From 8fca90b5d56892758a98a0ece15b5845009893e3 Mon Sep 17 00:00:00 2001
From: Yuri Kunde Schlesner <yuriks@yuriks.net>
Date: Sat, 28 Jan 2017 12:34:31 -0800
Subject: [PATCH] VideoCore: Split geometry pipeline regs from Regs struct

---
 src/video_core/CMakeLists.txt         |   1 +
 src/video_core/command_processor.cpp  |  51 +++---
 src/video_core/pica.cpp               |   2 +-
 src/video_core/pica.h                 | 230 ++------------------------
 src/video_core/primitive_assembly.cpp |  18 +-
 src/video_core/primitive_assembly.h   |   9 +-
 src/video_core/regs_pipeline.h        | 224 +++++++++++++++++++++++++
 src/video_core/vertex_loader.cpp      |  15 +-
 src/video_core/vertex_loader.h        |   6 +-
 9 files changed, 292 insertions(+), 264 deletions(-)
 create mode 100644 src/video_core/regs_pipeline.h

diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index 3c6ea42be..c8d2675ae 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -34,6 +34,7 @@ set(HEADERS
             rasterizer_interface.h
             regs_framebuffer.h
             regs_lighting.h
+            regs_pipeline.h
             regs_rasterizer.h
             regs_texturing.h
             renderer_base.h
diff --git a/src/video_core/command_processor.cpp b/src/video_core/command_processor.cpp
index fa3432f60..c8064bf6a 100644
--- a/src/video_core/command_processor.cpp
+++ b/src/video_core/command_processor.cpp
@@ -74,23 +74,23 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
         Service::GSP::SignalInterrupt(Service::GSP::InterruptId::P3D);
         break;
 
-    case PICA_REG_INDEX_WORKAROUND(triangle_topology, 0x25E):
-        g_state.primitive_assembler.Reconfigure(regs.triangle_topology);
+    case PICA_REG_INDEX(pipeline.triangle_topology):
+        g_state.primitive_assembler.Reconfigure(regs.pipeline.triangle_topology);
         break;
 
-    case PICA_REG_INDEX_WORKAROUND(restart_primitive, 0x25F):
+    case PICA_REG_INDEX(pipeline.restart_primitive):
         g_state.primitive_assembler.Reset();
         break;
 
-    case PICA_REG_INDEX_WORKAROUND(vs_default_attributes_setup.index, 0x232):
+    case PICA_REG_INDEX(pipeline.vs_default_attributes_setup.index):
         g_state.immediate.current_attribute = 0;
         default_attr_counter = 0;
         break;
 
     // Load default vertex input attributes
-    case PICA_REG_INDEX_WORKAROUND(vs_default_attributes_setup.set_value[0], 0x233):
-    case PICA_REG_INDEX_WORKAROUND(vs_default_attributes_setup.set_value[1], 0x234):
-    case PICA_REG_INDEX_WORKAROUND(vs_default_attributes_setup.set_value[2], 0x235): {
+    case PICA_REG_INDEX_WORKAROUND(pipeline.vs_default_attributes_setup.set_value[0], 0x233):
+    case PICA_REG_INDEX_WORKAROUND(pipeline.vs_default_attributes_setup.set_value[1], 0x234):
+    case PICA_REG_INDEX_WORKAROUND(pipeline.vs_default_attributes_setup.set_value[2], 0x235): {
         // TODO: Does actual hardware indeed keep an intermediate buffer or does
         //       it directly write the values?
         default_attr_write_buffer[default_attr_counter++] = value;
@@ -102,7 +102,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
         if (default_attr_counter >= 3) {
             default_attr_counter = 0;
 
-            auto& setup = regs.vs_default_attributes_setup;
+            auto& setup = regs.pipeline.vs_default_attributes_setup;
 
             if (setup.index >= 16) {
                 LOG_ERROR(HW_GPU, "Invalid VS default attribute index %d", (int)setup.index);
@@ -137,7 +137,7 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
 
                 immediate_input.attr[immediate_attribute_id] = attribute;
 
-                if (immediate_attribute_id < regs.max_input_attrib_index) {
+                if (immediate_attribute_id < regs.pipeline.max_input_attrib_index) {
                     immediate_attribute_id += 1;
                 } else {
                     MICROPROFILE_SCOPE(GPU_Drawing);
@@ -173,8 +173,8 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
         break;
     }
 
-    case PICA_REG_INDEX(gpu_mode):
-        if (regs.gpu_mode == Regs::GPUMode::Configuring) {
+    case PICA_REG_INDEX(pipeline.gpu_mode):
+        if (regs.pipeline.gpu_mode == PipelineRegs::GPUMode::Configuring) {
             MICROPROFILE_SCOPE(GPU_Drawing);
 
             // Draw immediate mode triangles when GPU Mode is set to GPUMode::Configuring
@@ -186,19 +186,20 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
         }
         break;
 
-    case PICA_REG_INDEX_WORKAROUND(command_buffer.trigger[0], 0x23c):
-    case PICA_REG_INDEX_WORKAROUND(command_buffer.trigger[1], 0x23d): {
-        unsigned index = static_cast<unsigned>(id - PICA_REG_INDEX(command_buffer.trigger[0]));
-        u32* head_ptr =
-            (u32*)Memory::GetPhysicalPointer(regs.command_buffer.GetPhysicalAddress(index));
+    case PICA_REG_INDEX_WORKAROUND(pipeline.command_buffer.trigger[0], 0x23c):
+    case PICA_REG_INDEX_WORKAROUND(pipeline.command_buffer.trigger[1], 0x23d): {
+        unsigned index =
+            static_cast<unsigned>(id - PICA_REG_INDEX(pipeline.command_buffer.trigger[0]));
+        u32* head_ptr = (u32*)Memory::GetPhysicalPointer(
+            regs.pipeline.command_buffer.GetPhysicalAddress(index));
         g_state.cmd_list.head_ptr = g_state.cmd_list.current_ptr = head_ptr;
-        g_state.cmd_list.length = regs.command_buffer.GetSize(index) / sizeof(u32);
+        g_state.cmd_list.length = regs.pipeline.command_buffer.GetSize(index) / sizeof(u32);
         break;
     }
 
     // It seems like these trigger vertex rendering
-    case PICA_REG_INDEX(trigger_draw):
-    case PICA_REG_INDEX(trigger_draw_indexed): {
+    case PICA_REG_INDEX(pipeline.trigger_draw):
+    case PICA_REG_INDEX(pipeline.trigger_draw_indexed): {
         MICROPROFILE_SCOPE(GPU_Drawing);
 
 #if PICA_LOG_TEV
@@ -210,13 +211,13 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
         // Processes information about internal vertex attributes to figure out how a vertex is
         // loaded.
         // Later, these can be compiled and cached.
-        const u32 base_address = regs.vertex_attributes.GetPhysicalBaseAddress();
-        VertexLoader loader(regs);
+        const u32 base_address = regs.pipeline.vertex_attributes.GetPhysicalBaseAddress();
+        VertexLoader loader(regs.pipeline);
 
         // Load vertices
-        bool is_indexed = (id == PICA_REG_INDEX(trigger_draw_indexed));
+        bool is_indexed = (id == PICA_REG_INDEX(pipeline.trigger_draw_indexed));
 
-        const auto& index_info = regs.index_array;
+        const auto& index_info = regs.pipeline.index_array;
         const u8* index_address_8 = Memory::GetPhysicalPointer(base_address + index_info.offset);
         const u16* index_address_16 = reinterpret_cast<const u16*>(index_address_8);
         bool index_u16 = index_info.format != 0;
@@ -254,11 +255,11 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
 
         shader_engine->SetupBatch(g_state.vs, regs.vs.main_offset);
 
-        for (unsigned int index = 0; index < regs.num_vertices; ++index) {
+        for (unsigned int index = 0; index < regs.pipeline.num_vertices; ++index) {
             // Indexed rendering doesn't use the start offset
             unsigned int vertex =
                 is_indexed ? (index_u16 ? index_address_16[index] : index_address_8[index])
-                           : (index + regs.vertex_offset);
+                           : (index + regs.pipeline.vertex_offset);
 
             // -1 is a common special value used for primitive restart. Since it's unknown if
             // the PICA supports it, and it would mess up the caching, guard against it here.
diff --git a/src/video_core/pica.cpp b/src/video_core/pica.cpp
index b4a77c632..6604ce83c 100644
--- a/src/video_core/pica.cpp
+++ b/src/video_core/pica.cpp
@@ -513,6 +513,6 @@ void State::Reset() {
     Zero(gs);
     Zero(cmd_list);
     Zero(immediate);
-    primitive_assembler.Reconfigure(Regs::TriangleTopology::List);
+    primitive_assembler.Reconfigure(PipelineRegs::TriangleTopology::List);
 }
 }
diff --git a/src/video_core/pica.h b/src/video_core/pica.h
index 371bcdb84..765fa5dd4 100644
--- a/src/video_core/pica.h
+++ b/src/video_core/pica.h
@@ -20,6 +20,7 @@
 #include "common/vector_math.h"
 #include "video_core/regs_framebuffer.h"
 #include "video_core/regs_lighting.h"
+#include "video_core/regs_pipeline.h"
 #include "video_core/regs_rasterizer.h"
 #include "video_core/regs_texturing.h"
 
@@ -55,210 +56,7 @@ struct Regs {
     TexturingRegs texturing;
     FramebufferRegs framebuffer;
     LightingRegs lighting;
-
-    enum class VertexAttributeFormat : u64 {
-        BYTE = 0,
-        UBYTE = 1,
-        SHORT = 2,
-        FLOAT = 3,
-    };
-
-    struct {
-        BitField<0, 29, u32> base_address;
-
-        u32 GetPhysicalBaseAddress() const {
-            return DecodeAddressRegister(base_address);
-        }
-
-        // Descriptor for internal vertex attributes
-        union {
-            BitField<0, 2, VertexAttributeFormat> format0; // size of one element
-            BitField<2, 2, u64> size0;                     // number of elements minus 1
-            BitField<4, 2, VertexAttributeFormat> format1;
-            BitField<6, 2, u64> size1;
-            BitField<8, 2, VertexAttributeFormat> format2;
-            BitField<10, 2, u64> size2;
-            BitField<12, 2, VertexAttributeFormat> format3;
-            BitField<14, 2, u64> size3;
-            BitField<16, 2, VertexAttributeFormat> format4;
-            BitField<18, 2, u64> size4;
-            BitField<20, 2, VertexAttributeFormat> format5;
-            BitField<22, 2, u64> size5;
-            BitField<24, 2, VertexAttributeFormat> format6;
-            BitField<26, 2, u64> size6;
-            BitField<28, 2, VertexAttributeFormat> format7;
-            BitField<30, 2, u64> size7;
-            BitField<32, 2, VertexAttributeFormat> format8;
-            BitField<34, 2, u64> size8;
-            BitField<36, 2, VertexAttributeFormat> format9;
-            BitField<38, 2, u64> size9;
-            BitField<40, 2, VertexAttributeFormat> format10;
-            BitField<42, 2, u64> size10;
-            BitField<44, 2, VertexAttributeFormat> format11;
-            BitField<46, 2, u64> size11;
-
-            BitField<48, 12, u64> attribute_mask;
-
-            // number of total attributes minus 1
-            BitField<60, 4, u64> max_attribute_index;
-        };
-
-        inline VertexAttributeFormat GetFormat(int n) const {
-            VertexAttributeFormat formats[] = {format0, format1, format2,  format3,
-                                               format4, format5, format6,  format7,
-                                               format8, format9, format10, format11};
-            return formats[n];
-        }
-
-        inline int GetNumElements(int n) const {
-            u64 sizes[] = {size0, size1, size2, size3, size4,  size5,
-                           size6, size7, size8, size9, size10, size11};
-            return (int)sizes[n] + 1;
-        }
-
-        inline int GetElementSizeInBytes(int n) const {
-            return (GetFormat(n) == VertexAttributeFormat::FLOAT)
-                       ? 4
-                       : (GetFormat(n) == VertexAttributeFormat::SHORT) ? 2 : 1;
-        }
-
-        inline int GetStride(int n) const {
-            return GetNumElements(n) * GetElementSizeInBytes(n);
-        }
-
-        inline bool IsDefaultAttribute(int id) const {
-            return (id >= 12) || (attribute_mask & (1ULL << id)) != 0;
-        }
-
-        inline int GetNumTotalAttributes() const {
-            return (int)max_attribute_index + 1;
-        }
-
-        // Attribute loaders map the source vertex data to input attributes
-        // This e.g. allows to load different attributes from different memory locations
-        struct {
-            // Source attribute data offset from the base address
-            u32 data_offset;
-
-            union {
-                BitField<0, 4, u64> comp0;
-                BitField<4, 4, u64> comp1;
-                BitField<8, 4, u64> comp2;
-                BitField<12, 4, u64> comp3;
-                BitField<16, 4, u64> comp4;
-                BitField<20, 4, u64> comp5;
-                BitField<24, 4, u64> comp6;
-                BitField<28, 4, u64> comp7;
-                BitField<32, 4, u64> comp8;
-                BitField<36, 4, u64> comp9;
-                BitField<40, 4, u64> comp10;
-                BitField<44, 4, u64> comp11;
-
-                // bytes for a single vertex in this loader
-                BitField<48, 8, u64> byte_count;
-
-                BitField<60, 4, u64> component_count;
-            };
-
-            inline int GetComponent(int n) const {
-                u64 components[] = {comp0, comp1, comp2, comp3, comp4,  comp5,
-                                    comp6, comp7, comp8, comp9, comp10, comp11};
-                return (int)components[n];
-            }
-        } attribute_loaders[12];
-    } vertex_attributes;
-
-    struct {
-        enum IndexFormat : u32 {
-            BYTE = 0,
-            SHORT = 1,
-        };
-
-        union {
-            BitField<0, 31, u32> offset; // relative to base attribute address
-            BitField<31, 1, IndexFormat> format;
-        };
-    } index_array;
-
-    // Number of vertices to render
-    u32 num_vertices;
-
-    INSERT_PADDING_WORDS(0x1);
-
-    // The index of the first vertex to render
-    u32 vertex_offset;
-
-    INSERT_PADDING_WORDS(0x3);
-
-    // These two trigger rendering of triangles
-    u32 trigger_draw;
-    u32 trigger_draw_indexed;
-
-    INSERT_PADDING_WORDS(0x2);
-
-    // These registers are used to setup the default "fall-back" vertex shader attributes
-    struct {
-        // Index of the current default attribute
-        u32 index;
-
-        // Writing to these registers sets the "current" default attribute.
-        u32 set_value[3];
-    } vs_default_attributes_setup;
-
-    INSERT_PADDING_WORDS(0x2);
-
-    struct {
-        // There are two channels that can be used to configure the next command buffer, which
-        // can be then executed by writing to the "trigger" registers. There are two reasons why a
-        // game might use this feature:
-        //  1) With this, an arbitrary number of additional command buffers may be executed in
-        //     sequence without requiring any intervention of the CPU after the initial one is
-        //     kicked off.
-        //  2) Games can configure these registers to provide a command list subroutine mechanism.
-
-        BitField<0, 20, u32> size[2]; ///< Size (in bytes / 8) of each channel's command buffer
-        BitField<0, 28, u32> addr[2]; ///< Physical address / 8 of each channel's command buffer
-        u32 trigger[2]; ///< Triggers execution of the channel's command buffer when written to
-
-        unsigned GetSize(unsigned index) const {
-            ASSERT(index < 2);
-            return 8 * size[index];
-        }
-
-        PAddr GetPhysicalAddress(unsigned index) const {
-            ASSERT(index < 2);
-            return (PAddr)(8 * addr[index]);
-        }
-    } command_buffer;
-
-    INSERT_PADDING_WORDS(4);
-
-    /// Number of input attributes to the vertex shader minus 1
-    BitField<0, 4, u32> max_input_attrib_index;
-
-    INSERT_PADDING_WORDS(2);
-
-    enum class GPUMode : u32 {
-        Drawing = 0,
-        Configuring = 1,
-    };
-
-    GPUMode gpu_mode;
-
-    INSERT_PADDING_WORDS(0x18);
-
-    enum class TriangleTopology : u32 {
-        List = 0,
-        Strip = 1,
-        Fan = 2,
-        Shader = 3, // Programmable setup unit implemented in a geometry shader
-    };
-
-    BitField<8, 2, TriangleTopology> triangle_topology;
-
-    u32 restart_primitive;
-
-    INSERT_PADDING_WORDS(0x20);
+    PipelineRegs pipeline;
 
     struct ShaderConfig {
         BitField<0, 16, u32> bool_uniforms;
@@ -430,17 +228,19 @@ ASSERT_REG_POSITION(framebuffer.framebuffer, 0x110);
 
 ASSERT_REG_POSITION(lighting, 0x140);
 
-ASSERT_REG_POSITION(vertex_attributes, 0x200);
-ASSERT_REG_POSITION(index_array, 0x227);
-ASSERT_REG_POSITION(num_vertices, 0x228);
-ASSERT_REG_POSITION(vertex_offset, 0x22a);
-ASSERT_REG_POSITION(trigger_draw, 0x22e);
-ASSERT_REG_POSITION(trigger_draw_indexed, 0x22f);
-ASSERT_REG_POSITION(vs_default_attributes_setup, 0x232);
-ASSERT_REG_POSITION(command_buffer, 0x238);
-ASSERT_REG_POSITION(gpu_mode, 0x245);
-ASSERT_REG_POSITION(triangle_topology, 0x25e);
-ASSERT_REG_POSITION(restart_primitive, 0x25f);
+ASSERT_REG_POSITION(pipeline, 0x200);
+ASSERT_REG_POSITION(pipeline.vertex_attributes, 0x200);
+ASSERT_REG_POSITION(pipeline.index_array, 0x227);
+ASSERT_REG_POSITION(pipeline.num_vertices, 0x228);
+ASSERT_REG_POSITION(pipeline.vertex_offset, 0x22a);
+ASSERT_REG_POSITION(pipeline.trigger_draw, 0x22e);
+ASSERT_REG_POSITION(pipeline.trigger_draw_indexed, 0x22f);
+ASSERT_REG_POSITION(pipeline.vs_default_attributes_setup, 0x232);
+ASSERT_REG_POSITION(pipeline.command_buffer, 0x238);
+ASSERT_REG_POSITION(pipeline.gpu_mode, 0x245);
+ASSERT_REG_POSITION(pipeline.triangle_topology, 0x25e);
+ASSERT_REG_POSITION(pipeline.restart_primitive, 0x25f);
+
 ASSERT_REG_POSITION(gs, 0x280);
 ASSERT_REG_POSITION(vs, 0x2b0);
 
diff --git a/src/video_core/primitive_assembly.cpp b/src/video_core/primitive_assembly.cpp
index e71ff5719..acd2ac5e2 100644
--- a/src/video_core/primitive_assembly.cpp
+++ b/src/video_core/primitive_assembly.cpp
@@ -3,14 +3,14 @@
 // Refer to the license.txt file included.
 
 #include "common/logging/log.h"
-#include "video_core/pica.h"
 #include "video_core/primitive_assembly.h"
+#include "video_core/regs_pipeline.h"
 #include "video_core/shader/shader.h"
 
 namespace Pica {
 
 template <typename VertexType>
-PrimitiveAssembler<VertexType>::PrimitiveAssembler(Regs::TriangleTopology topology)
+PrimitiveAssembler<VertexType>::PrimitiveAssembler(PipelineRegs::TriangleTopology topology)
     : topology(topology), buffer_index(0) {}
 
 template <typename VertexType>
@@ -18,8 +18,8 @@ void PrimitiveAssembler<VertexType>::SubmitVertex(const VertexType& vtx,
                                                   TriangleHandler triangle_handler) {
     switch (topology) {
     // TODO: Figure out what's different with TriangleTopology::Shader.
-    case Regs::TriangleTopology::List:
-    case Regs::TriangleTopology::Shader:
+    case PipelineRegs::TriangleTopology::List:
+    case PipelineRegs::TriangleTopology::Shader:
         if (buffer_index < 2) {
             buffer[buffer_index++] = vtx;
         } else {
@@ -29,8 +29,8 @@ void PrimitiveAssembler<VertexType>::SubmitVertex(const VertexType& vtx,
         }
         break;
 
-    case Regs::TriangleTopology::Strip:
-    case Regs::TriangleTopology::Fan:
+    case PipelineRegs::TriangleTopology::Strip:
+    case PipelineRegs::TriangleTopology::Fan:
         if (strip_ready)
             triangle_handler(buffer[0], buffer[1], vtx);
 
@@ -38,9 +38,9 @@ void PrimitiveAssembler<VertexType>::SubmitVertex(const VertexType& vtx,
 
         strip_ready |= (buffer_index == 1);
 
-        if (topology == Regs::TriangleTopology::Strip)
+        if (topology == PipelineRegs::TriangleTopology::Strip)
             buffer_index = !buffer_index;
-        else if (topology == Regs::TriangleTopology::Fan)
+        else if (topology == PipelineRegs::TriangleTopology::Fan)
             buffer_index = 1;
         break;
 
@@ -57,7 +57,7 @@ void PrimitiveAssembler<VertexType>::Reset() {
 }
 
 template <typename VertexType>
-void PrimitiveAssembler<VertexType>::Reconfigure(Regs::TriangleTopology topology) {
+void PrimitiveAssembler<VertexType>::Reconfigure(PipelineRegs::TriangleTopology topology) {
     Reset();
     this->topology = topology;
 }
diff --git a/src/video_core/primitive_assembly.h b/src/video_core/primitive_assembly.h
index 24da47382..e8eccdf27 100644
--- a/src/video_core/primitive_assembly.h
+++ b/src/video_core/primitive_assembly.h
@@ -5,7 +5,7 @@
 #pragma once
 
 #include <functional>
-#include "video_core/pica.h"
+#include "video_core/regs_pipeline.h"
 
 namespace Pica {
 
@@ -18,7 +18,8 @@ struct PrimitiveAssembler {
     using TriangleHandler =
         std::function<void(const VertexType& v0, const VertexType& v1, const VertexType& v2)>;
 
-    PrimitiveAssembler(Regs::TriangleTopology topology = Regs::TriangleTopology::List);
+    PrimitiveAssembler(
+        PipelineRegs::TriangleTopology topology = PipelineRegs::TriangleTopology::List);
 
     /*
      * Queues a vertex, builds primitives from the vertex queue according to the given
@@ -36,10 +37,10 @@ struct PrimitiveAssembler {
     /**
      * Reconfigures the PrimitiveAssembler to use a different triangle topology.
      */
-    void Reconfigure(Regs::TriangleTopology topology);
+    void Reconfigure(PipelineRegs::TriangleTopology topology);
 
 private:
-    Regs::TriangleTopology topology;
+    PipelineRegs::TriangleTopology topology;
 
     int buffer_index;
     VertexType buffer[2];
diff --git a/src/video_core/regs_pipeline.h b/src/video_core/regs_pipeline.h
new file mode 100644
index 000000000..5844a66ee
--- /dev/null
+++ b/src/video_core/regs_pipeline.h
@@ -0,0 +1,224 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include <array>
+
+#include "common/assert.h"
+#include "common/bit_field.h"
+#include "common/common_funcs.h"
+#include "common/common_types.h"
+
+namespace Pica {
+
+struct PipelineRegs {
+    enum class VertexAttributeFormat : u64 {
+        BYTE = 0,
+        UBYTE = 1,
+        SHORT = 2,
+        FLOAT = 3,
+    };
+
+    struct {
+        BitField<0, 29, u32> base_address;
+
+        PAddr GetPhysicalBaseAddress() const {
+            return base_address * 8;
+        }
+
+        // Descriptor for internal vertex attributes
+        union {
+            BitField<0, 2, VertexAttributeFormat> format0; // size of one element
+            BitField<2, 2, u64> size0;                     // number of elements minus 1
+            BitField<4, 2, VertexAttributeFormat> format1;
+            BitField<6, 2, u64> size1;
+            BitField<8, 2, VertexAttributeFormat> format2;
+            BitField<10, 2, u64> size2;
+            BitField<12, 2, VertexAttributeFormat> format3;
+            BitField<14, 2, u64> size3;
+            BitField<16, 2, VertexAttributeFormat> format4;
+            BitField<18, 2, u64> size4;
+            BitField<20, 2, VertexAttributeFormat> format5;
+            BitField<22, 2, u64> size5;
+            BitField<24, 2, VertexAttributeFormat> format6;
+            BitField<26, 2, u64> size6;
+            BitField<28, 2, VertexAttributeFormat> format7;
+            BitField<30, 2, u64> size7;
+            BitField<32, 2, VertexAttributeFormat> format8;
+            BitField<34, 2, u64> size8;
+            BitField<36, 2, VertexAttributeFormat> format9;
+            BitField<38, 2, u64> size9;
+            BitField<40, 2, VertexAttributeFormat> format10;
+            BitField<42, 2, u64> size10;
+            BitField<44, 2, VertexAttributeFormat> format11;
+            BitField<46, 2, u64> size11;
+
+            BitField<48, 12, u64> attribute_mask;
+
+            // number of total attributes minus 1
+            BitField<60, 4, u64> max_attribute_index;
+        };
+
+        inline VertexAttributeFormat GetFormat(int n) const {
+            VertexAttributeFormat formats[] = {format0, format1, format2,  format3,
+                                               format4, format5, format6,  format7,
+                                               format8, format9, format10, format11};
+            return formats[n];
+        }
+
+        inline int GetNumElements(int n) const {
+            u64 sizes[] = {size0, size1, size2, size3, size4,  size5,
+                           size6, size7, size8, size9, size10, size11};
+            return (int)sizes[n] + 1;
+        }
+
+        inline int GetElementSizeInBytes(int n) const {
+            return (GetFormat(n) == VertexAttributeFormat::FLOAT)
+                       ? 4
+                       : (GetFormat(n) == VertexAttributeFormat::SHORT) ? 2 : 1;
+        }
+
+        inline int GetStride(int n) const {
+            return GetNumElements(n) * GetElementSizeInBytes(n);
+        }
+
+        inline bool IsDefaultAttribute(int id) const {
+            return (id >= 12) || (attribute_mask & (1ULL << id)) != 0;
+        }
+
+        inline int GetNumTotalAttributes() const {
+            return (int)max_attribute_index + 1;
+        }
+
+        // Attribute loaders map the source vertex data to input attributes
+        // This e.g. allows to load different attributes from different memory locations
+        struct {
+            // Source attribute data offset from the base address
+            u32 data_offset;
+
+            union {
+                BitField<0, 4, u64> comp0;
+                BitField<4, 4, u64> comp1;
+                BitField<8, 4, u64> comp2;
+                BitField<12, 4, u64> comp3;
+                BitField<16, 4, u64> comp4;
+                BitField<20, 4, u64> comp5;
+                BitField<24, 4, u64> comp6;
+                BitField<28, 4, u64> comp7;
+                BitField<32, 4, u64> comp8;
+                BitField<36, 4, u64> comp9;
+                BitField<40, 4, u64> comp10;
+                BitField<44, 4, u64> comp11;
+
+                // bytes for a single vertex in this loader
+                BitField<48, 8, u64> byte_count;
+
+                BitField<60, 4, u64> component_count;
+            };
+
+            inline int GetComponent(int n) const {
+                u64 components[] = {comp0, comp1, comp2, comp3, comp4,  comp5,
+                                    comp6, comp7, comp8, comp9, comp10, comp11};
+                return (int)components[n];
+            }
+        } attribute_loaders[12];
+    } vertex_attributes;
+
+    struct {
+        enum IndexFormat : u32 {
+            BYTE = 0,
+            SHORT = 1,
+        };
+
+        union {
+            BitField<0, 31, u32> offset; // relative to base attribute address
+            BitField<31, 1, IndexFormat> format;
+        };
+    } index_array;
+
+    // Number of vertices to render
+    u32 num_vertices;
+
+    INSERT_PADDING_WORDS(0x1);
+
+    // The index of the first vertex to render
+    u32 vertex_offset;
+
+    INSERT_PADDING_WORDS(0x3);
+
+    // These two trigger rendering of triangles
+    u32 trigger_draw;
+    u32 trigger_draw_indexed;
+
+    INSERT_PADDING_WORDS(0x2);
+
+    // These registers are used to setup the default "fall-back" vertex shader attributes
+    struct {
+        // Index of the current default attribute
+        u32 index;
+
+        // Writing to these registers sets the "current" default attribute.
+        u32 set_value[3];
+    } vs_default_attributes_setup;
+
+    INSERT_PADDING_WORDS(0x2);
+
+    struct {
+        // There are two channels that can be used to configure the next command buffer, which can
+        // be then executed by writing to the "trigger" registers. There are two reasons why a game
+        // might use this feature:
+        //  1) With this, an arbitrary number of additional command buffers may be executed in
+        //     sequence without requiring any intervention of the CPU after the initial one is
+        //     kicked off.
+        //  2) Games can configure these registers to provide a command list subroutine mechanism.
+
+        BitField<0, 20, u32> size[2]; ///< Size (in bytes / 8) of each channel's command buffer
+        BitField<0, 28, u32> addr[2]; ///< Physical address / 8 of each channel's command buffer
+        u32 trigger[2]; ///< Triggers execution of the channel's command buffer when written to
+
+        unsigned GetSize(unsigned index) const {
+            ASSERT(index < 2);
+            return 8 * size[index];
+        }
+
+        PAddr GetPhysicalAddress(unsigned index) const {
+            ASSERT(index < 2);
+            return (PAddr)(8 * addr[index]);
+        }
+    } command_buffer;
+
+    INSERT_PADDING_WORDS(4);
+
+    /// Number of input attributes to the vertex shader minus 1
+    BitField<0, 4, u32> max_input_attrib_index;
+
+    INSERT_PADDING_WORDS(2);
+
+    enum class GPUMode : u32 {
+        Drawing = 0,
+        Configuring = 1,
+    };
+
+    GPUMode gpu_mode;
+
+    INSERT_PADDING_WORDS(0x18);
+
+    enum class TriangleTopology : u32 {
+        List = 0,
+        Strip = 1,
+        Fan = 2,
+        Shader = 3, // Programmable setup unit implemented in a geometry shader
+    };
+
+    BitField<8, 2, TriangleTopology> triangle_topology;
+
+    u32 restart_primitive;
+
+    INSERT_PADDING_WORDS(0x20);
+};
+
+static_assert(sizeof(PipelineRegs) == 0x80 * sizeof(u32), "PipelineRegs struct has incorrect size");
+
+} // namespace Pica
diff --git a/src/video_core/vertex_loader.cpp b/src/video_core/vertex_loader.cpp
index bf83b61ca..20e2370be 100644
--- a/src/video_core/vertex_loader.cpp
+++ b/src/video_core/vertex_loader.cpp
@@ -16,7 +16,7 @@
 
 namespace Pica {
 
-void VertexLoader::Setup(const Pica::Regs& regs) {
+void VertexLoader::Setup(const PipelineRegs& regs) {
     ASSERT_MSG(!is_setup, "VertexLoader is not intended to be setup more than once.");
 
     const auto& attribute_config = regs.vertex_attributes;
@@ -85,15 +85,16 @@ void VertexLoader::LoadVertex(u32 base_address, int index, int vertex,
                 memory_accesses.AddAccess(
                     source_addr,
                     vertex_attribute_elements[i] *
-                        ((vertex_attribute_formats[i] == Regs::VertexAttributeFormat::FLOAT)
+                        ((vertex_attribute_formats[i] == PipelineRegs::VertexAttributeFormat::FLOAT)
                              ? 4
-                             : (vertex_attribute_formats[i] == Regs::VertexAttributeFormat::SHORT)
+                             : (vertex_attribute_formats[i] ==
+                                PipelineRegs::VertexAttributeFormat::SHORT)
                                    ? 2
                                    : 1));
             }
 
             switch (vertex_attribute_formats[i]) {
-            case Regs::VertexAttributeFormat::BYTE: {
+            case PipelineRegs::VertexAttributeFormat::BYTE: {
                 const s8* srcdata =
                     reinterpret_cast<const s8*>(Memory::GetPhysicalPointer(source_addr));
                 for (unsigned int comp = 0; comp < vertex_attribute_elements[i]; ++comp) {
@@ -101,7 +102,7 @@ void VertexLoader::LoadVertex(u32 base_address, int index, int vertex,
                 }
                 break;
             }
-            case Regs::VertexAttributeFormat::UBYTE: {
+            case PipelineRegs::VertexAttributeFormat::UBYTE: {
                 const u8* srcdata =
                     reinterpret_cast<const u8*>(Memory::GetPhysicalPointer(source_addr));
                 for (unsigned int comp = 0; comp < vertex_attribute_elements[i]; ++comp) {
@@ -109,7 +110,7 @@ void VertexLoader::LoadVertex(u32 base_address, int index, int vertex,
                 }
                 break;
             }
-            case Regs::VertexAttributeFormat::SHORT: {
+            case PipelineRegs::VertexAttributeFormat::SHORT: {
                 const s16* srcdata =
                     reinterpret_cast<const s16*>(Memory::GetPhysicalPointer(source_addr));
                 for (unsigned int comp = 0; comp < vertex_attribute_elements[i]; ++comp) {
@@ -117,7 +118,7 @@ void VertexLoader::LoadVertex(u32 base_address, int index, int vertex,
                 }
                 break;
             }
-            case Regs::VertexAttributeFormat::FLOAT: {
+            case PipelineRegs::VertexAttributeFormat::FLOAT: {
                 const float* srcdata =
                     reinterpret_cast<const float*>(Memory::GetPhysicalPointer(source_addr));
                 for (unsigned int comp = 0; comp < vertex_attribute_elements[i]; ++comp) {
diff --git a/src/video_core/vertex_loader.h b/src/video_core/vertex_loader.h
index 51f3d45b4..7815715bc 100644
--- a/src/video_core/vertex_loader.h
+++ b/src/video_core/vertex_loader.h
@@ -17,11 +17,11 @@ struct AttributeBuffer;
 class VertexLoader {
 public:
     VertexLoader() = default;
-    explicit VertexLoader(const Pica::Regs& regs) {
+    explicit VertexLoader(const PipelineRegs& regs) {
         Setup(regs);
     }
 
-    void Setup(const Pica::Regs& regs);
+    void Setup(const PipelineRegs& regs);
     void LoadVertex(u32 base_address, int index, int vertex, Shader::AttributeBuffer& input,
                     DebugUtils::MemoryAccessTracker& memory_accesses);
 
@@ -32,7 +32,7 @@ public:
 private:
     std::array<u32, 16> vertex_attribute_sources;
     std::array<u32, 16> vertex_attribute_strides{};
-    std::array<Regs::VertexAttributeFormat, 16> vertex_attribute_formats;
+    std::array<PipelineRegs::VertexAttributeFormat, 16> vertex_attribute_formats;
     std::array<u32, 16> vertex_attribute_elements{};
     std::array<bool, 16> vertex_attribute_is_default;
     int num_total_attributes = 0;