GPU: Implemented more depth buffer formats.

This fixes the horizontal lines in Picross E, Cubic Ninja, Cave Story 3D and possibly others

src/core/hw/gpu.cpp

@@ -81,9 +81,9 @@ inline void Write(u32 addr, const T data) {
             if (config.fill_24bit) {
                 // fill with 24-bit values
                 for (u8* ptr = start; ptr < end; ptr += 3) {
-                    ptr[0] = config.value_24bit_b;
+                    ptr[0] = config.value_24bit_r;
                     ptr[1] = config.value_24bit_g;
-                    ptr[2] = config.value_24bit_r;
+                    ptr[2] = config.value_24bit_b;
                 }
             } else if (config.fill_32bit) {
                 // fill with 32-bit values

src/core/hw/gpu.h

@@ -100,10 +100,10 @@ struct Regs {
             // Set to 1 upon completion.
             BitField<0, 1, u32> finished;
 
-            // 0: fill with 16- or 32-bit wide values; 1: fill with 24-bit wide values
+            // If both of these bits are unset, then it will fill the memory with a 16 bit value
+            // 1: fill with 24-bit wide values
             BitField<8, 1, u32> fill_24bit;
-
-            // 0: fill with 16-bit wide values; 1: fill with 32-bit wide values
+            // 1: fill with 32-bit wide values
             BitField<9, 1, u32> fill_32bit;
         };
 

src/video_core/color.h

@@ -100,6 +100,33 @@ inline const Math::Vec4<u8> DecodeRGBA4(const u8* bytes) {
         Convert4To8((pixel >> 4) & 0xF), Convert4To8(pixel & 0xF) };
 }
 
+/**
+ * Decode a depth value stored in D16 format
+ * @param bytes Pointer to encoded source value
+ * @return Depth value as an u32
+ */
+inline const u32 DecodeD16(const u8* bytes) {
+    return *reinterpret_cast<const u16_le*>(bytes);
+}
+
+/**
+ * Decode a depth value stored in D24 format
+ * @param bytes Pointer to encoded source value
+ * @return Depth value as an u32
+ */
+inline const u32 DecodeD24(const u8* bytes) {
+    return (bytes[2] << 16) | (bytes[1] << 8) | bytes[0];
+}
+
+/**
+ * Decode a depth value and a stencil value stored in D24S8 format
+ * @param bytes Pointer to encoded source values
+ * @return Resulting values stored as a Math::Vec2
+ */
+inline const Math::Vec2<u32> DecodeD24S8(const u8* bytes) {
+    return { (bytes[2] << 16) | (bytes[1] << 8) | bytes[0], bytes[3] };
+}
+
 /**
  * Encode a color as RGBA8 format
  * @param color Source color to encode
@@ -153,4 +180,34 @@ inline void EncodeRGBA4(const Math::Vec4<u8>& color, u8* bytes) {
         (Convert8To4(color.g()) << 8) | (Convert8To4(color.b()) << 4) | Convert8To4(color.a());
 }
 
+/**
+ * Encode a depth value as D16 format
+ * @param value Source depth value to encode
+ * @param bytes Pointer where to store the encoded value
+ */
+inline void EncodeD16(u32 value, u8* bytes) {
+    *reinterpret_cast<u16_le*>(bytes) = value & 0xFFFF;
+}
+
+/**
+ * Encode a depth value as D24 format
+ * @param value Source depth value to encode
+ * @param bytes Pointer where to store the encoded value
+ */
+inline void EncodeD24(u32 value, u8* bytes) {
+    bytes[0] = value & 0xFF;
+    bytes[1] = (value >> 8) & 0xFF;
+    bytes[2] = (value >> 16) & 0xFF;
+}
+
+/**
+ * Encode depth and stencil values as D24S8 format
+ * @param depth Source depth values to encode
+ * @param stencil Source stencil value to encode
+ * @param bytes Pointer where to store the encoded value
+ */
+inline void EncodeD24S8(u32 depth, u8 stencil, u8* bytes) {
+    *reinterpret_cast<u32_le*>(bytes) = (stencil << 24) | depth;
+}
+
 } // namespace

src/video_core/pica.h

@@ -418,6 +418,13 @@ struct Regs {
             RGBA4    = 4,
         };
 
+        enum DepthFormat : u32 {
+            D16     = 0,
+        
+            D24     = 2,
+            D24S8   = 3
+        };
+
         INSERT_PADDING_WORDS(0x6);
 
         u32 depth_format;

src/video_core/rasterizer.cpp

@@ -100,23 +100,65 @@ static u32 GetDepth(int x, int y) {
     y = (registers.framebuffer.height - y);
     
     const u32 coarse_y = y & ~7;
-    u32 stride = registers.framebuffer.width * 2;
 
-    // Assuming 16-bit depth buffer format until actual format handling is implemented
-    return *(u16*)(depth_buffer + VideoCore::GetMortonOffset(x, y, 2) + coarse_y * stride);
+    switch (registers.framebuffer.depth_format) {
+        case registers.framebuffer.D16:
+        {
+            u32 stride = registers.framebuffer.width * 2;
+            return Color::DecodeD16(depth_buffer + VideoCore::GetMortonOffset(x, y, 2) + coarse_y * stride);
+        }
+        case registers.framebuffer.D24:
+        {
+            u32 stride = registers.framebuffer.width * 3;
+            u8* address = depth_buffer + VideoCore::GetMortonOffset(x, y, 3) + coarse_y * stride;
+            return Color::DecodeD24(address);
+        }
+        case registers.framebuffer.D24S8:
+        {
+            u32 stride = registers.framebuffer.width * 4;
+            return Color::DecodeD24S8(depth_buffer + VideoCore::GetMortonOffset(x, y, 4) + coarse_y * stride).x;
+        }
+        default:
+            LOG_CRITICAL(HW_GPU, "Unimplemented depth format %u", registers.framebuffer.depth_format);
+            UNIMPLEMENTED();
+            return 0;
+    }
 }
 
-static void SetDepth(int x, int y, u16 value) {
+static void SetDepth(int x, int y, u32 value) {
     const PAddr addr = registers.framebuffer.GetDepthBufferPhysicalAddress();
     u8* depth_buffer = Memory::GetPointer(PAddrToVAddr(addr));
 
     y = (registers.framebuffer.height - y);
 
     const u32 coarse_y = y & ~7;
-    u32 stride = registers.framebuffer.width * 2;
 
-    // Assuming 16-bit depth buffer format until actual format handling is implemented
-    *(u16*)(depth_buffer + VideoCore::GetMortonOffset(x, y, 2) + coarse_y * stride) = value;
+    switch (registers.framebuffer.depth_format) {
+        case registers.framebuffer.D16:
+        {
+            u32 stride = registers.framebuffer.width * 2;
+            Color::EncodeD16(value, depth_buffer + VideoCore::GetMortonOffset(x, y, 2) + coarse_y * stride);
+            break;
+        }
+        case registers.framebuffer.D24:
+        {
+            u32 stride = registers.framebuffer.width * 3;
+            u8* address = depth_buffer + VideoCore::GetMortonOffset(x, y, 3) + coarse_y * stride;
+            Color::EncodeD24(value, address);
+            break;
+        }
+        case registers.framebuffer.D24S8:
+        {
+            u32 stride = registers.framebuffer.width * 4;
+            // TODO(Subv): Implement the stencil buffer
+            Color::EncodeD24S8(value, 0, depth_buffer + VideoCore::GetMortonOffset(x, y, 4) + coarse_y * stride);
+            break;
+        }
+        default:
+            LOG_CRITICAL(HW_GPU, "Unimplemented depth format %u", registers.framebuffer.depth_format);
+            UNIMPLEMENTED();
+            break;
+    }
 }
 
 // NOTE: Assuming that rasterizer coordinates are 12.4 fixed-point values
@@ -595,7 +637,7 @@ static void ProcessTriangleInternal(const VertexShader::OutputVertex& v0,
                 u16 z = (u16)((v0.screenpos[2].ToFloat32() * w0 +
                             v1.screenpos[2].ToFloat32() * w1 +
                             v2.screenpos[2].ToFloat32() * w2) * 65535.f / wsum);
-                u16 ref_z = GetDepth(x >> 4, y >> 4);
+                u32 ref_z = GetDepth(x >> 4, y >> 4);
 
                 bool pass = false;