +180,9 @@ private:
     /// When true, signals that a reschedule should happen
     bool reschedule_pending{};
 
+    /// Service manager
+    std::shared_ptr<Service::SM::ServiceManager> service_manager;
+
     /// Telemetry session for this emulation session
     std::unique_ptr<Core::TelemetrySession> telemetry_session;
 

src/core/hle/kernel/resource_limit.cpp

@@ -34,57 +34,57 @@ SharedPtr<ResourceLimit> ResourceLimit::GetForCategory(ResourceLimitCategory cat
     }
 }
 
-s32 ResourceLimit::GetCurrentResourceValue(u32 resource) const {
+s32 ResourceLimit::GetCurrentResourceValue(ResourceType resource) const {
     switch (resource) {
-    case COMMIT:
+    case ResourceType::Commit:
         return current_commit;
-    case THREAD:
+    case ResourceType::Thread:
         return current_threads;
-    case EVENT:
+    case ResourceType::Event:
         return current_events;
-    case MUTEX:
+    case ResourceType::Mutex:
         return current_mutexes;
-    case SEMAPHORE:
+    case ResourceType::Semaphore:
         return current_semaphores;
-    case TIMER:
+    case ResourceType::Timer:
         return current_timers;
-    case SHARED_MEMORY:
+    case ResourceType::SharedMemory:
         return current_shared_mems;
-    case ADDRESS_ARBITER:
+    case ResourceType::AddressArbiter:
         return current_address_arbiters;
-    case CPU_TIME:
+    case ResourceType::CPUTime:
         return current_cpu_time;
     default:
-        LOG_ERROR(Kernel, "Unknown resource type=%08X", resource);
+        LOG_ERROR(Kernel, "Unknown resource type=%08X", static_cast<u32>(resource));
         UNIMPLEMENTED();
         return 0;
     }
 }
 
-u32 ResourceLimit::GetMaxResourceValue(u32 resource) const {
+u32 ResourceLimit::GetMaxResourceValue(ResourceType resource) const {
     switch (resource) {
-    case PRIORITY:
+    case ResourceType::Priority:
         return max_priority;
-    case COMMIT:
+    case ResourceType::Commit:
         return max_commit;
-    case THREAD:
+    case ResourceType::Thread:
         return max_threads;
-    case EVENT:
+    case ResourceType::Event:
         return max_events;
-    case MUTEX:
+    case ResourceType::Mutex:
         return max_mutexes;
-    case SEMAPHORE:
+    case ResourceType::Semaphore:
         return max_semaphores;
-    case TIMER:
+    case ResourceType::Timer:
         return max_timers;
-    case SHARED_MEMORY:
+    case ResourceType::SharedMemory:
         return max_shared_mems;
-    case ADDRESS_ARBITER:
+    case ResourceType::AddressArbiter:
         return max_address_arbiters;
-    case CPU_TIME:
+    case ResourceType::CPUTime:
         return max_cpu_time;
     default:
-        LOG_ERROR(Kernel, "Unknown resource type=%08X", resource);
+        LOG_ERROR(Kernel, "Unknown resource type=%08X", static_cast<u32>(resource));
         UNIMPLEMENTED();
         return 0;
     }

src/core/hle/kernel/resource_limit.h

@@ -16,17 +16,17 @@ enum class ResourceLimitCategory : u8 {
     OTHER = 3
 };
 
-enum ResourceTypes {
-    PRIORITY = 0,
-    COMMIT = 1,
-    THREAD = 2,
-    EVENT = 3,
-    MUTEX = 4,
-    SEMAPHORE = 5,
-    TIMER = 6,
-    SHARED_MEMORY = 7,
-    ADDRESS_ARBITER = 8,
-    CPU_TIME = 9,
+enum class ResourceType {
+    Priority = 0,
+    Commit = 1,
+    Thread = 2,
+    Event = 3,
+    Mutex = 4,
+    Semaphore = 5,
+    Timer = 6,
+    SharedMemory = 7,
+    AddressArbiter = 8,
+    CPUTime = 9,
 };
 
 class ResourceLimit final : public Object {
@@ -60,14 +60,14 @@ public:
      * @param resource Requested resource type
      * @returns The current value of the resource type
      */
-    s32 GetCurrentResourceValue(u32 resource) const;
+    s32 GetCurrentResourceValue(ResourceType resource) const;
 
     /**
      * Gets the max value for the specified resource.
      * @param resource Requested resource type
      * @returns The max value of the resource type
      */
-    u32 GetMaxResourceValue(u32 resource) const;
+    u32 GetMaxResourceValue(ResourceType resource) const;
 
     /// Name of resource limit object.
     std::string name;

src/core/hle/kernel/svc.cpp

@@ -407,7 +407,7 @@ static ResultCode SetThreadPriority(Handle handle, u32 priority) {
     // Note: The kernel uses the current process's resource limit instead of
     // the one from the thread owner's resource limit.
     SharedPtr<ResourceLimit>& resource_limit = Core::CurrentProcess()->resource_limit;
-    if (resource_limit->GetMaxResourceValue(ResourceTypes::PRIORITY) > priority) {
+    if (resource_limit->GetMaxResourceValue(ResourceType::Priority) > priority) {
         return ERR_NOT_AUTHORIZED;
     }
 
@@ -541,7 +541,7 @@ static ResultCode CreateThread(Handle* out_handle, VAddr entry_point, u64 arg, V
     }
 
     SharedPtr<ResourceLimit>& resource_limit = Core::CurrentProcess()->resource_limit;
-    if (resource_limit->GetMaxResourceValue(ResourceTypes::PRIORITY) > priority) {
+    if (resource_limit->GetMaxResourceValue(ResourceType::Priority) > priority) {
         return ERR_NOT_AUTHORIZED;
     }
 

src/core/hle/service/service.cpp

@@ -145,7 +145,7 @@ ResultCode ServiceFrameworkBase::HandleSyncRequest(Kernel::HLERequestContext& co
         return ResultCode(ErrorModule::HIPC, ErrorDescription::RemoteProcessDead);
     }
     case IPC::CommandType::Control: {
-        SM::g_service_manager->InvokeControlRequest(context);
+        Core::System::GetInstance().ServiceManager().InvokeControlRequest(context);
         break;
     }
     case IPC::CommandType::Request: {
@@ -170,42 +170,40 @@ void AddNamedPort(std::string name, SharedPtr<ClientPort> port) {
 }
 
 /// Initialize ServiceManager
-void Init() {
+void Init(std::shared_ptr<SM::ServiceManager>& sm) {
     // NVFlinger needs to be accessed by several services like Vi and AppletOE so we instantiate it
     // here and pass it into the respective InstallInterfaces functions.
     auto nv_flinger = std::make_shared<NVFlinger::NVFlinger>();
 
-    SM::g_service_manager = std::make_shared<SM::ServiceManager>();
-    SM::ServiceManager::InstallInterfaces(SM::g_service_manager);
+    SM::ServiceManager::InstallInterfaces(sm);
 
-    Account::InstallInterfaces(*SM::g_service_manager);
-    AM::InstallInterfaces(*SM::g_service_manager, nv_flinger);
-    AOC::InstallInterfaces(*SM::g_service_manager);
-    APM::InstallInterfaces(*SM::g_service_manager);
-    Audio::InstallInterfaces(*SM::g_service_manager);
-    Fatal::InstallInterfaces(*SM::g_service_manager);
-    FileSystem::InstallInterfaces(*SM::g_service_manager);
-    Friend::InstallInterfaces(*SM::g_service_manager);
-    HID::InstallInterfaces(*SM::g_service_manager);
-    LM::InstallInterfaces(*SM::g_service_manager);
-    NFP::InstallInterfaces(*SM::g_service_manager);
-    NIFM::InstallInterfaces(*SM::g_service_manager);
-    NS::InstallInterfaces(*SM::g_service_manager);
-    Nvidia::InstallInterfaces(*SM::g_service_manager);
-    PCTL::InstallInterfaces(*SM::g_service_manager);
-    Sockets::InstallInterfaces(*SM::g_service_manager);
-    SPL::InstallInterfaces(*SM::g_service_manager);
-    SSL::InstallInterfaces(*SM::g_service_manager);
-    Time::InstallInterfaces(*SM::g_service_manager);
-    VI::InstallInterfaces(*SM::g_service_manager, nv_flinger);
-    Set::InstallInterfaces(*SM::g_service_manager);
+    Account::InstallInterfaces(*sm);
+    AM::InstallInterfaces(*sm, nv_flinger);
+    AOC::InstallInterfaces(*sm);
+    APM::InstallInterfaces(*sm);
+    Audio::InstallInterfaces(*sm);
+    Fatal::InstallInterfaces(*sm);
+    FileSystem::InstallInterfaces(*sm);
+    Friend::InstallInterfaces(*sm);
+    HID::InstallInterfaces(*sm);
+    LM::InstallInterfaces(*sm);
+    NFP::InstallInterfaces(*sm);
+    NIFM::InstallInterfaces(*sm);
+    NS::InstallInterfaces(*sm);
+    Nvidia::InstallInterfaces(*sm);
+    PCTL::InstallInterfaces(*sm);
+    Sockets::InstallInterfaces(*sm);
+    SPL::InstallInterfaces(*sm);
+    SSL::InstallInterfaces(*sm);
+    Time::InstallInterfaces(*sm);
+    VI::InstallInterfaces(*sm, nv_flinger);
+    Set::InstallInterfaces(*sm);
 
     LOG_DEBUG(Service, "initialized OK");
 }
 
 /// Shutdown ServiceManager
 void Shutdown() {
-    SM::g_service_manager = nullptr;
     g_kernel_named_ports.clear();
     LOG_DEBUG(Service, "shutdown OK");
 }

src/core/hle/service/service.h

@@ -178,7 +178,7 @@ private:
 };
 
 /// Initialize ServiceManager
-void Init();
+void Init(std::shared_ptr<SM::ServiceManager>& sm);
 
 /// Shutdown ServiceManager
 void Shutdown();

src/core/hle/service/sm/sm.cpp

@@ -14,6 +14,8 @@
 
 namespace Service::SM {
 
+ServiceManager::~ServiceManager() = default;
+
 void ServiceManager::InvokeControlRequest(Kernel::HLERequestContext& context) {
     controller_interface->InvokeRequest(context);
 }
@@ -72,7 +74,7 @@ ResultVal<Kernel::SharedPtr<Kernel::ClientSession>> ServiceManager::ConnectToSer
     return client_port->Connect();
 }
 
-std::shared_ptr<ServiceManager> g_service_manager;
+SM::~SM() = default;
 
 /**
  * SM::Initialize service function

src/core/hle/service/sm/sm.h

@@ -23,7 +23,7 @@ namespace Service::SM {
 class SM final : public ServiceFramework<SM> {
 public:
     SM(std::shared_ptr<ServiceManager> service_manager);
-    ~SM() = default;
+    ~SM() override;
 
 private:
     void Initialize(Kernel::HLERequestContext& ctx);
@@ -44,6 +44,8 @@ class ServiceManager {
 public:
     static void InstallInterfaces(std::shared_ptr<ServiceManager> self);
 
+    ~ServiceManager();
+
     ResultVal<Kernel::SharedPtr<Kernel::ServerPort>> RegisterService(std::string name,
                                                                      unsigned int max_sessions);
     ResultVal<Kernel::SharedPtr<Kernel::ClientPort>> GetServicePort(const std::string& name);
@@ -59,6 +61,4 @@ private:
     std::unordered_map<std::string, Kernel::SharedPtr<Kernel::ClientPort>> registered_services;
 };
 
-extern std::shared_ptr<ServiceManager> g_service_manager;
-
 } // namespace Service::SM

src/video_core/engines/shader_bytecode.h

@@ -4,15 +4,24 @@
 
 #pragma once
 
+#include <bitset>
 #include <cstring>
 #include <map>
 #include <string>
+#include <vector>
+
+#include <boost/optional.hpp>
+
 #include "common/bit_field.h"
+#include "common/common_types.h"
 
 namespace Tegra {
 namespace Shader {
 
 struct Register {
+    // Register 255 is special cased to always be 0
+    static constexpr size_t ZeroIndex = 255;
+
     constexpr Register() = default;
 
     constexpr Register(u64 value) : value(value) {}
@@ -86,181 +95,12 @@ union Uniform {
     BitField<34, 5, u64> index;
 };
 
-union OpCode {
-    enum class Id : u64 {
-        TEXS = 0x6C,
-        IPA = 0xE0,
-        FMUL32_IMM = 0x1E,
-        FFMA_IMM = 0x65,
-        FFMA_CR = 0x93,
-        FFMA_RC = 0xA3,
-        FFMA_RR = 0xB3,
-
-        FADD_C = 0x98B,
-        FMUL_C = 0x98D,
-        MUFU = 0xA10,
-        FADD_R = 0xB8B,
-        FMUL_R = 0xB8D,
-        LD_A = 0x1DFB,
-        ST_A = 0x1DFE,
-
-        FSETP_R = 0x5BB,
-        FSETP_C = 0x4BB,
-        EXIT = 0xE30,
-        KIL = 0xE33,
-
-        FMUL_IMM = 0x70D,
-        FMUL_IMM_x = 0x72D,
-        FADD_IMM = 0x70B,
-        FADD_IMM_x = 0x72B,
-    };
-
-    enum class Type {
-        Trivial,
-        Arithmetic,
-        Ffma,
-        Flow,
-        Memory,
-        Unknown,
-    };
-
-    struct Info {
-        Type type;
-        std::string name;
-    };
-
-    OpCode() = default;
-
-    constexpr OpCode(Id value) : value(static_cast<u64>(value)) {}
-
-    constexpr OpCode(u64 value) : value{value} {}
-
-    constexpr Id EffectiveOpCode() const {
-        switch (op1) {
-        case Id::TEXS:
-            return op1;
-        }
-
-        switch (op2) {
-        case Id::IPA:
-        case Id::FMUL32_IMM:
-            return op2;
-        }
-
-        switch (op3) {
-        case Id::FFMA_IMM:
-        case Id::FFMA_CR:
-        case Id::FFMA_RC:
-        case Id::FFMA_RR:
-            return op3;
-        }
-
-        switch (op4) {
-        case Id::EXIT:
-        case Id::FSETP_R:
-        case Id::FSETP_C:
-        case Id::KIL:
-            return op4;
-        }
-
-        switch (op5) {
-        case Id::MUFU:
-        case Id::LD_A:
-        case Id::ST_A:
-        case Id::FADD_R:
-        case Id::FADD_C:
-        case Id::FMUL_R:
-        case Id::FMUL_C:
-            return op5;
-
-        case Id::FMUL_IMM:
-        case Id::FMUL_IMM_x:
-            return Id::FMUL_IMM;
-
-        case Id::FADD_IMM:
-        case Id::FADD_IMM_x:
-            return Id::FADD_IMM;
-        }
-
-        return static_cast<Id>(value);
-    }
-
-    static const Info& GetInfo(const OpCode& opcode) {
-        static const std::map<Id, Info> info_table{BuildInfoTable()};
-        const auto& search{info_table.find(opcode.EffectiveOpCode())};
-        if (search != info_table.end()) {
-            return search->second;
-        }
-
-        static const Info unknown{Type::Unknown, "UNK"};
-        return unknown;
-    }
-
-    constexpr operator Id() const {
-        return static_cast<Id>(value);
-    }
-
-    constexpr OpCode operator<<(size_t bits) const {
-        return value << bits;
-    }
-
-    constexpr OpCode operator>>(size_t bits) const {
-        return value >> bits;
-    }
-
-    template <typename T>
-    constexpr u64 operator-(const T& oth) const {
-        return value - oth;
-    }
-
-    constexpr u64 operator&(const OpCode& oth) const {
-        return value & oth.value;
-    }
-
-    constexpr u64 operator~() const {
-        return ~value;
-    }
-
-    static std::map<Id, Info> BuildInfoTable() {
-        std::map<Id, Info> info_table;
-        info_table[Id::TEXS] = {Type::Memory, "texs"};
-        info_table[Id::LD_A] = {Type::Memory, "ld_a"};
-        info_table[Id::ST_A] = {Type::Memory, "st_a"};
-        info_table[Id::MUFU] = {Type::Arithmetic, "mufu"};
-        info_table[Id::FFMA_IMM] = {Type::Ffma, "ffma_imm"};
-        info_table[Id::FFMA_CR] = {Type::Ffma, "ffma_cr"};
-        info_table[Id::FFMA_RC] = {Type::Ffma, "ffma_rc"};
-        info_table[Id::FFMA_RR] = {Type::Ffma, "ffma_rr"};
-        info_table[Id::FADD_R] = {Type::Arithmetic, "fadd_r"};
-        info_table[Id::FADD_C] = {Type::Arithmetic, "fadd_c"};
-        info_table[Id::FADD_IMM] = {Type::Arithmetic, "fadd_imm"};
-        info_table[Id::FMUL_R] = {Type::Arithmetic, "fmul_r"};
-        info_table[Id::FMUL_C] = {Type::Arithmetic, "fmul_c"};
-        info_table[Id::FMUL_IMM] = {Type::Arithmetic, "fmul_imm"};
-        info_table[Id::FMUL32_IMM] = {Type::Arithmetic, "fmul32_imm"};
-        info_table[Id::FSETP_C] = {Type::Arithmetic, "fsetp_c"};
-        info_table[Id::FSETP_R] = {Type::Arithmetic, "fsetp_r"};
-        info_table[Id::EXIT] = {Type::Trivial, "exit"};
-        info_table[Id::IPA] = {Type::Trivial, "ipa"};
-        info_table[Id::KIL] = {Type::Flow, "kil"};
-        return info_table;
-    }
-
-    BitField<57, 7, Id> op1;
-    BitField<56, 8, Id> op2;
-    BitField<55, 9, Id> op3;
-    BitField<52, 12, Id> op4;
-    BitField<51, 13, Id> op5;
-    u64 value{};
-};
-static_assert(sizeof(OpCode) == 0x8, "Incorrect structure size");
-
 } // namespace Shader
 } // namespace Tegra
 
 namespace std {
 
-// TODO(bunne): The below is forbidden by the C++ standard, but works fine. See #330.
+// TODO(bunnei): The below is forbidden by the C++ standard, but works fine. See #330.
 template <>
 struct make_unsigned<Tegra::Shader::Attribute> {
     using type = Tegra::Shader::Attribute;
@@ -271,11 +111,6 @@ struct make_unsigned<Tegra::Shader::Register> {
     using type = Tegra::Shader::Register;
 };
 
-template <>
-struct make_unsigned<Tegra::Shader::OpCode> {
-    using type = Tegra::Shader::OpCode;
-};
-
 } // namespace std
 
 namespace Tegra {
@@ -283,7 +118,23 @@ namespace Shader {
 
 enum class Pred : u64 {
     UnusedIndex = 0x7,
-    NeverExecute = 0xf,
+    NeverExecute = 0xF,
+};
+
+enum class PredCondition : u64 {
+    LessThan = 1,
+    Equal = 2,
+    LessEqual = 3,
+    GreaterThan = 4,
+    NotEqual = 5,
+    GreaterEqual = 6,
+    // TODO(Subv): Other condition types
+};
+
+enum class PredOperation : u64 {
+    And = 0,
+    Or = 1,
+    Xor = 2,
 };
 
 enum class SubOp : u64 {
@@ -298,18 +149,24 @@ enum class SubOp : u64 {
 
 union Instruction {
     Instruction& operator=(const Instruction& instr) {
-        hex = instr.hex;
+        value = instr.value;
         return *this;
     }
 
-    OpCode opcode;
+    constexpr Instruction(u64 value) : value{value} {}
+
     BitField<0, 8, Register> gpr0;
     BitField<8, 8, Register> gpr8;
-    BitField<16, 4, Pred> pred;
+    union {
+        BitField<16, 4, Pred> full_pred;
+        BitField<16, 3, u64> pred_index;
+    } pred;
+    BitField<19, 1, u64> negate_pred;
     BitField<20, 8, Register> gpr20;
     BitField<20, 7, SubOp> sub_op;
     BitField<28, 8, Register> gpr28;
     BitField<39, 8, Register> gpr39;
+    BitField<48, 16, u64> opcode;
 
     union {
         BitField<20, 19, u64> imm20_19;
@@ -343,6 +200,20 @@ union Instruction {
         BitField<49, 1, u64> negate_c;
     } ffma;
 
+    union {
+        BitField<0, 3, u64> pred0;
+        BitField<3, 3, u64> pred3;
+        BitField<7, 1, u64> abs_a;
+        BitField<39, 3, u64> pred39;
+        BitField<42, 1, u64> neg_pred;
+        BitField<43, 1, u64> neg_a;
+        BitField<44, 1, u64> abs_b;
+        BitField<45, 2, PredOperation> op;
+        BitField<47, 1, u64> ftz;
+        BitField<48, 4, PredCondition> cond;
+        BitField<56, 1, u64> neg_b;
+    } fsetp;
+
     BitField<61, 1, u64> is_b_imm;
     BitField<60, 1, u64> is_b_gpr;
     BitField<59, 1, u64> is_c_gpr;
@@ -351,11 +222,218 @@ union Instruction {
     Uniform uniform;
     Sampler sampler;
 
-    u64 hex;
+    u64 value;
 };
 static_assert(sizeof(Instruction) == 0x8, "Incorrect structure size");
 static_assert(std::is_standard_layout<Instruction>::value,
               "Structure does not have standard layout");
 
+class OpCode {
+public:
+    enum class Id {
+        KIL,
+        LD_A,
+        ST_A,
+        TEXQ, // Texture Query
+        TEXS, // Texture Fetch with scalar/non-vec4 source/destinations
+        TLDS, // Texture Load with scalar/non-vec4 source/destinations
+        EXIT,
+        IPA,
+        FFMA_IMM, // Fused Multiply and Add
+        FFMA_CR,
+        FFMA_RC,
+        FFMA_RR,
+        FADD_C,
+        FADD_R,
+        FADD_IMM,
+        FMUL_C,
+        FMUL_R,
+        FMUL_IMM,
+        FMUL32_IMM,
+        MUFU, // Multi-Function Operator
+        RRO,  // Range Reduction Operator
+        F2F_C,
+        F2F_R,
+        F2F_IMM,
+        F2I_C,
+        F2I_R,
+        F2I_IMM,
+        I2F_C,
+        I2F_R,
+        I2F_IMM,
+        LOP32I,
+        MOV_C,
+        MOV_R,
+        MOV_IMM,
+        MOV32I,
+        SHR_C,
+        SHR_R,
+        SHR_IMM,
+        FSETP_C, // Set Predicate
+        FSETP_R,
+        FSETP_IMM,
+        ISETP_C,
+        ISETP_IMM,
+        ISETP_R,
+    };
+
+    enum class Type {
+        Trivial,
+        Arithmetic,
+        Ffma,
+        Flow,
+        Memory,
+        FloatPredicate,
+        IntegerPredicate,
+        Unknown,
+    };
+
+    class Matcher {
+    public:
+        Matcher(const char* const name, u16 mask, u16 expected, OpCode::Id id, OpCode::Type type)
+            : name{name}, mask{mask}, expected{expected}, id{id}, type{type} {}
+
+        const char* GetName() const {
+            return name;
+        }
+
+        u16 GetMask() const {
+            return mask;
+        }
+
+        Id GetId() const {
+            return id;
+        }
+
+        Type GetType() const {
+            return type;
+        }
+
+        /**
+         * Tests to see if the given instruction is the instruction this matcher represents.
+         * @param instruction The instruction to test
+         * @returns true if the given instruction matches.
+         */
+        bool Matches(u16 instruction) const {
+            return (instruction & mask) == expected;
+        }
+
+    private:
+        const char* name;
+        u16 mask;
+        u16 expected;
+        Id id;
+        Type type;
+    };
+
+    static boost::optional<const Matcher&> Decode(Instruction instr) {
+        static const auto table{GetDecodeTable()};
+
+        const auto matches_instruction = [instr](const auto& matcher) {
+            return matcher.Matches(static_cast<u16>(instr.opcode));
+        };
+
+        auto iter = std::find_if(table.begin(), table.end(), matches_instruction);
+        return iter != table.end() ? boost::optional<const Matcher&>(*iter) : boost::none;
+    }
+
+private:
+    struct Detail {
+    private:
+        static constexpr size_t opcode_bitsize = 16;
+
+        /**
+         * Generates the mask and the expected value after masking from a given bitstring.
+         * A '0' in a bitstring indicates that a zero must be present at that bit position.
+         * A '1' in a bitstring indicates that a one must be present at that bit position.
+         */
+        static auto GetMaskAndExpect(const char* const bitstring) {
+            u16 mask = 0, expect = 0;
+            for (size_t i = 0; i < opcode_bitsize; i++) {
+                const size_t bit_position = opcode_bitsize - i - 1;
+                switch (bitstring[i]) {
+                case '0':
+                    mask |= 1 << bit_position;
+                    break;
+                case '1':
+                    expect |= 1 << bit_position;
+                    mask |= 1 << bit_position;
+                    break;
+                default:
+                    // Ignore
+                    break;
+                }
+            }
+            return std::make_tuple(mask, expect);
+        }
+
+    public:
+        /// Creates a matcher that can match and parse instructions based on bitstring.
+        static auto GetMatcher(const char* const bitstring, OpCode::Id op, OpCode::Type type,
+                               const char* const name) {
+            const auto mask_expect = GetMaskAndExpect(bitstring);
+            return Matcher(name, std::get<0>(mask_expect), std::get<1>(mask_expect), op, type);
+        }
+    };
+
+    static std::vector<Matcher> GetDecodeTable() {
+        std::vector<Matcher> table = {
+#define INST(bitstring, op, type, name) Detail::GetMatcher(bitstring, op, type, name)
+            INST("111000110011----", Id::KIL, Type::Flow, "KIL"),
+            INST("1110111111011---", Id::LD_A, Type::Memory, "LD_A"),
+            INST("1110111111110---", Id::ST_A, Type::Memory, "ST_A"),
+            INST("1101111101001---", Id::TEXQ, Type::Memory, "TEXQ"),
+            INST("1101100---------", Id::TEXS, Type::Memory, "TEXS"),
+            INST("1101101---------", Id::TLDS, Type::Memory, "TLDS"),
+            INST("111000110000----", Id::EXIT, Type::Trivial, "EXIT"),
+            INST("11100000--------", Id::IPA, Type::Trivial, "IPA"),
+            INST("001100101-------", Id::FFMA_IMM, Type::Ffma, "FFMA_IMM"),
+            INST("010010011-------", Id::FFMA_CR, Type::Ffma, "FFMA_CR"),
+            INST("010100011-------", Id::FFMA_RC, Type::Ffma, "FFMA_RC"),
+            INST("010110011-------", Id::FFMA_RR, Type::Ffma, "FFMA_RR"),
+            INST("0100110001011---", Id::FADD_C, Type::Arithmetic, "FADD_C"),
+            INST("0101110001011---", Id::FADD_R, Type::Arithmetic, "FADD_R"),
+            INST("0011100-01011---", Id::FADD_IMM, Type::Arithmetic, "FADD_IMM"),
+            INST("0100110001101---", Id::FMUL_C, Type::Arithmetic, "FMUL_C"),
+            INST("0101110001101---", Id::FMUL_R, Type::Arithmetic, "FMUL_R"),
+            INST("0011100-01101---", Id::FMUL_IMM, Type::Arithmetic, "FMUL_IMM"),
+            INST("00011110--------", Id::FMUL32_IMM, Type::Arithmetic, "FMUL32_IMM"),
+            INST("0101000010000---", Id::MUFU, Type::Arithmetic, "MUFU"),
+            INST("0101110010010---", Id::RRO, Type::Arithmetic, "RRO"),
+            INST("0100110010101---", Id::F2F_C, Type::Arithmetic, "F2F_C"),
+            INST("0101110010101---", Id::F2F_R, Type::Arithmetic, "F2F_R"),
+            INST("0011100-10101---", Id::F2F_IMM, Type::Arithmetic, "F2F_IMM"),
+            INST("0100110010110---", Id::F2I_C, Type::Arithmetic, "F2I_C"),
+            INST("0101110010110---", Id::F2I_R, Type::Arithmetic, "F2I_R"),
+            INST("0011100-10110---", Id::F2I_IMM, Type::Arithmetic, "F2I_IMM"),
+            INST("0100110010111---", Id::I2F_C, Type::Arithmetic, "I2F_C"),
+            INST("0101110010111---", Id::I2F_R, Type::Arithmetic, "I2F_R"),
+            INST("0011100-10111---", Id::I2F_IMM, Type::Arithmetic, "I2F_IMM"),
+            INST("000001----------", Id::LOP32I, Type::Arithmetic, "LOP32I"),
+            INST("0100110010011---", Id::MOV_C, Type::Arithmetic, "MOV_C"),
+            INST("0101110010011---", Id::MOV_R, Type::Arithmetic, "MOV_R"),
+            INST("0011100-10011---", Id::MOV_IMM, Type::Arithmetic, "MOV_IMM"),
+            INST("000000010000----", Id::MOV32I, Type::Arithmetic, "MOV32I"),
+            INST("0100110000101---", Id::SHR_C, Type::Arithmetic, "SHR_C"),
+            INST("0101110000101---", Id::SHR_R, Type::Arithmetic, "SHR_R"),
+            INST("0011100-00101---", Id::SHR_IMM, Type::Arithmetic, "SHR_IMM"),
+            INST("010010111011----", Id::FSETP_C, Type::FloatPredicate, "FSETP_C"),
+            INST("010110111011----", Id::FSETP_R, Type::FloatPredicate, "FSETP_R"),
+            INST("0011011-1011----", Id::FSETP_IMM, Type::FloatPredicate, "FSETP_IMM"),
+            INST("010010110110----", Id::ISETP_C, Type::IntegerPredicate, "ISETP_C"),
+            INST("010110110110----", Id::ISETP_R, Type::IntegerPredicate, "ISETP_R"),
+            INST("0011011-0110----", Id::ISETP_IMM, Type::IntegerPredicate, "ISETP_IMM"),
+        };
+#undef INST
+        std::stable_sort(table.begin(), table.end(), [](const auto& a, const auto& b) {
+            // If a matcher has more bits in its mask it is more specific, so it
+            // should come first.
+            return std::bitset<16>(a.GetMask()).count() > std::bitset<16>(b.GetMask()).count();
+        });
+
+        return table;
+    }
+};
+
 } // namespace Shader
 } // namespace Tegra

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -14,6 +14,7 @@
 #include "common/math_util.h"
 #include "common/microprofile.h"
 #include "common/scope_exit.h"
+#include "common/vector_math.h"
 #include "core/core.h"
 #include "core/hle/kernel/process.h"
 #include "core/settings.h"

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -6,10 +6,15 @@
 
 #include <array>
 #include <cstddef>
+#include <cstring>
 #include <memory>
+#include <unordered_map>
 #include <vector>
 #include <glad/glad.h>
+#include "common/bit_field.h"
 #include "common/common_types.h"
+#include "common/hash.h"
+#include "common/vector_math.h"
 #include "video_core/engines/maxwell_3d.h"
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_opengl/gl_rasterizer_cache.h"

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -7,6 +7,7 @@
 #include <cstring>
 #include <iterator>
 #include <memory>
+#include <unordered_set>
 #include <utility>
 #include <vector>
 #include <boost/optional.hpp>
@@ -19,6 +20,7 @@
 #include "common/math_util.h"
 #include "common/microprofile.h"
 #include "common/scope_exit.h"
+#include "common/vector_math.h"
 #include "core/core.h"
 #include "core/frontend/emu_window.h"
 #include "core/hle/kernel/process.h"
@@ -670,7 +672,8 @@ void CachedSurface::DownloadGLTexture(const MathUtil::Rectangle<u32>& rect, GLui
     glPixelStorei(GL_PACK_ROW_LENGTH, 0);
 }
 
-enum MatchFlags {
+enum class MatchFlags {
+    None = 0,
     Invalid = 1,      // Flag that can be applied to other match types, invalid matches require
                       // validation before they can be used
     Exact = 1 << 1,   // Surfaces perfectly match
@@ -684,6 +687,10 @@ constexpr MatchFlags operator|(MatchFlags lhs, MatchFlags rhs) {
     return static_cast<MatchFlags>(static_cast<int>(lhs) | static_cast<int>(rhs));
 }
 
+constexpr MatchFlags operator&(MatchFlags lhs, MatchFlags rhs) {
+    return static_cast<MatchFlags>(static_cast<int>(lhs) & static_cast<int>(rhs));
+}
+
 /// Get the best surface match (and its match type) for the given flags
 template <MatchFlags find_flags>
 Surface FindMatch(const SurfaceCache& surface_cache, const SurfaceParams& params,
@@ -701,15 +708,15 @@ Surface FindMatch(const SurfaceCache& surface_cache, const SurfaceParams& params
                                          : (params.res_scale <= surface->res_scale);
             // validity will be checked in GetCopyableInterval
             bool is_valid =
-                find_flags & MatchFlags::Copy
+                (find_flags & MatchFlags::Copy) != MatchFlags::None
                     ? true
                     : surface->IsRegionValid(validate_interval.value_or(params.GetInterval()));
 
-            if (!(find_flags & MatchFlags::Invalid) && !is_valid)
+            if ((find_flags & MatchFlags::Invalid) == MatchFlags::None && !is_valid)
                 continue;
 
             auto IsMatch_Helper = [&](auto check_type, auto match_fn) {
-                if (!(find_flags & check_type))
+                if ((find_flags & check_type) == MatchFlags::None)
                     return;
 
                 bool matched;

src/video_core/renderer_opengl/gl_resource_manager.h

@@ -14,13 +14,13 @@ class OGLTexture : private NonCopyable {
 public:
     OGLTexture() = default;
 
-    OGLTexture(OGLTexture&& o) : handle(std::exchange(o.handle, 0)) {}
+    OGLTexture(OGLTexture&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLTexture() {
         Release();
     }
 
-    OGLTexture& operator=(OGLTexture&& o) {
+    OGLTexture& operator=(OGLTexture&& o) noexcept {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;
@@ -49,13 +49,13 @@ class OGLSampler : private NonCopyable {
 public:
     OGLSampler() = default;
 
-    OGLSampler(OGLSampler&& o) : handle(std::exchange(o.handle, 0)) {}
+    OGLSampler(OGLSampler&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLSampler() {
         Release();
     }
 
-    OGLSampler& operator=(OGLSampler&& o) {
+    OGLSampler& operator=(OGLSampler&& o) noexcept {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;
@@ -84,13 +84,13 @@ class OGLShader : private NonCopyable {
 public:
     OGLShader() = default;
 
-    OGLShader(OGLShader&& o) : handle(std::exchange(o.handle, 0)) {}
+    OGLShader(OGLShader&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLShader() {
         Release();
     }
 
-    OGLShader& operator=(OGLShader&& o) {
+    OGLShader& operator=(OGLShader&& o) noexcept {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;
@@ -118,13 +118,13 @@ class OGLProgram : private NonCopyable {
 public:
     OGLProgram() = default;
 
-    OGLProgram(OGLProgram&& o) : handle(std::exchange(o.handle, 0)) {}
+    OGLProgram(OGLProgram&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLProgram() {
         Release();
     }
 
-    OGLProgram& operator=(OGLProgram&& o) {
+    OGLProgram& operator=(OGLProgram&& o) noexcept {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;
@@ -165,13 +165,12 @@ public:
 class OGLPipeline : private NonCopyable {
 public:
     OGLPipeline() = default;
-    OGLPipeline(OGLPipeline&& o) {
-        handle = std::exchange<GLuint>(o.handle, 0);
-    }
+    OGLPipeline(OGLPipeline&& o) noexcept : handle{std::exchange<GLuint>(o.handle, 0)} {}
+
     ~OGLPipeline() {
         Release();
     }
-    OGLPipeline& operator=(OGLPipeline&& o) {
+    OGLPipeline& operator=(OGLPipeline&& o) noexcept {
         handle = std::exchange<GLuint>(o.handle, 0);
         return *this;
     }
@@ -199,13 +198,13 @@ class OGLBuffer : private NonCopyable {
 public:
     OGLBuffer() = default;
 
-    OGLBuffer(OGLBuffer&& o) : handle(std::exchange(o.handle, 0)) {}
+    OGLBuffer(OGLBuffer&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLBuffer() {
         Release();
     }
 
-    OGLBuffer& operator=(OGLBuffer&& o) {
+    OGLBuffer& operator=(OGLBuffer&& o) noexcept {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;
@@ -234,12 +233,12 @@ class OGLSync : private NonCopyable {
 public:
     OGLSync() = default;
 
-    OGLSync(OGLSync&& o) : handle(std::exchange(o.handle, nullptr)) {}
+    OGLSync(OGLSync&& o) noexcept : handle(std::exchange(o.handle, nullptr)) {}
 
     ~OGLSync() {
         Release();
     }
-    OGLSync& operator=(OGLSync&& o) {
+    OGLSync& operator=(OGLSync&& o) noexcept {
         Release();
         handle = std::exchange(o.handle, nullptr);
         return *this;
@@ -267,13 +266,13 @@ class OGLVertexArray : private NonCopyable {
 public:
     OGLVertexArray() = default;
 
-    OGLVertexArray(OGLVertexArray&& o) : handle(std::exchange(o.handle, 0)) {}
+    OGLVertexArray(OGLVertexArray&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLVertexArray() {
         Release();
     }
 
-    OGLVertexArray& operator=(OGLVertexArray&& o) {
+    OGLVertexArray& operator=(OGLVertexArray&& o) noexcept {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;
@@ -302,13 +301,13 @@ class OGLFramebuffer : private NonCopyable {
 public:
     OGLFramebuffer() = default;
 
-    OGLFramebuffer(OGLFramebuffer&& o) : handle(std::exchange(o.handle, 0)) {}
+    OGLFramebuffer(OGLFramebuffer&& o) noexcept : handle(std::exchange(o.handle, 0)) {}
 
     ~OGLFramebuffer() {
         Release();
     }
 
-    OGLFramebuffer& operator=(OGLFramebuffer&& o) {
+    OGLFramebuffer& operator=(OGLFramebuffer&& o) noexcept {
         Release();
         handle = std::exchange(o.handle, 0);
         return *this;

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -97,11 +97,12 @@ private:
             return exit_method;
 
         for (u32 offset = begin; offset != end && offset != PROGRAM_END; ++offset) {
-            const Instruction instr = {program_code[offset]};
-            switch (instr.opcode.EffectiveOpCode()) {
-            case OpCode::Id::EXIT: {
-                return exit_method = ExitMethod::AlwaysEnd;
-            }
+            if (const auto opcode = OpCode::Decode({program_code[offset]})) {
+                switch (opcode->GetId()) {
+                case OpCode::Id::EXIT: {
+                    return exit_method = ExitMethod::AlwaysEnd;
+                }
+                }
             }
         }
         return exit_method = ExitMethod::AlwaysReturn;
@@ -220,6 +221,8 @@ private:
 
     /// Generates code representing a temporary (GPR) register.
     std::string GetRegister(const Register& reg, unsigned elem = 0) {
+        if (reg == Register::ZeroIndex)
+            return "0";
         if (stage == Maxwell3D::Regs::ShaderStage::Fragment && reg < 4) {
             // GPRs 0-3 are output color for the fragment shader
             return std::string{"color."} + "rgba"[(reg + elem) & 3];
@@ -276,6 +279,52 @@ private:
         shader.AddLine(dest + " = " + src + ";");
     }
 
+    /*
+     * Writes code that assigns a predicate boolean variable.
+     * @param pred The id of the predicate to write to.
+     * @param value The expression value to assign to the predicate.
+     */
+    void SetPredicate(u64 pred, const std::string& value) {
+        using Tegra::Shader::Pred;
+        // Can't assign to the constant predicate.
+        ASSERT(pred != static_cast<u64>(Pred::UnusedIndex));
+
+        std::string variable = 'p' + std::to_string(pred);
+        shader.AddLine(variable + " = " + value + ';');
+        declr_predicates.insert(std::move(variable));
+    }
+
+    /*
+     * Returns the condition to use in the 'if' for a predicated instruction.
+     * @param instr Instruction to generate the if condition for.
+     * @returns string containing the predicate condition.
+     */
+    std::string GetPredicateCondition(Instruction instr) const {
+        using Tegra::Shader::Pred;
+        ASSERT(instr.pred.pred_index != static_cast<u64>(Pred::UnusedIndex));
+
+        std::string variable =
+            'p' + std::to_string(static_cast<u64>(instr.pred.pred_index.Value()));
+
+        if (instr.negate_pred) {
+            return "!(" + variable + ')';
+        }
+
+        return variable;
+    }
+
+    /*
+     * Returns whether the instruction at the specified offset is a 'sched' instruction.
+     * Sched instructions always appear before a sequence of 3 instructions.
+     */
+    bool IsSchedInstruction(u32 offset) const {
+        // sched instructions appear once every 4 instructions.
+        static constexpr size_t SchedPeriod = 4;
+        u32 absolute_offset = offset - main_offset;
+
+        return (absolute_offset % SchedPeriod) == 0;
+    }
+
     /**
      * Compiles a single instruction from Tegra to GLSL.
      * @param offset the offset of the Tegra shader instruction.
@@ -283,11 +332,33 @@ private:
      * + 1. If the current instruction always terminates the program, returns PROGRAM_END.
      */
     u32 CompileInstr(u32 offset) {
+        // Ignore sched instructions when generating code.
+        if (IsSchedInstruction(offset)) {
+            return offset + 1;
+        }
+
         const Instruction instr = {program_code[offset]};
+        const auto opcode = OpCode::Decode(instr);
 
-        shader.AddLine("// " + std::to_string(offset) + ": " + OpCode::GetInfo(instr.opcode).name);
+        // Decoding failure
+        if (!opcode) {
+            NGLOG_CRITICAL(HW_GPU, "Unhandled instruction: {0:x}", instr.value);
+            UNREACHABLE();
+        }
 
-        switch (OpCode::GetInfo(instr.opcode).type) {
+        shader.AddLine("// " + std::to_string(offset) + ": " + opcode->GetName());
+
+        using Tegra::Shader::Pred;
+        ASSERT_MSG(instr.pred.full_pred != Pred::NeverExecute,
+                   "NeverExecute predicate not implemented");
+
+        if (instr.pred.pred_index != static_cast<u64>(Pred::UnusedIndex)) {
+            shader.AddLine("if (" + GetPredicateCondition(instr) + ')');
+            shader.AddLine('{');
+            ++shader.scope;
+        }
+
+        switch (opcode->GetType()) {
         case OpCode::Type::Arithmetic: {
             std::string dest = GetRegister(instr.gpr0);
             std::string op_a = instr.alu.negate_a ? "-" : "";
@@ -312,7 +383,7 @@ private:
                 op_b = "abs(" + op_b + ")";
             }
 
-            switch (instr.opcode.EffectiveOpCode()) {
+            switch (opcode->GetId()) {
             case OpCode::Id::FMUL_C:
             case OpCode::Id::FMUL_R:
             case OpCode::Id::FMUL_IMM: {
@@ -354,16 +425,18 @@ private:
                     SetDest(0, dest, "min(" + op_a + "," + op_b + ")", 1, 1, instr.alu.abs_d);
                     break;
                 default:
-                    NGLOG_CRITICAL(HW_GPU, "Unhandled MUFU sub op: {}",
+                    NGLOG_CRITICAL(HW_GPU, "Unhandled MUFU sub op: {0:x}",
                                    static_cast<unsigned>(instr.sub_op.Value()));
                     UNREACHABLE();
                 }
                 break;
             }
+            case OpCode::Id::RRO: {
+                NGLOG_DEBUG(HW_GPU, "Skipping RRO instruction");
+                break;
+            }
             default: {
-                NGLOG_CRITICAL(HW_GPU, "Unhandled arithmetic instruction: {} ({}): {}",
-                               static_cast<unsigned>(instr.opcode.EffectiveOpCode()),
-                               OpCode::GetInfo(instr.opcode).name, instr.hex);
+                NGLOG_CRITICAL(HW_GPU, "Unhandled arithmetic instruction: {}", opcode->GetName());
                 UNREACHABLE();
             }
             }
@@ -375,7 +448,7 @@ private:
             std::string op_b = instr.ffma.negate_b ? "-" : "";
             std::string op_c = instr.ffma.negate_c ? "-" : "";
 
-            switch (instr.opcode.EffectiveOpCode()) {
+            switch (opcode->GetId()) {
             case OpCode::Id::FFMA_CR: {
                 op_b += GetUniform(instr.uniform);
                 op_c += GetRegister(instr.gpr39);
@@ -397,9 +470,7 @@ private:
                 break;
             }
             default: {
-                NGLOG_CRITICAL(HW_GPU, "Unhandled FFMA instruction: {} ({}): {}",
-                               static_cast<unsigned>(instr.opcode.EffectiveOpCode()),
-                               OpCode::GetInfo(instr.opcode).name, instr.hex);
+                NGLOG_CRITICAL(HW_GPU, "Unhandled FFMA instruction: {}", opcode->GetName());
                 UNREACHABLE();
             }
             }
@@ -411,7 +482,7 @@ private:
             std::string gpr0 = GetRegister(instr.gpr0);
             const Attribute::Index attribute = instr.attribute.fmt20.index;
 
-            switch (instr.opcode.EffectiveOpCode()) {
+            switch (opcode->GetId()) {
             case OpCode::Id::LD_A: {
                 ASSERT_MSG(instr.attribute.fmt20.size == 0, "untested");
                 SetDest(instr.attribute.fmt20.element, gpr0, GetInputAttribute(attribute), 1, 4);
@@ -442,22 +513,76 @@ private:
                 break;
             }
             default: {
-                NGLOG_CRITICAL(HW_GPU, "Unhandled memory instruction: {} ({}): {}",
-                               static_cast<unsigned>(instr.opcode.EffectiveOpCode()),
-                               OpCode::GetInfo(instr.opcode).name, instr.hex);
+                NGLOG_CRITICAL(HW_GPU, "Unhandled memory instruction: {}", opcode->GetName());
                 UNREACHABLE();
             }
             }
             break;
         }
+        case OpCode::Type::FloatPredicate: {
+            std::string op_a = instr.fsetp.neg_a ? "-" : "";
+            op_a += GetRegister(instr.gpr8);
 
+            if (instr.fsetp.abs_a) {
+                op_a = "abs(" + op_a + ')';
+            }
+
+            std::string op_b{};
+
+            if (instr.is_b_imm) {
+                if (instr.fsetp.neg_b) {
+                    // Only the immediate version of fsetp has a neg_b bit.
+                    op_b += '-';
+                }
+                op_b += '(' + GetImmediate19(instr) + ')';
+            } else {
+                if (instr.is_b_gpr) {
+                    op_b += GetRegister(instr.gpr20);
+                } else {
+                    op_b += GetUniform(instr.uniform);
+                }
+            }
+
+            if (instr.fsetp.abs_b) {
+                op_b = "abs(" + op_b + ')';
+            }
+
+            using Tegra::Shader::Pred;
+            ASSERT_MSG(instr.fsetp.pred0 == static_cast<u64>(Pred::UnusedIndex) &&
+                           instr.fsetp.pred39 == static_cast<u64>(Pred::UnusedIndex),
+                       "Compound predicates are not implemented");
+
+            // We can't use the constant predicate as destination.
+            ASSERT(instr.fsetp.pred3 != static_cast<u64>(Pred::UnusedIndex));
+
+            using Tegra::Shader::PredCondition;
+            switch (instr.fsetp.cond) {
+            case PredCondition::LessThan:
+                SetPredicate(instr.fsetp.pred3, '(' + op_a + ") < (" + op_b + ')');
+                break;
+            case PredCondition::Equal:
+                SetPredicate(instr.fsetp.pred3, '(' + op_a + ") == (" + op_b + ')');
+                break;
+            default:
+                NGLOG_CRITICAL(HW_GPU, "Unhandled predicate condition: {} (a: {}, b: {})",
+                               static_cast<unsigned>(instr.fsetp.cond.Value()), op_a, op_b);
+                UNREACHABLE();
+            }
+            break;
+        }
         default: {
-            switch (instr.opcode.EffectiveOpCode()) {
+            switch (opcode->GetId()) {
             case OpCode::Id::EXIT: {
+                ASSERT_MSG(instr.pred.pred_index == static_cast<u64>(Pred::UnusedIndex),
+                           "Predicated exits not implemented");
                 shader.AddLine("return true;");
                 offset = PROGRAM_END - 1;
                 break;
             }
+            case OpCode::Id::KIL: {
+                shader.AddLine("discard;");
+                break;
+            }
             case OpCode::Id::IPA: {
                 const auto& attribute = instr.attribute.fmt28;
                 std::string dest = GetRegister(instr.gpr0);
@@ -465,9 +590,7 @@ private:
                 break;
             }
             default: {
-                NGLOG_CRITICAL(HW_GPU, "Unhandled instruction: {} ({}): {}",
-                               static_cast<unsigned>(instr.opcode.EffectiveOpCode()),
-                               OpCode::GetInfo(instr.opcode).name, instr.hex);
+                NGLOG_CRITICAL(HW_GPU, "Unhandled instruction: {}", opcode->GetName());
                 UNREACHABLE();
             }
             }
@@ -476,6 +599,12 @@ private:
         }
         }
 
+        // Close the predicate condition scope.
+        if (instr.pred.pred_index != static_cast<u64>(Pred::UnusedIndex)) {
+            --shader.scope;
+            shader.AddLine('}');
+        }
+
         return offset + 1;
     }
 
@@ -605,6 +734,12 @@ private:
             declarations.AddNewLine();
             ++const_buffer_layout;
         }
+
+        declarations.AddNewLine();
+        for (const auto& pred : declr_predicates) {
+            declarations.AddLine("bool " + pred + " = false;");
+        }
+        declarations.AddNewLine();
     }
 
 private:
@@ -618,6 +753,7 @@ private:
 
     // Declarations
     std::set<std::string> declr_register;
+    std::set<std::string> declr_predicates;
     std::set<Attribute::Index> declr_input_attribute;
     std::set<Attribute::Index> declr_output_attribute;
     std::array<ConstBufferEntry, Maxwell3D::Regs::MaxConstBuffers> declr_const_buffers;

src/video_core/renderer_opengl/renderer_opengl.cpp

@@ -9,10 +9,13 @@
 #include <memory>
 #include <glad/glad.h>
 #include "common/assert.h"
+#include "common/bit_field.h"
 #include "common/logging/log.h"
 #include "core/core.h"
 #include "core/core_timing.h"
 #include "core/frontend/emu_window.h"
+#include "core/hw/hw.h"
+#include "core/hw/lcd.h"
 #include "core/memory.h"
 #include "core/settings.h"
 #include "core/tracer/recorder.h"