Pica: Initial support for multitexturing.

4 files changed, 115 insertions, 31 deletions

diff --git a/src/citra_qt/debugger/graphics_cmdlists.cpp b/src/citra_qt/debugger/graphics_cmdlists.cpp
index 7f97cf143..bdd676470 100644
--- a/src/citra_qt/debugger/graphics_cmdlists.cpp
+++ b/src/citra_qt/debugger/graphics_cmdlists.cpp
@@ -223,9 +223,21 @@ void GPUCommandListModel::OnPicaTraceFinished(const Pica::DebugUtils::PicaTrace&
 
 void GPUCommandListWidget::OnCommandDoubleClicked(const QModelIndex& index) {
     const int command_id = list_widget->model()->data(index, GPUCommandListModel::CommandIdRole).toInt();
-    if (COMMAND_IN_RANGE(command_id, texture0)) {
-        auto info = Pica::DebugUtils::TextureInfo::FromPicaRegister(Pica::registers.texture0,
-                                                                    Pica::registers.texture0_format);
+    if (COMMAND_IN_RANGE(command_id, texture0) ||
+        COMMAND_IN_RANGE(command_id, texture1) ||
+        COMMAND_IN_RANGE(command_id, texture2)) {
+
+        unsigned index;
+        if (COMMAND_IN_RANGE(command_id, texture0)) {
+            index = 0;
+        } else if (COMMAND_IN_RANGE(command_id, texture1)) {
+            index = 1;
+        } else {
+            index = 2;
+        }
+        auto config = Pica::registers.GetTextures()[index].config;
+        auto format = Pica::registers.GetTextures()[index].format;
+        auto info = Pica::DebugUtils::TextureInfo::FromPicaRegister(config, format);
 
         // TODO: Instead, emit a signal here to be caught by the main window widget.
         auto main_window = static_cast<QMainWindow*>(parent());
@@ -237,10 +249,23 @@ void GPUCommandListWidget::SetCommandInfo(const QModelIndex& index) {
     QWidget* new_info_widget;
 
     const int command_id = list_widget->model()->data(index, GPUCommandListModel::CommandIdRole).toInt();
-    if (COMMAND_IN_RANGE(command_id, texture0)) {
-        u8* src = Memory::GetPointer(Pica::registers.texture0.GetPhysicalAddress());
-        auto info = Pica::DebugUtils::TextureInfo::FromPicaRegister(Pica::registers.texture0,
-                                                                    Pica::registers.texture0_format);
+    if (COMMAND_IN_RANGE(command_id, texture0) ||
+        COMMAND_IN_RANGE(command_id, texture1) ||
+        COMMAND_IN_RANGE(command_id, texture2)) {
+
+        unsigned index;
+        if (COMMAND_IN_RANGE(command_id, texture0)) {
+            index = 0;
+        } else if (COMMAND_IN_RANGE(command_id, texture1)) {
+            index = 1;
+        } else {
+            index = 2;
+        }
+        auto config = Pica::registers.GetTextures()[index].config;
+        auto format = Pica::registers.GetTextures()[index].format;
+
+        auto info = Pica::DebugUtils::TextureInfo::FromPicaRegister(config, format);
+        u8* src = Memory::GetPointer(config.GetPhysicalAddress());
         new_info_widget = new TextureInfoWidget(src, info);
     } else {
         new_info_widget = new QWidget;
diff --git a/src/video_core/pica.h b/src/video_core/pica.h
index 4c3791ad9..92a87c086 100644
--- a/src/video_core/pica.h
+++ b/src/video_core/pica.h
@@ -155,12 +155,34 @@ struct Regs {
         }
     }
 
-    BitField< 0, 1, u32> texturing_enable;
+    union {
+        BitField< 0, 1, u32> texture0_enable;
+        BitField< 1, 1, u32> texture1_enable;
+        BitField< 2, 1, u32> texture2_enable;
+    };
     TextureConfig texture0;
     INSERT_PADDING_WORDS(0x8);
     BitField<0, 4, TextureFormat> texture0_format;
-
-    INSERT_PADDING_WORDS(0x31);
+    INSERT_PADDING_WORDS(0x2);
+    TextureConfig texture1;
+    BitField<0, 4, TextureFormat> texture1_format;
+    INSERT_PADDING_WORDS(0x2);
+    TextureConfig texture2;
+    BitField<0, 4, TextureFormat> texture2_format;
+    INSERT_PADDING_WORDS(0x21);
+
+    struct FullTextureConfig {
+        const bool enabled;
+        const TextureConfig config;
+        const TextureFormat format;
+    };
+    const std::array<FullTextureConfig, 3> GetTextures() const {
+        return {{
+                   { static_cast<bool>(texture0_enable), texture0, texture0_format },
+                   { static_cast<bool>(texture1_enable), texture1, texture1_format },
+                   { static_cast<bool>(texture2_enable), texture2, texture2_format }
+               }};
+    }
 
     // 0xc0-0xff: Texture Combiner (akin to glTexEnv)
     struct TevStageConfig {
@@ -556,9 +578,13 @@ struct Regs {
         ADD_FIELD(viewport_depth_range);
         ADD_FIELD(viewport_depth_far_plane);
         ADD_FIELD(viewport_corner);
-        ADD_FIELD(texturing_enable);
+        ADD_FIELD(texture0_enable);
         ADD_FIELD(texture0);
         ADD_FIELD(texture0_format);
+        ADD_FIELD(texture1);
+        ADD_FIELD(texture1_format);
+        ADD_FIELD(texture2);
+        ADD_FIELD(texture2_format);
         ADD_FIELD(tev_stage0);
         ADD_FIELD(tev_stage1);
         ADD_FIELD(tev_stage2);
@@ -622,9 +648,13 @@ ASSERT_REG_POSITION(viewport_depth_far_plane, 0x4e);
 ASSERT_REG_POSITION(vs_output_attributes[0], 0x50);
 ASSERT_REG_POSITION(vs_output_attributes[1], 0x51);
 ASSERT_REG_POSITION(viewport_corner, 0x68);
-ASSERT_REG_POSITION(texturing_enable, 0x80);
+ASSERT_REG_POSITION(texture0_enable, 0x80);
 ASSERT_REG_POSITION(texture0, 0x81);
 ASSERT_REG_POSITION(texture0_format, 0x8e);
+ASSERT_REG_POSITION(texture1, 0x91);
+ASSERT_REG_POSITION(texture1_format, 0x96);
+ASSERT_REG_POSITION(texture2, 0x99);
+ASSERT_REG_POSITION(texture2_format, 0x9e);
 ASSERT_REG_POSITION(tev_stage0, 0xc0);
 ASSERT_REG_POSITION(tev_stage1, 0xc8);
 ASSERT_REG_POSITION(tev_stage2, 0xd0);
diff --git a/src/video_core/rasterizer.cpp b/src/video_core/rasterizer.cpp
index b7e04a560..2ff6d19a6 100644
--- a/src/video_core/rasterizer.cpp
+++ b/src/video_core/rasterizer.cpp
@@ -167,10 +167,22 @@ void ProcessTriangle(const VertexShader::OutputVertex& v0,
                 (u8)(GetInterpolatedAttribute(v0.color.a(), v1.color.a(), v2.color.a()).ToFloat32() * 255)
             };
 
-            Math::Vec4<u8> texture_color{};
-            float24 u = GetInterpolatedAttribute(v0.tc0.u(), v1.tc0.u(), v2.tc0.u());
-            float24 v = GetInterpolatedAttribute(v0.tc0.v(), v1.tc0.v(), v2.tc0.v());
-            if (registers.texturing_enable) {
+            Math::Vec2<float24> uv[3];
+            uv[0].u() = GetInterpolatedAttribute(v0.tc0.u(), v1.tc0.u(), v2.tc0.u());
+            uv[0].v() = GetInterpolatedAttribute(v0.tc0.v(), v1.tc0.v(), v2.tc0.v());
+            uv[1].u() = GetInterpolatedAttribute(v0.tc1.u(), v1.tc1.u(), v2.tc1.u());
+            uv[1].v() = GetInterpolatedAttribute(v0.tc1.v(), v1.tc1.v(), v2.tc1.v());
+            uv[2].u() = GetInterpolatedAttribute(v0.tc2.u(), v1.tc2.u(), v2.tc2.u());
+            uv[2].v() = GetInterpolatedAttribute(v0.tc2.v(), v1.tc2.v(), v2.tc2.v());
+
+            Math::Vec4<u8> texture_color[3]{};
+            for (int i = 0; i < 3; ++i) {
+                auto texture = registers.GetTextures()[i];
+                if (!texture.enabled)
+                    continue;
+
+                _dbg_assert_(GPU, 0 != texture.config.address);
+
                 // Images are split into 8x8 tiles. Each tile is composed of four 4x4 subtiles each
                 // of which is composed of four 2x2 subtiles each of which is composed of four texels.
                 // Each structure is embedded into the next-bigger one in a diagonal pattern, e.g.
@@ -189,14 +201,11 @@ void ProcessTriangle(const VertexShader::OutputVertex& v0,
                 // 02 03 06 07 18 19 22 23
                 // 00 01 04 05 16 17 20 21
 
-                // TODO: This is currently hardcoded for RGB8
-                u32* texture_data = (u32*)Memory::GetPointer(registers.texture0.GetPhysicalAddress());
-
                 // TODO(neobrain): Not sure if this swizzling pattern is used for all textures.
                 // To be flexible in case different but similar patterns are used, we keep this
                 // somewhat inefficient code around for now.
-                int s = (int)(u * float24::FromFloat32(static_cast<float>(registers.texture0.width))).ToFloat32();
-                int t = (int)(v * float24::FromFloat32(static_cast<float>(registers.texture0.height))).ToFloat32();
+                int s = (int)(uv[i].u() * float24::FromFloat32(static_cast<float>(texture.config.width))).ToFloat32();
+                int t = (int)(uv[i].v() * float24::FromFloat32(static_cast<float>(texture.config.height))).ToFloat32();
                 int texel_index_within_tile = 0;
                 for (int block_size_index = 0; block_size_index < 3; ++block_size_index) {
                     int sub_tile_width = 1 << block_size_index;
@@ -213,14 +222,17 @@ void ProcessTriangle(const VertexShader::OutputVertex& v0,
                 int coarse_s = (s / block_width) * block_width;
                 int coarse_t = (t / block_height) * block_height;
 
-                const int row_stride = registers.texture0.width * 3;
+                // TODO: This is currently hardcoded for RGB8
+                u32* texture_data = (u32*)Memory::GetPointer(texture.config.GetPhysicalAddress());
+
+                const int row_stride = texture.config.width * 3;
                 u8* source_ptr = (u8*)texture_data + coarse_s * block_height * 3 + coarse_t * row_stride + texel_index_within_tile * 3;
-                texture_color.r() = source_ptr[2];
-                texture_color.g() = source_ptr[1];
-                texture_color.b() = source_ptr[0];
-                texture_color.a() = 0xFF;
+                texture_color[i].r() = source_ptr[2];
+                texture_color[i].g() = source_ptr[1];
+                texture_color[i].b() = source_ptr[0];
+                texture_color[i].a() = 0xFF;
 
-                DebugUtils::DumpTexture(registers.texture0, (u8*)texture_data);
+                DebugUtils::DumpTexture(texture.config, (u8*)texture_data);
             }
 
             // Texture environment - consists of 6 stages of color and alpha combining.
@@ -243,7 +255,13 @@ void ProcessTriangle(const VertexShader::OutputVertex& v0,
                         return primary_color.rgb();
 
                     case Source::Texture0:
-                        return texture_color.rgb();
+                        return texture_color[0].rgb();
+
+                    case Source::Texture1:
+                        return texture_color[1].rgb();
+
+                    case Source::Texture2:
+                        return texture_color[2].rgb();
 
                     case Source::Constant:
                         return {tev_stage.const_r, tev_stage.const_g, tev_stage.const_b};
@@ -263,7 +281,13 @@ void ProcessTriangle(const VertexShader::OutputVertex& v0,
                         return primary_color.a();
 
                     case Source::Texture0:
-                        return texture_color.a();
+                        return texture_color[0].a();
+
+                    case Source::Texture1:
+                        return texture_color[1].a();
+
+                    case Source::Texture2:
+                        return texture_color[2].a();
 
                     case Source::Constant:
                         return tev_stage.const_a;
diff --git a/src/video_core/vertex_shader.h b/src/video_core/vertex_shader.h
index bfb6fb6e3..c1292fc2d 100644
--- a/src/video_core/vertex_shader.h
+++ b/src/video_core/vertex_shader.h
@@ -27,15 +27,18 @@ struct OutputVertex {
     Math::Vec4<float24> dummy; // quaternions (not implemented, yet)
     Math::Vec4<float24> color;
     Math::Vec2<float24> tc0;
+    Math::Vec2<float24> tc1;
+    float24 pad[6];
+    Math::Vec2<float24> tc2;
 
     // Padding for optimal alignment
-    float24 pad[14];
+    float24 pad2[4];
 
     // Attributes used to store intermediate results
 
     // position after perspective divide
     Math::Vec3<float24> screenpos;
-    float24 pad2;
+    float24 pad3;
 
     // Linear interpolation
     // factor: 0=this, 1=vtx
@@ -44,6 +47,8 @@ struct OutputVertex {
 
         // TODO: Should perform perspective correct interpolation here...
         tc0 = tc0 * factor + vtx.tc0 * (float24::FromFloat32(1) - factor);
+        tc1 = tc1 * factor + vtx.tc1 * (float24::FromFloat32(1) - factor);
+        tc2 = tc2 * factor + vtx.tc2 * (float24::FromFloat32(1) - factor);
 
         screenpos = screenpos * factor + vtx.screenpos * (float24::FromFloat32(1) - factor);