21 files changed, 1037 insertions, 215 deletions

diff --git a/src/citra_qt/debugger/graphics_cmdlists.cpp b/src/citra_qt/debugger/graphics_cmdlists.cpp
index 7ac3ea542..35a3140b2 100644
--- a/src/citra_qt/debugger/graphics_cmdlists.cpp
+++ b/src/citra_qt/debugger/graphics_cmdlists.cpp
@@ -73,7 +73,7 @@ TextureInfoDockWidget::TextureInfoDockWidget(const Pica::DebugUtils::TextureInfo
     format_choice->addItem(tr("RGB565"));
     format_choice->addItem(tr("RGBA4"));
     format_choice->addItem(tr("IA8"));
-    format_choice->addItem(tr("UNK6"));
+    format_choice->addItem(tr("RG8"));
     format_choice->addItem(tr("I8"));
     format_choice->addItem(tr("A8"));
     format_choice->addItem(tr("IA4"));
@@ -175,29 +175,29 @@ int GPUCommandListModel::rowCount(const QModelIndex& parent) const {
 }
 
 int GPUCommandListModel::columnCount(const QModelIndex& parent) const {
-    return 3;
+    return 4;
 }
 
 QVariant GPUCommandListModel::data(const QModelIndex& index, int role) const {
     if (!index.isValid())
         return QVariant();
 
-    const auto& writes = pica_trace.writes;
-    const Pica::CommandProcessor::CommandHeader cmd{writes[index.row()].Id()};
-    const u32 val{writes[index.row()].Value()};
+    const auto& write = pica_trace.writes[index.row()];
 
     if (role == Qt::DisplayRole) {
         QString content;
         switch ( index.column() ) {
         case 0:
-            return QString::fromLatin1(Pica::Regs::GetCommandName(cmd.cmd_id).c_str());
+            return QString::fromLatin1(Pica::Regs::GetCommandName(write.cmd_id).c_str());
         case 1:
-            return QString("%1").arg(cmd.cmd_id, 3, 16, QLatin1Char('0'));
+            return QString("%1").arg(write.cmd_id, 3, 16, QLatin1Char('0'));
         case 2:
-            return QString("%1").arg(val, 8, 16, QLatin1Char('0'));
+            return QString("%1").arg(write.mask, 4, 2, QLatin1Char('0'));
+        case 3:
+            return QString("%1").arg(write.value, 8, 16, QLatin1Char('0'));
         }
     } else if (role == CommandIdRole) {
-        return QVariant::fromValue<int>(cmd.cmd_id.Value());
+        return QVariant::fromValue<int>(write.cmd_id);
     }
 
     return QVariant();
@@ -213,6 +213,8 @@ QVariant GPUCommandListModel::headerData(int section, Qt::Orientation orientatio
         case 1:
             return tr("Register");
         case 2:
+            return tr("Mask");
+        case 3:
             return tr("New Value");
         }
 
@@ -260,7 +262,7 @@ void GPUCommandListWidget::OnCommandDoubleClicked(const QModelIndex& index) {
 }
 
 void GPUCommandListWidget::SetCommandInfo(const QModelIndex& index) {
-    QWidget* new_info_widget;
+    QWidget* new_info_widget = nullptr;
 
     const unsigned int command_id = list_widget->model()->data(index, GPUCommandListModel::CommandIdRole).toUInt();
     if (COMMAND_IN_RANGE(command_id, texture0) ||
@@ -281,14 +283,15 @@ void GPUCommandListWidget::SetCommandInfo(const QModelIndex& index) {
         auto info = Pica::DebugUtils::TextureInfo::FromPicaRegister(config, format);
         u8* src = Memory::GetPhysicalPointer(config.GetPhysicalAddress());
         new_info_widget = new TextureInfoWidget(src, info);
-    } else {
-        new_info_widget = new QWidget;
     }
-
-    widget()->layout()->removeWidget(command_info_widget);
-    delete command_info_widget;
-    widget()->layout()->addWidget(new_info_widget);
-    command_info_widget = new_info_widget;
+    if (command_info_widget) {
+        delete command_info_widget;
+        command_info_widget = nullptr;
+    }
+    if (new_info_widget) {
+        widget()->layout()->addWidget(new_info_widget);
+        command_info_widget = new_info_widget;
+    }
 }
 #undef COMMAND_IN_RANGE
 
@@ -300,7 +303,9 @@ GPUCommandListWidget::GPUCommandListWidget(QWidget* parent) : QDockWidget(tr("Pi
 
     list_widget = new QTreeView;
     list_widget->setModel(model);
-    list_widget->setFont(QFont("monospace"));
+    QFont font("monospace");
+    font.setStyleHint(QFont::Monospace); // Automatic fallback to a monospace font on on platforms without a font called "monospace"
+    list_widget->setFont(font);
     list_widget->setRootIsDecorated(false);
     list_widget->setUniformRowHeights(true);
 
@@ -324,7 +329,7 @@ GPUCommandListWidget::GPUCommandListWidget(QWidget* parent) : QDockWidget(tr("Pi
 
     connect(copy_all, SIGNAL(clicked()), this, SLOT(CopyAllToClipboard()));
 
-    command_info_widget = new QWidget;
+    command_info_widget = nullptr;
 
     QVBoxLayout* main_layout = new QVBoxLayout;
     main_layout->addWidget(list_widget);
@@ -334,7 +339,6 @@ GPUCommandListWidget::GPUCommandListWidget(QWidget* parent) : QDockWidget(tr("Pi
         sub_layout->addWidget(copy_all);
         main_layout->addLayout(sub_layout);
     }
-    main_layout->addWidget(command_info_widget);
     main_widget->setLayout(main_layout);
 
     setWidget(main_widget);
diff --git a/src/citra_qt/debugger/graphics_vertex_shader.cpp b/src/citra_qt/debugger/graphics_vertex_shader.cpp
index 302e22d7a..0c17edee0 100644
--- a/src/citra_qt/debugger/graphics_vertex_shader.cpp
+++ b/src/citra_qt/debugger/graphics_vertex_shader.cpp
@@ -6,9 +6,16 @@
 #include <sstream>
 
 #include <QBoxLayout>
+#include <QFileDialog>
+#include <QGroupBox>
+#include <QLabel>
+#include <QLineEdit>
+#include <QPushButton>
+#include <QSignalMapper>
+#include <QSpinBox>
 #include <QTreeView>
 
-#include "video_core/shader/shader_interpreter.h"
+#include "video_core/shader/shader.h"
 
 #include "graphics_vertex_shader.h"
 
@@ -17,7 +24,7 @@ using nihstro::Instruction;
 using nihstro::SourceRegister;
 using nihstro::SwizzlePattern;
 
-GraphicsVertexShaderModel::GraphicsVertexShaderModel(QObject* parent): QAbstractItemModel(parent) {
+GraphicsVertexShaderModel::GraphicsVertexShaderModel(GraphicsVertexShaderWidget* parent): QAbstractItemModel(parent), par(parent) {
 
 }
 
@@ -34,7 +41,7 @@ int GraphicsVertexShaderModel::columnCount(const QModelIndex& parent) const {
 }
 
 int GraphicsVertexShaderModel::rowCount(const QModelIndex& parent) const {
-    return static_cast<int>(info.code.size());
+    return static_cast<int>(par->info.code.size());
 }
 
 QVariant GraphicsVertexShaderModel::headerData(int section, Qt::Orientation orientation, int role) const {
@@ -62,21 +69,21 @@ QVariant GraphicsVertexShaderModel::data(const QModelIndex& index, int role) con
     {
         switch (index.column()) {
         case 0:
-            if (info.HasLabel(index.row()))
-                return QString::fromStdString(info.GetLabel(index.row()));
+            if (par->info.HasLabel(index.row()))
+                return QString::fromStdString(par->info.GetLabel(index.row()));
 
             return QString("%1").arg(4*index.row(), 4, 16, QLatin1Char('0'));
 
         case 1:
-            return QString("%1").arg(info.code[index.row()].hex, 8, 16, QLatin1Char('0'));
+            return QString("%1").arg(par->info.code[index.row()].hex, 8, 16, QLatin1Char('0'));
 
         case 2:
         {
             std::stringstream output;
             output.flags(std::ios::hex);
 
-            Instruction instr = info.code[index.row()];
-            const SwizzlePattern& swizzle = info.swizzle_info[instr.common.operand_desc_id].pattern;
+            Instruction instr = par->info.code[index.row()];
+            const SwizzlePattern& swizzle = par->info.swizzle_info[instr.common.operand_desc_id].pattern;
 
             // longest known instruction name: "setemit "
             output << std::setw(8) << std::left << instr.opcode.Value().GetInfo().name;
@@ -130,13 +137,13 @@ QVariant GraphicsVertexShaderModel::data(const QModelIndex& index, int role) con
 
                     print_input_indexed_compact(output, src1, swizzle.negate_src1, swizzle.SelectorToString(false).substr(0,1), instr.common.AddressRegisterName());
                     output << " " << instr.common.compare_op.ToString(instr.common.compare_op.x) << " ";
-                    print_input(output, src2, swizzle.negate_src2, swizzle.SelectorToString(false).substr(0,1));
+                    print_input(output, src2, swizzle.negate_src2, swizzle.SelectorToString(true).substr(0,1));
 
                     output << ", ";
 
                     print_input_indexed_compact(output, src1, swizzle.negate_src1, swizzle.SelectorToString(false).substr(1,1), instr.common.AddressRegisterName());
                     output << " " << instr.common.compare_op.ToString(instr.common.compare_op.y) << " ";
-                    print_input(output, src2, swizzle.negate_src2, swizzle.SelectorToString(false).substr(1,1));
+                    print_input(output, src2, swizzle.negate_src2, swizzle.SelectorToString(true).substr(1,1));
 
                     break;
                 }
@@ -167,7 +174,7 @@ QVariant GraphicsVertexShaderModel::data(const QModelIndex& index, int role) con
                     // TODO: In some cases, the Address Register is used as an index for SRC2 instead of SRC1
                     if (instr.opcode.Value().GetInfo().subtype & OpCode::Info::Src2) {
                         SourceRegister src2 = instr.common.GetSrc2(src_is_inverted);
-                        print_input(output, src2, swizzle.negate_src2, swizzle.SelectorToString(false));
+                        print_input(output, src2, swizzle.negate_src2, swizzle.SelectorToString(true));
                     }
                     break;
                 }
@@ -240,6 +247,18 @@ QVariant GraphicsVertexShaderModel::data(const QModelIndex& index, int role) con
     case Qt::FontRole:
         return QFont("monospace");
 
+    case Qt::BackgroundRole:
+        // Highlight instructions which have no debug data associated to them
+        for (const auto& record : par->debug_data.records)
+            if (index.row() == record.instruction_offset)
+                return QVariant();
+
+        return QBrush(QColor(255, 255, 127));
+
+
+    // TODO: Draw arrows for each "reachable" instruction to visualize control flow
+
+
     default:
         break;
     }
@@ -247,53 +266,232 @@ QVariant GraphicsVertexShaderModel::data(const QModelIndex& index, int role) con
     return QVariant();
 }
 
-void GraphicsVertexShaderModel::OnUpdate()
-{
-    beginResetModel();
-
-    info.Clear();
-
-    for (auto instr : Pica::g_state.vs.program_code)
-        info.code.push_back({instr});
+void GraphicsVertexShaderWidget::DumpShader() {
+    QString filename = QFileDialog::getSaveFileName(this, tr("Save Shader Dump"), "shader_dump.shbin",
+                                                    tr("Shader Binary (*.shbin)"));
 
-    for (auto pattern : Pica::g_state.vs.swizzle_data)
-        info.swizzle_info.push_back({pattern});
+    if (filename.isEmpty()) {
+        // If the user canceled the dialog, don't dump anything.
+        return;
+    }
 
-    info.labels.insert({ Pica::g_state.regs.vs.main_offset, "main" });
+    auto& setup  = Pica::g_state.vs;
+    auto& config = Pica::g_state.regs.vs;
 
-    endResetModel();
+    Pica::DebugUtils::DumpShader(filename.toStdString(), config, setup, Pica::g_state.regs.vs_output_attributes);
 }
 
-
 GraphicsVertexShaderWidget::GraphicsVertexShaderWidget(std::shared_ptr< Pica::DebugContext > debug_context,
                                                        QWidget* parent)
         : BreakPointObserverDock(debug_context, "Pica Vertex Shader", parent) {
     setObjectName("PicaVertexShader");
 
-    auto binary_model = new GraphicsVertexShaderModel(this);
-    auto binary_list = new QTreeView;
-    binary_list->setModel(binary_model);
+    auto input_data_mapper = new QSignalMapper(this);
+
+    // TODO: Support inputting data in hexadecimal raw format
+    for (unsigned i = 0; i < ARRAY_SIZE(input_data); ++i) {
+        input_data[i] = new QLineEdit;
+        input_data[i]->setValidator(new QDoubleValidator(input_data[i]));
+    }
+
+    breakpoint_warning = new QLabel(tr("(data only available at VertexLoaded breakpoints)"));
+
+    // TODO: Add some button for jumping to the shader entry point
+
+    model = new GraphicsVertexShaderModel(this);
+    binary_list = new QTreeView;
+    binary_list->setModel(model);
     binary_list->setRootIsDecorated(false);
     binary_list->setAlternatingRowColors(true);
 
-    connect(this, SIGNAL(Update()), binary_model, SLOT(OnUpdate()));
+    auto dump_shader = new QPushButton(QIcon::fromTheme("document-save"), tr("Dump"));
+
+    instruction_description = new QLabel;
+
+    cycle_index = new QSpinBox;
+
+    connect(this, SIGNAL(SelectCommand(const QModelIndex&, QItemSelectionModel::SelectionFlags)),
+            binary_list->selectionModel(), SLOT(select(const QModelIndex&, QItemSelectionModel::SelectionFlags)));
+
+    connect(dump_shader, SIGNAL(clicked()), this, SLOT(DumpShader()));
+
+    connect(cycle_index, SIGNAL(valueChanged(int)), this, SLOT(OnCycleIndexChanged(int)));
+
+    for (unsigned i = 0; i < ARRAY_SIZE(input_data); ++i) {
+        connect(input_data[i], SIGNAL(textEdited(const QString&)), input_data_mapper, SLOT(map()));
+        input_data_mapper->setMapping(input_data[i], i);
+    }
+    connect(input_data_mapper, SIGNAL(mapped(int)), this, SLOT(OnInputAttributeChanged(int)));
 
     auto main_widget = new QWidget;
     auto main_layout = new QVBoxLayout;
     {
+        auto input_data_group = new QGroupBox(tr("Input Data"));
+
+        // For each vertex attribute, add a QHBoxLayout consisting of:
+        // - A QLabel denoting the source attribute index
+        // - Four QLineEdits for showing and manipulating attribute data
+        // - A QLabel denoting the shader input attribute index
+        auto sub_layout = new QVBoxLayout;
+        for (unsigned i = 0; i < 16; ++i) {
+            // Create an HBoxLayout to store the widgets used to specify a particular attribute
+            // and store it in a QWidget to allow for easy hiding and unhiding.
+            auto row_layout = new QHBoxLayout;
+            row_layout->addWidget(new QLabel(tr("Attribute %1").arg(i, 2)));
+            for (unsigned comp = 0; comp < 4; ++comp)
+                row_layout->addWidget(input_data[4 * i + comp]);
+
+            row_layout->addWidget(input_data_mapping[i] = new QLabel);
+
+            input_data_container[i] = new QWidget;
+            input_data_container[i]->setLayout(row_layout);
+            input_data_container[i]->hide();
+
+            sub_layout->addWidget(input_data_container[i]);
+        }
+
+        sub_layout->addWidget(breakpoint_warning);
+        breakpoint_warning->hide();
+
+        input_data_group->setLayout(sub_layout);
+        main_layout->addWidget(input_data_group);
+    }
+    {
         auto sub_layout = new QHBoxLayout;
         sub_layout->addWidget(binary_list);
         main_layout->addLayout(sub_layout);
     }
+    main_layout->addWidget(dump_shader);
+    {
+        auto sub_layout = new QHBoxLayout;
+        sub_layout->addWidget(new QLabel(tr("Cycle Index:")));
+        sub_layout->addWidget(cycle_index);
+        main_layout->addLayout(sub_layout);
+    }
+    main_layout->addWidget(instruction_description);
+    main_layout->addStretch();
     main_widget->setLayout(main_layout);
     setWidget(main_widget);
+
+    widget()->setEnabled(false);
 }
 
 void GraphicsVertexShaderWidget::OnBreakPointHit(Pica::DebugContext::Event event, void* data) {
-    emit Update();
+    auto input = static_cast<Pica::Shader::InputVertex*>(data);
+    if (event == Pica::DebugContext::Event::VertexLoaded) {
+        Reload(true, data);
+    } else {
+        // No vertex data is retrievable => invalidate currently stored vertex data
+        Reload(true, nullptr);
+    }
     widget()->setEnabled(true);
 }
 
+void GraphicsVertexShaderWidget::Reload(bool replace_vertex_data, void* vertex_data) {
+    model->beginResetModel();
+
+    if (replace_vertex_data) {
+        if (vertex_data) {
+            memcpy(&input_vertex, vertex_data, sizeof(input_vertex));
+            for (unsigned attr = 0; attr < 16; ++attr) {
+                for (unsigned comp = 0; comp < 4; ++comp) {
+                    input_data[4 * attr + comp]->setText(QString("%1").arg(input_vertex.attr[attr][comp].ToFloat32()));
+                }
+            }
+            breakpoint_warning->hide();
+        } else {
+            for (unsigned attr = 0; attr < 16; ++attr) {
+                for (unsigned comp = 0; comp < 4; ++comp) {
+                    input_data[4 * attr + comp]->setText(QString("???"));
+                }
+            }
+            breakpoint_warning->show();
+        }
+    }
+
+    // Reload shader code
+    info.Clear();
+
+    auto& shader_setup = Pica::g_state.vs;
+    auto& shader_config = Pica::g_state.regs.vs;
+    for (auto instr : shader_setup.program_code)
+        info.code.push_back({instr});
+
+    for (auto pattern : shader_setup.swizzle_data)
+        info.swizzle_info.push_back({pattern});
+
+    u32 entry_point = Pica::g_state.regs.vs.main_offset;
+    info.labels.insert({ entry_point, "main" });
+
+    // Generate debug information
+    debug_data = Pica::Shader::ProduceDebugInfo(input_vertex, 1, shader_config, shader_setup);
+
+    // Reload widget state
+
+    // Only show input attributes which are used as input to the shader
+    for (unsigned int attr = 0; attr < 16; ++attr) {
+        input_data_container[attr]->setVisible(false);
+    }
+    for (unsigned int attr = 0; attr < Pica::g_state.regs.vertex_attributes.GetNumTotalAttributes(); ++attr) {
+        unsigned source_attr = shader_config.input_register_map.GetRegisterForAttribute(attr);
+        input_data_mapping[source_attr]->setText(QString("-> v%1").arg(attr));
+        input_data_container[source_attr]->setVisible(true);
+    }
+
+    // Initialize debug info text for current cycle count
+    cycle_index->setMaximum(debug_data.records.size() - 1);
+    OnCycleIndexChanged(cycle_index->value());
+
+    model->endResetModel();
+}
+
 void GraphicsVertexShaderWidget::OnResumed() {
     widget()->setEnabled(false);
 }
+
+void GraphicsVertexShaderWidget::OnInputAttributeChanged(int index) {
+    float value = input_data[index]->text().toFloat();
+    Reload();
+}
+
+void GraphicsVertexShaderWidget::OnCycleIndexChanged(int index) {
+    QString text;
+
+    auto& record = debug_data.records[index];
+    if (record.mask & Pica::Shader::DebugDataRecord::SRC1)
+        text += tr("SRC1: %1, %2, %3, %4\n").arg(record.src1.x.ToFloat32()).arg(record.src1.y.ToFloat32()).arg(record.src1.z.ToFloat32()).arg(record.src1.w.ToFloat32());
+    if (record.mask & Pica::Shader::DebugDataRecord::SRC2)
+        text += tr("SRC2: %1, %2, %3, %4\n").arg(record.src2.x.ToFloat32()).arg(record.src2.y.ToFloat32()).arg(record.src2.z.ToFloat32()).arg(record.src2.w.ToFloat32());
+    if (record.mask & Pica::Shader::DebugDataRecord::SRC3)
+        text += tr("SRC3: %1, %2, %3, %4\n").arg(record.src3.x.ToFloat32()).arg(record.src3.y.ToFloat32()).arg(record.src3.z.ToFloat32()).arg(record.src3.w.ToFloat32());
+    if (record.mask & Pica::Shader::DebugDataRecord::DEST_IN)
+        text += tr("DEST_IN: %1, %2, %3, %4\n").arg(record.dest_in.x.ToFloat32()).arg(record.dest_in.y.ToFloat32()).arg(record.dest_in.z.ToFloat32()).arg(record.dest_in.w.ToFloat32());
+    if (record.mask & Pica::Shader::DebugDataRecord::DEST_OUT)
+        text += tr("DEST_OUT: %1, %2, %3, %4\n").arg(record.dest_out.x.ToFloat32()).arg(record.dest_out.y.ToFloat32()).arg(record.dest_out.z.ToFloat32()).arg(record.dest_out.w.ToFloat32());
+
+    if (record.mask & Pica::Shader::DebugDataRecord::ADDR_REG_OUT)
+        text += tr("Addres Registers: %1, %2\n").arg(record.address_registers[0]).arg(record.address_registers[1]);
+    if (record.mask & Pica::Shader::DebugDataRecord::CMP_RESULT)
+        text += tr("Compare Result: %1, %2\n").arg(record.conditional_code[0] ? "true" : "false").arg(record.conditional_code[1] ? "true" : "false");
+
+    if (record.mask & Pica::Shader::DebugDataRecord::COND_BOOL_IN)
+        text += tr("Static Condition: %1\n").arg(record.cond_bool ? "true" : "false");
+    if (record.mask & Pica::Shader::DebugDataRecord::COND_CMP_IN)
+        text += tr("Dynamic Conditions: %1, %2\n").arg(record.cond_cmp[0] ? "true" : "false").arg(record.cond_cmp[1] ? "true" : "false");
+    if (record.mask & Pica::Shader::DebugDataRecord::LOOP_INT_IN)
+        text += tr("Loop Parameters: %1 (repeats), %2 (initializer), %3 (increment), %4\n").arg(record.loop_int.x).arg(record.loop_int.y).arg(record.loop_int.z).arg(record.loop_int.w);
+
+    text += tr("Instruction offset: 0x%1").arg(4 * record.instruction_offset, 4, 16, QLatin1Char('0'));
+    if (record.mask & Pica::Shader::DebugDataRecord::NEXT_INSTR) {
+        text += tr(" -> 0x%2").arg(4 * record.next_instruction, 4, 16, QLatin1Char('0'));
+    } else {
+        text += tr(" (last instruction)");
+    }
+
+    instruction_description->setText(text);
+
+    // Scroll to current instruction
+    const QModelIndex& instr_index = model->index(record.instruction_offset, 0);
+    emit SelectCommand(instr_index, QItemSelectionModel::ClearAndSelect | QItemSelectionModel::Rows);
+    binary_list->scrollTo(instr_index, QAbstractItemView::EnsureVisible);
+}
diff --git a/src/citra_qt/debugger/graphics_vertex_shader.h b/src/citra_qt/debugger/graphics_vertex_shader.h
index 38339dc05..1b3f1f7ec 100644
--- a/src/citra_qt/debugger/graphics_vertex_shader.h
+++ b/src/citra_qt/debugger/graphics_vertex_shader.h
@@ -10,11 +10,18 @@
 
 #include "nihstro/parser_shbin.h"
 
+#include "video_core/shader/shader.h"
+
+class QLabel;
+class QSpinBox;
+
+class GraphicsVertexShaderWidget;
+
 class GraphicsVertexShaderModel : public QAbstractItemModel {
     Q_OBJECT
 
 public:
-    GraphicsVertexShaderModel(QObject* parent);
+    GraphicsVertexShaderModel(GraphicsVertexShaderWidget* parent);
 
     QModelIndex index(int row, int column, const QModelIndex& parent = QModelIndex()) const override;
     QModelIndex parent(const QModelIndex& child) const override;
@@ -23,11 +30,10 @@ public:
     QVariant data(const QModelIndex& index, int role = Qt::DisplayRole) const override;
     QVariant headerData(int section, Qt::Orientation orientation, int role = Qt::DisplayRole) const override;
 
-public slots:
-    void OnUpdate();
-
 private:
-    nihstro::ShaderInfo info;
+    GraphicsVertexShaderWidget* par;
+
+    friend class GraphicsVertexShaderWidget;
 };
 
 class GraphicsVertexShaderWidget : public BreakPointObserverDock {
@@ -43,9 +49,42 @@ private slots:
     void OnBreakPointHit(Pica::DebugContext::Event event, void* data) override;
     void OnResumed() override;
 
+    void OnInputAttributeChanged(int index);
+
+    void OnCycleIndexChanged(int index);
+
+    void DumpShader();
+
+    /**
+     * Reload widget based on the current PICA200 state
+     * @param replace_vertex_data If true, invalidate all current vertex data
+     * @param vertex_data New vertex data to use, as passed to OnBreakPointHit. May be nullptr to specify that no valid vertex data can be retrieved currently. Only used if replace_vertex_data is true.
+     */
+    void Reload(bool replace_vertex_data = false, void* vertex_data = nullptr);
+
+
 signals:
-    void Update();
+    // Call this to change the current command selection in the disassembly view
+    void SelectCommand(const QModelIndex&, QItemSelectionModel::SelectionFlags);
 
 private:
+    QLabel* instruction_description;
+    QTreeView* binary_list;
+    GraphicsVertexShaderModel* model;
+
+    /// TODO: Move these into a single struct
+    std::array<QLineEdit*, 4*16> input_data;  // A text box for each of the 4 components of up to 16 vertex attributes
+    std::array<QWidget*, 16> input_data_container; // QWidget containing the QLayout containing each vertex attribute
+    std::array<QLabel*, 16> input_data_mapping; // A QLabel denoting the shader input attribute which the vertex attribute maps to
+
+    // Text to be shown when input vertex data is not retrievable
+    QLabel* breakpoint_warning;
+
+    QSpinBox* cycle_index;
+
+    nihstro::ShaderInfo info;
+    Pica::Shader::DebugData<true> debug_data;
+    Pica::Shader::InputVertex input_vertex;
 
+    friend class GraphicsVertexShaderModel;
 };
diff --git a/src/citra_qt/main.cpp b/src/citra_qt/main.cpp
index a1a4865bd..8bf2a3e13 100644
--- a/src/citra_qt/main.cpp
+++ b/src/citra_qt/main.cpp
@@ -287,6 +287,17 @@ void GMainWindow::ShutdownGame() {
     render_window->hide();
 }
 
+void GMainWindow::StoreRecentFile(const QString& filename)
+{
+    QSettings settings;
+    QStringList recent_files = settings.value("recentFiles").toStringList();
+    recent_files.prepend(filename);
+    recent_files.removeDuplicates();
+    settings.setValue("recentFiles", recent_files);
+
+    UpdateRecentFiles();
+}
+
 void GMainWindow::UpdateRecentFiles() {
     QSettings settings;
     QStringList recent_files = settings.value("recentFiles").toStringList();
@@ -297,6 +308,7 @@ void GMainWindow::UpdateRecentFiles() {
         QString text = QString("&%1. %2").arg(i + 1).arg(QFileInfo(recent_files[i]).fileName());
         actions_recent_files[i]->setText(text);
         actions_recent_files[i]->setData(recent_files[i]);
+        actions_recent_files[i]->setToolTip(recent_files[i]);
         actions_recent_files[i]->setVisible(true);
     }
 
@@ -319,11 +331,7 @@ void GMainWindow::OnMenuLoadFile() {
     QString filename = QFileDialog::getOpenFileName(this, tr("Load File"), rom_path, tr("3DS executable (*.3ds *.3dsx *.elf *.axf *.cci *.cxi)"));
     if (filename.size()) {
         settings.setValue("romsPath", QFileInfo(filename).path());
-        // Update recent files list
-        QStringList recent_files = settings.value("recentFiles").toStringList();
-        recent_files.prepend(filename);
-        settings.setValue("recentFiles", recent_files);
-        UpdateRecentFiles(); // Update UI
+        StoreRecentFile(filename);
 
         BootGame(filename.toLatin1().data());
     }
@@ -349,6 +357,7 @@ void GMainWindow::OnMenuRecentFile() {
     QFileInfo file_info(filename);
     if (file_info.exists()) {
         BootGame(filename.toLatin1().data());
+        StoreRecentFile(filename); // Put the filename on top of the list
     } else {
         // Display an error message and remove the file from the list.
         QMessageBox::information(this, tr("File not found"), tr("File \"%1\" not found").arg(filename));
@@ -357,12 +366,7 @@ void GMainWindow::OnMenuRecentFile() {
         QStringList recent_files = settings.value("recentFiles").toStringList();
         recent_files.removeOne(filename);
         settings.setValue("recentFiles", recent_files);
-
-        action->setVisible(false);
-        // Grey out the recent files menu if the list is empty
-        if (ui.menu_recent_files->isEmpty()) {
-            ui.menu_recent_files->setEnabled(false);
-        }
+        UpdateRecentFiles();
     }
 }
 
diff --git a/src/citra_qt/main.h b/src/citra_qt/main.h
index 4b260ae8b..6f1292295 100644
--- a/src/citra_qt/main.h
+++ b/src/citra_qt/main.h
@@ -60,6 +60,24 @@ private:
     void BootGame(const std::string& filename);
     void ShutdownGame();
 
+    /**
+     * Stores the filename in the recently loaded files list.
+     * The new filename is stored at the beginning of the recently loaded files list.
+     * After inserting the new entry, duplicates are removed meaning that if
+     * this was inserted from \a OnMenuRecentFile(), the entry will be put on top
+     * and remove from its previous position.
+     *
+     * Finally, this function calls \a UpdateRecentFiles() to update the UI.
+     *
+     * @param filename the filename to store
+     */
+    void StoreRecentFile(const QString& filename);
+
+    /**
+     * Updates the recent files menu.
+     * Menu entries are rebuilt from the configuration file.
+     * If there is no entry in the menu, the menu is greyed out.
+     */
     void UpdateRecentFiles();
 
     void closeEvent(QCloseEvent* event) override;
diff --git a/src/common/color.h b/src/common/color.h
index 9dafdca0c..eb199e308 100644
--- a/src/common/color.h
+++ b/src/common/color.h
@@ -69,6 +69,15 @@ inline const Math::Vec4<u8> DecodeRGB8(const u8* bytes) {
 }
 
 /**
+ * Decode a color stored in RG8 (aka HILO8) format
+ * @param bytes Pointer to encoded source color
+ * @return Result color decoded as Math::Vec4<u8>
+ */
+inline const Math::Vec4<u8> DecodeRG8(const u8* bytes) {
+    return { bytes[1], bytes[0], 0, 255 };
+}
+
+/**
  * Decode a color stored in RGB565 format
  * @param bytes Pointer to encoded source color
  * @return Result color decoded as Math::Vec4<u8>
@@ -152,6 +161,15 @@ inline void EncodeRGB8(const Math::Vec4<u8>& color, u8* bytes) {
 }
 
 /**
+ * Encode a color as RG8 (aka HILO8) format
+ * @param color Source color to encode
+ * @param bytes Destination pointer to store encoded color
+ */
+inline void EncodeRG8(const Math::Vec4<u8>& color, u8* bytes) {
+    bytes[1] = color.r();
+    bytes[0] = color.g();
+}
+/**
  * Encode a color as RGB565 format
  * @param color Source color to encode
  * @param bytes Destination pointer to store encoded color
diff --git a/src/core/hle/service/gsp_gpu.cpp b/src/core/hle/service/gsp_gpu.cpp
index e93c1b436..3c41e656c 100644
--- a/src/core/hle/service/gsp_gpu.cpp
+++ b/src/core/hle/service/gsp_gpu.cpp
@@ -418,7 +418,7 @@ static void ExecuteCommand(const Command& command, u32 thread_id) {
 
     case CommandId::SET_DISPLAY_TRANSFER:
     {
-        auto& params = command.image_copy;
+        auto& params = command.display_transfer;
         WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.input_address)),
                 Memory::VirtualToPhysicalAddress(params.in_buffer_address) >> 3);
         WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.output_address)),
@@ -433,17 +433,22 @@ static void ExecuteCommand(const Command& command, u32 thread_id) {
     // TODO: Check if texture copies are implemented correctly..
     case CommandId::SET_TEXTURE_COPY:
     {
-        auto& params = command.image_copy;
-        WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.input_address)),
+        auto& params = command.texture_copy;
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.input_address),
                 Memory::VirtualToPhysicalAddress(params.in_buffer_address) >> 3);
-        WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.output_address)),
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.output_address),
                 Memory::VirtualToPhysicalAddress(params.out_buffer_address) >> 3);
-        WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.input_size)), params.in_buffer_size);
-        WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.output_size)), params.out_buffer_size);
-        WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.flags)), params.flags);
-
-        // TODO: Should this register be set to 1 or should instead its value be OR-ed with 1?
-        WriteGPURegister(static_cast<u32>(GPU_REG_INDEX(display_transfer_config.trigger)), 1);
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.texture_copy.size),
+                params.size);
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.texture_copy.input_size),
+                params.in_width_gap);
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.texture_copy.output_size),
+                params.out_width_gap);
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.flags),
+                params.flags);
+
+        // NOTE: Actual GSP ORs 1 with current register instead of overwriting. Doesn't seem to matter.
+        WriteGPURegister((u32)GPU_REG_INDEX(display_transfer_config.trigger), 1);
         break;
     }
 
diff --git a/src/core/hle/service/gsp_gpu.h b/src/core/hle/service/gsp_gpu.h
index c89d0a467..8bcb30ad1 100644
--- a/src/core/hle/service/gsp_gpu.h
+++ b/src/core/hle/service/gsp_gpu.h
@@ -127,7 +127,16 @@ struct Command {
             u32 in_buffer_size;
             u32 out_buffer_size;
             u32 flags;
-        } image_copy;
+        } display_transfer;
+
+        struct {
+            u32 in_buffer_address;
+            u32 out_buffer_address;
+            u32 size;
+            u32 in_width_gap;
+            u32 out_width_gap;
+            u32 flags;
+        } texture_copy;
 
         u8 raw_data[0x1C];
     };
diff --git a/src/core/hw/gpu.cpp b/src/core/hw/gpu.cpp
index 3ccbc03b2..68ae38289 100644
--- a/src/core/hw/gpu.cpp
+++ b/src/core/hw/gpu.cpp
@@ -3,6 +3,7 @@
 // Refer to the license.txt file included.
 
 #include <cstring>
+#include <numeric>
 #include <type_traits>
 
 #include "common/color.h"
@@ -158,14 +159,59 @@ inline void Write(u32 addr, const T data) {
             u8* src_pointer = Memory::GetPhysicalPointer(config.GetPhysicalInputAddress());
             u8* dst_pointer = Memory::GetPhysicalPointer(config.GetPhysicalOutputAddress());
 
+            if (config.is_texture_copy) {
+                u32 input_width = config.texture_copy.input_width * 16;
+                u32 input_gap = config.texture_copy.input_gap * 16;
+                u32 output_width = config.texture_copy.output_width * 16;
+                u32 output_gap = config.texture_copy.output_gap * 16;
+
+                size_t contiguous_input_size = config.texture_copy.size / input_width * (input_width + input_gap);
+                VideoCore::g_renderer->hw_rasterizer->NotifyPreRead(config.GetPhysicalInputAddress(), contiguous_input_size);
+
+                u32 remaining_size = config.texture_copy.size;
+                u32 remaining_input = input_width;
+                u32 remaining_output = output_width;
+                while (remaining_size > 0) {
+                    u32 copy_size = std::min({ remaining_input, remaining_output, remaining_size });
+
+                    std::memcpy(dst_pointer, src_pointer, copy_size);
+                    src_pointer += copy_size;
+                    dst_pointer += copy_size;
+
+                    remaining_input -= copy_size;
+                    remaining_output -= copy_size;
+                    remaining_size -= copy_size;
+
+                    if (remaining_input == 0) {
+                        remaining_input = input_width;
+                        src_pointer += input_gap;
+                    }
+                    if (remaining_output == 0) {
+                        remaining_output = output_width;
+                        dst_pointer += output_gap;
+                    }
+                }
+
+                LOG_TRACE(HW_GPU, "TextureCopy: 0x%X bytes from 0x%08X(%u+%u)-> 0x%08X(%u+%u), flags 0x%08X",
+                    config.texture_copy.size,
+                    config.GetPhysicalInputAddress(), input_width, input_gap,
+                    config.GetPhysicalOutputAddress(), output_width, output_gap,
+                    config.flags);
+
+                size_t contiguous_output_size = config.texture_copy.size / output_width * (output_width + output_gap);
+                VideoCore::g_renderer->hw_rasterizer->NotifyFlush(config.GetPhysicalOutputAddress(), contiguous_output_size);
+
+                GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PPF);
+                break;
+            }
+
             if (config.scaling > config.ScaleXY) {
                 LOG_CRITICAL(HW_GPU, "Unimplemented display transfer scaling mode %u", config.scaling.Value());
                 UNIMPLEMENTED();
                 break;
             }
 
-            if (config.output_tiled &&
-                    (config.scaling == config.ScaleXY || config.scaling == config.ScaleX)) {
+            if (config.input_linear && config.scaling != config.NoScale) {
                 LOG_CRITICAL(HW_GPU, "Scaling is only implemented on tiled input");
                 UNIMPLEMENTED();
                 break;
@@ -182,23 +228,6 @@ inline void Write(u32 addr, const T data) {
 
             VideoCore::g_renderer->hw_rasterizer->NotifyPreRead(config.GetPhysicalInputAddress(), input_size);
 
-            if (config.raw_copy) {
-                // Raw copies do not perform color conversion nor tiled->linear / linear->tiled conversions
-                // TODO(Subv): Verify if raw copies perform scaling
-                memcpy(dst_pointer, src_pointer, output_size);
-
-                LOG_TRACE(HW_GPU, "DisplayTriggerTransfer: 0x%08x bytes from 0x%08x(%ux%u)-> 0x%08x(%ux%u), output format: %x, flags 0x%08X, Raw copy",
-                    output_size,
-                    config.GetPhysicalInputAddress(), config.input_width.Value(), config.input_height.Value(),
-                    config.GetPhysicalOutputAddress(), config.output_width.Value(), config.output_height.Value(),
-                    config.output_format.Value(), config.flags);
-
-                GSP_GPU::SignalInterrupt(GSP_GPU::InterruptId::PPF);
-
-                VideoCore::g_renderer->hw_rasterizer->NotifyFlush(config.GetPhysicalOutputAddress(), output_size);
-                break;
-            }
-
             for (u32 y = 0; y < output_height; ++y) {
                 for (u32 x = 0; x < output_width; ++x) {
                     Math::Vec4<u8> src_color;
@@ -220,7 +249,7 @@ inline void Write(u32 addr, const T data) {
                     u32 src_offset;
                     u32 dst_offset;
 
-                    if (config.output_tiled) {
+                    if (config.input_linear) {
                         if (!config.dont_swizzle) {
                             // Interpret the input as linear and the output as tiled
                             u32 coarse_y = y & ~7;
diff --git a/src/core/hw/gpu.h b/src/core/hw/gpu.h
index daad506fe..2e3a9f779 100644
--- a/src/core/hw/gpu.h
+++ b/src/core/hw/gpu.h
@@ -201,12 +201,14 @@ struct Regs {
             u32 flags;
 
             BitField< 0, 1, u32> flip_vertically;  // flips input data vertically
-            BitField< 1, 1, u32> output_tiled;     // Converts from linear to tiled format
-            BitField< 3, 1, u32> raw_copy;         // Copies the data without performing any processing
+            BitField< 1, 1, u32> input_linear;     // Converts from linear to tiled format
+            BitField< 2, 1, u32> crop_input_lines;
+            BitField< 3, 1, u32> is_texture_copy;  // Copies the data without performing any processing and respecting texture copy fields
             BitField< 5, 1, u32> dont_swizzle;
             BitField< 8, 3, PixelFormat> input_format;
             BitField<12, 3, PixelFormat> output_format;
-
+            /// Uses some kind of 32x32 block swizzling mode, instead of the usual 8x8 one.
+            BitField<16, 1, u32> block_32; // TODO(yuriks): unimplemented
             BitField<24, 2, ScalingMode> scaling; // Determines the scaling mode of the transfer
         };
 
@@ -214,10 +216,30 @@ struct Regs {
 
         // it seems that writing to this field triggers the display transfer
         u32 trigger;
+
+        INSERT_PADDING_WORDS(0x1);
+
+        struct {
+            u32 size;
+
+            union {
+                u32 input_size;
+
+                BitField< 0, 16, u32> input_width;
+                BitField<16, 16, u32> input_gap;
+            };
+
+            union {
+                u32 output_size;
+
+                BitField< 0, 16, u32> output_width;
+                BitField<16, 16, u32> output_gap;
+            };
+        } texture_copy;
     } display_transfer_config;
-    ASSERT_MEMBER_SIZE(display_transfer_config, 0x1c);
+    ASSERT_MEMBER_SIZE(display_transfer_config, 0x2c);
 
-    INSERT_PADDING_WORDS(0x331);
+    INSERT_PADDING_WORDS(0x32D);
 
     struct {
         // command list size (in bytes)
diff --git a/src/core/loader/loader.cpp b/src/core/loader/loader.cpp
index f5b349a77..062291006 100644
--- a/src/core/loader/loader.cpp
+++ b/src/core/loader/loader.cpp
@@ -77,6 +77,8 @@ static const char* GetFileTypeString(FileType type) {
         return "NCSD";
     case FileType::CXI:
         return "NCCH";
+    case FileType::CIA:
+        return "CIA";
     case FileType::ELF:
         return "ELF";
     case FileType::THREEDSX:
@@ -134,6 +136,10 @@ ResultStatus LoadFile(const std::string& filename) {
         break;
     }
 
+    // CIA file format...
+    case FileType::CIA:
+        return ResultStatus::ErrorNotImplemented;
+
     // Error occurred durring IdentifyFile...
     case FileType::Error:
 
diff --git a/src/video_core/command_processor.cpp b/src/video_core/command_processor.cpp
index 374c4748d..d82e20f86 100644
--- a/src/video_core/command_processor.cpp
+++ b/src/video_core/command_processor.cpp
@@ -35,7 +35,15 @@ static u32 default_attr_write_buffer[3];
 
 Common::Profiling::TimingCategory category_drawing("Drawing");
 
-static inline void WritePicaReg(u32 id, u32 value, u32 mask) {
+// Expand a 4-bit mask to 4-byte mask, e.g. 0b0101 -> 0x00FF00FF
+static const u32 expand_bits_to_bytes[] = {
+    0x00000000, 0x000000ff, 0x0000ff00, 0x0000ffff,
+    0x00ff0000, 0x00ff00ff, 0x00ffff00, 0x00ffffff,
+    0xff000000, 0xff0000ff, 0xff00ff00, 0xff00ffff,
+    0xffff0000, 0xffff00ff, 0xffffff00, 0xffffffff
+};
+
+static void WritePicaReg(u32 id, u32 value, u32 mask) {
     auto& regs = g_state.regs;
 
     if (id >= regs.NumIds())
@@ -47,13 +55,16 @@ static inline void WritePicaReg(u32 id, u32 value, u32 mask) {
 
     // TODO: Figure out how register masking acts on e.g. vs.uniform_setup.set_value
     u32 old_value = regs[id];
-    regs[id] = (old_value & ~mask) | (value & mask);
+
+    const u32 write_mask = expand_bits_to_bytes[mask];
+
+    regs[id] = (old_value & ~write_mask) | (value & write_mask);
+
+    DebugUtils::OnPicaRegWrite({ (u16)id, (u16)mask, regs[id] });
 
     if (g_debug_context)
         g_debug_context->OnEvent(DebugContext::Event::PicaCommandLoaded, reinterpret_cast<void*>(&id));
 
-    DebugUtils::OnPicaRegWrite(id, regs[id]);
-
     switch(id) {
         // Trigger IRQ
         case PICA_REG_INDEX(trigger_irq):
@@ -215,7 +226,7 @@ static inline void WritePicaReg(u32 id, u32 value, u32 mask) {
             unsigned int vertex_cache_pos = 0;
             vertex_cache_ids.fill(-1);
 
-            Shader::UnitState shader_unit;
+            Shader::UnitState<false> shader_unit;
             Shader::Setup(shader_unit);
 
             for (unsigned int index = 0; index < regs.num_vertices; ++index)
@@ -469,13 +480,6 @@ void ProcessCommandList(const u32* list, u32 size) {
     g_state.cmd_list.length = size / sizeof(u32);
 
     while (g_state.cmd_list.current_ptr < g_state.cmd_list.head_ptr + g_state.cmd_list.length) {
-        // Expand a 4-bit mask to 4-byte mask, e.g. 0b0101 -> 0x00FF00FF
-        static const u32 expand_bits_to_bytes[] = {
-            0x00000000, 0x000000ff, 0x0000ff00, 0x0000ffff,
-            0x00ff0000, 0x00ff00ff, 0x00ffff00, 0x00ffffff,
-            0xff000000, 0xff0000ff, 0xff00ff00, 0xff00ffff,
-            0xffff0000, 0xffff00ff, 0xffffff00, 0xffffffff
-        };
 
         // Align read pointer to 8 bytes
         if ((g_state.cmd_list.head_ptr - g_state.cmd_list.current_ptr) % 2 != 0)
@@ -483,14 +487,13 @@ void ProcessCommandList(const u32* list, u32 size) {
 
         u32 value = *g_state.cmd_list.current_ptr++;
         const CommandHeader header = { *g_state.cmd_list.current_ptr++ };
-        const u32 write_mask = expand_bits_to_bytes[header.parameter_mask];
         u32 cmd = header.cmd_id;
 
-        WritePicaReg(cmd, value, write_mask);
+        WritePicaReg(cmd, value, header.parameter_mask);
 
         for (unsigned i = 0; i < header.extra_data_length; ++i) {
             u32 cmd = header.cmd_id + (header.group_commands ? i + 1 : 0);
-            WritePicaReg(cmd, *g_state.cmd_list.current_ptr++, write_mask);
+            WritePicaReg(cmd, *g_state.cmd_list.current_ptr++, header.parameter_mask);
          }
     }
 }
diff --git a/src/video_core/debug_utils/debug_utils.cpp b/src/video_core/debug_utils/debug_utils.cpp
index 572b4fd62..8ad77f0c8 100644
--- a/src/video_core/debug_utils/debug_utils.cpp
+++ b/src/video_core/debug_utils/debug_utils.cpp
@@ -4,9 +4,10 @@
 
 #include <algorithm>
 #include <condition_variable>
+#include <cstring>
+#include <fstream>
 #include <list>
 #include <map>
-#include <fstream>
 #include <mutex>
 #include <string>
 
@@ -14,6 +15,7 @@
 #include <png.h>
 #endif
 
+#include <nihstro/float24.h>
 #include <nihstro/shader_binary.h>
 
 #include "common/assert.h"
@@ -63,7 +65,7 @@ void DebugContext::OnEvent(Event event, void* data) {
 
 void DebugContext::Resume() {
     {
-        std::unique_lock<std::mutex> lock(breakpoint_mutex);
+        std::lock_guard<std::mutex> lock(breakpoint_mutex);
 
         // Tell all observers that we are about to resume
         for (auto& breakpoint_observer : breakpoint_observers) {
@@ -110,8 +112,7 @@ void GeometryDumper::Dump() {
 }
 
 
-void DumpShader(const u32* binary_data, u32 binary_size, const u32* swizzle_data, u32 swizzle_size,
-                u32 main_offset, const Regs::VSOutputAttributes* output_attributes)
+void DumpShader(const std::string& filename, const Regs::ShaderConfig& config, const State::ShaderSetup& setup, const Regs::VSOutputAttributes* output_attributes)
 {
     struct StuffToWrite {
         u8* pointer;
@@ -131,11 +132,14 @@ void DumpShader(const u32* binary_data, u32 binary_size, const u32* swizzle_data
     // into shbin format (separate type and component mask).
     union OutputRegisterInfo {
         enum Type : u64 {
-            POSITION = 0,
-            COLOR = 2,
-            TEXCOORD0 = 3,
-            TEXCOORD1 = 5,
-            TEXCOORD2 = 6,
+            POSITION   = 0,
+            QUATERNION = 1,
+            COLOR      = 2,
+            TEXCOORD0  = 3,
+            TEXCOORD1  = 5,
+            TEXCOORD2  = 6,
+
+            VIEW       = 8,
         };
 
         BitField< 0, 64, u64> hex;
@@ -157,6 +161,10 @@ void DumpShader(const u32* binary_data, u32 binary_size, const u32* swizzle_data
                 { OutputAttributes::POSITION_Y, { OutputRegisterInfo::POSITION, 2} },
                 { OutputAttributes::POSITION_Z, { OutputRegisterInfo::POSITION, 4} },
                 { OutputAttributes::POSITION_W, { OutputRegisterInfo::POSITION, 8} },
+                { OutputAttributes::QUATERNION_X, { OutputRegisterInfo::QUATERNION, 1} },
+                { OutputAttributes::QUATERNION_Y, { OutputRegisterInfo::QUATERNION, 2} },
+                { OutputAttributes::QUATERNION_Z, { OutputRegisterInfo::QUATERNION, 4} },
+                { OutputAttributes::QUATERNION_W, { OutputRegisterInfo::QUATERNION, 8} },
                 { OutputAttributes::COLOR_R, { OutputRegisterInfo::COLOR, 1} },
                 { OutputAttributes::COLOR_G, { OutputRegisterInfo::COLOR, 2} },
                 { OutputAttributes::COLOR_B, { OutputRegisterInfo::COLOR, 4} },
@@ -166,7 +174,10 @@ void DumpShader(const u32* binary_data, u32 binary_size, const u32* swizzle_data
                 { OutputAttributes::TEXCOORD1_U, { OutputRegisterInfo::TEXCOORD1, 1} },
                 { OutputAttributes::TEXCOORD1_V, { OutputRegisterInfo::TEXCOORD1, 2} },
                 { OutputAttributes::TEXCOORD2_U, { OutputRegisterInfo::TEXCOORD2, 1} },
-                { OutputAttributes::TEXCOORD2_V, { OutputRegisterInfo::TEXCOORD2, 2} }
+                { OutputAttributes::TEXCOORD2_V, { OutputRegisterInfo::TEXCOORD2, 2} },
+                { OutputAttributes::VIEW_X, { OutputRegisterInfo::VIEW, 1} },
+                { OutputAttributes::VIEW_Y, { OutputRegisterInfo::VIEW, 2} },
+                { OutputAttributes::VIEW_Z, { OutputRegisterInfo::VIEW, 4} }
             };
 
             for (const auto& semantic : std::vector<OutputAttributes::Semantic>{
@@ -221,28 +232,69 @@ void DumpShader(const u32* binary_data, u32 binary_size, const u32* swizzle_data
 
     // TODO: Reduce the amount of binary code written to relevant portions
     dvlp.binary_offset = write_offset - dvlp_offset;
-    dvlp.binary_size_words = binary_size;
-    QueueForWriting((u8*)binary_data, binary_size * sizeof(u32));
+    dvlp.binary_size_words = setup.program_code.size();
+    QueueForWriting((u8*)setup.program_code.data(), setup.program_code.size() * sizeof(u32));
 
     dvlp.swizzle_info_offset = write_offset - dvlp_offset;
-    dvlp.swizzle_info_num_entries = swizzle_size;
+    dvlp.swizzle_info_num_entries = setup.swizzle_data.size();
     u32 dummy = 0;
-    for (unsigned int i = 0; i < swizzle_size; ++i) {
-        QueueForWriting((u8*)&swizzle_data[i], sizeof(swizzle_data[i]));
+    for (unsigned int i = 0; i < setup.swizzle_data.size(); ++i) {
+        QueueForWriting((u8*)&setup.swizzle_data[i], sizeof(setup.swizzle_data[i]));
         QueueForWriting((u8*)&dummy, sizeof(dummy));
     }
 
-    dvle.main_offset_words = main_offset;
+    dvle.main_offset_words = config.main_offset;
     dvle.output_register_table_offset = write_offset - dvlb.dvle_offset;
     dvle.output_register_table_size = static_cast<u32>(output_info_table.size());
     QueueForWriting((u8*)output_info_table.data(), static_cast<u32>(output_info_table.size() * sizeof(OutputRegisterInfo)));
 
     // TODO: Create a label table for "main"
 
+    std::vector<nihstro::ConstantInfo> constant_table;
+    for (unsigned i = 0; i < setup.uniforms.b.size(); ++i) {
+        nihstro::ConstantInfo constant;
+        memset(&constant, 0, sizeof(constant));
+        constant.type = nihstro::ConstantInfo::Bool;
+        constant.regid = i;
+        constant.b = setup.uniforms.b[i];
+        constant_table.emplace_back(constant);
+    }
+    for (unsigned i = 0; i < setup.uniforms.i.size(); ++i) {
+        nihstro::ConstantInfo constant;
+        memset(&constant, 0, sizeof(constant));
+        constant.type = nihstro::ConstantInfo::Int;
+        constant.regid = i;
+        constant.i.x = setup.uniforms.i[i].x;
+        constant.i.y = setup.uniforms.i[i].y;
+        constant.i.z = setup.uniforms.i[i].z;
+        constant.i.w = setup.uniforms.i[i].w;
+        constant_table.emplace_back(constant);
+    }
+    for (unsigned i = 0; i < sizeof(setup.uniforms.f) / sizeof(setup.uniforms.f[0]); ++i) {
+        nihstro::ConstantInfo constant;
+        memset(&constant, 0, sizeof(constant));
+        constant.type = nihstro::ConstantInfo::Float;
+        constant.regid = i;
+        constant.f.x = nihstro::to_float24(setup.uniforms.f[i].x.ToFloat32());
+        constant.f.y = nihstro::to_float24(setup.uniforms.f[i].y.ToFloat32());
+        constant.f.z = nihstro::to_float24(setup.uniforms.f[i].z.ToFloat32());
+        constant.f.w = nihstro::to_float24(setup.uniforms.f[i].w.ToFloat32());
+
+        // Store constant if it's different from zero..
+        if (setup.uniforms.f[i].x.ToFloat32() != 0.0 ||
+            setup.uniforms.f[i].y.ToFloat32() != 0.0 ||
+            setup.uniforms.f[i].z.ToFloat32() != 0.0 ||
+            setup.uniforms.f[i].w.ToFloat32() != 0.0)
+            constant_table.emplace_back(constant);
+    }
+    dvle.constant_table_offset = write_offset - dvlb.dvle_offset;
+    dvle.constant_table_size = constant_table.size();
+    for (const auto& constant : constant_table) {
+        QueueForWriting((uint8_t*)&constant, sizeof(constant));
+    }
 
     // Write data to file
     static int dump_index = 0;
-    std::string filename = std::string("shader_dump") + std::to_string(++dump_index) + std::string(".shbin");
     std::ofstream file(filename, std::ios_base::out | std::ios_base::binary);
 
     for (auto& chunk : writing_queue) {
@@ -261,11 +313,10 @@ void StartPicaTracing()
         return;
     }
 
-    pica_trace_mutex.lock();
+    std::lock_guard<std::mutex> lock(pica_trace_mutex);
     pica_trace = std::unique_ptr<PicaTrace>(new PicaTrace);
 
     is_pica_tracing = true;
-    pica_trace_mutex.unlock();
 }
 
 bool IsPicaTracing()
@@ -273,18 +324,18 @@ bool IsPicaTracing()
     return is_pica_tracing != 0;
 }
 
-void OnPicaRegWrite(u32 id, u32 value)
+void OnPicaRegWrite(PicaTrace::Write write)
 {
     // Double check for is_pica_tracing to avoid pointless locking overhead
     if (!is_pica_tracing)
         return;
 
-    std::unique_lock<std::mutex> lock(pica_trace_mutex);
+    std::lock_guard<std::mutex> lock(pica_trace_mutex);
 
     if (!is_pica_tracing)
         return;
 
-    pica_trace->writes.emplace_back(id, value);
+    pica_trace->writes.push_back(write);
 }
 
 std::unique_ptr<PicaTrace> FinishPicaTracing()
@@ -298,9 +349,9 @@ std::unique_ptr<PicaTrace> FinishPicaTracing()
     is_pica_tracing = false;
 
     // Wait until running tracing is finished
-    pica_trace_mutex.lock();
+    std::lock_guard<std::mutex> lock(pica_trace_mutex);
     std::unique_ptr<PicaTrace> ret(std::move(pica_trace));
-    pica_trace_mutex.unlock();
+
     return std::move(ret);
 }
 
@@ -359,6 +410,12 @@ const Math::Vec4<u8> LookupTexture(const u8* source, int x, int y, const Texture
         }
     }
 
+    case Regs::TextureFormat::RG8:
+    {
+        auto res = Color::DecodeRG8(source + VideoCore::GetMortonOffset(x, y, 2));
+        return { res.r(), res.g(), 0, 255 };
+    }
+
     case Regs::TextureFormat::I8:
     {
         const u8* source_ptr = source + VideoCore::GetMortonOffset(x, y, 1);
diff --git a/src/video_core/debug_utils/debug_utils.h b/src/video_core/debug_utils/debug_utils.h
index 81eea30a9..85762f5b4 100644
--- a/src/video_core/debug_utils/debug_utils.h
+++ b/src/video_core/debug_utils/debug_utils.h
@@ -158,7 +158,6 @@ extern std::shared_ptr<DebugContext> g_debug_context; // TODO: Get rid of this g
 namespace DebugUtils {
 
 #define PICA_DUMP_GEOMETRY 0
-#define PICA_DUMP_SHADERS 0
 #define PICA_DUMP_TEXTURES 0
 #define PICA_LOG_TEV 0
 
@@ -182,27 +181,23 @@ private:
     std::vector<Face> faces;
 };
 
-void DumpShader(const u32* binary_data, u32 binary_size, const u32* swizzle_data, u32 swizzle_size,
-                u32 main_offset, const Regs::VSOutputAttributes* output_attributes);
+void DumpShader(const std::string& filename, const Regs::ShaderConfig& config,
+                const State::ShaderSetup& setup, const Regs::VSOutputAttributes* output_attributes);
 
 
 // Utility class to log Pica commands.
 struct PicaTrace {
-    struct Write : public std::pair<u32,u32> {
-        Write(u32 id, u32 value) : std::pair<u32,u32>(id, value) {}
-
-        u32& Id() { return first; }
-        const u32& Id() const { return first; }
-
-        u32& Value() { return second; }
-        const u32& Value() const { return second; }
+    struct Write {
+        u16 cmd_id;
+        u16 mask;
+        u32 value;
     };
     std::vector<Write> writes;
 };
 
 void StartPicaTracing();
 bool IsPicaTracing();
-void OnPicaRegWrite(u32 id, u32 value);
+void OnPicaRegWrite(PicaTrace::Write write);
 std::unique_ptr<PicaTrace> FinishPicaTracing();
 
 struct TextureInfo {
diff --git a/src/video_core/pica.h b/src/video_core/pica.h
index 6ce90f95a..58b924f9e 100644
--- a/src/video_core/pica.h
+++ b/src/video_core/pica.h
@@ -80,6 +80,11 @@ struct Regs {
             POSITION_Z   =  2,
             POSITION_W   =  3,
 
+            QUATERNION_X =  4,
+            QUATERNION_Y =  5,
+            QUATERNION_Z =  6,
+            QUATERNION_W =  7,
+
             COLOR_R      =  8,
             COLOR_G      =  9,
             COLOR_B      = 10,
@@ -89,6 +94,12 @@ struct Regs {
             TEXCOORD0_V  = 13,
             TEXCOORD1_U  = 14,
             TEXCOORD1_V  = 15,
+
+            // TODO: Not verified
+            VIEW_X       = 18,
+            VIEW_Y       = 19,
+            VIEW_Z       = 20,
+
             TEXCOORD2_U  = 22,
             TEXCOORD2_V  = 23,
 
@@ -163,7 +174,7 @@ struct Regs {
         RGB565       =  3,
         RGBA4        =  4,
         IA8          =  5,
-
+        RG8          =  6,  ///< @note Also called HILO8 in 3DBrew.
         I8           =  7,
         A8           =  8,
         IA4          =  9,
@@ -204,6 +215,7 @@ struct Regs {
         case TextureFormat::RGB565:
         case TextureFormat::RGBA4:
         case TextureFormat::IA8:
+        case TextureFormat::RG8:
             return 4;
 
         case TextureFormat::I4:
diff --git a/src/video_core/shader/shader.cpp b/src/video_core/shader/shader.cpp
index 6a27a8015..4e9836c80 100644
--- a/src/video_core/shader/shader.cpp
+++ b/src/video_core/shader/shader.cpp
@@ -5,6 +5,8 @@
 #include <memory>
 #include <unordered_map>
 
+#include <boost/range/algorithm/fill.hpp>
+
 #include "common/hash.h"
 #include "common/make_unique.h"
 #include "common/profiler.h"
@@ -30,7 +32,7 @@ static JitCompiler jit;
 static CompiledShader* jit_shader;
 #endif // ARCHITECTURE_x86_64
 
-void Setup(UnitState& state) {
+void Setup(UnitState<false>& state) {
 #ifdef ARCHITECTURE_x86_64
     if (VideoCore::g_shader_jit_enabled) {
         u64 cache_key = (Common::ComputeHash64(&g_state.vs.program_code, sizeof(g_state.vs.program_code)) ^
@@ -54,9 +56,8 @@ void Shutdown() {
 
 static Common::Profiling::TimingCategory shader_category("Vertex Shader");
 
-OutputVertex Run(UnitState& state, const InputVertex& input, int num_attributes) {
+OutputVertex Run(UnitState<false>& state, const InputVertex& input, int num_attributes) {
     auto& config = g_state.regs.vs;
-    auto& setup = g_state.vs;
 
     Common::Profiling::ScopeTimer timer(shader_category);
 
@@ -67,6 +68,8 @@ OutputVertex Run(UnitState& state, const InputVertex& input, int num_attributes)
     // Setup input register table
     const auto& attribute_register_map = config.input_register_map;
 
+    // TODO: Instead of this cumbersome logic, just load the input data directly like
+    // for (int attr = 0; attr < num_attributes; ++attr) { input_attr[0] = state.registers.input[attribute_register_map.attribute0_register]; }
     if (num_attributes > 0) state.registers.input[attribute_register_map.attribute0_register] = input.attr[0];
     if (num_attributes > 1) state.registers.input[attribute_register_map.attribute1_register] = input.attr[1];
     if (num_attributes > 2) state.registers.input[attribute_register_map.attribute2_register] = input.attr[2];
@@ -96,12 +99,6 @@ OutputVertex Run(UnitState& state, const InputVertex& input, int num_attributes)
     RunInterpreter(state);
 #endif // ARCHITECTURE_x86_64
 
-#if PICA_DUMP_SHADERS
-    DebugUtils::DumpShader(setup.program_code.data(), state.debug.max_offset, setup.swizzle_data.data(),
-        state.debug.max_opdesc_id, config.main_offset,
-        g_state.regs.vs_output_attributes); // TODO: Don't hardcode VS here
-#endif
-
     // Setup output data
     OutputVertex ret;
     // TODO(neobrain): Under some circumstances, up to 16 attributes may be output. We need to
@@ -132,14 +129,52 @@ OutputVertex Run(UnitState& state, const InputVertex& input, int num_attributes)
             std::fmin(std::fabs(ret.color[i].ToFloat32()), 1.0f));
     }
 
-    LOG_TRACE(Render_Software, "Output vertex: pos (%.2f, %.2f, %.2f, %.2f), col(%.2f, %.2f, %.2f, %.2f), tc0(%.2f, %.2f)",
+    LOG_TRACE(Render_Software, "Output vertex: pos (%.2f, %.2f, %.2f, %.2f), quat (%.2f, %.2f, %.2f, %.2f), col(%.2f, %.2f, %.2f, %.2f), tc0(%.2f, %.2f)",
         ret.pos.x.ToFloat32(), ret.pos.y.ToFloat32(), ret.pos.z.ToFloat32(), ret.pos.w.ToFloat32(),
+        ret.quat.x.ToFloat32(), ret.quat.y.ToFloat32(), ret.quat.z.ToFloat32(), ret.quat.w.ToFloat32(),
         ret.color.x.ToFloat32(), ret.color.y.ToFloat32(), ret.color.z.ToFloat32(), ret.color.w.ToFloat32(),
         ret.tc0.u().ToFloat32(), ret.tc0.v().ToFloat32());
 
     return ret;
 }
 
+DebugData<true> ProduceDebugInfo(const InputVertex& input, int num_attributes, const Regs::ShaderConfig& config, const State::ShaderSetup& setup) {
+    UnitState<true> state;
+
+    const auto& shader_memory = setup.program_code;
+    state.program_counter = config.main_offset;
+    state.debug.max_offset = 0;
+    state.debug.max_opdesc_id = 0;
+
+    // Setup input register table
+    const auto& attribute_register_map = config.input_register_map;
+    float24 dummy_register;
+    boost::fill(state.registers.input, &dummy_register);
+
+    if (num_attributes > 0) state.registers.input[attribute_register_map.attribute0_register] = &input.attr[0].x;
+    if (num_attributes > 1) state.registers.input[attribute_register_map.attribute1_register] = &input.attr[1].x;
+    if (num_attributes > 2) state.registers.input[attribute_register_map.attribute2_register] = &input.attr[2].x;
+    if (num_attributes > 3) state.registers.input[attribute_register_map.attribute3_register] = &input.attr[3].x;
+    if (num_attributes > 4) state.registers.input[attribute_register_map.attribute4_register] = &input.attr[4].x;
+    if (num_attributes > 5) state.registers.input[attribute_register_map.attribute5_register] = &input.attr[5].x;
+    if (num_attributes > 6) state.registers.input[attribute_register_map.attribute6_register] = &input.attr[6].x;
+    if (num_attributes > 7) state.registers.input[attribute_register_map.attribute7_register] = &input.attr[7].x;
+    if (num_attributes > 8) state.registers.input[attribute_register_map.attribute8_register] = &input.attr[8].x;
+    if (num_attributes > 9) state.registers.input[attribute_register_map.attribute9_register] = &input.attr[9].x;
+    if (num_attributes > 10) state.registers.input[attribute_register_map.attribute10_register] = &input.attr[10].x;
+    if (num_attributes > 11) state.registers.input[attribute_register_map.attribute11_register] = &input.attr[11].x;
+    if (num_attributes > 12) state.registers.input[attribute_register_map.attribute12_register] = &input.attr[12].x;
+    if (num_attributes > 13) state.registers.input[attribute_register_map.attribute13_register] = &input.attr[13].x;
+    if (num_attributes > 14) state.registers.input[attribute_register_map.attribute14_register] = &input.attr[14].x;
+    if (num_attributes > 15) state.registers.input[attribute_register_map.attribute15_register] = &input.attr[15].x;
+
+    state.conditional_code[0] = false;
+    state.conditional_code[1] = false;
+
+    RunInterpreter(state);
+    return state.debug;
+}
+
 } // namespace Shader
 
 } // namespace Pica
diff --git a/src/video_core/shader/shader.h b/src/video_core/shader/shader.h
index 2007a2844..bac51ddd8 100644
--- a/src/video_core/shader/shader.h
+++ b/src/video_core/shader/shader.h
@@ -4,7 +4,10 @@
 
 #pragma once
 
+#include <vector>
+
 #include <boost/container/static_vector.hpp>
+
 #include <nihstro/shader_binary.h>
 
 #include "common/common_funcs.h"
@@ -30,7 +33,7 @@ struct OutputVertex {
 
     // VS output attributes
     Math::Vec4<float24> pos;
-    Math::Vec4<float24> dummy; // quaternions (not implemented, yet)
+    Math::Vec4<float24> quat;
     Math::Vec4<float24> color;
     Math::Vec2<float24> tc0;
     Math::Vec2<float24> tc1;
@@ -72,12 +75,185 @@ struct OutputVertex {
 static_assert(std::is_pod<OutputVertex>::value, "Structure is not POD");
 static_assert(sizeof(OutputVertex) == 32 * sizeof(float), "OutputVertex has invalid size");
 
+
+// Helper structure used to keep track of data useful for inspection of shader emulation
+template<bool full_debugging>
+struct DebugData;
+
+template<>
+struct DebugData<false> {
+    // TODO: Hide these behind and interface and move them to DebugData<true>
+    u32 max_offset; // maximum program counter ever reached
+    u32 max_opdesc_id; // maximum swizzle pattern index ever used
+};
+
+template<>
+struct DebugData<true> {
+    // Records store the input and output operands of a particular instruction.
+    struct Record {
+        enum Type {
+            // Floating point arithmetic operands
+            SRC1         = 0x1,
+            SRC2         = 0x2,
+            SRC3         = 0x4,
+
+            // Initial and final output operand value
+            DEST_IN      = 0x8,
+            DEST_OUT     = 0x10,
+
+            // Current and next instruction offset (in words)
+            CUR_INSTR    = 0x20,
+            NEXT_INSTR   = 0x40,
+
+            // Output address register value
+            ADDR_REG_OUT = 0x80,
+
+            // Result of a comparison instruction
+            CMP_RESULT   = 0x100,
+
+            // Input values for conditional flow control instructions
+            COND_BOOL_IN = 0x200,
+            COND_CMP_IN  = 0x400,
+
+            // Input values for a loop
+            LOOP_INT_IN  = 0x800,
+        };
+
+        Math::Vec4<float24> src1;
+        Math::Vec4<float24> src2;
+        Math::Vec4<float24> src3;
+
+        Math::Vec4<float24> dest_in;
+        Math::Vec4<float24> dest_out;
+
+        s32 address_registers[2];
+        bool conditional_code[2];
+        bool cond_bool;
+        bool cond_cmp[2];
+        Math::Vec4<u8> loop_int;
+
+        u32 instruction_offset;
+        u32 next_instruction;
+
+        // set of enabled fields (as a combination of Type flags)
+        unsigned mask = 0;
+    };
+
+    u32 max_offset; // maximum program counter ever reached
+    u32 max_opdesc_id; // maximum swizzle pattern index ever used
+
+    // List of records for each executed shader instruction
+    std::vector<DebugData<true>::Record> records;
+};
+
+// Type alias for better readability
+using DebugDataRecord = DebugData<true>::Record;
+
+// Helper function to set a DebugData<true>::Record field based on the template enum parameter.
+template<DebugDataRecord::Type type, typename ValueType>
+inline void SetField(DebugDataRecord& record, ValueType value);
+
+template<>
+inline void SetField<DebugDataRecord::SRC1>(DebugDataRecord& record, float24* value) {
+    record.src1.x = value[0];
+    record.src1.y = value[1];
+    record.src1.z = value[2];
+    record.src1.w = value[3];
+}
+
+template<>
+inline void SetField<DebugDataRecord::SRC2>(DebugDataRecord& record, float24* value) {
+    record.src2.x = value[0];
+    record.src2.y = value[1];
+    record.src2.z = value[2];
+    record.src2.w = value[3];
+}
+
+template<>
+inline void SetField<DebugDataRecord::SRC3>(DebugDataRecord& record, float24* value) {
+    record.src3.x = value[0];
+    record.src3.y = value[1];
+    record.src3.z = value[2];
+    record.src3.w = value[3];
+}
+
+template<>
+inline void SetField<DebugDataRecord::DEST_IN>(DebugDataRecord& record, float24* value) {
+    record.dest_in.x = value[0];
+    record.dest_in.y = value[1];
+    record.dest_in.z = value[2];
+    record.dest_in.w = value[3];
+}
+
+template<>
+inline void SetField<DebugDataRecord::DEST_OUT>(DebugDataRecord& record, float24* value) {
+    record.dest_out.x = value[0];
+    record.dest_out.y = value[1];
+    record.dest_out.z = value[2];
+    record.dest_out.w = value[3];
+}
+
+template<>
+inline void SetField<DebugDataRecord::ADDR_REG_OUT>(DebugDataRecord& record, s32* value) {
+    record.address_registers[0] = value[0];
+    record.address_registers[1] = value[1];
+}
+
+template<>
+inline void SetField<DebugDataRecord::CMP_RESULT>(DebugDataRecord& record, bool* value) {
+    record.conditional_code[0] = value[0];
+    record.conditional_code[1] = value[1];
+}
+
+template<>
+inline void SetField<DebugDataRecord::COND_BOOL_IN>(DebugDataRecord& record, bool value) {
+    record.cond_bool = value;
+}
+
+template<>
+inline void SetField<DebugDataRecord::COND_CMP_IN>(DebugDataRecord& record, bool* value) {
+    record.cond_cmp[0] = value[0];
+    record.cond_cmp[1] = value[1];
+}
+
+template<>
+inline void SetField<DebugDataRecord::LOOP_INT_IN>(DebugDataRecord& record, Math::Vec4<u8> value) {
+    record.loop_int = value;
+}
+
+template<>
+inline void SetField<DebugDataRecord::CUR_INSTR>(DebugDataRecord& record, u32 value) {
+    record.instruction_offset = value;
+}
+
+template<>
+inline void SetField<DebugDataRecord::NEXT_INSTR>(DebugDataRecord& record, u32 value) {
+    record.next_instruction = value;
+}
+
+// Helper function to set debug information on the current shader iteration.
+template<DebugDataRecord::Type type, typename ValueType>
+inline void Record(DebugData<false>& debug_data, u32 offset, ValueType value) {
+    // Debugging disabled => nothing to do
+}
+
+template<DebugDataRecord::Type type, typename ValueType>
+inline void Record(DebugData<true>& debug_data, u32 offset, ValueType value) {
+    if (offset >= debug_data.records.size())
+        debug_data.records.resize(offset + 1);
+
+   SetField<type, ValueType>(debug_data.records[offset], value);
+   debug_data.records[offset].mask |= type;
+}
+
+
 /**
  * This structure contains the state information that needs to be unique for a shader unit. The 3DS
  * has four shader units that process shaders in parallel. At the present, Citra only implements a
  * single shader unit that processes all shaders serially. Putting the state information in a struct
  * here will make it easier for us to parallelize the shader processing later.
  */
+template<bool Debug>
 struct UnitState {
     struct Registers {
         // The registers are accessed by the shader JIT using SSE instructions, and are therefore
@@ -111,10 +287,7 @@ struct UnitState {
     // TODO: Is there a maximal size for this?
     boost::container::static_vector<CallStackElement, 16> call_stack;
 
-    struct {
-        u32 max_offset; // maximum program counter ever reached
-        u32 max_opdesc_id; // maximum swizzle pattern index ever used
-    } debug;
+    DebugData<Debug> debug;
 
     static int InputOffset(const SourceRegister& reg) {
         switch (reg.GetRegisterType()) {
@@ -150,7 +323,7 @@ struct UnitState {
  * vertex, which would happen within the `Run` function).
  * @param state Shader unit state, must be setup per shader and per shader unit
  */
-void Setup(UnitState& state);
+void Setup(UnitState<false>& state);
 
 /// Performs any cleanup when the emulator is shutdown
 void Shutdown();
@@ -162,7 +335,17 @@ void Shutdown();
  * @param num_attributes The number of vertex shader attributes
  * @return The output vertex, after having been processed by the vertex shader
  */
-OutputVertex Run(UnitState& state, const InputVertex& input, int num_attributes);
+OutputVertex Run(UnitState<false>& state, const InputVertex& input, int num_attributes);
+
+/**
+ * Produce debug information based on the given shader and input vertex
+ * @param input Input vertex into the shader
+ * @param num_attributes The number of vertex shader attributes
+ * @param config Configuration object for the shader pipeline
+ * @param setup Setup object for the shader pipeline
+ * @return Debug information for this shader with regards to the given vertex
+ */
+DebugData<true> ProduceDebugInfo(const InputVertex& input, int num_attributes, const Regs::ShaderConfig& config, const State::ShaderSetup& setup);
 
 } // namespace Shader
 
diff --git a/src/video_core/shader/shader_interpreter.cpp b/src/video_core/shader/shader_interpreter.cpp
index c8489f920..063cc38f0 100644
--- a/src/video_core/shader/shader_interpreter.cpp
+++ b/src/video_core/shader/shader_interpreter.cpp
@@ -21,7 +21,8 @@ namespace Pica {
 
 namespace Shader {
 
-void RunInterpreter(UnitState& state) {
+template<bool Debug>
+void RunInterpreter(UnitState<Debug>& state) {
     const auto& uniforms = g_state.vs.uniforms;
     const auto& swizzle_data = g_state.vs.swizzle_data;
     const auto& program_code = g_state.vs.program_code;
@@ -29,7 +30,9 @@ void RunInterpreter(UnitState& state) {
     // Placeholder for invalid inputs
     static float24 dummy_vec4_float24[4];
 
-    while (true) {
+    unsigned iteration = 0;
+    bool exit_loop = false;
+    while (!exit_loop) {
         if (!state.call_stack.empty()) {
             auto& top = state.call_stack.back();
             if (state.program_counter == top.final_address) {
@@ -47,16 +50,19 @@ void RunInterpreter(UnitState& state) {
             }
         }
 
-        bool exit_loop = false;
         const Instruction instr = { program_code[state.program_counter] };
         const SwizzlePattern swizzle = { swizzle_data[instr.common.operand_desc_id] };
 
-        static auto call = [](UnitState& state, u32 offset, u32 num_instructions,
+        static auto call = [](UnitState<Debug>& state, u32 offset, u32 num_instructions,
                               u32 return_offset, u8 repeat_count, u8 loop_increment) {
             state.program_counter = offset - 1; // -1 to make sure when incrementing the PC we end up at the correct offset
             ASSERT(state.call_stack.size() < state.call_stack.capacity());
             state.call_stack.push_back({ offset + num_instructions, return_offset, repeat_count, loop_increment, offset });
         };
+        Record<DebugDataRecord::CUR_INSTR>(state.debug, iteration, state.program_counter);
+        if (iteration > 0)
+            Record<DebugDataRecord::NEXT_INSTR>(state.debug, iteration - 1, state.program_counter);
+
         state.debug.max_offset = std::max<u32>(state.debug.max_offset, 1 + state.program_counter);
 
         auto LookupSourceRegister = [&](const SourceRegister& source_reg) -> const float24* {
@@ -123,58 +129,78 @@ void RunInterpreter(UnitState& state) {
             switch (instr.opcode.Value().EffectiveOpCode()) {
             case OpCode::Id::ADD:
             {
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::SRC2>(state.debug, iteration, src2);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
 
                     dest[i] = src1[i] + src2[i];
                 }
-
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
                 break;
             }
 
             case OpCode::Id::MUL:
             {
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::SRC2>(state.debug, iteration, src2);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
 
                     dest[i] = src1[i] * src2[i];
                 }
-
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
                 break;
             }
 
             case OpCode::Id::FLR:
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
 
                     dest[i] = float24::FromFloat32(std::floor(src1[i].ToFloat32()));
                 }
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
                 break;
 
             case OpCode::Id::MAX:
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::SRC2>(state.debug, iteration, src2);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
 
                     dest[i] = std::max(src1[i], src2[i]);
                 }
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
                 break;
 
             case OpCode::Id::MIN:
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::SRC2>(state.debug, iteration, src2);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
 
                     dest[i] = std::min(src1[i], src2[i]);
                 }
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
                 break;
 
             case OpCode::Id::DP3:
             case OpCode::Id::DP4:
             {
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::SRC2>(state.debug, iteration, src2);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
                 float24 dot = float24::FromFloat32(0.f);
                 int num_components = (instr.opcode.Value() == OpCode::Id::DP3) ? 3 : 4;
                 for (int i = 0; i < num_components; ++i)
@@ -186,12 +212,15 @@ void RunInterpreter(UnitState& state) {
 
                     dest[i] = dot;
                 }
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
                 break;
             }
 
             // Reciprocal
             case OpCode::Id::RCP:
             {
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
@@ -200,13 +229,15 @@ void RunInterpreter(UnitState& state) {
                     // TODO: I think this might be wrong... we should only use one component here
                     dest[i] = float24::FromFloat32(1.0f / src1[i].ToFloat32());
                 }
-
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
                 break;
             }
 
             // Reciprocal Square Root
             case OpCode::Id::RSQ:
             {
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
@@ -215,12 +246,13 @@ void RunInterpreter(UnitState& state) {
                     // TODO: I think this might be wrong... we should only use one component here
                     dest[i] = float24::FromFloat32(1.0f / sqrt(src1[i].ToFloat32()));
                 }
-
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
                 break;
             }
 
             case OpCode::Id::MOVA:
             {
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
                 for (int i = 0; i < 2; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
@@ -228,32 +260,55 @@ void RunInterpreter(UnitState& state) {
                     // TODO: Figure out how the rounding is done on hardware
                     state.address_registers[i] = static_cast<s32>(src1[i].ToFloat32());
                 }
-
+                Record<DebugDataRecord::ADDR_REG_OUT>(state.debug, iteration, state.address_registers);
                 break;
             }
 
             case OpCode::Id::MOV:
             {
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
 
                     dest[i] = src1[i];
                 }
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
                 break;
             }
 
+            case OpCode::Id::SGE:
+            case OpCode::Id::SGEI:
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::SRC2>(state.debug, iteration, src2);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
+                for (int i = 0; i < 4; ++i) {
+                    if (!swizzle.DestComponentEnabled(i))
+                        continue;
+
+                    dest[i] = (src1[i] >= src2[i]) ? float24::FromFloat32(1.0f) : float24::FromFloat32(0.0f);
+                }
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
+                break;
+
             case OpCode::Id::SLT:
             case OpCode::Id::SLTI:
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::SRC2>(state.debug, iteration, src2);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
 
                     dest[i] = (src1[i] < src2[i]) ? float24::FromFloat32(1.0f) : float24::FromFloat32(0.0f);
                 }
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
                 break;
 
             case OpCode::Id::CMP:
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::SRC2>(state.debug, iteration, src2);
                 for (int i = 0; i < 2; ++i) {
                     // TODO: Can you restrict to one compare via dest masking?
 
@@ -261,27 +316,27 @@ void RunInterpreter(UnitState& state) {
                     auto op = (i == 0) ? compare_op.x.Value() : compare_op.y.Value();
 
                     switch (op) {
-                        case compare_op.Equal:
+                        case Instruction::Common::CompareOpType::Equal:
                             state.conditional_code[i] = (src1[i] == src2[i]);
                             break;
 
-                        case compare_op.NotEqual:
+                        case Instruction::Common::CompareOpType::NotEqual:
                             state.conditional_code[i] = (src1[i] != src2[i]);
                             break;
 
-                        case compare_op.LessThan:
+                        case Instruction::Common::CompareOpType::LessThan:
                             state.conditional_code[i] = (src1[i] <  src2[i]);
                             break;
 
-                        case compare_op.LessEqual:
+                        case Instruction::Common::CompareOpType::LessEqual:
                             state.conditional_code[i] = (src1[i] <= src2[i]);
                             break;
 
-                        case compare_op.GreaterThan:
+                        case Instruction::Common::CompareOpType::GreaterThan:
                             state.conditional_code[i] = (src1[i] >  src2[i]);
                             break;
 
-                        case compare_op.GreaterEqual:
+                        case Instruction::Common::CompareOpType::GreaterEqual:
                             state.conditional_code[i] = (src1[i] >= src2[i]);
                             break;
 
@@ -290,7 +345,44 @@ void RunInterpreter(UnitState& state) {
                             break;
                     }
                 }
+                Record<DebugDataRecord::CMP_RESULT>(state.debug, iteration, state.conditional_code);
+                break;
+
+            case OpCode::Id::EX2:
+            {
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
+
+                // EX2 only takes first component exp2 and writes it to all dest components
+                float24 ex2_res = float24::FromFloat32(std::exp2(src1[0].ToFloat32()));
+                for (int i = 0; i < 4; ++i) {
+                    if (!swizzle.DestComponentEnabled(i))
+                        continue;
+
+                    dest[i] = ex2_res;
+                }
+
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
                 break;
+            }
+
+            case OpCode::Id::LG2:
+            {
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
+
+                // LG2 only takes the first component log2 and writes it to all dest components
+                float24 lg2_res = float24::FromFloat32(std::log2(src1[0].ToFloat32()));
+                for (int i = 0; i < 4; ++i) {
+                    if (!swizzle.DestComponentEnabled(i))
+                        continue;
+
+                    dest[i] = lg2_res;
+                }
+
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
+                break;
+            }
 
             default:
                 LOG_ERROR(HW_GPU, "Unhandled arithmetic instruction: 0x%02x (%s): 0x%08x",
@@ -359,12 +451,17 @@ void RunInterpreter(UnitState& state) {
                             : (instr.mad.dest.Value() < 0x20) ? &state.registers.temporary[instr.mad.dest.Value().GetIndex()][0]
                             : dummy_vec4_float24;
 
+                Record<DebugDataRecord::SRC1>(state.debug, iteration, src1);
+                Record<DebugDataRecord::SRC2>(state.debug, iteration, src2);
+                Record<DebugDataRecord::SRC3>(state.debug, iteration, src3);
+                Record<DebugDataRecord::DEST_IN>(state.debug, iteration, dest);
                 for (int i = 0; i < 4; ++i) {
                     if (!swizzle.DestComponentEnabled(i))
                         continue;
 
                     dest[i] = src1[i] * src2[i] + src3[i];
                 }
+                Record<DebugDataRecord::DEST_OUT>(state.debug, iteration, dest);
             } else {
                 LOG_ERROR(HW_GPU, "Unhandled multiply-add instruction: 0x%02x (%s): 0x%08x",
                           (int)instr.opcode.Value().EffectiveOpCode(), instr.opcode.Value().GetInfo().name, instr.hex);
@@ -374,7 +471,7 @@ void RunInterpreter(UnitState& state) {
 
         default:
         {
-            static auto evaluate_condition = [](const UnitState& state, bool refx, bool refy, Instruction::FlowControlType flow_control) {
+            static auto evaluate_condition = [](const UnitState<Debug>& state, bool refx, bool refy, Instruction::FlowControlType flow_control) {
                 bool results[2] = { refx == state.conditional_code[0],
                                     refy == state.conditional_code[1] };
 
@@ -400,12 +497,14 @@ void RunInterpreter(UnitState& state) {
                 break;
 
             case OpCode::Id::JMPC:
+                Record<DebugDataRecord::COND_CMP_IN>(state.debug, iteration, state.conditional_code);
                 if (evaluate_condition(state, instr.flow_control.refx, instr.flow_control.refy, instr.flow_control)) {
                     state.program_counter = instr.flow_control.dest_offset - 1;
                 }
                 break;
 
             case OpCode::Id::JMPU:
+                Record<DebugDataRecord::COND_BOOL_IN>(state.debug, iteration, uniforms.b[instr.flow_control.bool_uniform_id]);
                 if (uniforms.b[instr.flow_control.bool_uniform_id]) {
                     state.program_counter = instr.flow_control.dest_offset - 1;
                 }
@@ -419,6 +518,7 @@ void RunInterpreter(UnitState& state) {
                 break;
 
             case OpCode::Id::CALLU:
+                Record<DebugDataRecord::COND_BOOL_IN>(state.debug, iteration, uniforms.b[instr.flow_control.bool_uniform_id]);
                 if (uniforms.b[instr.flow_control.bool_uniform_id]) {
                     call(state,
                         instr.flow_control.dest_offset,
@@ -428,6 +528,7 @@ void RunInterpreter(UnitState& state) {
                 break;
 
             case OpCode::Id::CALLC:
+                Record<DebugDataRecord::COND_CMP_IN>(state.debug, iteration, state.conditional_code);
                 if (evaluate_condition(state, instr.flow_control.refx, instr.flow_control.refy, instr.flow_control)) {
                     call(state,
                         instr.flow_control.dest_offset,
@@ -440,6 +541,7 @@ void RunInterpreter(UnitState& state) {
                 break;
 
             case OpCode::Id::IFU:
+                Record<DebugDataRecord::COND_BOOL_IN>(state.debug, iteration, uniforms.b[instr.flow_control.bool_uniform_id]);
                 if (uniforms.b[instr.flow_control.bool_uniform_id]) {
                     call(state,
                          state.program_counter + 1,
@@ -458,6 +560,7 @@ void RunInterpreter(UnitState& state) {
             {
                 // TODO: Do we need to consider swizzlers here?
 
+                Record<DebugDataRecord::COND_CMP_IN>(state.debug, iteration, state.conditional_code);
                 if (evaluate_condition(state, instr.flow_control.refx, instr.flow_control.refy, instr.flow_control)) {
                     call(state,
                          state.program_counter + 1,
@@ -475,14 +578,19 @@ void RunInterpreter(UnitState& state) {
 
             case OpCode::Id::LOOP:
             {
-                state.address_registers[2] = uniforms.i[instr.flow_control.int_uniform_id].y;
+                Math::Vec4<u8> loop_param(uniforms.i[instr.flow_control.int_uniform_id].x,
+                                          uniforms.i[instr.flow_control.int_uniform_id].y,
+                                          uniforms.i[instr.flow_control.int_uniform_id].z,
+                                          uniforms.i[instr.flow_control.int_uniform_id].w);
+                state.address_registers[2] = loop_param.y;
 
+                Record<DebugDataRecord::LOOP_INT_IN>(state.debug, iteration, loop_param);
                 call(state,
                      state.program_counter + 1,
                      instr.flow_control.dest_offset - state.program_counter + 1,
                      instr.flow_control.dest_offset + 1,
-                     uniforms.i[instr.flow_control.int_uniform_id].x,
-                     uniforms.i[instr.flow_control.int_uniform_id].z);
+                     loop_param.x,
+                     loop_param.z);
                 break;
             }
 
@@ -497,12 +605,14 @@ void RunInterpreter(UnitState& state) {
         }
 
         ++state.program_counter;
-
-        if (exit_loop)
-            break;
+        ++iteration;
     }
 }
 
+// Explicit instantiation
+template void RunInterpreter(UnitState<false>& state);
+template void RunInterpreter(UnitState<true>& state);
+
 } // namespace
 
 } // namespace
diff --git a/src/video_core/shader/shader_interpreter.h b/src/video_core/shader/shader_interpreter.h
index ad6e58e39..71bcad5ac 100644
--- a/src/video_core/shader/shader_interpreter.h
+++ b/src/video_core/shader/shader_interpreter.h
@@ -12,7 +12,8 @@ namespace Pica {
 
 namespace Shader {
 
-void RunInterpreter(UnitState& state);
+template<bool Debug>
+void RunInterpreter(UnitState<Debug>& state);
 
 } // namespace
 
diff --git a/src/video_core/shader/shader_jit_x64.cpp b/src/video_core/shader/shader_jit_x64.cpp
index ce47774d5..a1bdd8456 100644
--- a/src/video_core/shader/shader_jit_x64.cpp
+++ b/src/video_core/shader/shader_jit_x64.cpp
@@ -25,12 +25,12 @@ const JitFunction instr_table[64] = {
     &JitCompiler::Compile_DP4,      // dp4
     nullptr,                        // dph
     nullptr,                        // unknown
-    nullptr,                        // ex2
-    nullptr,                        // lg2
+    &JitCompiler::Compile_EX2,      // ex2
+    &JitCompiler::Compile_LG2,      // lg2
     nullptr,                        // unknown
     &JitCompiler::Compile_MUL,      // mul
-    nullptr,                        // lge
-    nullptr,                        // slt
+    &JitCompiler::Compile_SGE,      // sge
+    &JitCompiler::Compile_SLT,      // slt
     &JitCompiler::Compile_FLR,      // flr
     &JitCompiler::Compile_MAX,      // max
     &JitCompiler::Compile_MIN,      // min
@@ -46,8 +46,8 @@ const JitFunction instr_table[64] = {
     nullptr,                        // unknown
     nullptr,                        // dphi
     nullptr,                        // unknown
-    nullptr,                        // sgei
-    &JitCompiler::Compile_SLTI,     // slti
+    &JitCompiler::Compile_SGE,      // sgei
+    &JitCompiler::Compile_SLT,      // slti
     nullptr,                        // unknown
     nullptr,                        // unknown
     nullptr,                        // unknown
@@ -141,7 +141,7 @@ void JitCompiler::Compile_SwizzleSrc(Instruction instr, unsigned src_num, Source
         src_offset = src_reg.GetIndex() * sizeof(float24) * 4;
     } else {
         src_ptr = REGISTERS;
-        src_offset = UnitState::InputOffset(src_reg);
+        src_offset = UnitState<false>::InputOffset(src_reg);
     }
 
     unsigned operand_desc_id;
@@ -217,11 +217,11 @@ void JitCompiler::Compile_DestEnable(Instruction instr,X64Reg src) {
     // If all components are enabled, write the result to the destination register
     if (swiz.dest_mask == NO_DEST_REG_MASK) {
         // Store dest back to memory
-        MOVAPS(MDisp(REGISTERS, UnitState::OutputOffset(dest)), src);
+        MOVAPS(MDisp(REGISTERS, UnitState<false>::OutputOffset(dest)), src);
 
     } else {
         // Not all components are enabled, so mask the result when storing to the destination register...
-        MOVAPS(SCRATCH, MDisp(REGISTERS, UnitState::OutputOffset(dest)));
+        MOVAPS(SCRATCH, MDisp(REGISTERS, UnitState<false>::OutputOffset(dest)));
 
         if (Common::GetCPUCaps().sse4_1) {
             u8 mask = ((swiz.dest_mask & 1) << 3) | ((swiz.dest_mask & 8) >> 3) | ((swiz.dest_mask & 2) << 1) | ((swiz.dest_mask & 4) >> 1);
@@ -240,7 +240,7 @@ void JitCompiler::Compile_DestEnable(Instruction instr,X64Reg src) {
         }
 
         // Store dest back to memory
-        MOVAPS(MDisp(REGISTERS, UnitState::OutputOffset(dest)), SCRATCH);
+        MOVAPS(MDisp(REGISTERS, UnitState<false>::OutputOffset(dest)), SCRATCH);
     }
 }
 
@@ -280,6 +280,22 @@ void JitCompiler::Compile_UniformCondition(Instruction instr) {
     CMP(sizeof(bool) * 8, MDisp(UNIFORMS, offset), Imm8(0));
 }
 
+void JitCompiler::Compile_PushCallerSavedXMM() {
+#ifndef _WIN32
+    SUB(64, R(RSP), Imm8(2 * 16));
+    MOVUPS(MDisp(RSP, 16), ONE);
+    MOVUPS(MDisp(RSP, 0), NEGBIT);
+#endif
+}
+
+void JitCompiler::Compile_PopCallerSavedXMM() {
+#ifndef _WIN32
+    MOVUPS(NEGBIT, MDisp(RSP, 0));
+    MOVUPS(ONE, MDisp(RSP, 16));
+    ADD(64, R(RSP), Imm8(2 * 16));
+#endif
+}
+
 void JitCompiler::Compile_ADD(Instruction instr) {
     Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1);
     Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2);
@@ -331,6 +347,38 @@ void JitCompiler::Compile_DP4(Instruction instr) {
     Compile_DestEnable(instr, SRC1);
 }
 
+void JitCompiler::Compile_EX2(Instruction instr) {
+    Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1);
+    MOVSS(XMM0, R(SRC1));
+
+    // The following will actually break the stack alignment
+    ABI_PushAllCallerSavedRegsAndAdjustStack();
+    Compile_PushCallerSavedXMM();
+    ABI_CallFunction(reinterpret_cast<const void*>(exp2f));
+    Compile_PopCallerSavedXMM();
+    ABI_PopAllCallerSavedRegsAndAdjustStack();
+
+    SHUFPS(XMM0, R(XMM0), _MM_SHUFFLE(0, 0, 0, 0));
+    MOVAPS(SRC1, R(XMM0));
+    Compile_DestEnable(instr, SRC1);
+}
+
+void JitCompiler::Compile_LG2(Instruction instr) {
+    Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1);
+    MOVSS(XMM0, R(SRC1));
+
+    // The following will actually break the stack alignment
+    ABI_PushAllCallerSavedRegsAndAdjustStack();
+    Compile_PushCallerSavedXMM();
+    ABI_CallFunction(reinterpret_cast<const void*>(log2f));
+    Compile_PopCallerSavedXMM();
+    ABI_PopAllCallerSavedRegsAndAdjustStack();
+
+    SHUFPS(XMM0, R(XMM0), _MM_SHUFFLE(0, 0, 0, 0));
+    MOVAPS(SRC1, R(XMM0));
+    Compile_DestEnable(instr, SRC1);
+}
+
 void JitCompiler::Compile_MUL(Instruction instr) {
     Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1);
     Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2);
@@ -338,6 +386,36 @@ void JitCompiler::Compile_MUL(Instruction instr) {
     Compile_DestEnable(instr, SRC1);
 }
 
+void JitCompiler::Compile_SGE(Instruction instr) {
+    if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::SGEI) {
+        Compile_SwizzleSrc(instr, 1, instr.common.src1i, SRC1);
+        Compile_SwizzleSrc(instr, 2, instr.common.src2i, SRC2);
+    } else {
+        Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1);
+        Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2);
+    }
+
+    CMPPS(SRC1, R(SRC2), CMP_NLT);
+    ANDPS(SRC1, R(ONE));
+
+    Compile_DestEnable(instr, SRC1);
+}
+
+void JitCompiler::Compile_SLT(Instruction instr) {
+    if (instr.opcode.Value().EffectiveOpCode() == OpCode::Id::SLTI) {
+        Compile_SwizzleSrc(instr, 1, instr.common.src1i, SRC1);
+        Compile_SwizzleSrc(instr, 2, instr.common.src2i, SRC2);
+    } else {
+        Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1);
+        Compile_SwizzleSrc(instr, 2, instr.common.src2, SRC2);
+    }
+
+    CMPPS(SRC1, R(SRC2), CMP_LT);
+    ANDPS(SRC1, R(ONE));
+
+    Compile_DestEnable(instr, SRC1);
+}
+
 void JitCompiler::Compile_FLR(Instruction instr) {
     Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1);
 
@@ -415,16 +493,6 @@ void JitCompiler::Compile_MOV(Instruction instr) {
     Compile_DestEnable(instr, SRC1);
 }
 
-void JitCompiler::Compile_SLTI(Instruction instr) {
-    Compile_SwizzleSrc(instr, 1, instr.common.src1i, SRC1);
-    Compile_SwizzleSrc(instr, 1, instr.common.src2i, SRC2);
-
-    CMPSS(SRC1, R(SRC2), CMP_LT);
-    ANDPS(SRC1, R(ONE));
-
-    Compile_DestEnable(instr, SRC1);
-}
-
 void JitCompiler::Compile_RCP(Instruction instr) {
     Compile_SwizzleSrc(instr, 1, instr.common.src1, SRC1);
 
diff --git a/src/video_core/shader/shader_jit_x64.h b/src/video_core/shader/shader_jit_x64.h
index b88f2a0d2..b2aa5293c 100644
--- a/src/video_core/shader/shader_jit_x64.h
+++ b/src/video_core/shader/shader_jit_x64.h
@@ -37,7 +37,11 @@ public:
     void Compile_ADD(Instruction instr);
     void Compile_DP3(Instruction instr);
     void Compile_DP4(Instruction instr);
+    void Compile_EX2(Instruction instr);
+    void Compile_LG2(Instruction instr);
     void Compile_MUL(Instruction instr);
+    void Compile_SGE(Instruction instr);
+    void Compile_SLT(Instruction instr);
     void Compile_FLR(Instruction instr);
     void Compile_MAX(Instruction instr);
     void Compile_MIN(Instruction instr);
@@ -45,7 +49,6 @@ public:
     void Compile_RSQ(Instruction instr);
     void Compile_MOVA(Instruction instr);
     void Compile_MOV(Instruction instr);
-    void Compile_SLTI(Instruction instr);
     void Compile_NOP(Instruction instr);
     void Compile_END(Instruction instr);
     void Compile_CALL(Instruction instr);
@@ -67,6 +70,9 @@ private:
     void Compile_EvaluateCondition(Instruction instr);
     void Compile_UniformCondition(Instruction instr);
 
+    void Compile_PushCallerSavedXMM();
+    void Compile_PopCallerSavedXMM();
+
     /// Pointer to the variable that stores the current Pica code offset. Used to handle nested code blocks.
     unsigned* offset_ptr = nullptr;