gl_shader_decompiler: Fix un/signed mismatch with SHL.

GPU: Implemented the F2I_R shader instruction.

src/core/hle/service/am/am.cpp

@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <cinttypes>
+#include <stack>
 #include "core/file_sys/filesystem.h"
 #include "core/hle/ipc_helpers.h"
 #include "core/hle/kernel/event.h"
@@ -154,7 +155,7 @@ ISelfController::ISelfController(std::shared_ptr<NVFlinger::NVFlinger> nvflinger
     RegisterHandlers(functions);
 
     launchable_event =
-        Kernel::Event::Create(Kernel::ResetType::OneShot, "ISelfController:LaunchableEvent");
+        Kernel::Event::Create(Kernel::ResetType::Sticky, "ISelfController:LaunchableEvent");
 }
 
 void ISelfController::SetFocusHandlingMode(Kernel::HLERequestContext& ctx) {
@@ -348,19 +349,100 @@ void ICommonStateGetter::GetPerformanceMode(Kernel::HLERequestContext& ctx) {
     NGLOG_WARNING(Service_AM, "(STUBBED) called");
 }
 
+class IStorageAccessor final : public ServiceFramework<IStorageAccessor> {
+public:
+    explicit IStorageAccessor(std::vector<u8> buffer)
+        : ServiceFramework("IStorageAccessor"), buffer(std::move(buffer)) {
+        static const FunctionInfo functions[] = {
+            {0, &IStorageAccessor::GetSize, "GetSize"},
+            {10, &IStorageAccessor::Write, "Write"},
+            {11, &IStorageAccessor::Read, "Read"},
+        };
+        RegisterHandlers(functions);
+    }
+
+private:
+    std::vector<u8> buffer;
+
+    void GetSize(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 4};
+
+        rb.Push(RESULT_SUCCESS);
+        rb.Push(static_cast<u64>(buffer.size()));
+
+        NGLOG_DEBUG(Service_AM, "called");
+    }
+
+    void Write(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+
+        const u64 offset{rp.Pop<u64>()};
+        const std::vector<u8> data{ctx.ReadBuffer()};
+
+        ASSERT(offset + data.size() <= buffer.size());
+
+        std::memcpy(&buffer[offset], data.data(), data.size());
+
+        IPC::ResponseBuilder rb{rp.MakeBuilder(2, 0, 0)};
+        rb.Push(RESULT_SUCCESS);
+
+        NGLOG_DEBUG(Service_AM, "called, offset={}", offset);
+    }
+
+    void Read(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+
+        const u64 offset{rp.Pop<u64>()};
+        const size_t size{ctx.GetWriteBufferSize()};
+
+        ASSERT(offset + size <= buffer.size());
+
+        ctx.WriteBuffer(buffer.data() + offset, size);
+
+        IPC::ResponseBuilder rb{rp.MakeBuilder(2, 0, 0)};
+        rb.Push(RESULT_SUCCESS);
+
+        NGLOG_DEBUG(Service_AM, "called, offset={}", offset);
+    }
+};
+
+class IStorage final : public ServiceFramework<IStorage> {
+public:
+    explicit IStorage(std::vector<u8> buffer)
+        : ServiceFramework("IStorage"), buffer(std::move(buffer)) {
+        static const FunctionInfo functions[] = {
+            {0, &IStorage::Open, "Open"},
+            {1, nullptr, "OpenTransferStorage"},
+        };
+        RegisterHandlers(functions);
+    }
+
+private:
+    std::vector<u8> buffer;
+
+    void Open(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+
+        rb.Push(RESULT_SUCCESS);
+        rb.PushIpcInterface<AM::IStorageAccessor>(buffer);
+
+        NGLOG_DEBUG(Service_AM, "called");
+    }
+};
+
 class ILibraryAppletAccessor final : public ServiceFramework<ILibraryAppletAccessor> {
 public:
     explicit ILibraryAppletAccessor() : ServiceFramework("ILibraryAppletAccessor") {
         static const FunctionInfo functions[] = {
             {0, &ILibraryAppletAccessor::GetAppletStateChangedEvent, "GetAppletStateChangedEvent"},
             {1, nullptr, "IsCompleted"},
-            {10, nullptr, "Start"},
+            {10, &ILibraryAppletAccessor::Start, "Start"},
             {20, nullptr, "RequestExit"},
             {25, nullptr, "Terminate"},
-            {30, nullptr, "GetResult"},
+            {30, &ILibraryAppletAccessor::GetResult, "GetResult"},
             {50, nullptr, "SetOutOfFocusApplicationSuspendingEnabled"},
-            {100, nullptr, "PushInData"},
-            {101, nullptr, "PopOutData"},
+            {100, &ILibraryAppletAccessor::PushInData, "PushInData"},
+            {101, &ILibraryAppletAccessor::PopOutData, "PopOutData"},
             {102, nullptr, "PushExtraStorage"},
             {103, nullptr, "PushInteractiveInData"},
             {104, nullptr, "PopInteractiveOutData"},
@@ -388,6 +470,41 @@ private:
         NGLOG_WARNING(Service_AM, "(STUBBED) called");
     }
 
+    void GetResult(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+
+        NGLOG_WARNING(Service_AM, "(STUBBED) called");
+    }
+
+    void Start(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2};
+        rb.Push(RESULT_SUCCESS);
+
+        NGLOG_WARNING(Service_AM, "(STUBBED) called");
+    }
+
+    void PushInData(Kernel::HLERequestContext& ctx) {
+        IPC::RequestParser rp{ctx};
+        storage_stack.push(rp.PopIpcInterface<AM::IStorage>());
+
+        IPC::ResponseBuilder rb{rp.MakeBuilder(2, 0, 0)};
+        rb.Push(RESULT_SUCCESS);
+
+        NGLOG_DEBUG(Service_AM, "called");
+    }
+
+    void PopOutData(Kernel::HLERequestContext& ctx) {
+        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
+        rb.Push(RESULT_SUCCESS);
+        rb.PushIpcInterface<AM::IStorage>(std::move(storage_stack.top()));
+
+        storage_stack.pop();
+
+        NGLOG_DEBUG(Service_AM, "called");
+    }
+
+    std::stack<std::shared_ptr<AM::IStorage>> storage_stack;
     Kernel::SharedPtr<Kernel::Event> state_changed_event;
 };
 
@@ -396,7 +513,7 @@ ILibraryAppletCreator::ILibraryAppletCreator() : ServiceFramework("ILibraryApple
         {0, &ILibraryAppletCreator::CreateLibraryApplet, "CreateLibraryApplet"},
         {1, nullptr, "TerminateAllLibraryApplets"},
         {2, nullptr, "AreAnyLibraryAppletsLeft"},
-        {10, nullptr, "CreateStorage"},
+        {10, &ILibraryAppletCreator::CreateStorage, "CreateStorage"},
         {11, nullptr, "CreateTransferMemoryStorage"},
         {12, nullptr, "CreateHandleStorage"},
     };
@@ -412,72 +529,17 @@ void ILibraryAppletCreator::CreateLibraryApplet(Kernel::HLERequestContext& ctx)
     NGLOG_DEBUG(Service_AM, "called");
 }
 
-class IStorageAccessor final : public ServiceFramework<IStorageAccessor> {
-public:
-    explicit IStorageAccessor(std::vector<u8> buffer)
-        : ServiceFramework("IStorageAccessor"), buffer(std::move(buffer)) {
-        static const FunctionInfo functions[] = {
-            {0, &IStorageAccessor::GetSize, "GetSize"},
-            {10, nullptr, "Write"},
-            {11, &IStorageAccessor::Read, "Read"},
-        };
-        RegisterHandlers(functions);
-    }
+void ILibraryAppletCreator::CreateStorage(Kernel::HLERequestContext& ctx) {
+    IPC::RequestParser rp{ctx};
+    const u64 size{rp.Pop<u64>()};
+    std::vector<u8> buffer(size);
 
-private:
-    std::vector<u8> buffer;
+    IPC::ResponseBuilder rb{rp.MakeBuilder(2, 0, 1)};
+    rb.Push(RESULT_SUCCESS);
+    rb.PushIpcInterface<AM::IStorage>(std::move(buffer));
 
-    void GetSize(Kernel::HLERequestContext& ctx) {
-        IPC::ResponseBuilder rb{ctx, 4};
-
-        rb.Push(RESULT_SUCCESS);
-        rb.Push(static_cast<u64>(buffer.size()));
-
-        NGLOG_DEBUG(Service_AM, "called");
-    }
-
-    void Read(Kernel::HLERequestContext& ctx) {
-        IPC::RequestParser rp{ctx};
-
-        u64 offset = rp.Pop<u64>();
-
-        const size_t size{ctx.GetWriteBufferSize()};
-
-        ASSERT(offset + size <= buffer.size());
-
-        ctx.WriteBuffer(buffer.data() + offset, size);
-
-        IPC::ResponseBuilder rb{ctx, 2};
-
-        rb.Push(RESULT_SUCCESS);
-
-        NGLOG_DEBUG(Service_AM, "called");
-    }
-};
-
-class IStorage final : public ServiceFramework<IStorage> {
-public:
-    explicit IStorage(std::vector<u8> buffer)
-        : ServiceFramework("IStorage"), buffer(std::move(buffer)) {
-        static const FunctionInfo functions[] = {
-            {0, &IStorage::Open, "Open"},
-            {1, nullptr, "OpenTransferStorage"},
-        };
-        RegisterHandlers(functions);
-    }
-
-private:
-    std::vector<u8> buffer;
-
-    void Open(Kernel::HLERequestContext& ctx) {
-        IPC::ResponseBuilder rb{ctx, 2, 0, 1};
-
-        rb.Push(RESULT_SUCCESS);
-        rb.PushIpcInterface<AM::IStorageAccessor>(buffer);
-
-        NGLOG_DEBUG(Service_AM, "called");
-    }
-};
+    NGLOG_DEBUG(Service_AM, "called, size={}", size);
+}
 
 IApplicationFunctions::IApplicationFunctions() : ServiceFramework("IApplicationFunctions") {
     static const FunctionInfo functions[] = {

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

@@ -121,6 +121,7 @@ public:
 
 private:
     void CreateLibraryApplet(Kernel::HLERequestContext& ctx);
+    void CreateStorage(Kernel::HLERequestContext& ctx);
 };
 
 class IApplicationFunctions final : public ServiceFramework<IApplicationFunctions> {

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

@@ -26,11 +26,19 @@ public:
 private:
     enum class IoctlCommand : u32_le {
         IocSetNVMAPfdCommand = 0x40044801,
+        IocAllocGPFIFOCommand = 0x40084805,
         IocSetClientDataCommand = 0x40084714,
         IocGetClientDataCommand = 0x80084715,
         IocZCullBind = 0xc010480b,
         IocSetErrorNotifierCommand = 0xC018480C,
         IocChannelSetPriorityCommand = 0x4004480D,
+        IocEnableCommand = 0x0000480E,
+        IocDisableCommand = 0x0000480F,
+        IocPreemptCommand = 0x00004810,
+        IocForceResetCommand = 0x00004811,
+        IocEventIdControlCommand = 0x40084812,
+        IocGetErrorNotificationCommand = 0xC0104817,
+        IocAllocGPFIFOExCommand = 0x40204818,
         IocAllocGPFIFOEx2Command = 0xC020481A,
         IocAllocObjCtxCommand = 0xC0104809,
         IocChannelGetWaitbaseCommand = 0xC0080003,
@@ -56,6 +64,12 @@ private:
     };
     static_assert(sizeof(IoctlChannelSetTimeout) == 4, "IoctlChannelSetTimeout is incorrect size");
 
+    struct IoctlAllocGPFIFO {
+        u32_le num_entries;
+        u32_le flags;
+    };
+    static_assert(sizeof(IoctlAllocGPFIFO) == 8, "IoctlAllocGPFIFO is incorrect size");
+
     struct IoctlClientData {
         u64_le data;
     };
@@ -76,12 +90,45 @@ private:
     };
     static_assert(sizeof(IoctlSetErrorNotifier) == 24, "IoctlSetErrorNotifier is incorrect size");
 
+    struct IoctlChannelSetPriority {
+        u32_le priority;
+    };
+    static_assert(sizeof(IoctlChannelSetPriority) == 4,
+                  "IoctlChannelSetPriority is incorrect size");
+
+    struct IoctlEventIdControl {
+        u32_le cmd; // 0=disable, 1=enable, 2=clear
+        u32_le id;
+    };
+    static_assert(sizeof(IoctlEventIdControl) == 8, "IoctlEventIdControl is incorrect size");
+
+    struct IoctlGetErrorNotification {
+        u64_le timestamp;
+        u32_le info32;
+        u16_le info16;
+        u16_le status; // always 0xFFFF
+    };
+    static_assert(sizeof(IoctlGetErrorNotification) == 16,
+                  "IoctlGetErrorNotification is incorrect size");
+
     struct IoctlFence {
         u32_le id;
         u32_le value;
     };
     static_assert(sizeof(IoctlFence) == 8, "IoctlFence is incorrect size");
 
+    struct IoctlAllocGpfifoEx {
+        u32_le num_entries;
+        u32_le flags;
+        u32_le unk0;
+        u32_le unk1;
+        u32_le unk2;
+        u32_le unk3;
+        u32_le unk4;
+        u32_le unk5;
+    };
+    static_assert(sizeof(IoctlAllocGpfifoEx) == 32, "IoctlAllocGpfifoEx is incorrect size");
+
     struct IoctlAllocGpfifoEx2 {
         u32_le num_entries;   // in
         u32_le flags;         // in

src/video_core/engines/maxwell_3d.h

@@ -354,10 +354,35 @@ public:
                     f32 scale_x;
                     f32 scale_y;
                     f32 scale_z;
-                    u32 translate_x;
-                    u32 translate_y;
-                    u32 translate_z;
+                    f32 translate_x;
+                    f32 translate_y;
+                    f32 translate_z;
                     INSERT_PADDING_WORDS(2);
+
+                    MathUtil::Rectangle<s32> GetRect() const {
+                        return {
+                            GetX(),               // left
+                            GetY() + GetHeight(), // top
+                            GetX() + GetWidth(),  // right
+                            GetY()                // bottom
+                        };
+                    };
+
+                    s32 GetX() const {
+                        return static_cast<s32>(std::max(0.0f, translate_x - std::fabs(scale_x)));
+                    }
+
+                    s32 GetY() const {
+                        return static_cast<s32>(std::max(0.0f, translate_y - std::fabs(scale_y)));
+                    }
+
+                    s32 GetWidth() const {
+                        return static_cast<s32>(translate_x + std::fabs(scale_x)) - GetX();
+                    }
+
+                    s32 GetHeight() const {
+                        return static_cast<s32>(translate_y + std::fabs(scale_y)) - GetY();
+                    }
                 } viewport_transform[NumViewports];
 
                 struct {
@@ -371,15 +396,6 @@ public:
                     };
                     float depth_range_near;
                     float depth_range_far;
-
-                    MathUtil::Rectangle<s32> GetRect() const {
-                        return {
-                            static_cast<s32>(x),          // left
-                            static_cast<s32>(y + height), // top
-                            static_cast<s32>(x + width),  // right
-                            static_cast<s32>(y)           // bottom
-                        };
-                    };
                 } viewport[NumViewports];
 
                 INSERT_PADDING_WORDS(0x1D);

src/video_core/engines/shader_bytecode.h

@@ -156,6 +156,13 @@ enum class PredOperation : u64 {
     Xor = 2,
 };
 
+enum class LogicOperation : u64 {
+    And = 0,
+    Or = 1,
+    Xor = 2,
+    PassB = 3,
+};
+
 enum class SubOp : u64 {
     Cos = 0x0,
     Sin = 0x1,
@@ -166,6 +173,13 @@ enum class SubOp : u64 {
     Min = 0x8,
 };
 
+enum class FloatRoundingOp : u64 {
+    None = 0,
+    Floor = 1,
+    Ceil = 2,
+    Trunc = 3,
+};
+
 union Instruction {
     Instruction& operator=(const Instruction& instr) {
         value = instr.value;
@@ -202,6 +216,12 @@ union Instruction {
             BitField<42, 1, u64> negate_pred;
         } fmnmx;
 
+        union {
+            BitField<53, 2, LogicOperation> operation;
+            BitField<55, 1, u64> invert_a;
+            BitField<56, 1, u64> invert_b;
+        } lop;
+
         float GetImm20_19() const {
             float result{};
             u32 imm{static_cast<u32>(imm20_19)};
@@ -217,8 +237,21 @@ union Instruction {
             std::memcpy(&result, &imm, sizeof(imm));
             return result;
         }
+
+        s32 GetSignedImm20_20() const {
+            u32 immediate = static_cast<u32>(imm20_19 | (negate_imm << 19));
+            // Sign extend the 20-bit value.
+            u32 mask = 1U << (20 - 1);
+            return static_cast<s32>((immediate ^ mask) - mask);
+        }
     } alu;
 
+    union {
+        BitField<39, 5, u64> shift_amount;
+        BitField<48, 1, u64> negate_b;
+        BitField<49, 1, u64> negate_a;
+    } iscadd;
+
     union {
         BitField<48, 1, u64> negate_b;
         BitField<49, 1, u64> negate_c;
@@ -238,6 +271,16 @@ union Instruction {
         BitField<56, 1, u64> neg_b;
     } fsetp;
 
+    union {
+        BitField<0, 3, u64> pred0;
+        BitField<3, 3, u64> pred3;
+        BitField<39, 3, u64> pred39;
+        BitField<42, 1, u64> neg_pred;
+        BitField<45, 2, PredOperation> op;
+        BitField<48, 1, u64> is_signed;
+        BitField<49, 3, PredCondition> cond;
+    } isetp;
+
     union {
         BitField<39, 3, u64> pred39;
         BitField<42, 1, u64> neg_pred;
@@ -245,20 +288,29 @@ union Instruction {
         BitField<44, 1, u64> abs_b;
         BitField<45, 2, PredOperation> op;
         BitField<48, 4, PredCondition> cond;
+        BitField<52, 1, u64> bf;
         BitField<53, 1, u64> neg_b;
         BitField<54, 1, u64> abs_a;
-        BitField<52, 1, u64> bf;
         BitField<55, 1, u64> ftz;
         BitField<56, 1, u64> neg_imm;
     } fset;
 
     union {
         BitField<10, 2, Register::Size> size;
-        BitField<13, 1, u64> is_signed;
+        BitField<12, 1, u64> is_output_signed;
+        BitField<13, 1, u64> is_input_signed;
         BitField<41, 2, u64> selector;
         BitField<45, 1, u64> negate_a;
         BitField<49, 1, u64> abs_a;
         BitField<50, 1, u64> saturate_a;
+
+        union {
+            BitField<39, 2, FloatRoundingOp> rounding;
+        } f2i;
+
+        union {
+            BitField<39, 4, u64> rounding;
+        } f2f;
     } conversion;
 
     union {
@@ -289,15 +341,16 @@ union Instruction {
     } texs;
 
     union {
-        BitField<20, 5, u64> target;
+        BitField<20, 24, u64> target;
         BitField<5, 1, u64> constant_buffer;
 
         s32 GetBranchTarget() const {
             // Sign extend the branch target offset
-            u32 mask = 1U << (5 - 1);
+            u32 mask = 1U << (24 - 1);
             u32 value = static_cast<u32>(target);
-            // The branch offset is relative to the next instruction, so add 1 to it.
-            return static_cast<s32>((value ^ mask) - mask) + 1;
+            // The branch offset is relative to the next instruction and is stored in bytes, so
+            // divide it by the size of an instruction and add 1 to it.
+            return static_cast<s32>((value ^ mask) - mask) / sizeof(Instruction) + 1;
         }
     } bra;
 
@@ -339,6 +392,9 @@ public:
         FMUL_R,
         FMUL_IMM,
         FMUL32_IMM,
+        ISCADD_C, // Scale and Add
+        ISCADD_R,
+        ISCADD_IMM,
         MUFU,  // Multi-Function Operator
         RRO_C, // Range Reduction Operator
         RRO_R,
@@ -360,6 +416,9 @@ public:
         MOV_R,
         MOV_IMM,
         MOV32_IMM,
+        SHL_C,
+        SHL_R,
+        SHL_IMM,
         SHR_C,
         SHR_R,
         SHR_IMM,
@@ -381,6 +440,9 @@ public:
     enum class Type {
         Trivial,
         Arithmetic,
+        Logic,
+        Shift,
+        ScaledAdd,
         Ffma,
         Flow,
         Memory,
@@ -504,6 +566,9 @@ private:
             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("0100110000011---", Id::ISCADD_C, Type::ScaledAdd, "ISCADD_C"),
+            INST("0101110000011---", Id::ISCADD_R, Type::ScaledAdd, "ISCADD_R"),
+            INST("0011100-00011---", Id::ISCADD_IMM, Type::ScaledAdd, "ISCADD_IMM"),
             INST("0101000010000---", Id::MUFU, Type::Arithmetic, "MUFU"),
             INST("0100110010010---", Id::RRO_C, Type::Arithmetic, "RRO_C"),
             INST("0101110010010---", Id::RRO_R, Type::Arithmetic, "RRO_R"),
@@ -511,20 +576,23 @@ private:
             INST("0100110010101---", Id::F2F_C, Type::Conversion, "F2F_C"),
             INST("0101110010101---", Id::F2F_R, Type::Conversion, "F2F_R"),
             INST("0011100-10101---", Id::F2F_IMM, Type::Conversion, "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("000001----------", Id::LOP32I, Type::Arithmetic, "LOP32I"),
+            INST("0100110010110---", Id::F2I_C, Type::Conversion, "F2I_C"),
+            INST("0101110010110---", Id::F2I_R, Type::Conversion, "F2I_R"),
+            INST("0011100-10110---", Id::F2I_IMM, Type::Conversion, "F2I_IMM"),
             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::MOV32_IMM, Type::Arithmetic, "MOV32_IMM"),
-            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("0100110001100---", Id::FMNMX_C, Type::Arithmetic, "FMNMX_C"),
             INST("0101110001100---", Id::FMNMX_R, Type::Arithmetic, "FMNMX_R"),
             INST("0011100-01100---", Id::FMNMX_IMM, Type::Arithmetic, "FMNMX_IMM"),
+            INST("000001----------", Id::LOP32I, Type::Logic, "LOP32I"),
+            INST("0100110001001---", Id::SHL_C, Type::Shift, "SHL_C"),
+            INST("0101110001001---", Id::SHL_R, Type::Shift, "SHL_R"),
+            INST("0011100-01001---", Id::SHL_IMM, Type::Shift, "SHL_IMM"),
+            INST("0100110000101---", Id::SHR_C, Type::Shift, "SHR_C"),
+            INST("0101110000101---", Id::SHR_R, Type::Shift, "SHR_R"),
+            INST("0011100-00101---", Id::SHR_IMM, Type::Shift, "SHR_IMM"),
             INST("0100110011100---", Id::I2I_C, Type::Conversion, "I2I_C"),
             INST("0101110011100---", Id::I2I_R, Type::Conversion, "I2I_R"),
             INST("01110001-1000---", Id::I2I_IMM, Type::Conversion, "I2I_IMM"),

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -298,7 +298,7 @@ void RasterizerOpenGL::DrawArrays() {
     const bool has_stencil = false;
     const bool using_color_fb = true;
     const bool using_depth_fb = false;
-    const MathUtil::Rectangle<s32> viewport_rect{regs.viewport[0].GetRect()};
+    const MathUtil::Rectangle<s32> viewport_rect{regs.viewport_transform[0].GetRect()};
 
     const bool write_color_fb =
         state.color_mask.red_enabled == GL_TRUE || state.color_mask.green_enabled == GL_TRUE ||
@@ -702,7 +702,7 @@ void RasterizerOpenGL::BindFramebufferSurfaces(const Surface& color_surface,
 
 void RasterizerOpenGL::SyncViewport(const MathUtil::Rectangle<u32>& surfaces_rect, u16 res_scale) {
     const auto& regs = Core::System().GetInstance().GPU().Maxwell3D().regs;
-    const MathUtil::Rectangle<s32> viewport_rect{regs.viewport[0].GetRect()};
+    const MathUtil::Rectangle<s32> viewport_rect{regs.viewport_transform[0].GetRect()};
 
     state.viewport.x = static_cast<GLint>(surfaces_rect.left) + viewport_rect.left * res_scale;
     state.viewport.y = static_cast<GLint>(surfaces_rect.bottom) + viewport_rect.bottom * res_scale;

src/video_core/renderer_opengl/gl_rasterizer_cache.cpp

@@ -933,7 +933,8 @@ Surface RasterizerCacheOpenGL::GetSurface(const SurfaceParams& params, ScaleMatc
     // Use GetSurfaceSubRect instead
     ASSERT(params.width == params.stride);
 
-    ASSERT(!params.is_tiled || (params.width % 8 == 0 && params.height % 8 == 0));
+    ASSERT(!params.is_tiled ||
+           (params.GetActualWidth() % 8 == 0 && params.GetActualHeight() % 8 == 0));
 
     // Check for an exact match in existing surfaces
     Surface surface =

src/video_core/renderer_opengl/gl_shader_decompiler.cpp

@@ -115,7 +115,16 @@ private:
             if (const auto opcode = OpCode::Decode(instr)) {
                 switch (opcode->GetId()) {
                 case OpCode::Id::EXIT: {
-                    return exit_method = ExitMethod::AlwaysEnd;
+                    // The EXIT instruction can be predicated, which means that the shader can
+                    // conditionally end on this instruction. We have to consider the case where the
+                    // condition is not met and check the exit method of that other basic block.
+                    using Tegra::Shader::Pred;
+                    if (instr.pred.pred_index == static_cast<u64>(Pred::UnusedIndex)) {
+                        return exit_method = ExitMethod::AlwaysEnd;
+                    } else {
+                        ExitMethod not_met = Scan(offset + 1, end, labels);
+                        return exit_method = ParallelExit(ExitMethod::AlwaysEnd, not_met);
+                    }
                 }
                 case OpCode::Id::BRA: {
                     u32 target = offset + instr.bra.GetBranchTarget();
@@ -219,6 +228,11 @@ public:
         return active_type == Type::Integer;
     }
 
+    /// Returns the current active type of the register
+    Type GetActiveType() const {
+        return active_type;
+    }
+
     /// Returns the index of the register
     size_t GetIndex() const {
         return index;
@@ -350,22 +364,28 @@ public:
         shader.AddLine(dest + " = " + src + ';');
     }
 
-    /// Generates code representing a uniform (C buffer) register.
-    std::string GetUniform(const Uniform& uniform, const Register& dest_reg) {
+    /// Generates code representing a uniform (C buffer) register, interpreted as the input type.
+    std::string GetUniform(const Uniform& uniform, GLSLRegister::Type type) {
         declr_const_buffers[uniform.index].MarkAsUsed(static_cast<unsigned>(uniform.index),
                                                       static_cast<unsigned>(uniform.offset), stage);
         std::string value =
             'c' + std::to_string(uniform.index) + '[' + std::to_string(uniform.offset) + ']';
 
-        if (regs[dest_reg].IsFloat()) {
+        if (type == GLSLRegister::Type::Float) {
             return value;
-        } else if (regs[dest_reg].IsInteger()) {
+        } else if (type == GLSLRegister::Type::Integer) {
             return "floatBitsToInt(" + value + ')';
         } else {
             UNREACHABLE();
         }
     }
 
+    /// Generates code representing a uniform (C buffer) register, interpreted as the type of the
+    /// destination register.
+    std::string GetUniform(const Uniform& uniform, const Register& dest_reg) {
+        return GetUniform(uniform, regs[dest_reg].GetActiveType());
+    }
+
     /// Add declarations for registers
     void GenerateDeclarations() {
         for (const auto& reg : regs) {
@@ -634,9 +654,9 @@ private:
     std::string GetPredicateComparison(Tegra::Shader::PredCondition condition) const {
         using Tegra::Shader::PredCondition;
         static const std::unordered_map<PredCondition, const char*> PredicateComparisonStrings = {
-            {PredCondition::LessThan, "<"},      {PredCondition::Equal, "=="},
-            {PredCondition::LessEqual, "<="},    {PredCondition::GreaterThan, ">"},
-            {PredCondition::GreaterEqual, ">="},
+            {PredCondition::LessThan, "<"},   {PredCondition::Equal, "=="},
+            {PredCondition::LessEqual, "<="}, {PredCondition::GreaterThan, ">"},
+            {PredCondition::NotEqual, "!="},  {PredCondition::GreaterEqual, ">="},
         };
 
         auto comparison = PredicateComparisonStrings.find(condition);
@@ -830,6 +850,102 @@ private:
             }
             break;
         }
+        case OpCode::Type::Logic: {
+            std::string op_a = regs.GetRegisterAsInteger(instr.gpr8, 0, false);
+
+            if (instr.alu.lop.invert_a)
+                op_a = "~(" + op_a + ')';
+
+            switch (opcode->GetId()) {
+            case OpCode::Id::LOP32I: {
+                u32 imm = static_cast<u32>(instr.alu.imm20_32.Value());
+
+                if (instr.alu.lop.invert_b)
+                    imm = ~imm;
+
+                switch (instr.alu.lop.operation) {
+                case Tegra::Shader::LogicOperation::And: {
+                    regs.SetRegisterToInteger(instr.gpr0, false, 0,
+                                              '(' + op_a + " & " + std::to_string(imm) + ')', 1, 1);
+                    break;
+                }
+                case Tegra::Shader::LogicOperation::Or: {
+                    regs.SetRegisterToInteger(instr.gpr0, false, 0,
+                                              '(' + op_a + " | " + std::to_string(imm) + ')', 1, 1);
+                    break;
+                }
+                case Tegra::Shader::LogicOperation::Xor: {
+                    regs.SetRegisterToInteger(instr.gpr0, false, 0,
+                                              '(' + op_a + " ^ " + std::to_string(imm) + ')', 1, 1);
+                    break;
+                }
+                default:
+                    NGLOG_CRITICAL(HW_GPU, "Unimplemented lop32i operation: {}",
+                                   static_cast<u32>(instr.alu.lop.operation.Value()));
+                    UNREACHABLE();
+                }
+                break;
+            }
+            default: {
+                NGLOG_CRITICAL(HW_GPU, "Unhandled logic instruction: {}", opcode->GetName());
+                UNREACHABLE();
+            }
+            }
+            break;
+        }
+
+        case OpCode::Type::Shift: {
+            std::string op_a = regs.GetRegisterAsInteger(instr.gpr8, 0, true);
+            std::string op_b;
+
+            if (instr.is_b_imm) {
+                op_b += '(' + std::to_string(instr.alu.GetSignedImm20_20()) + ')';
+            } else {
+                if (instr.is_b_gpr) {
+                    op_b += regs.GetRegisterAsInteger(instr.gpr20);
+                } else {
+                    op_b += regs.GetUniform(instr.uniform, GLSLRegister::Type::Integer);
+                }
+            }
+
+            switch (opcode->GetId()) {
+            case OpCode::Id::SHL_C:
+            case OpCode::Id::SHL_R:
+            case OpCode::Id::SHL_IMM:
+                regs.SetRegisterToInteger(instr.gpr0, true, 0, op_a + " << " + op_b, 1, 1);
+                break;
+            default: {
+                NGLOG_CRITICAL(HW_GPU, "Unhandled shift instruction: {}", opcode->GetName());
+                UNREACHABLE();
+            }
+            }
+            break;
+        }
+
+        case OpCode::Type::ScaledAdd: {
+            std::string op_a = regs.GetRegisterAsInteger(instr.gpr8);
+
+            if (instr.iscadd.negate_a)
+                op_a = '-' + op_a;
+
+            std::string op_b = instr.iscadd.negate_b ? "-" : "";
+
+            if (instr.is_b_imm) {
+                op_b += '(' + std::to_string(instr.alu.GetSignedImm20_20()) + ')';
+            } else {
+                if (instr.is_b_gpr) {
+                    op_b += regs.GetRegisterAsInteger(instr.gpr20);
+                } else {
+                    op_b += regs.GetUniform(instr.uniform, GLSLRegister::Type::Integer);
+                }
+            }
+
+            std::string shift = std::to_string(instr.iscadd.shift_amount.Value());
+
+            regs.SetRegisterToInteger(instr.gpr0, true, 0,
+                                      "((" + op_a + " << " + shift + ") + " + op_b + ')', 1, 1);
+            break;
+        }
         case OpCode::Type::Ffma: {
             std::string op_a = regs.GetRegisterAsFloat(instr.gpr8);
             std::string op_b = instr.ffma.negate_b ? "-" : "";
@@ -871,21 +987,35 @@ private:
             ASSERT_MSG(!instr.conversion.saturate_a, "Unimplemented");
 
             switch (opcode->GetId()) {
-            case OpCode::Id::I2I_R:
-            case OpCode::Id::I2F_R: {
+            case OpCode::Id::I2I_R: {
                 ASSERT_MSG(!instr.conversion.selector, "Unimplemented");
 
                 std::string op_a =
-                    regs.GetRegisterAsInteger(instr.gpr20, 0, instr.conversion.is_signed);
+                    regs.GetRegisterAsInteger(instr.gpr20, 0, instr.conversion.is_input_signed);
 
                 if (instr.conversion.abs_a) {
                     op_a = "abs(" + op_a + ')';
                 }
 
-                regs.SetRegisterToInteger(instr.gpr0, instr.conversion.is_signed, 0, op_a, 1, 1);
+                regs.SetRegisterToInteger(instr.gpr0, instr.conversion.is_output_signed, 0, op_a, 1,
+                                          1);
+                break;
+            }
+            case OpCode::Id::I2F_R: {
+                ASSERT_MSG(!instr.conversion.selector, "Unimplemented");
+                std::string op_a =
+                    regs.GetRegisterAsInteger(instr.gpr20, 0, instr.conversion.is_input_signed);
+
+                if (instr.conversion.abs_a) {
+                    op_a = "abs(" + op_a + ')';
+                }
+
+                regs.SetRegisterToFloat(instr.gpr0, 0, op_a, 1, 1);
                 break;
             }
             case OpCode::Id::F2F_R: {
+                // TODO(Subv): Implement rounding operations.
+                ASSERT_MSG(instr.conversion.f2f.rounding == 0, "Unimplemented rounding operation");
                 std::string op_a = regs.GetRegisterAsFloat(instr.gpr20);
 
                 if (instr.conversion.abs_a) {
@@ -895,6 +1025,43 @@ private:
                 regs.SetRegisterToFloat(instr.gpr0, 0, op_a, 1, 1);
                 break;
             }
+            case OpCode::Id::F2I_R: {
+                std::string op_a = regs.GetRegisterAsFloat(instr.gpr20);
+
+                if (instr.conversion.abs_a) {
+                    op_a = "abs(" + op_a + ')';
+                }
+
+                using Tegra::Shader::FloatRoundingOp;
+                switch (instr.conversion.f2i.rounding) {
+                case FloatRoundingOp::None:
+                    break;
+                case FloatRoundingOp::Floor:
+                    op_a = "floor(" + op_a + ')';
+                    break;
+                case FloatRoundingOp::Ceil:
+                    op_a = "ceil(" + op_a + ')';
+                    break;
+                case FloatRoundingOp::Trunc:
+                    op_a = "trunc(" + op_a + ')';
+                    break;
+                default:
+                    NGLOG_CRITICAL(HW_GPU, "Unimplemented f2i rounding mode {}",
+                                   static_cast<u32>(instr.conversion.f2i.rounding.Value()));
+                    UNREACHABLE();
+                    break;
+                }
+
+                if (instr.conversion.is_output_signed) {
+                    op_a = "int(" + op_a + ')';
+                } else {
+                    op_a = "uint(" + op_a + ')';
+                }
+
+                regs.SetRegisterToInteger(instr.gpr0, instr.conversion.is_output_signed, 0, op_a, 1,
+                                          1);
+                break;
+            }
             default: {
                 NGLOG_CRITICAL(HW_GPU, "Unhandled conversion instruction: {}", opcode->GetName());
                 UNREACHABLE();
@@ -1008,7 +1175,7 @@ private:
                 if (instr.is_b_gpr) {
                     op_b += regs.GetRegisterAsFloat(instr.gpr20);
                 } else {
-                    op_b += regs.GetUniform(instr.uniform, instr.gpr0);
+                    op_b += regs.GetUniform(instr.uniform, GLSLRegister::Type::Float);
                 }
             }
 
@@ -1039,6 +1206,42 @@ private:
             }
             break;
         }
+        case OpCode::Type::IntegerSetPredicate: {
+            std::string op_a = regs.GetRegisterAsInteger(instr.gpr8, 0, instr.isetp.is_signed);
+
+            std::string op_b{};
+
+            ASSERT_MSG(!instr.is_b_imm, "ISETP_IMM not implemented");
+
+            if (instr.is_b_gpr) {
+                op_b += regs.GetRegisterAsInteger(instr.gpr20, 0, instr.isetp.is_signed);
+            } else {
+                op_b += regs.GetUniform(instr.uniform, GLSLRegister::Type::Integer);
+            }
+
+            using Tegra::Shader::Pred;
+            // We can't use the constant predicate as destination.
+            ASSERT(instr.isetp.pred3 != static_cast<u64>(Pred::UnusedIndex));
+
+            std::string second_pred =
+                GetPredicateCondition(instr.isetp.pred39, instr.isetp.neg_pred != 0);
+
+            std::string comparator = GetPredicateComparison(instr.isetp.cond);
+            std::string combiner = GetPredicateCombiner(instr.isetp.op);
+
+            std::string predicate = '(' + op_a + ") " + comparator + " (" + op_b + ')';
+            // Set the primary predicate to the result of Predicate OP SecondPredicate
+            SetPredicate(instr.isetp.pred3,
+                         '(' + predicate + ") " + combiner + " (" + second_pred + ')');
+
+            if (instr.isetp.pred0 != static_cast<u64>(Pred::UnusedIndex)) {
+                // Set the secondary predicate to the result of !Predicate OP SecondPredicate,
+                // if enabled
+                SetPredicate(instr.isetp.pred0,
+                             "!(" + predicate + ") " + combiner + " (" + second_pred + ')');
+            }
+            break;
+        }
         case OpCode::Type::FloatSet: {
             std::string op_a = instr.fset.neg_a ? "-" : "";
             op_a += regs.GetRegisterAsFloat(instr.gpr8);
@@ -1059,7 +1262,7 @@ private:
                 if (instr.is_b_gpr) {
                     op_b += regs.GetRegisterAsFloat(instr.gpr20);
                 } else {
-                    op_b += regs.GetUniform(instr.uniform, instr.gpr0);
+                    op_b += regs.GetUniform(instr.uniform, GLSLRegister::Type::Float);
                 }
             }
 
@@ -1078,15 +1281,17 @@ private:
             std::string predicate = "(((" + op_a + ") " + comparator + " (" + op_b + ")) " +
                                     combiner + " (" + second_pred + "))";
 
-            regs.SetRegisterToFloat(instr.gpr0, 0, predicate + " ? 1.0 : 0.0", 1, 1);
+            if (instr.fset.bf) {
+                regs.SetRegisterToFloat(instr.gpr0, 0, predicate + " ? 1.0 : 0.0", 1, 1);
+            } else {
+                regs.SetRegisterToInteger(instr.gpr0, false, 0, predicate + " ? 0xFFFFFFFF : 0", 1,
+                                          1);
+            }
             break;
         }
         default: {
             switch (opcode->GetId()) {
             case OpCode::Id::EXIT: {
-                ASSERT_MSG(instr.pred.pred_index == static_cast<u64>(Pred::UnusedIndex),
-                           "Predicated exits not implemented");
-
                 // Final color output is currently hardcoded to GPR0-3 for fragment shaders
                 if (stage == Maxwell3D::Regs::ShaderStage::Fragment) {
                     shader.AddLine("color.r = " + regs.GetRegisterAsFloat(0) + ';');
@@ -1096,7 +1301,12 @@ private:
                 }
 
                 shader.AddLine("return true;");
-                offset = PROGRAM_END - 1;
+                if (instr.pred.pred_index == static_cast<u64>(Pred::UnusedIndex)) {
+                    // If this is an unconditional exit then just end processing here, otherwise we
+                    // have to account for the possibility of the condition not being met, so
+                    // continue processing the next instruction.
+                    offset = PROGRAM_END - 1;
+                }
                 break;
             }
             case OpCode::Id::KIL: {

src/yuzu/main.cpp

@@ -335,6 +335,24 @@ void GMainWindow::OnDisplayTitleBars(bool show) {
     }
 }
 
+bool GMainWindow::SupportsRequiredGLExtensions() {
+    QStringList unsupported_ext;
+
+    if (!GLAD_GL_ARB_program_interface_query)
+        unsupported_ext.append("ARB_program_interface_query");
+    if (!GLAD_GL_ARB_separate_shader_objects)
+        unsupported_ext.append("ARB_separate_shader_objects");
+    if (!GLAD_GL_ARB_shader_storage_buffer_object)
+        unsupported_ext.append("ARB_shader_storage_buffer_object");
+    if (!GLAD_GL_ARB_vertex_attrib_binding)
+        unsupported_ext.append("ARB_vertex_attrib_binding");
+
+    for (const QString& ext : unsupported_ext)
+        NGLOG_CRITICAL(Frontend, "Unsupported GL extension: {}", ext.toStdString());
+
+    return unsupported_ext.empty();
+}
+
 bool GMainWindow::LoadROM(const QString& filename) {
     // Shutdown previous session if the emu thread is still active...
     if (emu_thread != nullptr)
@@ -350,6 +368,14 @@ bool GMainWindow::LoadROM(const QString& filename) {
         return false;
     }
 
+    if (!SupportsRequiredGLExtensions()) {
+        QMessageBox::critical(
+            this, tr("Error while initializing OpenGL Core!"),
+            tr("Your GPU may not support one or more required OpenGL extensions. Please "
+               "ensure you have the latest graphics driver. See the log for more details."));
+        return false;
+    }
+
     Core::System& system{Core::System::GetInstance()};
 
     system.SetGPUDebugContext(debug_context);

src/yuzu/main.h

@@ -79,6 +79,7 @@ private:
     void ConnectWidgetEvents();
     void ConnectMenuEvents();
 
+    bool SupportsRequiredGLExtensions();
     bool LoadROM(const QString& filename);
     void BootGame(const QString& filename);
     void ShutdownGame();

src/yuzu_cmd/emu_window/emu_window_sdl2.cpp

@@ -78,6 +78,24 @@ void EmuWindow_SDL2::Fullscreen() {
     SDL_MaximizeWindow(render_window);
 }
 
+bool EmuWindow_SDL2::SupportsRequiredGLExtensions() {
+    std::vector<std::string> unsupported_ext;
+
+    if (!GLAD_GL_ARB_program_interface_query)
+        unsupported_ext.push_back("ARB_program_interface_query");
+    if (!GLAD_GL_ARB_separate_shader_objects)
+        unsupported_ext.push_back("ARB_separate_shader_objects");
+    if (!GLAD_GL_ARB_shader_storage_buffer_object)
+        unsupported_ext.push_back("ARB_shader_storage_buffer_object");
+    if (!GLAD_GL_ARB_vertex_attrib_binding)
+        unsupported_ext.push_back("ARB_vertex_attrib_binding");
+
+    for (const std::string& ext : unsupported_ext)
+        NGLOG_CRITICAL(Frontend, "Unsupported GL extension: {}", ext);
+
+    return unsupported_ext.empty();
+}
+
 EmuWindow_SDL2::EmuWindow_SDL2(bool fullscreen) {
     InputCommon::Init();
 
@@ -128,6 +146,11 @@ EmuWindow_SDL2::EmuWindow_SDL2(bool fullscreen) {
         exit(1);
     }
 
+    if (!SupportsRequiredGLExtensions()) {
+        NGLOG_CRITICAL(Frontend, "GPU does not support all required OpenGL extensions! Exiting...");
+        exit(1);
+    }
+
     OnResize();
     OnMinimalClientAreaChangeRequest(GetActiveConfig().min_client_area_size);
     SDL_PumpEvents();

src/yuzu_cmd/emu_window/emu_window_sdl2.h

@@ -46,6 +46,9 @@ private:
     /// Called when user passes the fullscreen parameter flag
     void Fullscreen();
 
+    /// Whether the GPU and driver supports the OpenGL extension required
+    bool SupportsRequiredGLExtensions();
+
     /// Called when a configuration change affects the minimal size of the window
     void OnMinimalClientAreaChangeRequest(
         const std::pair<unsigned, unsigned>& minimal_size) override;