src/plugin/graphic/src/resource/Texture.hpp (3)

47-64: Consider combining identical switch cases.

The RGBA8UnormSrgb and RGBA8Unorm cases have identical handling. Use fallthrough to reduce duplication.

♻️ Proposed fix

         switch (data->format)
         {
         case wgpu::TextureFormat::RGBA8UnormSrgb:
-            pixel.r = mapped[i * 4 + 0];
-            pixel.g = mapped[i * 4 + 1];
-            pixel.b = mapped[i * 4 + 2];
-            pixel.a = mapped[i * 4 + 3];
-            break;
-        case wgpu::TextureFormat::RGBA8Unorm:
+        case wgpu::TextureFormat::RGBA8Unorm: // fallthrough
             pixel.r = mapped[i * 4 + 0];
             pixel.g = mapped[i * 4 + 1];
             pixel.b = mapped[i * 4 + 2];
             pixel.a = mapped[i * 4 + 3];
             break;

---

65-82: Endianness assumption in bit_cast.

The std::bit_cast from std::array<uint8_t, N> to numeric types assumes little-endian byte order. This is typically fine for WebGPU targets but worth noting in a comment for maintainability.

---

208-216: Consider adding Depth16Unorm to depth format check.

The depth format detection is missing wgpu::TextureFormat::Depth16Unorm. If this format is used, the copy would fail or produce incorrect results.

♻️ Proposed fix

         if (_webgpuTexture.getFormat() == wgpu::TextureFormat::Depth24Plus ||
             _webgpuTexture.getFormat() == wgpu::TextureFormat::Depth24PlusStencil8 ||
             _webgpuTexture.getFormat() == wgpu::TextureFormat::Depth32Float ||
-            _webgpuTexture.getFormat() == wgpu::TextureFormat::Depth32FloatStencil8)
+            _webgpuTexture.getFormat() == wgpu::TextureFormat::Depth32FloatStencil8 ||
+            _webgpuTexture.getFormat() == wgpu::TextureFormat::Depth16Unorm)

src/plugin/default-pipeline/src/utils/DefaultRenderPass.hpp (1)

224-230: Stale comment: Update to reflect actual buffer layout.

The comment says "mat4 modelMatrix (64 bytes) + 3 * vec4 normalMatrix columns (48 bytes) = 112 bytes" but the actual layout is now two mat4 matrices (128 bytes total), matching the sizeof(glm::mat4) + sizeof(glm::mat4) on line 228.

📝 Suggested fix

-        // Model buffer contains: mat4 modelMatrix (64 bytes) + 3 * vec4 normalMatrix columns (48 bytes) = 112 bytes
+        // Model buffer contains: mat4 modelMatrix (64 bytes) + mat4 normalMatrix (64 bytes) = 128 bytes
         auto modelLayout = Graphic::Utils::BindGroupLayout("ModelLayout")

src/plugin/default-pipeline/tests/shader/GBufferTest.cpp (1)

29-47: Consider adding basic assertions to validate extracted textures.

The extracted textures (normalImage, albedoImage, depthImage) are retrieved but never validated. Even for a smoke test, consider adding minimal assertions to ensure the textures have expected dimensions or are non-empty.

💡 Example assertions

void ExtractTextures(Engine::Core &core)
{
    auto &context = core.GetResource<Graphic::Resource::Context>();
    auto &textures = core.GetResource<Graphic::Resource::TextureContainer>();

    auto &normalTexture = textures.Get(DefaultPipeline::Resource::GBUFFER_PASS_OUTPUT_NORMAL_ID);
    auto normalImage = normalTexture.RetrieveImage(context);
+   EXPECT_GT(normalImage.width, 0);
+   EXPECT_GT(normalImage.height, 0);

    auto &albedoTexture = textures.Get(DefaultPipeline::Resource::GBUFFER_PASS_OUTPUT_ALBEDO_ID);
    auto albedoImage = albedoTexture.RetrieveImage(context);
+   EXPECT_GT(albedoImage.width, 0);

    auto &depthTexture = textures.Get(DefaultPipeline::Resource::GBUFFER_PASS_OUTPUT_DEPTH_ID);
    auto depthImage = depthTexture.RetrieveImage(context);
+   EXPECT_GT(depthImage.width, 0);

    // Uncomment to save the images
    // ...
}

src/plugin/default-pipeline/src/resource/pass/GBuffer.hpp (1)

158-164: Stale comment: same issue as in DefaultRenderPass.hpp.

The comment references the old layout "3 * vec4 normalMatrix columns (48 bytes) = 112 bytes" but the actual buffer is now two mat4 matrices (128 bytes).

📝 Suggested fix

-        // Model buffer contains: mat4 modelMatrix (64 bytes) + 3 * vec4 normalMatrix columns (48 bytes) = 112 bytes
+        // Model buffer contains: mat4 modelMatrix (64 bytes) + mat4 normalMatrix (64 bytes) = 128 bytes
         auto modelLayout = Graphic::Utils::BindGroupLayout("Model").addEntry(

src/plugin/default-pipeline/src/system/initialization/Create3DGraph.cpp (2)

42-47: Stale TODO comment.

The TODO says "resize on window resize" but resize handling is already implemented via the OnResize callbacks below.

📝 Suggested fix

 static void CreateGBufferTextures(Engine::Core &core)
 {
-    // TODO: resize on window resize
     const auto &context = core.GetResource<Graphic::Resource::Context>();

---

50-60: Consider extracting hardcoded texture size to a constant.

The initial texture size {800, 800} is hardcoded in three places. Consider extracting to a named constant for maintainability and to make it clear this is an initial/default size before the first resize event.

📝 Example

static constexpr glm::uvec2 DEFAULT_GBUFFER_SIZE{800, 800};

// Then use:
auto descriptor = CreateGBufferPassOutputNormalTextureDescriptor(DEFAULT_GBUFFER_SIZE);

Also applies to: 62-64, 74-76

src/plugin/default-pipeline/tests/shader/GBufferTest.cpp (1)

61-66: Make the smoke test fail if the GBuffer pass is missing.
Right now the test succeeds even when the pass isn’t registered, so regressions won’t be caught.

💡 Suggested tweak

     core.RegisterSystem<RenderingPipeline::CommandCreation>([](Engine::Core &core) {
         auto &renderPassContainer = core.GetResource<Graphic::Resource::RenderGraphContainer>();
-        if (renderPassContainer.Contains(DefaultPipeline::Resource::GBUFFER_PASS_ID))
-        {
-            renderPassContainer.Get(DefaultPipeline::Resource::GBUFFER_PASS_ID).Execute(core);
-        }
+        ASSERT_TRUE(renderPassContainer.Contains(DefaultPipeline::Resource::GBUFFER_PASS_ID))
+            << "GBuffer pass not registered";
+        renderPassContainer.Get(DefaultPipeline::Resource::GBUFFER_PASS_ID).Execute(core);
     });

src/plugin/default-pipeline/src/plugin/PluginDefaultPipeline.cpp (1)

45-47: Registration order looks correct.

Based on the current implementation in Create3DGraph.cpp (line 183), Create3DGraph calls renderPassContainer.Add(GBUFFER_PASS_ID, ...) rather than SetDefault(), so it adds a named render graph without overwriting the default set by CreateDefaultRenderPipeline. This is the expected behavior for having both a default render pipeline and a GBuffer-based 3D graph available.

src/plugin/default-pipeline/src/system/initialization/Create3DGraph.cpp (2)

54-64: LGTM! Depth texture now includes TextureBinding usage.

The depth texture descriptor correctly includes TextureBinding | RenderAttachment | CopySrc | CopyDst, enabling sampling for deferred shading passes.

---

75-119: LGTM! GBuffer texture creation and resize handling looks correct.

The implementation now:

• Uses window.GetSize() for initial texture dimensions instead of hardcoded values
• Correctly uses the matching texture IDs in each resize callback (NORMAL, ALBEDO, DEPTH respectively)

src/plugin/default-pipeline/src/resource/pass/GBuffer.hpp (2)

33-93: LGTM! GBuffer shader is well-structured.

The shader correctly:

• Uses mat4x4<f32> for the normal matrix
• Transforms normals with vec4(normal, 0.0) to exclude translation
• Outputs normalized world-space normals and albedo to the G-buffer

---

177-257: LGTM! CreateShader is correctly configured.

The shader descriptor properly sets up:

• Camera, model (128 bytes), and material bind group layouts
• Vertex layout with position/normal/uv attributes
• RGBA16Float for normal output, BGRA8Unorm for albedo
• Depth32Float with depth write enabled

src/plugin/graphic/src/system/preparation/PrepareEndRenderTexture.cpp (1)

108-120: Consider using designated initializers for consistency.

The existing EnsurePlaceholderEndRenderTexture function (line 16) uses designated initializers for wgpu::Extent3D:

textureDesc.size = {.width = 1920u, .height = 1080u, .depthOrArrayLayers = 1};

For consistency and improved readability, consider using the same style here.

♻️ Suggested change

-    textureDesc.size = {requiredSize.x, requiredSize.y, 1};
+    textureDesc.size = {.width = requiredSize.x, .height = requiredSize.y, .depthOrArrayLayers = 1};