From a1c9ac7845395c250a78fc8df93a9ffed29f3d5b Mon Sep 17 00:00:00 2001
From: Yuri Kunde Schlesner <yuriks@yuriks.net>
Date: Thu, 5 Jan 2017 20:11:23 -0200
Subject: [PATCH 1/3] VideoCore: Move LookupTexture out of debug_utils.h

---
 .../debugger/graphics/graphics_cmdlists.cpp   |  10 +-
 .../debugger/graphics/graphics_surface.cpp    |   7 +-
 src/video_core/CMakeLists.txt                 |  34 +--
 src/video_core/debug_utils/debug_utils.cpp    | 256 +----------------
 src/video_core/debug_utils/debug_utils.h      |  25 --
 src/video_core/rasterizer.cpp                 |   5 +-
 .../renderer_opengl/gl_rasterizer_cache.cpp   |  11 +-
 src/video_core/texture/texture_decode.cpp     | 270 ++++++++++++++++++
 src/video_core/texture/texture_decode.h       |  40 +++
 9 files changed, 350 insertions(+), 308 deletions(-)
 create mode 100644 src/video_core/texture/texture_decode.cpp
 create mode 100644 src/video_core/texture/texture_decode.h

diff --git a/src/citra_qt/debugger/graphics/graphics_cmdlists.cpp b/src/citra_qt/debugger/graphics/graphics_cmdlists.cpp
index f5a2ec761..ee79f0edf 100644
--- a/src/citra_qt/debugger/graphics/graphics_cmdlists.cpp
+++ b/src/citra_qt/debugger/graphics/graphics_cmdlists.cpp
@@ -20,13 +20,14 @@
 #include "video_core/debug_utils/debug_utils.h"
 #include "video_core/pica.h"
 #include "video_core/pica_state.h"
+#include "video_core/texture/texture_decode.h"
 
 namespace {
-QImage LoadTexture(const u8* src, const Pica::DebugUtils::TextureInfo& info) {
+QImage LoadTexture(const u8* src, const Pica::Texture::TextureInfo& info) {
     QImage decoded_image(info.width, info.height, QImage::Format_ARGB32);
     for (int y = 0; y < info.height; ++y) {
         for (int x = 0; x < info.width; ++x) {
-            Math::Vec4<u8> color = Pica::DebugUtils::LookupTexture(src, x, y, info, true);
+            Math::Vec4<u8> color = Pica::Texture::LookupTexture(src, x, y, info, true);
             decoded_image.setPixel(x, y, qRgba(color.r(), color.g(), color.b(), color.a()));
         }
     }
@@ -36,9 +37,10 @@ QImage LoadTexture(const u8* src, const Pica::DebugUtils::TextureInfo& info) {
 
 class TextureInfoWidget : public QWidget {
 public:
-    TextureInfoWidget(const u8* src, const Pica::DebugUtils::TextureInfo& info,
+    TextureInfoWidget(const u8* src, const Pica::Texture::TextureInfo& info,
                       QWidget* parent = nullptr)
         : QWidget(parent) {
+
         QLabel* image_widget = new QLabel;
         QPixmap image_pixmap = QPixmap::fromImage(LoadTexture(src, info));
         image_pixmap = image_pixmap.scaled(200, 100, Qt::KeepAspectRatio, Qt::SmoothTransformation);
@@ -160,7 +162,7 @@ void GPUCommandListWidget::SetCommandInfo(const QModelIndex& index) {
         const auto config = texture.config;
         const auto format = texture.format;
 
-        const auto info = Pica::DebugUtils::TextureInfo::FromPicaRegister(config, format);
+        const auto info = Pica::Texture::TextureInfo::FromPicaRegister(config, format);
         const u8* src = Memory::GetPhysicalPointer(config.GetPhysicalAddress());
         new_info_widget = new TextureInfoWidget(src, info);
     }
diff --git a/src/citra_qt/debugger/graphics/graphics_surface.cpp b/src/citra_qt/debugger/graphics/graphics_surface.cpp
index 4efd95d3c..c0a72a6ef 100644
--- a/src/citra_qt/debugger/graphics/graphics_surface.cpp
+++ b/src/citra_qt/debugger/graphics/graphics_surface.cpp
@@ -18,6 +18,7 @@
 #include "core/memory.h"
 #include "video_core/pica.h"
 #include "video_core/pica_state.h"
+#include "video_core/texture/texture_decode.h"
 #include "video_core/utils.h"
 
 SurfacePicture::SurfacePicture(QWidget* parent, GraphicsSurfaceWidget* surface_widget_)
@@ -512,7 +513,7 @@ void GraphicsSurfaceWidget::OnUpdate() {
         }
 
         const auto texture = Pica::g_state.regs.GetTextures()[texture_index];
-        auto info = Pica::DebugUtils::TextureInfo::FromPicaRegister(texture.config, texture.format);
+        auto info = Pica::Texture::TextureInfo::FromPicaRegister(texture.config, texture.format);
 
         surface_address = info.physical_address;
         surface_width = info.width;
@@ -574,7 +575,7 @@ void GraphicsSurfaceWidget::OnUpdate() {
     if (surface_format <= Format::MaxTextureFormat) {
 
         // Generate a virtual texture
-        Pica::DebugUtils::TextureInfo info;
+        Pica::Texture::TextureInfo info;
         info.physical_address = surface_address;
         info.width = surface_width;
         info.height = surface_height;
@@ -583,7 +584,7 @@ void GraphicsSurfaceWidget::OnUpdate() {
 
         for (unsigned int y = 0; y < surface_height; ++y) {
             for (unsigned int x = 0; x < surface_width; ++x) {
-                Math::Vec4<u8> color = Pica::DebugUtils::LookupTexture(buffer, x, y, info, true);
+                Math::Vec4<u8> color = Pica::Texture::LookupTexture(buffer, x, y, info, true);
                 decoded_image.setPixel(x, y, qRgba(color.r(), color.g(), color.b(), color.a()));
             }
         }
diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index d55b84ce0..ecb424cfd 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -1,36 +1,29 @@
 set(SRCS
+            clipper.cpp
+            command_processor.cpp
+            debug_utils/debug_utils.cpp
+            pica.cpp
+            primitive_assembly.cpp
+            rasterizer.cpp
+            renderer_base.cpp
             renderer_opengl/gl_rasterizer.cpp
             renderer_opengl/gl_rasterizer_cache.cpp
             renderer_opengl/gl_shader_gen.cpp
             renderer_opengl/gl_shader_util.cpp
             renderer_opengl/gl_state.cpp
             renderer_opengl/renderer_opengl.cpp
-            debug_utils/debug_utils.cpp
-            clipper.cpp
-            command_processor.cpp
-            pica.cpp
-            primitive_assembly.cpp
-            rasterizer.cpp
-            renderer_base.cpp
             shader/shader.cpp
             shader/shader_interpreter.cpp
             swrasterizer.cpp
+            texture/texture_decode.cpp
             vertex_loader.cpp
             video_core.cpp
             )
 
 set(HEADERS
-            debug_utils/debug_utils.h
-            renderer_opengl/gl_rasterizer.h
-            renderer_opengl/gl_rasterizer_cache.h
-            renderer_opengl/gl_resource_manager.h
-            renderer_opengl/gl_shader_gen.h
-            renderer_opengl/gl_shader_util.h
-            renderer_opengl/gl_state.h
-            renderer_opengl/pica_to_gl.h
-            renderer_opengl/renderer_opengl.h
             clipper.h
             command_processor.h
+            debug_utils/debug_utils.h
             gpu_debugger.h
             pica.h
             pica_state.h
@@ -39,10 +32,19 @@ set(HEADERS
             rasterizer.h
             rasterizer_interface.h
             renderer_base.h
+            renderer_opengl/gl_rasterizer.h
+            renderer_opengl/gl_rasterizer_cache.h
+            renderer_opengl/gl_resource_manager.h
+            renderer_opengl/gl_shader_gen.h
+            renderer_opengl/gl_shader_util.h
+            renderer_opengl/gl_state.h
+            renderer_opengl/pica_to_gl.h
+            renderer_opengl/renderer_opengl.h
             shader/debug_data.h
             shader/shader.h
             shader/shader_interpreter.h
             swrasterizer.h
+            texture/texture_decode.h
             utils.h
             vertex_loader.h
             video_core.h
diff --git a/src/video_core/debug_utils/debug_utils.cpp b/src/video_core/debug_utils/debug_utils.cpp
index c44b3d95a..2d40f7d4f 100644
--- a/src/video_core/debug_utils/debug_utils.cpp
+++ b/src/video_core/debug_utils/debug_utils.cpp
@@ -35,6 +35,7 @@
 #include "video_core/rasterizer_interface.h"
 #include "video_core/renderer_base.h"
 #include "video_core/shader/shader.h"
+#include "video_core/texture/texture_decode.h"
 #include "video_core/utils.h"
 #include "video_core/video_core.h"
 
@@ -315,257 +316,6 @@ std::unique_ptr<PicaTrace> FinishPicaTracing() {
     return ret;
 }
 
-const Math::Vec4<u8> LookupTexture(const u8* source, int x, int y, const TextureInfo& info,
-                                   bool disable_alpha) {
-    const unsigned int coarse_x = x & ~7;
-    const unsigned int coarse_y = y & ~7;
-
-    if (info.format != Regs::TextureFormat::ETC1 && info.format != Regs::TextureFormat::ETC1A4) {
-        // TODO(neobrain): Fix code design to unify vertical block offsets!
-        source += coarse_y * info.stride;
-    }
-
-    // TODO: Assert that width/height are multiples of block dimensions
-
-    switch (info.format) {
-    case Regs::TextureFormat::RGBA8: {
-        auto res = Color::DecodeRGBA8(source + VideoCore::GetMortonOffset(x, y, 4));
-        return {res.r(), res.g(), res.b(), static_cast<u8>(disable_alpha ? 255 : res.a())};
-    }
-
-    case Regs::TextureFormat::RGB8: {
-        auto res = Color::DecodeRGB8(source + VideoCore::GetMortonOffset(x, y, 3));
-        return {res.r(), res.g(), res.b(), 255};
-    }
-
-    case Regs::TextureFormat::RGB5A1: {
-        auto res = Color::DecodeRGB5A1(source + VideoCore::GetMortonOffset(x, y, 2));
-        return {res.r(), res.g(), res.b(), static_cast<u8>(disable_alpha ? 255 : res.a())};
-    }
-
-    case Regs::TextureFormat::RGB565: {
-        auto res = Color::DecodeRGB565(source + VideoCore::GetMortonOffset(x, y, 2));
-        return {res.r(), res.g(), res.b(), 255};
-    }
-
-    case Regs::TextureFormat::RGBA4: {
-        auto res = Color::DecodeRGBA4(source + VideoCore::GetMortonOffset(x, y, 2));
-        return {res.r(), res.g(), res.b(), static_cast<u8>(disable_alpha ? 255 : res.a())};
-    }
-
-    case Regs::TextureFormat::IA8: {
-        const u8* source_ptr = source + VideoCore::GetMortonOffset(x, y, 2);
-
-        if (disable_alpha) {
-            // Show intensity as red, alpha as green
-            return {source_ptr[1], source_ptr[0], 0, 255};
-        } else {
-            return {source_ptr[1], source_ptr[1], source_ptr[1], source_ptr[0]};
-        }
-    }
-
-    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);
-        return {*source_ptr, *source_ptr, *source_ptr, 255};
-    }
-
-    case Regs::TextureFormat::A8: {
-        const u8* source_ptr = source + VideoCore::GetMortonOffset(x, y, 1);
-
-        if (disable_alpha) {
-            return {*source_ptr, *source_ptr, *source_ptr, 255};
-        } else {
-            return {0, 0, 0, *source_ptr};
-        }
-    }
-
-    case Regs::TextureFormat::IA4: {
-        const u8* source_ptr = source + VideoCore::GetMortonOffset(x, y, 1);
-
-        u8 i = Color::Convert4To8(((*source_ptr) & 0xF0) >> 4);
-        u8 a = Color::Convert4To8((*source_ptr) & 0xF);
-
-        if (disable_alpha) {
-            // Show intensity as red, alpha as green
-            return {i, a, 0, 255};
-        } else {
-            return {i, i, i, a};
-        }
-    }
-
-    case Regs::TextureFormat::I4: {
-        u32 morton_offset = VideoCore::GetMortonOffset(x, y, 1);
-        const u8* source_ptr = source + morton_offset / 2;
-
-        u8 i = (morton_offset % 2) ? ((*source_ptr & 0xF0) >> 4) : (*source_ptr & 0xF);
-        i = Color::Convert4To8(i);
-
-        return {i, i, i, 255};
-    }
-
-    case Regs::TextureFormat::A4: {
-        u32 morton_offset = VideoCore::GetMortonOffset(x, y, 1);
-        const u8* source_ptr = source + morton_offset / 2;
-
-        u8 a = (morton_offset % 2) ? ((*source_ptr & 0xF0) >> 4) : (*source_ptr & 0xF);
-        a = Color::Convert4To8(a);
-
-        if (disable_alpha) {
-            return {a, a, a, 255};
-        } else {
-            return {0, 0, 0, a};
-        }
-    }
-
-    case Regs::TextureFormat::ETC1:
-    case Regs::TextureFormat::ETC1A4: {
-        bool has_alpha = (info.format == Regs::TextureFormat::ETC1A4);
-
-        // ETC1 further subdivides each 8x8 tile into four 4x4 subtiles
-        const int subtile_width = 4;
-        const int subtile_height = 4;
-
-        int subtile_index = ((x / subtile_width) & 1) + 2 * ((y / subtile_height) & 1);
-        unsigned subtile_bytes = has_alpha ? 2 : 1; // TODO: Name...
-
-        const u64* source_ptr = (const u64*)(source + coarse_x * subtile_bytes * 4 +
-                                             coarse_y * subtile_bytes * 4 * (info.width / 8) +
-                                             subtile_index * subtile_bytes * 8);
-        u64 alpha = 0xFFFFFFFFFFFFFFFF;
-        if (has_alpha) {
-            alpha = *source_ptr;
-            source_ptr++;
-        }
-
-        union ETC1Tile {
-            // Each of these two is a collection of 16 bits (one per lookup value)
-            BitField<0, 16, u64> table_subindexes;
-            BitField<16, 16, u64> negation_flags;
-
-            unsigned GetTableSubIndex(unsigned index) const {
-                return (table_subindexes >> index) & 1;
-            }
-
-            bool GetNegationFlag(unsigned index) const {
-                return ((negation_flags >> index) & 1) == 1;
-            }
-
-            BitField<32, 1, u64> flip;
-            BitField<33, 1, u64> differential_mode;
-
-            BitField<34, 3, u64> table_index_2;
-            BitField<37, 3, u64> table_index_1;
-
-            union {
-                // delta value + base value
-                BitField<40, 3, s64> db;
-                BitField<43, 5, u64> b;
-
-                BitField<48, 3, s64> dg;
-                BitField<51, 5, u64> g;
-
-                BitField<56, 3, s64> dr;
-                BitField<59, 5, u64> r;
-            } differential;
-
-            union {
-                BitField<40, 4, u64> b2;
-                BitField<44, 4, u64> b1;
-
-                BitField<48, 4, u64> g2;
-                BitField<52, 4, u64> g1;
-
-                BitField<56, 4, u64> r2;
-                BitField<60, 4, u64> r1;
-            } separate;
-
-            const Math::Vec3<u8> GetRGB(int x, int y) const {
-                int texel = 4 * x + y;
-
-                if (flip)
-                    std::swap(x, y);
-
-                // Lookup base value
-                Math::Vec3<int> ret;
-                if (differential_mode) {
-                    ret.r() = static_cast<int>(differential.r);
-                    ret.g() = static_cast<int>(differential.g);
-                    ret.b() = static_cast<int>(differential.b);
-                    if (x >= 2) {
-                        ret.r() += static_cast<int>(differential.dr);
-                        ret.g() += static_cast<int>(differential.dg);
-                        ret.b() += static_cast<int>(differential.db);
-                    }
-                    ret.r() = Color::Convert5To8(ret.r());
-                    ret.g() = Color::Convert5To8(ret.g());
-                    ret.b() = Color::Convert5To8(ret.b());
-                } else {
-                    if (x < 2) {
-                        ret.r() = Color::Convert4To8(static_cast<u8>(separate.r1));
-                        ret.g() = Color::Convert4To8(static_cast<u8>(separate.g1));
-                        ret.b() = Color::Convert4To8(static_cast<u8>(separate.b1));
-                    } else {
-                        ret.r() = Color::Convert4To8(static_cast<u8>(separate.r2));
-                        ret.g() = Color::Convert4To8(static_cast<u8>(separate.g2));
-                        ret.b() = Color::Convert4To8(static_cast<u8>(separate.b2));
-                    }
-                }
-
-                // Add modifier
-                unsigned table_index =
-                    static_cast<int>((x < 2) ? table_index_1.Value() : table_index_2.Value());
-
-                static const std::array<std::array<u8, 2>, 8> etc1_modifier_table = {{
-                    {{2, 8}},
-                    {{5, 17}},
-                    {{9, 29}},
-                    {{13, 42}},
-                    {{18, 60}},
-                    {{24, 80}},
-                    {{33, 106}},
-                    {{47, 183}},
-                }};
-
-                int modifier = etc1_modifier_table.at(table_index).at(GetTableSubIndex(texel));
-                if (GetNegationFlag(texel))
-                    modifier *= -1;
-
-                ret.r() = MathUtil::Clamp(ret.r() + modifier, 0, 255);
-                ret.g() = MathUtil::Clamp(ret.g() + modifier, 0, 255);
-                ret.b() = MathUtil::Clamp(ret.b() + modifier, 0, 255);
-
-                return ret.Cast<u8>();
-            }
-        } const* etc1_tile = reinterpret_cast<const ETC1Tile*>(source_ptr);
-
-        alpha >>= 4 * ((x & 3) * 4 + (y & 3));
-        return Math::MakeVec(etc1_tile->GetRGB(x & 3, y & 3),
-                             disable_alpha ? (u8)255 : Color::Convert4To8(alpha & 0xF));
-    }
-
-    default:
-        LOG_ERROR(HW_GPU, "Unknown texture format: %x", (u32)info.format);
-        DEBUG_ASSERT(false);
-        return {};
-    }
-}
-
-TextureInfo TextureInfo::FromPicaRegister(const Regs::TextureConfig& config,
-                                          const Regs::TextureFormat& format) {
-    TextureInfo info;
-    info.physical_address = config.GetPhysicalAddress();
-    info.width = config.width;
-    info.height = config.height;
-    info.format = format;
-    info.stride = Pica::Regs::NibblesPerPixel(info.format) * info.width / 2;
-    return info;
-}
-
 #ifdef HAVE_PNG
 // Adapter functions to libpng to write/flush to File::IOFile instances.
 static void WriteIOFile(png_structp png_ptr, png_bytep data, png_size_t length) {
@@ -642,12 +392,12 @@ void DumpTexture(const Pica::Regs::TextureConfig& texture_config, u8* data) {
     buf = new u8[row_stride * texture_config.height];
     for (unsigned y = 0; y < texture_config.height; ++y) {
         for (unsigned x = 0; x < texture_config.width; ++x) {
-            TextureInfo info;
+            Pica::Texture::TextureInfo info;
             info.width = texture_config.width;
             info.height = texture_config.height;
             info.stride = row_stride;
             info.format = g_state.regs.texture0_format;
-            Math::Vec4<u8> texture_color = LookupTexture(data, x, y, info);
+            Math::Vec4<u8> texture_color = Pica::Texture::LookupTexture(data, x, y, info);
             buf[3 * x + y * row_stride] = texture_color.r();
             buf[3 * x + y * row_stride + 1] = texture_color.g();
             buf[3 * x + y * row_stride + 2] = texture_color.b();
diff --git a/src/video_core/debug_utils/debug_utils.h b/src/video_core/debug_utils/debug_utils.h
index 46ea8d9c7..938a2e1b5 100644
--- a/src/video_core/debug_utils/debug_utils.h
+++ b/src/video_core/debug_utils/debug_utils.h
@@ -205,31 +205,6 @@ inline bool IsPicaTracing() {
 void OnPicaRegWrite(PicaTrace::Write write);
 std::unique_ptr<PicaTrace> FinishPicaTracing();
 
-struct TextureInfo {
-    PAddr physical_address;
-    int width;
-    int height;
-    int stride;
-    Pica::Regs::TextureFormat format;
-
-    static TextureInfo FromPicaRegister(const Pica::Regs::TextureConfig& config,
-                                        const Pica::Regs::TextureFormat& format);
-};
-
-/**
- * Lookup texel located at the given coordinates and return an RGBA vector of its color.
- * @param source Source pointer to read data from
- * @param s,t Texture coordinates to read from
- * @param info TextureInfo object describing the texture setup
- * @param disable_alpha This is used for debug widgets which use this method to display textures
- * without providing a good way to visualize alpha by themselves. If true, this will return 255 for
- * the alpha component, and either drop the information entirely or store it in an "unused" color
- * channel.
- * @todo Eventually we should get rid of the disable_alpha parameter.
- */
-const Math::Vec4<u8> LookupTexture(const u8* source, int s, int t, const TextureInfo& info,
-                                   bool disable_alpha = false);
-
 void DumpTexture(const Pica::Regs::TextureConfig& texture_config, u8* data);
 
 std::string GetTevStageConfigColorCombinerString(const Pica::Regs::TevStageConfig& tev_stage);
diff --git a/src/video_core/rasterizer.cpp b/src/video_core/rasterizer.cpp
index b9f5d4533..c034c12d3 100644
--- a/src/video_core/rasterizer.cpp
+++ b/src/video_core/rasterizer.cpp
@@ -21,6 +21,7 @@
 #include "video_core/pica_types.h"
 #include "video_core/rasterizer.h"
 #include "video_core/shader/shader.h"
+#include "video_core/texture/texture_decode.h"
 #include "video_core/utils.h"
 
 namespace Pica {
@@ -579,10 +580,10 @@ static void ProcessTriangleInternal(const Shader::OutputVertex& v0, const Shader
                     u8* texture_data =
                         Memory::GetPhysicalPointer(texture.config.GetPhysicalAddress());
                     auto info =
-                        DebugUtils::TextureInfo::FromPicaRegister(texture.config, texture.format);
+                        Texture::TextureInfo::FromPicaRegister(texture.config, texture.format);
 
                     // TODO: Apply the min and mag filters to the texture
-                    texture_color[i] = DebugUtils::LookupTexture(texture_data, s, t, info);
+                    texture_color[i] = Texture::LookupTexture(texture_data, s, t, info);
 #if PICA_DUMP_TEXTURES
                     DebugUtils::DumpTexture(texture.config, texture_data);
 #endif
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 1e7eedecb..4167d3161 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -17,10 +17,10 @@
 #include "common/vector_math.h"
 #include "core/frontend/emu_window.h"
 #include "core/memory.h"
-#include "video_core/debug_utils/debug_utils.h"
 #include "video_core/pica_state.h"
 #include "video_core/renderer_opengl/gl_rasterizer_cache.h"
 #include "video_core/renderer_opengl/gl_state.h"
+#include "video_core/texture/texture_decode.h"
 #include "video_core/utils.h"
 #include "video_core/video_core.h"
 
@@ -339,7 +339,7 @@ CachedSurface* RasterizerCacheOpenGL::GetSurface(const CachedSurface& params, bo
 
                 std::vector<Math::Vec4<u8>> tex_buffer(params.width * params.height);
 
-                Pica::DebugUtils::TextureInfo tex_info;
+                Pica::Texture::TextureInfo tex_info;
                 tex_info.width = params.width;
                 tex_info.height = params.height;
                 tex_info.stride =
@@ -349,7 +349,7 @@ CachedSurface* RasterizerCacheOpenGL::GetSurface(const CachedSurface& params, bo
 
                 for (unsigned y = 0; y < params.height; ++y) {
                     for (unsigned x = 0; x < params.width; ++x) {
-                        tex_buffer[x + params.width * y] = Pica::DebugUtils::LookupTexture(
+                        tex_buffer[x + params.width * y] = Pica::Texture::LookupTexture(
                             texture_src_data, x, params.height - 1 - y, tex_info);
                     }
                 }
@@ -512,8 +512,9 @@ CachedSurface* RasterizerCacheOpenGL::GetSurfaceRect(const CachedSurface& params
 
 CachedSurface* RasterizerCacheOpenGL::GetTextureSurface(
     const Pica::Regs::FullTextureConfig& config) {
-    Pica::DebugUtils::TextureInfo info =
-        Pica::DebugUtils::TextureInfo::FromPicaRegister(config.config, config.format);
+
+    Pica::Texture::TextureInfo info =
+        Pica::Texture::TextureInfo::FromPicaRegister(config.config, config.format);
 
     CachedSurface params;
     params.addr = info.physical_address;
diff --git a/src/video_core/texture/texture_decode.cpp b/src/video_core/texture/texture_decode.cpp
new file mode 100644
index 000000000..a3b05fe81
--- /dev/null
+++ b/src/video_core/texture/texture_decode.cpp
@@ -0,0 +1,270 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/assert.h"
+#include "common/color.h"
+#include "common/logging/log.h"
+#include "common/math_util.h"
+#include "common/vector_math.h"
+#include "video_core/texture/texture_decode.h"
+#include "video_core/utils.h"
+
+namespace Pica {
+namespace Texture {
+
+Math::Vec4<u8> LookupTexture(const u8* source, int x, int y, const TextureInfo& info,
+                             bool disable_alpha) {
+    const unsigned int coarse_x = x & ~7;
+    const unsigned int coarse_y = y & ~7;
+
+    if (info.format != Regs::TextureFormat::ETC1 && info.format != Regs::TextureFormat::ETC1A4) {
+        // TODO(neobrain): Fix code design to unify vertical block offsets!
+        source += coarse_y * info.stride;
+    }
+
+    // TODO: Assert that width/height are multiples of block dimensions
+
+    switch (info.format) {
+    case Regs::TextureFormat::RGBA8: {
+        auto res = Color::DecodeRGBA8(source + VideoCore::GetMortonOffset(x, y, 4));
+        return {res.r(), res.g(), res.b(), static_cast<u8>(disable_alpha ? 255 : res.a())};
+    }
+
+    case Regs::TextureFormat::RGB8: {
+        auto res = Color::DecodeRGB8(source + VideoCore::GetMortonOffset(x, y, 3));
+        return {res.r(), res.g(), res.b(), 255};
+    }
+
+    case Regs::TextureFormat::RGB5A1: {
+        auto res = Color::DecodeRGB5A1(source + VideoCore::GetMortonOffset(x, y, 2));
+        return {res.r(), res.g(), res.b(), static_cast<u8>(disable_alpha ? 255 : res.a())};
+    }
+
+    case Regs::TextureFormat::RGB565: {
+        auto res = Color::DecodeRGB565(source + VideoCore::GetMortonOffset(x, y, 2));
+        return {res.r(), res.g(), res.b(), 255};
+    }
+
+    case Regs::TextureFormat::RGBA4: {
+        auto res = Color::DecodeRGBA4(source + VideoCore::GetMortonOffset(x, y, 2));
+        return {res.r(), res.g(), res.b(), static_cast<u8>(disable_alpha ? 255 : res.a())};
+    }
+
+    case Regs::TextureFormat::IA8: {
+        const u8* source_ptr = source + VideoCore::GetMortonOffset(x, y, 2);
+
+        if (disable_alpha) {
+            // Show intensity as red, alpha as green
+            return {source_ptr[1], source_ptr[0], 0, 255};
+        } else {
+            return {source_ptr[1], source_ptr[1], source_ptr[1], source_ptr[0]};
+        }
+    }
+
+    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);
+        return {*source_ptr, *source_ptr, *source_ptr, 255};
+    }
+
+    case Regs::TextureFormat::A8: {
+        const u8* source_ptr = source + VideoCore::GetMortonOffset(x, y, 1);
+
+        if (disable_alpha) {
+            return {*source_ptr, *source_ptr, *source_ptr, 255};
+        } else {
+            return {0, 0, 0, *source_ptr};
+        }
+    }
+
+    case Regs::TextureFormat::IA4: {
+        const u8* source_ptr = source + VideoCore::GetMortonOffset(x, y, 1);
+
+        u8 i = Color::Convert4To8(((*source_ptr) & 0xF0) >> 4);
+        u8 a = Color::Convert4To8((*source_ptr) & 0xF);
+
+        if (disable_alpha) {
+            // Show intensity as red, alpha as green
+            return {i, a, 0, 255};
+        } else {
+            return {i, i, i, a};
+        }
+    }
+
+    case Regs::TextureFormat::I4: {
+        u32 morton_offset = VideoCore::GetMortonOffset(x, y, 1);
+        const u8* source_ptr = source + morton_offset / 2;
+
+        u8 i = (morton_offset % 2) ? ((*source_ptr & 0xF0) >> 4) : (*source_ptr & 0xF);
+        i = Color::Convert4To8(i);
+
+        return {i, i, i, 255};
+    }
+
+    case Regs::TextureFormat::A4: {
+        u32 morton_offset = VideoCore::GetMortonOffset(x, y, 1);
+        const u8* source_ptr = source + morton_offset / 2;
+
+        u8 a = (morton_offset % 2) ? ((*source_ptr & 0xF0) >> 4) : (*source_ptr & 0xF);
+        a = Color::Convert4To8(a);
+
+        if (disable_alpha) {
+            return {a, a, a, 255};
+        } else {
+            return {0, 0, 0, a};
+        }
+    }
+
+    case Regs::TextureFormat::ETC1:
+    case Regs::TextureFormat::ETC1A4: {
+        bool has_alpha = (info.format == Regs::TextureFormat::ETC1A4);
+
+        // ETC1 further subdivides each 8x8 tile into four 4x4 subtiles
+        const int subtile_width = 4;
+        const int subtile_height = 4;
+
+        int subtile_index = ((x / subtile_width) & 1) + 2 * ((y / subtile_height) & 1);
+        unsigned subtile_bytes = has_alpha ? 2 : 1; // TODO: Name...
+
+        const u64* source_ptr = (const u64*)(source + coarse_x * subtile_bytes * 4 +
+                                             coarse_y * subtile_bytes * 4 * (info.width / 8) +
+                                             subtile_index * subtile_bytes * 8);
+        u64 alpha = 0xFFFFFFFFFFFFFFFF;
+        if (has_alpha) {
+            alpha = *source_ptr;
+            source_ptr++;
+        }
+
+        union ETC1Tile {
+            // Each of these two is a collection of 16 bits (one per lookup value)
+            BitField<0, 16, u64> table_subindexes;
+            BitField<16, 16, u64> negation_flags;
+
+            unsigned GetTableSubIndex(unsigned index) const {
+                return (table_subindexes >> index) & 1;
+            }
+
+            bool GetNegationFlag(unsigned index) const {
+                return ((negation_flags >> index) & 1) == 1;
+            }
+
+            BitField<32, 1, u64> flip;
+            BitField<33, 1, u64> differential_mode;
+
+            BitField<34, 3, u64> table_index_2;
+            BitField<37, 3, u64> table_index_1;
+
+            union {
+                // delta value + base value
+                BitField<40, 3, s64> db;
+                BitField<43, 5, u64> b;
+
+                BitField<48, 3, s64> dg;
+                BitField<51, 5, u64> g;
+
+                BitField<56, 3, s64> dr;
+                BitField<59, 5, u64> r;
+            } differential;
+
+            union {
+                BitField<40, 4, u64> b2;
+                BitField<44, 4, u64> b1;
+
+                BitField<48, 4, u64> g2;
+                BitField<52, 4, u64> g1;
+
+                BitField<56, 4, u64> r2;
+                BitField<60, 4, u64> r1;
+            } separate;
+
+            const Math::Vec3<u8> GetRGB(int x, int y) const {
+                int texel = 4 * x + y;
+
+                if (flip)
+                    std::swap(x, y);
+
+                // Lookup base value
+                Math::Vec3<int> ret;
+                if (differential_mode) {
+                    ret.r() = static_cast<int>(differential.r);
+                    ret.g() = static_cast<int>(differential.g);
+                    ret.b() = static_cast<int>(differential.b);
+                    if (x >= 2) {
+                        ret.r() += static_cast<int>(differential.dr);
+                        ret.g() += static_cast<int>(differential.dg);
+                        ret.b() += static_cast<int>(differential.db);
+                    }
+                    ret.r() = Color::Convert5To8(ret.r());
+                    ret.g() = Color::Convert5To8(ret.g());
+                    ret.b() = Color::Convert5To8(ret.b());
+                } else {
+                    if (x < 2) {
+                        ret.r() = Color::Convert4To8(static_cast<u8>(separate.r1));
+                        ret.g() = Color::Convert4To8(static_cast<u8>(separate.g1));
+                        ret.b() = Color::Convert4To8(static_cast<u8>(separate.b1));
+                    } else {
+                        ret.r() = Color::Convert4To8(static_cast<u8>(separate.r2));
+                        ret.g() = Color::Convert4To8(static_cast<u8>(separate.g2));
+                        ret.b() = Color::Convert4To8(static_cast<u8>(separate.b2));
+                    }
+                }
+
+                // Add modifier
+                unsigned table_index =
+                    static_cast<int>((x < 2) ? table_index_1.Value() : table_index_2.Value());
+
+                static const std::array<std::array<u8, 2>, 8> etc1_modifier_table = {{
+                    {{2, 8}},
+                    {{5, 17}},
+                    {{9, 29}},
+                    {{13, 42}},
+                    {{18, 60}},
+                    {{24, 80}},
+                    {{33, 106}},
+                    {{47, 183}},
+                }};
+
+                int modifier = etc1_modifier_table.at(table_index).at(GetTableSubIndex(texel));
+                if (GetNegationFlag(texel))
+                    modifier *= -1;
+
+                ret.r() = MathUtil::Clamp(ret.r() + modifier, 0, 255);
+                ret.g() = MathUtil::Clamp(ret.g() + modifier, 0, 255);
+                ret.b() = MathUtil::Clamp(ret.b() + modifier, 0, 255);
+
+                return ret.Cast<u8>();
+            }
+        } const* etc1_tile = reinterpret_cast<const ETC1Tile*>(source_ptr);
+
+        alpha >>= 4 * ((x & 3) * 4 + (y & 3));
+        return Math::MakeVec(etc1_tile->GetRGB(x & 3, y & 3),
+                             disable_alpha ? (u8)255 : Color::Convert4To8(alpha & 0xF));
+    }
+
+    default:
+        LOG_ERROR(HW_GPU, "Unknown texture format: %x", (u32)info.format);
+        DEBUG_ASSERT(false);
+        return {};
+    }
+}
+
+TextureInfo TextureInfo::FromPicaRegister(const Regs::TextureConfig& config,
+                                          const Regs::TextureFormat& format) {
+    TextureInfo info;
+    info.physical_address = config.GetPhysicalAddress();
+    info.width = config.width;
+    info.height = config.height;
+    info.format = format;
+    info.stride = Pica::Regs::NibblesPerPixel(info.format) * info.width / 2;
+    return info;
+}
+
+} // namespace Texture
+} // namespace Pica
diff --git a/src/video_core/texture/texture_decode.h b/src/video_core/texture/texture_decode.h
new file mode 100644
index 000000000..0c1438b0f
--- /dev/null
+++ b/src/video_core/texture/texture_decode.h
@@ -0,0 +1,40 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+#include "common/vector_math.h"
+#include "video_core/pica.h"
+
+namespace Pica {
+namespace Texture {
+
+struct TextureInfo {
+    PAddr physical_address;
+    int width;
+    int height;
+    int stride;
+    Pica::Regs::TextureFormat format;
+
+    static TextureInfo FromPicaRegister(const Pica::Regs::TextureConfig& config,
+                                        const Pica::Regs::TextureFormat& format);
+};
+
+/**
+ * Lookup texel located at the given coordinates and return an RGBA vector of its color.
+ * @param source Source pointer to read data from
+ * @param s,t Texture coordinates to read from
+ * @param info TextureInfo object describing the texture setup
+ * @param disable_alpha This is used for debug widgets which use this method to display textures
+ * without providing a good way to visualize alpha by themselves. If true, this will return 255 for
+ * the alpha component, and either drop the information entirely or store it in an "unused" color
+ * channel.
+ * @todo Eventually we should get rid of the disable_alpha parameter.
+ */
+Math::Vec4<u8> LookupTexture(const u8* source, int s, int t, const TextureInfo& info,
+                             bool disable_alpha = false);
+
+} // namespace Texture
+} // namespace Pica

From 09a750e8662e5d4d608177fdfb69b398c3202cd6 Mon Sep 17 00:00:00 2001
From: Yuri Kunde Schlesner <yuriks@yuriks.net>
Date: Fri, 6 Jan 2017 01:19:06 -0200
Subject: [PATCH 2/3] Pica/Texture: Simplify/cleanup texture tile addressing

---
 .../debugger/graphics/graphics_surface.cpp    |  13 +--
 src/video_core/pica.h                         |   5 +-
 .../renderer_opengl/gl_rasterizer_cache.cpp   |   3 +-
 src/video_core/texture/texture_decode.cpp     | 108 +++++++++++++-----
 src/video_core/texture/texture_decode.h       |  32 +++++-
 5 files changed, 117 insertions(+), 44 deletions(-)

diff --git a/src/citra_qt/debugger/graphics/graphics_surface.cpp b/src/citra_qt/debugger/graphics/graphics_surface.cpp
index c0a72a6ef..bd82b00d4 100644
--- a/src/citra_qt/debugger/graphics/graphics_surface.cpp
+++ b/src/citra_qt/debugger/graphics/graphics_surface.cpp
@@ -568,19 +568,14 @@ void GraphicsSurfaceWidget::OnUpdate() {
 
     surface_picture_label->show();
 
-    unsigned nibbles_per_pixel = GraphicsSurfaceWidget::NibblesPerPixel(surface_format);
-    unsigned stride = nibbles_per_pixel * surface_width / 2;
-
-    // We handle depth formats here because DebugUtils only supports TextureFormats
     if (surface_format <= Format::MaxTextureFormat) {
-
         // Generate a virtual texture
         Pica::Texture::TextureInfo info;
         info.physical_address = surface_address;
         info.width = surface_width;
         info.height = surface_height;
         info.format = static_cast<Pica::Regs::TextureFormat>(surface_format);
-        info.stride = stride;
+        info.SetDefaultStride();
 
         for (unsigned int y = 0; y < surface_height; ++y) {
             for (unsigned int x = 0; x < surface_width; ++x) {
@@ -588,8 +583,12 @@ void GraphicsSurfaceWidget::OnUpdate() {
                 decoded_image.setPixel(x, y, qRgba(color.r(), color.g(), color.b(), color.a()));
             }
         }
-
     } else {
+        // We handle depth formats here because DebugUtils only supports TextureFormats
+
+        // TODO(yuriks): Convert to newer tile-based addressing
+        unsigned nibbles_per_pixel = GraphicsSurfaceWidget::NibblesPerPixel(surface_format);
+        unsigned stride = nibbles_per_pixel * surface_width / 2;
 
         ASSERT_MSG(nibbles_per_pixel >= 2,
                    "Depth decoder only supports formats with at least one byte per pixel");
diff --git a/src/video_core/pica.h b/src/video_core/pica.h
index b2db609ec..4ab4f1f40 100644
--- a/src/video_core/pica.h
+++ b/src/video_core/pica.h
@@ -275,8 +275,11 @@ struct Regs {
         case TextureFormat::I8:
         case TextureFormat::A8:
         case TextureFormat::IA4:
-        default: // placeholder for yet unknown formats
             return 2;
+
+        default: // placeholder for yet unknown formats
+            UNIMPLEMENTED();
+            return 0;
         }
     }
 
diff --git a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
index 4167d3161..60380257a 100644
--- a/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
+++ b/src/video_core/renderer_opengl/gl_rasterizer_cache.cpp
@@ -342,9 +342,8 @@ CachedSurface* RasterizerCacheOpenGL::GetSurface(const CachedSurface& params, bo
                 Pica::Texture::TextureInfo tex_info;
                 tex_info.width = params.width;
                 tex_info.height = params.height;
-                tex_info.stride =
-                    params.width * CachedSurface::GetFormatBpp(params.pixel_format) / 8;
                 tex_info.format = (Pica::Regs::TextureFormat)params.pixel_format;
+                tex_info.SetDefaultStride();
                 tex_info.physical_address = params.addr;
 
                 for (unsigned y = 0; y < params.height; ++y) {
diff --git a/src/video_core/texture/texture_decode.cpp b/src/video_core/texture/texture_decode.cpp
index a3b05fe81..f13d6e577 100644
--- a/src/video_core/texture/texture_decode.cpp
+++ b/src/video_core/texture/texture_decode.cpp
@@ -9,52 +9,104 @@
 #include "common/logging/log.h"
 #include "common/math_util.h"
 #include "common/vector_math.h"
+#include "video_core/pica.h"
 #include "video_core/texture/texture_decode.h"
 #include "video_core/utils.h"
 
+using TextureFormat = Pica::Regs::TextureFormat;
+
 namespace Pica {
 namespace Texture {
 
-Math::Vec4<u8> LookupTexture(const u8* source, int x, int y, const TextureInfo& info,
-                             bool disable_alpha) {
-    const unsigned int coarse_x = x & ~7;
-    const unsigned int coarse_y = y & ~7;
+constexpr size_t TILE_SIZE = 8 * 8;
+constexpr size_t ETC1_SUBTILES = 2 * 2;
 
-    if (info.format != Regs::TextureFormat::ETC1 && info.format != Regs::TextureFormat::ETC1A4) {
-        // TODO(neobrain): Fix code design to unify vertical block offsets!
-        source += coarse_y * info.stride;
+size_t CalculateTileSize(TextureFormat format) {
+    switch (format) {
+    case TextureFormat::RGBA8:
+        return 4 * TILE_SIZE;
+
+    case TextureFormat::RGB8:
+        return 3 * TILE_SIZE;
+
+    case TextureFormat::RGB5A1:
+    case TextureFormat::RGB565:
+    case TextureFormat::RGBA4:
+    case TextureFormat::IA8:
+    case TextureFormat::RG8:
+        return 2 * TILE_SIZE;
+
+    case TextureFormat::I8:
+    case TextureFormat::A8:
+    case TextureFormat::IA4:
+        return 1 * TILE_SIZE;
+
+    case TextureFormat::I4:
+    case TextureFormat::A4:
+        return TILE_SIZE / 2;
+
+    case TextureFormat::ETC1:
+        return ETC1_SUBTILES * 8;
+
+    case TextureFormat::ETC1A4:
+        return ETC1_SUBTILES * 16;
+
+    default: // placeholder for yet unknown formats
+        UNIMPLEMENTED();
+        return 0;
     }
+}
 
-    // TODO: Assert that width/height are multiples of block dimensions
+Math::Vec4<u8> LookupTexture(const u8* source, unsigned int x, unsigned int y,
+                             const TextureInfo& info, bool disable_alpha) {
+    // Coordinate in tiles
+    const unsigned int coarse_x = x / 8;
+    const unsigned int coarse_y = y / 8;
+
+    // Coordinate inside the tile
+    const unsigned int fine_x = x % 8;
+    const unsigned int fine_y = y % 8;
+
+    const u8* line = source + coarse_y * info.stride;
+    const u8* tile = line + coarse_x * CalculateTileSize(info.format);
+    return LookupTexelInTile(tile, fine_x, fine_y, info, disable_alpha);
+}
+
+Math::Vec4<u8> LookupTexelInTile(const u8* source, unsigned int x, unsigned int y,
+                                 const TextureInfo& info, bool disable_alpha) {
+    DEBUG_ASSERT(x < 8);
+    DEBUG_ASSERT(y < 8);
+
+    using VideoCore::MortonInterleave;
 
     switch (info.format) {
     case Regs::TextureFormat::RGBA8: {
-        auto res = Color::DecodeRGBA8(source + VideoCore::GetMortonOffset(x, y, 4));
+        auto res = Color::DecodeRGBA8(source + MortonInterleave(x, y) * 4);
         return {res.r(), res.g(), res.b(), static_cast<u8>(disable_alpha ? 255 : res.a())};
     }
 
     case Regs::TextureFormat::RGB8: {
-        auto res = Color::DecodeRGB8(source + VideoCore::GetMortonOffset(x, y, 3));
+        auto res = Color::DecodeRGB8(source + MortonInterleave(x, y) * 3);
         return {res.r(), res.g(), res.b(), 255};
     }
 
     case Regs::TextureFormat::RGB5A1: {
-        auto res = Color::DecodeRGB5A1(source + VideoCore::GetMortonOffset(x, y, 2));
+        auto res = Color::DecodeRGB5A1(source + MortonInterleave(x, y) * 2);
         return {res.r(), res.g(), res.b(), static_cast<u8>(disable_alpha ? 255 : res.a())};
     }
 
     case Regs::TextureFormat::RGB565: {
-        auto res = Color::DecodeRGB565(source + VideoCore::GetMortonOffset(x, y, 2));
+        auto res = Color::DecodeRGB565(source + MortonInterleave(x, y) * 2);
         return {res.r(), res.g(), res.b(), 255};
     }
 
     case Regs::TextureFormat::RGBA4: {
-        auto res = Color::DecodeRGBA4(source + VideoCore::GetMortonOffset(x, y, 2));
+        auto res = Color::DecodeRGBA4(source + MortonInterleave(x, y) * 2);
         return {res.r(), res.g(), res.b(), static_cast<u8>(disable_alpha ? 255 : res.a())};
     }
 
     case Regs::TextureFormat::IA8: {
-        const u8* source_ptr = source + VideoCore::GetMortonOffset(x, y, 2);
+        const u8* source_ptr = source + MortonInterleave(x, y) * 2;
 
         if (disable_alpha) {
             // Show intensity as red, alpha as green
@@ -65,17 +117,17 @@ Math::Vec4<u8> LookupTexture(const u8* source, int x, int y, const TextureInfo&
     }
 
     case Regs::TextureFormat::RG8: {
-        auto res = Color::DecodeRG8(source + VideoCore::GetMortonOffset(x, y, 2));
+        auto res = Color::DecodeRG8(source + MortonInterleave(x, y) * 2);
         return {res.r(), res.g(), 0, 255};
     }
 
     case Regs::TextureFormat::I8: {
-        const u8* source_ptr = source + VideoCore::GetMortonOffset(x, y, 1);
+        const u8* source_ptr = source + MortonInterleave(x, y);
         return {*source_ptr, *source_ptr, *source_ptr, 255};
     }
 
     case Regs::TextureFormat::A8: {
-        const u8* source_ptr = source + VideoCore::GetMortonOffset(x, y, 1);
+        const u8* source_ptr = source + MortonInterleave(x, y);
 
         if (disable_alpha) {
             return {*source_ptr, *source_ptr, *source_ptr, 255};
@@ -85,7 +137,7 @@ Math::Vec4<u8> LookupTexture(const u8* source, int x, int y, const TextureInfo&
     }
 
     case Regs::TextureFormat::IA4: {
-        const u8* source_ptr = source + VideoCore::GetMortonOffset(x, y, 1);
+        const u8* source_ptr = source + MortonInterleave(x, y);
 
         u8 i = Color::Convert4To8(((*source_ptr) & 0xF0) >> 4);
         u8 a = Color::Convert4To8((*source_ptr) & 0xF);
@@ -99,7 +151,7 @@ Math::Vec4<u8> LookupTexture(const u8* source, int x, int y, const TextureInfo&
     }
 
     case Regs::TextureFormat::I4: {
-        u32 morton_offset = VideoCore::GetMortonOffset(x, y, 1);
+        u32 morton_offset = MortonInterleave(x, y);
         const u8* source_ptr = source + morton_offset / 2;
 
         u8 i = (morton_offset % 2) ? ((*source_ptr & 0xF0) >> 4) : (*source_ptr & 0xF);
@@ -109,7 +161,7 @@ Math::Vec4<u8> LookupTexture(const u8* source, int x, int y, const TextureInfo&
     }
 
     case Regs::TextureFormat::A4: {
-        u32 morton_offset = VideoCore::GetMortonOffset(x, y, 1);
+        u32 morton_offset = MortonInterleave(x, y);
         const u8* source_ptr = source + morton_offset / 2;
 
         u8 a = (morton_offset % 2) ? ((*source_ptr & 0xF0) >> 4) : (*source_ptr & 0xF);
@@ -127,15 +179,15 @@ Math::Vec4<u8> LookupTexture(const u8* source, int x, int y, const TextureInfo&
         bool has_alpha = (info.format == Regs::TextureFormat::ETC1A4);
 
         // ETC1 further subdivides each 8x8 tile into four 4x4 subtiles
-        const int subtile_width = 4;
-        const int subtile_height = 4;
+        constexpr unsigned int subtile_width = 4;
+        constexpr unsigned int subtile_height = 4;
 
-        int subtile_index = ((x / subtile_width) & 1) + 2 * ((y / subtile_height) & 1);
-        unsigned subtile_bytes = has_alpha ? 2 : 1; // TODO: Name...
+        unsigned int subtile_index = (x / subtile_width) + 2 * (y / subtile_height);
+        size_t subtile_size = has_alpha ? 16 : 8;
+
+        // TODO(yuriks): Use memcpy instead of reinterpret_cast
+        const u64* source_ptr = reinterpret_cast<const u64*>(source + subtile_index * subtile_size);
 
-        const u64* source_ptr = (const u64*)(source + coarse_x * subtile_bytes * 4 +
-                                             coarse_y * subtile_bytes * 4 * (info.width / 8) +
-                                             subtile_index * subtile_bytes * 8);
         u64 alpha = 0xFFFFFFFFFFFFFFFF;
         if (has_alpha) {
             alpha = *source_ptr;
@@ -262,7 +314,7 @@ TextureInfo TextureInfo::FromPicaRegister(const Regs::TextureConfig& config,
     info.width = config.width;
     info.height = config.height;
     info.format = format;
-    info.stride = Pica::Regs::NibblesPerPixel(info.format) * info.width / 2;
+    info.SetDefaultStride();
     return info;
 }
 
diff --git a/src/video_core/texture/texture_decode.h b/src/video_core/texture/texture_decode.h
index 0c1438b0f..5c636939a 100644
--- a/src/video_core/texture/texture_decode.h
+++ b/src/video_core/texture/texture_decode.h
@@ -11,21 +11,29 @@
 namespace Pica {
 namespace Texture {
 
+/// Returns the byte size of a 8*8 tile of the specified texture format.
+size_t CalculateTileSize(Pica::Regs::TextureFormat format);
+
 struct TextureInfo {
     PAddr physical_address;
-    int width;
-    int height;
-    int stride;
+    unsigned int width;
+    unsigned int height;
+    ptrdiff_t stride;
     Pica::Regs::TextureFormat format;
 
     static TextureInfo FromPicaRegister(const Pica::Regs::TextureConfig& config,
                                         const Pica::Regs::TextureFormat& format);
+
+    /// Calculates stride from format and width, assuming that the entire texture is contiguous.
+    void SetDefaultStride() {
+        stride = Pica::Texture::CalculateTileSize(format) * (width / 8);
+    }
 };
 
 /**
  * Lookup texel located at the given coordinates and return an RGBA vector of its color.
  * @param source Source pointer to read data from
- * @param s,t Texture coordinates to read from
+ * @param x,y Texture coordinates to read from
  * @param info TextureInfo object describing the texture setup
  * @param disable_alpha This is used for debug widgets which use this method to display textures
  * without providing a good way to visualize alpha by themselves. If true, this will return 255 for
@@ -33,8 +41,20 @@ struct TextureInfo {
  * channel.
  * @todo Eventually we should get rid of the disable_alpha parameter.
  */
-Math::Vec4<u8> LookupTexture(const u8* source, int s, int t, const TextureInfo& info,
-                             bool disable_alpha = false);
+Math::Vec4<u8> LookupTexture(const u8* source, unsigned int x, unsigned int y,
+                             const TextureInfo& info, bool disable_alpha = false);
+
+/**
+ * Looks up a texel from a single 8x8 texture tile.
+ *
+ * @param source Pointer to the beginning of the tile.
+ * @param x, y In-tile coordinates to read from. Must be < 8.
+ * @param info TextureInfo describing the texture format.
+ * @param disable_alpha Used for debugging. Sets the result alpha to 255 and either discards the
+ *                      real alpha or inserts it in an otherwise unused channel.
+ */
+Math::Vec4<u8> LookupTexelInTile(const u8* source, unsigned int x, unsigned int y,
+                                 const TextureInfo& info, bool disable_alpha);
 
 } // namespace Texture
 } // namespace Pica

From c74787a11c44bd1fa444425d36ac3049dffe1496 Mon Sep 17 00:00:00 2001
From: Yuri Kunde Schlesner <yuriks@yuriks.net>
Date: Fri, 6 Jan 2017 01:59:34 -0200
Subject: [PATCH 3/3] Pica/Texture: Move part of ETC1 decoding to new file and
 cleanups

---
 src/video_core/CMakeLists.txt             |   2 +
 src/video_core/texture/etc1.cpp           | 124 +++++++++++++++++++++
 src/video_core/texture/etc1.h             |  16 +++
 src/video_core/texture/texture_decode.cpp | 125 +++-------------------
 4 files changed, 158 insertions(+), 109 deletions(-)
 create mode 100644 src/video_core/texture/etc1.cpp
 create mode 100644 src/video_core/texture/etc1.h

diff --git a/src/video_core/CMakeLists.txt b/src/video_core/CMakeLists.txt
index ecb424cfd..ad984cd94 100644
--- a/src/video_core/CMakeLists.txt
+++ b/src/video_core/CMakeLists.txt
@@ -15,6 +15,7 @@ set(SRCS
             shader/shader.cpp
             shader/shader_interpreter.cpp
             swrasterizer.cpp
+            texture/etc1.cpp
             texture/texture_decode.cpp
             vertex_loader.cpp
             video_core.cpp
@@ -44,6 +45,7 @@ set(HEADERS
             shader/shader.h
             shader/shader_interpreter.h
             swrasterizer.h
+            texture/etc1.h
             texture/texture_decode.h
             utils.h
             vertex_loader.h
diff --git a/src/video_core/texture/etc1.cpp b/src/video_core/texture/etc1.cpp
new file mode 100644
index 000000000..af60cde1e
--- /dev/null
+++ b/src/video_core/texture/etc1.cpp
@@ -0,0 +1,124 @@
+// 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/bit_field.h"
+#include "common/color.h"
+#include "common/common_types.h"
+#include "common/math_util.h"
+#include "common/vector_math.h"
+#include "video_core/texture/etc1.h"
+
+namespace Pica {
+namespace Texture {
+
+namespace {
+
+constexpr std::array<u8[2], 8> etc1_modifier_table = {{
+    {2, 8}, {5, 17}, {9, 29}, {13, 42}, {18, 60}, {24, 80}, {33, 106}, {47, 183},
+}};
+
+union ETC1Tile {
+    u64 raw;
+
+    // Each of these two is a collection of 16 bits (one per lookup value)
+    BitField<0, 16, u64> table_subindexes;
+    BitField<16, 16, u64> negation_flags;
+
+    unsigned GetTableSubIndex(unsigned index) const {
+        return (table_subindexes >> index) & 1;
+    }
+
+    bool GetNegationFlag(unsigned index) const {
+        return ((negation_flags >> index) & 1) == 1;
+    }
+
+    BitField<32, 1, u64> flip;
+    BitField<33, 1, u64> differential_mode;
+
+    BitField<34, 3, u64> table_index_2;
+    BitField<37, 3, u64> table_index_1;
+
+    union {
+        // delta value + base value
+        BitField<40, 3, s64> db;
+        BitField<43, 5, u64> b;
+
+        BitField<48, 3, s64> dg;
+        BitField<51, 5, u64> g;
+
+        BitField<56, 3, s64> dr;
+        BitField<59, 5, u64> r;
+    } differential;
+
+    union {
+        BitField<40, 4, u64> b2;
+        BitField<44, 4, u64> b1;
+
+        BitField<48, 4, u64> g2;
+        BitField<52, 4, u64> g1;
+
+        BitField<56, 4, u64> r2;
+        BitField<60, 4, u64> r1;
+    } separate;
+
+    const Math::Vec3<u8> GetRGB(unsigned int x, unsigned int y) const {
+        int texel = 4 * x + y;
+
+        if (flip)
+            std::swap(x, y);
+
+        // Lookup base value
+        Math::Vec3<int> ret;
+        if (differential_mode) {
+            ret.r() = static_cast<int>(differential.r);
+            ret.g() = static_cast<int>(differential.g);
+            ret.b() = static_cast<int>(differential.b);
+            if (x >= 2) {
+                ret.r() += static_cast<int>(differential.dr);
+                ret.g() += static_cast<int>(differential.dg);
+                ret.b() += static_cast<int>(differential.db);
+            }
+            ret.r() = Color::Convert5To8(ret.r());
+            ret.g() = Color::Convert5To8(ret.g());
+            ret.b() = Color::Convert5To8(ret.b());
+        } else {
+            if (x < 2) {
+                ret.r() = Color::Convert4To8(static_cast<u8>(separate.r1));
+                ret.g() = Color::Convert4To8(static_cast<u8>(separate.g1));
+                ret.b() = Color::Convert4To8(static_cast<u8>(separate.b1));
+            } else {
+                ret.r() = Color::Convert4To8(static_cast<u8>(separate.r2));
+                ret.g() = Color::Convert4To8(static_cast<u8>(separate.g2));
+                ret.b() = Color::Convert4To8(static_cast<u8>(separate.b2));
+            }
+        }
+
+        // Add modifier
+        unsigned table_index =
+            static_cast<int>((x < 2) ? table_index_1.Value() : table_index_2.Value());
+
+        int modifier = etc1_modifier_table[table_index][GetTableSubIndex(texel)];
+        if (GetNegationFlag(texel))
+            modifier *= -1;
+
+        ret.r() = MathUtil::Clamp(ret.r() + modifier, 0, 255);
+        ret.g() = MathUtil::Clamp(ret.g() + modifier, 0, 255);
+        ret.b() = MathUtil::Clamp(ret.b() + modifier, 0, 255);
+
+        return ret.Cast<u8>();
+    }
+};
+
+} // anonymous namespace
+
+Math::Vec3<u8> SampleETC1Subtile(u64 value, unsigned int x, unsigned int y) {
+    ETC1Tile tile{value};
+    return tile.GetRGB(x, y);
+}
+
+} // namespace Texture
+} // namespace Pica
diff --git a/src/video_core/texture/etc1.h b/src/video_core/texture/etc1.h
new file mode 100644
index 000000000..e188b19df
--- /dev/null
+++ b/src/video_core/texture/etc1.h
@@ -0,0 +1,16 @@
+// Copyright 2017 Citra Emulator Project
+// Licensed under GPLv2 or any later version
+// Refer to the license.txt file included.
+
+#pragma once
+
+#include "common/common_types.h"
+#include "common/vector_math.h"
+
+namespace Pica {
+namespace Texture {
+
+Math::Vec3<u8> SampleETC1Subtile(u64 value, unsigned int x, unsigned int y);
+
+} // namespace Texture
+} // namespace Pica
diff --git a/src/video_core/texture/texture_decode.cpp b/src/video_core/texture/texture_decode.cpp
index f13d6e577..f611a1aa9 100644
--- a/src/video_core/texture/texture_decode.cpp
+++ b/src/video_core/texture/texture_decode.cpp
@@ -8,8 +8,10 @@
 #include "common/color.h"
 #include "common/logging/log.h"
 #include "common/math_util.h"
+#include "common/swap.h"
 #include "common/vector_math.h"
 #include "video_core/pica.h"
+#include "video_core/texture/etc1.h"
 #include "video_core/texture/texture_decode.h"
 #include "video_core/utils.h"
 
@@ -177,127 +179,32 @@ Math::Vec4<u8> LookupTexelInTile(const u8* source, unsigned int x, unsigned int
     case Regs::TextureFormat::ETC1:
     case Regs::TextureFormat::ETC1A4: {
         bool has_alpha = (info.format == Regs::TextureFormat::ETC1A4);
+        size_t subtile_size = has_alpha ? 16 : 8;
 
         // ETC1 further subdivides each 8x8 tile into four 4x4 subtiles
         constexpr unsigned int subtile_width = 4;
         constexpr unsigned int subtile_height = 4;
 
         unsigned int subtile_index = (x / subtile_width) + 2 * (y / subtile_height);
-        size_t subtile_size = has_alpha ? 16 : 8;
+        x %= subtile_width;
+        y %= subtile_height;
 
-        // TODO(yuriks): Use memcpy instead of reinterpret_cast
-        const u64* source_ptr = reinterpret_cast<const u64*>(source + subtile_index * subtile_size);
+        const u8* subtile_ptr = source + subtile_index * subtile_size;
 
-        u64 alpha = 0xFFFFFFFFFFFFFFFF;
+        u8 alpha = 255;
         if (has_alpha) {
-            alpha = *source_ptr;
-            source_ptr++;
+            u64_le packed_alpha;
+            memcpy(&packed_alpha, subtile_ptr, sizeof(u64));
+            subtile_ptr += sizeof(u64);
+
+            alpha = Color::Convert4To8((packed_alpha >> (4 * (x * subtile_width + y))) & 0xF);
         }
 
-        union ETC1Tile {
-            // Each of these two is a collection of 16 bits (one per lookup value)
-            BitField<0, 16, u64> table_subindexes;
-            BitField<16, 16, u64> negation_flags;
+        u64_le subtile_data;
+        memcpy(&subtile_data, subtile_ptr, sizeof(u64));
 
-            unsigned GetTableSubIndex(unsigned index) const {
-                return (table_subindexes >> index) & 1;
-            }
-
-            bool GetNegationFlag(unsigned index) const {
-                return ((negation_flags >> index) & 1) == 1;
-            }
-
-            BitField<32, 1, u64> flip;
-            BitField<33, 1, u64> differential_mode;
-
-            BitField<34, 3, u64> table_index_2;
-            BitField<37, 3, u64> table_index_1;
-
-            union {
-                // delta value + base value
-                BitField<40, 3, s64> db;
-                BitField<43, 5, u64> b;
-
-                BitField<48, 3, s64> dg;
-                BitField<51, 5, u64> g;
-
-                BitField<56, 3, s64> dr;
-                BitField<59, 5, u64> r;
-            } differential;
-
-            union {
-                BitField<40, 4, u64> b2;
-                BitField<44, 4, u64> b1;
-
-                BitField<48, 4, u64> g2;
-                BitField<52, 4, u64> g1;
-
-                BitField<56, 4, u64> r2;
-                BitField<60, 4, u64> r1;
-            } separate;
-
-            const Math::Vec3<u8> GetRGB(int x, int y) const {
-                int texel = 4 * x + y;
-
-                if (flip)
-                    std::swap(x, y);
-
-                // Lookup base value
-                Math::Vec3<int> ret;
-                if (differential_mode) {
-                    ret.r() = static_cast<int>(differential.r);
-                    ret.g() = static_cast<int>(differential.g);
-                    ret.b() = static_cast<int>(differential.b);
-                    if (x >= 2) {
-                        ret.r() += static_cast<int>(differential.dr);
-                        ret.g() += static_cast<int>(differential.dg);
-                        ret.b() += static_cast<int>(differential.db);
-                    }
-                    ret.r() = Color::Convert5To8(ret.r());
-                    ret.g() = Color::Convert5To8(ret.g());
-                    ret.b() = Color::Convert5To8(ret.b());
-                } else {
-                    if (x < 2) {
-                        ret.r() = Color::Convert4To8(static_cast<u8>(separate.r1));
-                        ret.g() = Color::Convert4To8(static_cast<u8>(separate.g1));
-                        ret.b() = Color::Convert4To8(static_cast<u8>(separate.b1));
-                    } else {
-                        ret.r() = Color::Convert4To8(static_cast<u8>(separate.r2));
-                        ret.g() = Color::Convert4To8(static_cast<u8>(separate.g2));
-                        ret.b() = Color::Convert4To8(static_cast<u8>(separate.b2));
-                    }
-                }
-
-                // Add modifier
-                unsigned table_index =
-                    static_cast<int>((x < 2) ? table_index_1.Value() : table_index_2.Value());
-
-                static const std::array<std::array<u8, 2>, 8> etc1_modifier_table = {{
-                    {{2, 8}},
-                    {{5, 17}},
-                    {{9, 29}},
-                    {{13, 42}},
-                    {{18, 60}},
-                    {{24, 80}},
-                    {{33, 106}},
-                    {{47, 183}},
-                }};
-
-                int modifier = etc1_modifier_table.at(table_index).at(GetTableSubIndex(texel));
-                if (GetNegationFlag(texel))
-                    modifier *= -1;
-
-                ret.r() = MathUtil::Clamp(ret.r() + modifier, 0, 255);
-                ret.g() = MathUtil::Clamp(ret.g() + modifier, 0, 255);
-                ret.b() = MathUtil::Clamp(ret.b() + modifier, 0, 255);
-
-                return ret.Cast<u8>();
-            }
-        } const* etc1_tile = reinterpret_cast<const ETC1Tile*>(source_ptr);
-
-        alpha >>= 4 * ((x & 3) * 4 + (y & 3));
-        return Math::MakeVec(etc1_tile->GetRGB(x & 3, y & 3),
-                             disable_alpha ? (u8)255 : Color::Convert4To8(alpha & 0xF));
+        return Math::MakeVec(SampleETC1Subtile(subtile_data, x, y),
+                             disable_alpha ? (u8)255 : alpha);
     }
 
     default: