diff --git a/Runtime/ConsoleVariables/FileStoreManager.cpp b/Runtime/ConsoleVariables/FileStoreManager.cpp
index bbc0f0852..722f0d0a5 100644
--- a/Runtime/ConsoleVariables/FileStoreManager.cpp
+++ b/Runtime/ConsoleVariables/FileStoreManager.cpp
@@ -71,6 +71,7 @@ FileStoreManager::FileStoreManager(std::string_view org, std::string_view domain
 #endif
   } else {
     m_storeRoot = std::string(prefPath);
+    SDL_free(prefPath);
   }
   g_instance = this;
 }
diff --git a/Runtime/Graphics/CCubeMaterial.cpp b/Runtime/Graphics/CCubeMaterial.cpp
index 549846b50..f7d25faa7 100644
--- a/Runtime/Graphics/CCubeMaterial.cpp
+++ b/Runtime/Graphics/CCubeMaterial.cpp
@@ -220,7 +220,7 @@ void CCubeMaterial::SetCurrent(const CModelFlags& flags, const CCubeSurface& sur
   }
 
   // SetNumIndStages(numIndStages);
-  // SetNumTevStages(finalTevCount);
+  aurora::gfx::disable_tev_stage(static_cast<ERglTevStage>(finalTevCount));
   // SetNumTexGens(tcgCount);
   // SetNumColorChans(finalNumColorChans);
 }
diff --git a/Runtime/Graphics/CCubeModel.cpp b/Runtime/Graphics/CCubeModel.cpp
index e67d742de..991a59860 100644
--- a/Runtime/Graphics/CCubeModel.cpp
+++ b/Runtime/Graphics/CCubeModel.cpp
@@ -25,8 +25,8 @@ static zeus::CVector3f sPlayerPosition;
 
 CCubeModel::CCubeModel(std::vector<CCubeSurface>* surfaces, std::vector<TCachedToken<CTexture>>* textures,
                        u8* materialData, std::vector<zeus::CVector3f>* positions, std::vector<zeus::CColor>* colors,
-                       std::vector<zeus::CVector3f>* normals, std::vector<zeus::CVector2f>* texCoords,
-                       std::vector<zeus::CVector2f>* packedTexCoords, const zeus::CAABox& aabb, u8 flags, bool b1,
+                       std::vector<zeus::CVector3f>* normals, std::vector<aurora::Vec2<float>>* texCoords,
+                       std::vector<aurora::Vec2<float>>* packedTexCoords, const zeus::CAABox& aabb, u8 flags, bool b1,
                        u32 idx)
 : x0_modelInstance(surfaces, materialData, positions, colors, normals, texCoords, packedTexCoords)
 , x1c_textures(textures)
@@ -256,12 +256,8 @@ void CCubeModel::EnableShadowMaps(const CTexture& shadowTex, const zeus::CTransf
 void CCubeModel::DisableShadowMaps() { sRenderModelShadow = false; }
 
 void CCubeModel::SetArraysCurrent() {
-  if (x0_modelInstance.GetVertexPointer() != nullptr) {
-    aurora::gfx::model::set_vertex_buffer(*x0_modelInstance.GetVertexPointer());
-  }
-  if (x0_modelInstance.GetNormalPointer() != nullptr) {
-    aurora::gfx::model::set_normal_buffer(*x0_modelInstance.GetNormalPointer());
-  }
+  aurora::gfx::model::set_vertex_buffer(x0_modelInstance.GetVertexPointer());
+  aurora::gfx::model::set_normal_buffer(x0_modelInstance.GetNormalPointer());
   SetStaticArraysCurrent();
 }
 
@@ -281,8 +277,8 @@ void CCubeModel::SetRenderModelBlack(bool v) {
 }
 
 void CCubeModel::SetSkinningArraysCurrent(TVectorRef positions, TVectorRef normals) {
-  aurora::gfx::model::set_vertex_buffer(*positions);
-  aurora::gfx::model::set_normal_buffer(*normals);
+  aurora::gfx::model::set_vertex_buffer(positions);
+  aurora::gfx::model::set_normal_buffer(normals);
   // colors unused
   SetStaticArraysCurrent();
 }
@@ -295,22 +291,20 @@ void CCubeModel::SetStaticArraysCurrent() {
     sUsingPackedLightmaps = false;
   }
   if (sUsingPackedLightmaps) {
-    aurora::gfx::model::set_tex0_tc_buffer(*packedTexCoords);
-  } else if (texCoords != nullptr) {
-    aurora::gfx::model::set_tex0_tc_buffer(*packedTexCoords);
-  }
-  if (texCoords != nullptr) {
-    aurora::gfx::model::set_tc_buffer(*texCoords);
+    aurora::gfx::model::set_tex0_tc_buffer(packedTexCoords);
+  } else {
+    aurora::gfx::model::set_tex0_tc_buffer(texCoords);
   }
+  aurora::gfx::model::set_tc_buffer(texCoords);
   CCubeMaterial::KillCachedViewDepState();
 }
 
 void CCubeModel::SetUsingPackedLightmaps(bool v) {
   sUsingPackedLightmaps = v;
   if (v) {
-    aurora::gfx::model::set_tex0_tc_buffer(*x0_modelInstance.GetPackedTCPointer());
+    aurora::gfx::model::set_tex0_tc_buffer(x0_modelInstance.GetPackedTCPointer());
   } else {
-    aurora::gfx::model::set_tex0_tc_buffer(*x0_modelInstance.GetTCPointer());
+    aurora::gfx::model::set_tex0_tc_buffer(x0_modelInstance.GetTCPointer());
   }
 }
 
diff --git a/Runtime/Graphics/CCubeModel.hpp b/Runtime/Graphics/CCubeModel.hpp
index f4b437e42..4b7a5148b 100644
--- a/Runtime/Graphics/CCubeModel.hpp
+++ b/Runtime/Graphics/CCubeModel.hpp
@@ -33,13 +33,13 @@ private:
     std::vector<zeus::CVector3f>* x8_positions;        // was void*
     std::vector<zeus::CVector3f>* xc_normals;          // was void*
     std::vector<zeus::CColor>* x10_colors;             // was void*
-    std::vector<zeus::CVector2f>* x14_texCoords;       // was void*
-    std::vector<zeus::CVector2f>* x18_packedTexCoords; // was void*
+    std::vector<aurora::Vec2<float>>* x14_texCoords;       // was void*
+    std::vector<aurora::Vec2<float>>* x18_packedTexCoords; // was void*
 
   public:
     ModelInstance(std::vector<CCubeSurface>* surfaces, u8* material, std::vector<zeus::CVector3f>* positions,
                   std::vector<zeus::CColor>* colors, std::vector<zeus::CVector3f>* normals,
-                  std::vector<zeus::CVector2f>* texCoords, std::vector<zeus::CVector2f>* packedTexCoords)
+                  std::vector<aurora::Vec2<float>>* texCoords, std::vector<aurora::Vec2<float>>* packedTexCoords)
     : x0_surfacePtrs(surfaces)
     , x4_materialData(material)
     , x8_positions(positions)
@@ -58,8 +58,8 @@ private:
     [[nodiscard]] TVectorRef GetVertexPointer() const { return x8_positions; }
     [[nodiscard]] TVectorRef GetNormalPointer() const { return xc_normals; }
     [[nodiscard]] std::vector<zeus::CColor>* GetColorPointer() const { return x10_colors; }
-    [[nodiscard]] std::vector<zeus::CVector2f>* GetTCPointer() const { return x14_texCoords; }
-    [[nodiscard]] std::vector<zeus::CVector2f>* GetPackedTCPointer() const { return x18_packedTexCoords; }
+    [[nodiscard]] std::vector<aurora::Vec2<float>>* GetTCPointer() const { return x14_texCoords; }
+    [[nodiscard]] std::vector<aurora::Vec2<float>>* GetPackedTCPointer() const { return x18_packedTexCoords; }
   };
 
   ModelInstance x0_modelInstance;
@@ -75,8 +75,8 @@ private:
 public:
   CCubeModel(std::vector<CCubeSurface>* surfaces, std::vector<TCachedToken<CTexture>>* textures, u8* materialData,
              std::vector<zeus::CVector3f>* positions, std::vector<zeus::CColor>* colors,
-             std::vector<zeus::CVector3f>* normals, std::vector<zeus::CVector2f>* texCoords,
-             std::vector<zeus::CVector2f>* packedTexCoords, const zeus::CAABox& aabb, u8 flags, bool b1, u32 idx);
+             std::vector<zeus::CVector3f>* normals, std::vector<aurora::Vec2<float>>* texCoords,
+             std::vector<aurora::Vec2<float>>* packedTexCoords, const zeus::CAABox& aabb, u8 flags, bool b1, u32 idx);
 
   CCubeMaterial GetMaterialByIndex(u32 idx);
   bool TryLockTextures();
diff --git a/Runtime/Graphics/CCubeRenderer.cpp b/Runtime/Graphics/CCubeRenderer.cpp
index 638d6c1f7..622a123fe 100644
--- a/Runtime/Graphics/CCubeRenderer.cpp
+++ b/Runtime/Graphics/CCubeRenderer.cpp
@@ -233,8 +233,8 @@ void CCubeRenderer::AddStaticGeometry(const std::vector<CMetroidModelInstance>*
                                          const_cast<std::vector<zeus::CVector3f>*>(inst.GetVertexPointer()),
                                          const_cast<std::vector<zeus::CColor>*>(inst.GetColorPointer()),
                                          const_cast<std::vector<zeus::CVector3f>*>(inst.GetNormalPointer()),
-                                         const_cast<std::vector<zeus::CVector2f>*>(inst.GetTCPointer()),
-                                         const_cast<std::vector<zeus::CVector2f>*>(inst.GetPackedTCPointer()),
+                                         const_cast<std::vector<aurora::Vec2<float>>*>(inst.GetTCPointer()),
+                                         const_cast<std::vector<aurora::Vec2<float>>*>(inst.GetPackedTCPointer()),
                                          inst.GetBoundingBox(), inst.GetFlags(), false, instIdx));
         ++instIdx;
       }
diff --git a/Runtime/Graphics/CMetroidModelInstance.hpp b/Runtime/Graphics/CMetroidModelInstance.hpp
index c8eb6a024..087c1aff9 100644
--- a/Runtime/Graphics/CMetroidModelInstance.hpp
+++ b/Runtime/Graphics/CMetroidModelInstance.hpp
@@ -22,8 +22,8 @@ class CMetroidModelInstance {
   std::vector<zeus::CVector3f> x60_positions;       // was void*
   std::vector<zeus::CVector3f> x64_normals;         // was void*
   std::vector<zeus::CColor> x68_colors;             // was void*
-  std::vector<zeus::CVector2f> x6c_texCoords;       // was void*
-  std::vector<zeus::CVector2f> x70_packedTexCoords; // was void*
+  std::vector<aurora::Vec2<float>> x6c_texCoords;       // was void*
+  std::vector<aurora::Vec2<float>> x70_packedTexCoords; // was void*
 
 public:
   CMetroidModelInstance() = default;
@@ -40,8 +40,8 @@ public:
   [[nodiscard]] TVectorRef GetVertexPointer() const { return &x60_positions; }
   [[nodiscard]] TVectorRef GetNormalPointer() const { return &x64_normals; }
   [[nodiscard]] const std::vector<zeus::CColor>* GetColorPointer() const { return &x68_colors; }
-  [[nodiscard]] const std::vector<zeus::CVector2f>* GetTCPointer() const { return &x6c_texCoords; }
-  [[nodiscard]] const std::vector<zeus::CVector2f>* GetPackedTCPointer() const { return &x70_packedTexCoords; }
+  [[nodiscard]] const std::vector<aurora::Vec2<float>>* GetTCPointer() const { return &x6c_texCoords; }
+  [[nodiscard]] const std::vector<aurora::Vec2<float>>* GetPackedTCPointer() const { return &x70_packedTexCoords; }
 };
 
 } // namespace metaforce
diff --git a/Runtime/Graphics/CModel.hpp b/Runtime/Graphics/CModel.hpp
index ea3cc0b74..97bb6d04a 100644
--- a/Runtime/Graphics/CModel.hpp
+++ b/Runtime/Graphics/CModel.hpp
@@ -87,8 +87,8 @@ private:
   std::vector<zeus::CVector3f> m_positions;
   std::vector<zeus::CVector3f> m_normals;
   std::vector<zeus::CColor> m_colors;
-  std::vector<zeus::CVector2f> m_floatUVs;
-  std::vector<zeus::CVector2f> m_shortUVs;
+  std::vector<aurora::Vec2<float>> m_floatUVs;
+  std::vector<aurora::Vec2<float>> m_shortUVs;
 
 public:
   CModel(std::unique_ptr<u8[]> in, u32 dataLen, IObjectStore* store);
diff --git a/aurora/include/aurora/common.hpp b/aurora/include/aurora/common.hpp
index 90306b0be..7a0305b48 100644
--- a/aurora/include/aurora/common.hpp
+++ b/aurora/include/aurora/common.hpp
@@ -7,7 +7,55 @@
 
 #include "Runtime/RetroTypes.hpp"
 
+#include <zeus/CColor.hpp>
+
 namespace aurora {
+template <typename T>
+struct Vec2 {
+  T x{};
+  T y{};
+
+  constexpr Vec2(T x, T y) : x(x), y(y) {}
+  constexpr Vec2(const zeus::CVector2f& vec) : x(vec.x()), y(vec.y()) {}
+};
+template <typename T>
+struct Vec3 {
+  T x{};
+  T y{};
+  T z{};
+
+  constexpr Vec3(T x, T y, T z) : x(x), y(y), z(z) {}
+  constexpr Vec3(const zeus::CVector3f& vec) : x(vec.x()), y(vec.y()), z(vec.z()) {}
+};
+template <typename T>
+struct Vec4 {
+  T x{};
+  T y{};
+  T z{};
+  T w{};
+
+  constexpr Vec4(T x, T y, T z, T w) : x(x), y(y), z(z), w(w) {}
+  constexpr Vec4(const zeus::CVector4f& vec) : x(vec.x()), y(vec.y()), z(vec.z()), w(vec.w()) {}
+  constexpr Vec4(const zeus::CColor& color) : x(color.r()), y(color.g()), z(color.b()), w(color.a()) {}
+};
+template <typename T>
+struct Mat4x4 {
+  Vec4<T> m0{};
+  Vec4<T> m1{};
+  Vec4<T> m2{};
+  Vec4<T> m3{};
+
+  constexpr Mat4x4(const Vec4<T>& m0, const Vec4<T>& m1, const Vec4<T>& m2, const Vec4<T>& m3)
+  : m0(m0), m1(m1), m2(m2), m3(m3) {}
+  constexpr Mat4x4(const zeus::CMatrix4f& m) : m0(m[0]), m1(m[1]), m2(m[2]), m3(m[3]) {}
+};
+constexpr Mat4x4<float> Mat4x4_Identity{
+    Vec4<float>{1.f, 0.f, 0.f, 0.f},
+    Vec4<float>{0.f, 1.f, 0.f, 0.f},
+    Vec4<float>{0.f, 0.f, 1.f, 0.f},
+    Vec4<float>{0.f, 0.f, 0.f, 1.f},
+};
+
 template <typename T>
 class ArrayRef {
 public:
diff --git a/aurora/include/aurora/model.hpp b/aurora/include/aurora/model.hpp
index 9d0c02f20..e3533b9c1 100644
--- a/aurora/include/aurora/model.hpp
+++ b/aurora/include/aurora/model.hpp
@@ -3,10 +3,10 @@
 #include "common.hpp"
 
 namespace aurora::gfx::model {
-void set_vertex_buffer(const std::vector<zeus::CVector3f>& data) noexcept;
-void set_normal_buffer(const std::vector<zeus::CVector3f>& norm) noexcept;
-void set_tex0_tc_buffer(const std::vector<zeus::CVector2f>& tcs) noexcept; // Tex coords for TEX0
-void set_tc_buffer(const std::vector<zeus::CVector2f>& tcs) noexcept; // Tex coords for the TEX1-7
+void set_vertex_buffer(const std::vector<zeus::CVector3f>* data) noexcept;
+void set_normal_buffer(const std::vector<zeus::CVector3f>* norm) noexcept;
+void set_tex0_tc_buffer(const std::vector<Vec2<float>>* tcs) noexcept; // Tex coords for TEX0
+void set_tc_buffer(const std::vector<Vec2<float>>* tcs) noexcept; // Tex coords for the TEX1-7
 
 void set_vtx_desc_compressed(u32 vtxDesc) noexcept;
 void queue_surface(const u8* dlStart, u32 dlSize) noexcept;
diff --git a/aurora/lib/gfx/common.hpp b/aurora/lib/gfx/common.hpp
index 9879b87e9..0e2b99175 100644
--- a/aurora/lib/gfx/common.hpp
+++ b/aurora/lib/gfx/common.hpp
@@ -95,52 +95,6 @@ private:
     m_capacity = size;
   }
 };
-
-template <typename T>
-struct Vec2 {
-  T x{};
-  T y{};
-
-  constexpr Vec2(T x, T y) : x(x), y(y) {}
-  constexpr Vec2(const zeus::CVector2f& vec) : x(vec.x()), y(vec.y()) {}
-};
-template <typename T>
-struct Vec3 {
-  T x{};
-  T y{};
-  T z{};
-
-  constexpr Vec3(T x, T y, T z) : x(x), y(y), z(z) {}
-  constexpr Vec3(const zeus::CVector3f& vec) : x(vec.x()), y(vec.y()), z(vec.z()) {}
-};
-template <typename T>
-struct Vec4 {
-  T x{};
-  T y{};
-  T z{};
-  T w{};
-
-  constexpr Vec4(T x, T y, T z, T w) : x(x), y(y), z(z), w(w) {}
-  constexpr Vec4(const zeus::CVector4f& vec) : x(vec.x()), y(vec.y()), z(vec.z()), w(vec.w()) {}
-  constexpr Vec4(const zeus::CColor& color) : x(color.r()), y(color.g()), z(color.b()), w(color.a()) {}
-};
-template <typename T>
-struct Mat4x4 {
-  Vec4<T> m0{};
-  Vec4<T> m1{};
-  Vec4<T> m2{};
-  Vec4<T> m3{};
-
-  constexpr Mat4x4(const Vec4<T>& m0, const Vec4<T>& m1, const Vec4<T>& m2, const Vec4<T>& m3)
-  : m0(m0), m1(m1), m2(m2), m3(m3) {}
-  constexpr Mat4x4(const zeus::CMatrix4f& m) : m0(m[0]), m1(m[1]), m2(m[2]), m3(m[3]) {}
-};
-constexpr Mat4x4<float> Mat4x4_Identity{
-    Vec4<float>{1.f, 0.f, 0.f, 0.f},
-    Vec4<float>{0.f, 1.f, 0.f, 0.f},
-    Vec4<float>{0.f, 0.f, 1.f, 0.f},
-    Vec4<float>{0.f, 0.f, 0.f, 1.f},
-};
 } // namespace aurora
 
 namespace aurora::gfx {
diff --git a/aurora/lib/gfx/gx.cpp b/aurora/lib/gfx/gx.cpp
index aeb2c42a5..64dbd09f4 100644
--- a/aurora/lib/gfx/gx.cpp
+++ b/aurora/lib/gfx/gx.cpp
@@ -68,7 +68,7 @@ constexpr zeus::CMatrix4f DepthCorrect{
     0.f, 0.f, 0.f, 1.f,
     // clang-format on
 };
-void update_projection(const zeus::CMatrix4f& proj) noexcept { gx::g_proj = DepthCorrect * proj; }
+void update_projection(const zeus::CMatrix4f& proj) noexcept { gx::g_proj = /*DepthCorrect **/ proj; }
 void update_fog_state(const metaforce::CFogState& state) noexcept { gx::g_fogState = state; }
 
 void disable_tev_stage(metaforce::ERglTevStage stage) noexcept { gx::g_tevStages[static_cast<size_t>(stage)].reset(); }
@@ -295,11 +295,11 @@ wgpu::RenderPipeline build_pipeline(const PipelineConfig& config, const ShaderIn
       .targetCount = colorTargets.size(),
       .targets = colorTargets.data(),
   };
-  const auto layouts = build_bind_group_layouts(info, config.shaderConfig);
+  auto layouts = build_bind_group_layouts(info, config.shaderConfig);
   const std::array bindGroupLayouts{
-      layouts.uniformLayout,
-      layouts.samplerLayout,
-      layouts.textureLayout,
+      std::move(layouts.uniformLayout),
+      std::move(layouts.samplerLayout),
+      std::move(layouts.textureLayout),
   };
   const auto pipelineLayoutDescriptor = wgpu::PipelineLayoutDescriptor{
       .label = "GX Pipeline Layout",
@@ -368,6 +368,12 @@ Range build_uniform(const ShaderInfo& info) noexcept {
     const auto xf = get_combined_matrix();
     uniBuf.append(&xf, 64);
   }
+  for (int i = 0; i < info.usesTevReg.size(); ++i) {
+    if (!info.usesTevReg.test(i)) {
+      continue;
+    }
+    uniBuf.append(&g_colorRegs[i], 16);
+  }
   for (int i = 0; i < info.sampledColorChannels.size(); ++i) {
     if (!info.sampledColorChannels.test(i)) {
       continue;
diff --git a/aurora/lib/gfx/gx.hpp b/aurora/lib/gfx/gx.hpp
index 8aad4caf5..ccc75d06e 100644
--- a/aurora/lib/gfx/gx.hpp
+++ b/aurora/lib/gfx/gx.hpp
@@ -108,6 +108,7 @@ struct ShaderInfo {
   std::bitset<maxTextures> sampledTextures;
   std::bitset<4> sampledKcolors;
   std::bitset<2> sampledColorChannels;
+  std::bitset<3> usesTevReg;
   u32 uniformSize = 0;
   bool usesVtxColor : 1 = false;
   bool usesNormal : 1 = false;
@@ -143,21 +144,29 @@ struct DlVert {
 
 namespace aurora {
 template <>
+inline void xxh3_update(XXH3_state_t& state, const metaforce::CTevCombiners::CTevOp& input) {
+  XXH3_64bits_update(&state, &input.x0_clamp, sizeof(bool));
+  XXH3_64bits_update(&state, &input.x4_op, sizeof(metaforce::CTevCombiners::CTevOp::x4_op));
+  XXH3_64bits_update(&state, &input.x8_bias, sizeof(metaforce::CTevCombiners::CTevOp::x8_bias));
+  XXH3_64bits_update(&state, &input.xc_scale, sizeof(metaforce::CTevCombiners::CTevOp::xc_scale));
+  XXH3_64bits_update(&state, &input.x10_regId, sizeof(metaforce::CTevCombiners::CTevOp::x10_regId));
+}
+template <>
 inline void xxh3_update(XXH3_state_t& state, const gfx::gx::STevStage& input) {
-  XXH3_64bits_update(&state, &input.colorPass, sizeof(metaforce::CTevCombiners::ColorPass));
-  XXH3_64bits_update(&state, &input.alphaPass, sizeof(metaforce::CTevCombiners::AlphaPass));
-  XXH3_64bits_update(&state, &input.colorOp, sizeof(metaforce::CTevCombiners::CTevOp));
-  XXH3_64bits_update(&state, &input.alphaOp, sizeof(metaforce::CTevCombiners::CTevOp));
-  XXH3_64bits_update(&state, &input.kcSel, sizeof(GX::TevKColorSel));
-  XXH3_64bits_update(&state, &input.kaSel, sizeof(GX::TevKAlphaSel));
-  XXH3_64bits_update(&state, &input.texCoordId, sizeof(GX::TexCoordID));
-  XXH3_64bits_update(&state, &input.texMapId, sizeof(GX::TexMapID));
-  XXH3_64bits_update(&state, &input.channelId, sizeof(GX::ChannelID));
+  XXH3_64bits_update(&state, &input.colorPass, sizeof(gfx::gx::STevStage::colorPass));
+  XXH3_64bits_update(&state, &input.alphaPass, sizeof(gfx::gx::STevStage::alphaPass));
+  xxh3_update(state, input.colorOp);
+  xxh3_update(state, input.alphaOp);
+  XXH3_64bits_update(&state, &input.kcSel, sizeof(gfx::gx::STevStage::kcSel));
+  XXH3_64bits_update(&state, &input.kaSel, sizeof(gfx::gx::STevStage::kaSel));
+  XXH3_64bits_update(&state, &input.texCoordId, sizeof(gfx::gx::STevStage::texCoordId));
+  XXH3_64bits_update(&state, &input.texMapId, sizeof(gfx::gx::STevStage::texMapId));
+  XXH3_64bits_update(&state, &input.channelId, sizeof(gfx::gx::STevStage::channelId));
 }
 template <>
 inline XXH64_hash_t xxh3_hash(const gfx::gx::ShaderConfig& input, XXH64_hash_t seed) {
   XXH3_state_t state;
-  XXH3_INITSTATE(&state);
+  memset(&state, 0, sizeof(XXH3_state_t));
   XXH3_64bits_reset_withSeed(&state, seed);
   for (const auto& item : input.tevStages) {
     if (!item) {
@@ -165,7 +174,9 @@ inline XXH64_hash_t xxh3_hash(const gfx::gx::ShaderConfig& input, XXH64_hash_t s
     }
     xxh3_update(state, *item);
   }
-  XXH3_64bits_update(&state, input.channelMatSrcs.data(), input.channelMatSrcs.size() * sizeof(GX::ColorSrc));
+  for (const auto& item : input.channelMatSrcs) {
+    XXH3_64bits_update(&state, &item, sizeof(GX::ColorSrc));
+  }
   XXH3_64bits_update(&state, &input.alphaDiscard, sizeof(bool));
   XXH3_64bits_update(&state, &input.denormalizedVertexAttributes, sizeof(bool));
   XXH3_64bits_update(&state, &input.denormalizedHasNrm, sizeof(bool));
diff --git a/aurora/lib/gfx/gx_shader.cpp b/aurora/lib/gfx/gx_shader.cpp
index bb7789383..d447af852 100644
--- a/aurora/lib/gfx/gx_shader.cpp
+++ b/aurora/lib/gfx/gx_shader.cpp
@@ -19,13 +19,23 @@ static std::string color_arg_reg(GX::TevColorArg arg, size_t stageIdx, const STe
   case GX::CC_APREV:
     return "prev.a";
   case GX::CC_C0:
+    info.usesTevReg.set(0);
+    return "tevreg0.rgb";
   case GX::CC_A0:
+    info.usesTevReg.set(0);
+    return "tevreg0.a";
   case GX::CC_C1:
+    info.usesTevReg.set(1);
+    return "tevreg1.rgb";
   case GX::CC_A1:
+    info.usesTevReg.set(1);
+    return "tevreg1.a";
   case GX::CC_C2:
+    info.usesTevReg.set(2);
+    return "tevreg2.rgb";
   case GX::CC_A2:
-    Log.report(logvisor::Fatal, FMT_STRING("TODO {}"), arg);
-    unreachable();
+    info.usesTevReg.set(2);
+    return "tevreg2.a";
   case GX::CC_TEXC: {
     if (stage.texMapId == GX::TEXMAP_NULL) {
       Log.report(logvisor::Fatal, FMT_STRING("unmapped texture for stage {}"), stageIdx);
@@ -161,6 +171,9 @@ static std::string color_arg_reg(GX::TevColorArg arg, size_t stageIdx, const STe
   }
   case GX::CC_ZERO:
     return "0.0";
+  default:
+    Log.report(logvisor::Fatal, FMT_STRING("invalid color arg {}"), arg);
+    unreachable();
   }
 }
 
@@ -169,10 +182,14 @@ static std::string alpha_arg_reg(GX::TevAlphaArg arg, size_t stageIdx, const STe
   case GX::CA_APREV:
     return "prev.a";
   case GX::CA_A0:
+    info.usesTevReg.set(0);
+    return "tevreg0.a";
   case GX::CA_A1:
+    info.usesTevReg.set(1);
+    return "tevreg1.a";
   case GX::CA_A2:
-    Log.report(logvisor::Fatal, FMT_STRING("TODO {}"), arg);
-    unreachable();
+    info.usesTevReg.set(2);
+    return "tevreg2.a";
   case GX::CA_TEXA: {
     if (stage.texMapId == GX::TEXMAP_NULL) {
       Log.report(logvisor::Fatal, FMT_STRING("unmapped texture for stage {}"), stageIdx);
@@ -269,6 +286,9 @@ static std::string alpha_arg_reg(GX::TevAlphaArg arg, size_t stageIdx, const STe
   }
   case GX::CA_ZERO:
     return "0.0";
+  default:
+    Log.report(logvisor::Fatal, FMT_STRING("invalid alpha arg {}"), arg);
+    unreachable();
   }
 }
 
@@ -292,6 +312,9 @@ static std::string_view tev_bias(GX::TevBias bias) {
     return " + 0.5";
   case GX::TB_SUBHALF:
     return " - 0.5";
+  default:
+    Log.report(logvisor::Fatal, FMT_STRING("invalid bias {}"), bias);
+    unreachable();
   }
 }
 
@@ -305,6 +328,9 @@ static std::string_view tev_scale(GX::TevScale scale) {
     return " * 4.0";
   case GX::CS_DIVIDE_2:
     return " / 2.0";
+  default:
+    Log.report(logvisor::Fatal, FMT_STRING("invalid scale {}"), scale);
+    unreachable();
   }
 }
 
@@ -379,7 +405,7 @@ std::pair<wgpu::ShaderModule, ShaderInfo> build_shader(const ShaderConfig& confi
   } else {
     uniformBindings += R"""(
 struct Vec3Block {
-    data: array<vec3<f32>>;
+    data: array<vec4<f32>>;
 };
 struct Vec2Block {
     data: array<vec2<f32>>;
@@ -398,7 +424,7 @@ var<storage, read> v_packed_uvs: Vec2Block;
         "\n    , @location(1) in_uv_0_4_idx: vec4<i32>"
         "\n    , @location(2) in_uv_5_7_idx: vec4<i32>";
     vtxOutAttrs += "\n    @builtin(position) pos: vec4<f32>;";
-    vtxXfrAttrsPre += "\n    out.pos = ubuf.xf * vec4<f32>(v_verts.data[in_pos_nrm_idx[0]], 1.0);";
+    vtxXfrAttrsPre += "\n    out.pos = ubuf.xf * vec4<f32>(v_verts.data[in_pos_nrm_idx[0]].xyz, 1.0);";
   }
 
   std::string fragmentFnPre;
@@ -413,6 +439,18 @@ var<storage, read> v_packed_uvs: Vec2Block;
       case GX::TevRegID::TEVPREV:
         outReg = "prev";
         break;
+      case GX::TEVREG0:
+        outReg = "tevreg0";
+        info.usesTevReg.set(0);
+        break;
+      case GX::TEVREG1:
+        outReg = "tevreg1";
+        info.usesTevReg.set(1);
+        break;
+      case GX::TEVREG2:
+        outReg = "tevreg2";
+        info.usesTevReg.set(2);
+        break;
       default:
         Log.report(logvisor::Fatal, FMT_STRING("TODO: colorOp outReg {}"), stage->colorOp.x10_regId);
       }
@@ -423,6 +461,9 @@ var<storage, read> v_packed_uvs: Vec2Block;
                       color_arg_reg(stage->colorPass.x8_c, idx, *stage, info),
                       color_arg_reg(stage->colorPass.xc_d, idx, *stage, info), tev_op(stage->colorOp.x4_op),
                       tev_bias(stage->colorOp.x8_bias), tev_scale(stage->colorOp.xc_scale));
+      if (stage->colorOp.x0_clamp) {
+        op = fmt::format(FMT_STRING("clamp({}, vec3<f32>(0.0), vec3<f32>(1.0))"), op);
+      }
       fragmentFn += fmt::format(FMT_STRING("\n    {0} = vec4<f32>({1}, {0}.a);"), outReg, op);
     }
     {
@@ -431,6 +472,18 @@ var<storage, read> v_packed_uvs: Vec2Block;
       case GX::TevRegID::TEVPREV:
         outReg = "prev.a";
         break;
+      case GX::TEVREG0:
+        outReg = "tevreg0.a";
+        info.usesTevReg.set(0);
+        break;
+      case GX::TEVREG1:
+        outReg = "tevreg1.a";
+        info.usesTevReg.set(1);
+        break;
+      case GX::TEVREG2:
+        outReg = "tevreg2.a";
+        info.usesTevReg.set(2);
+        break;
       default:
         Log.report(logvisor::Fatal, FMT_STRING("TODO: alphaOp outReg {}"), stage->alphaOp.x10_regId);
       }
@@ -441,10 +494,20 @@ var<storage, read> v_packed_uvs: Vec2Block;
                       alpha_arg_reg(stage->alphaPass.x8_c, idx, *stage, info),
                       alpha_arg_reg(stage->alphaPass.xc_d, idx, *stage, info), tev_op(stage->alphaOp.x4_op),
                       tev_bias(stage->alphaOp.x8_bias), tev_scale(stage->alphaOp.xc_scale));
+      if (stage->alphaOp.x0_clamp) {
+        op = fmt::format(FMT_STRING("clamp({}, 0.0, 1.0)"), op);
+      }
       fragmentFn += fmt::format(FMT_STRING("\n    {0} = {1};"), outReg, op);
     }
     idx++;
   }
+  for (int i = 0; i < info.usesTevReg.size(); ++i) {
+    if (!info.usesTevReg.test(i)) {
+      continue;
+    }
+    uniBufAttrs += fmt::format(FMT_STRING("\n    tevreg{}: vec4<f32>;"), i);
+    fragmentFnPre += fmt::format(FMT_STRING("\n    var tevreg{0} = ubuf.tevreg{0};"), i);
+  }
   for (int i = 0; i < info.sampledColorChannels.size(); ++i) {
     if (!info.sampledColorChannels.test(i)) {
       continue;
@@ -499,12 +562,12 @@ var<storage, read> v_packed_uvs: Vec2Block;
       vtxOutAttrs += fmt::format(FMT_STRING("\n    @location({}) tex{}_uv: vec2<f32>;"), locIdx, i);
       if (i < 4) {
         if (i == 0) {
-          vtxXfrAttrs += fmt::format(FMT_STRING("\n    out.tex{}_uv = v_packed_uvs.data[in_uv_0_4[{}]];"), i, i);
+          vtxXfrAttrs += fmt::format(FMT_STRING("\n    out.tex{}_uv = v_packed_uvs.data[in_uv_0_4_idx[{}]];"), i, i);
         } else {
-          vtxXfrAttrs += fmt::format(FMT_STRING("\n    out.tex{}_uv = v_uvs.data[in_uv_0_4[{}]];"), i, i);
+          vtxXfrAttrs += fmt::format(FMT_STRING("\n    out.tex{}_uv = v_uvs.data[in_uv_0_4_idx[{}]];"), i, i);
         }
       } else {
-        vtxXfrAttrs += fmt::format(FMT_STRING("\n    out.tex{}_uv = v_uvs.data[in_uv_5_7[{}]];"), i, i - 4);
+        vtxXfrAttrs += fmt::format(FMT_STRING("\n    out.tex{}_uv = v_uvs.data[in_uv_5_7_idx[{}]];"), i, i - 4);
       }
     }
     fragmentFnPre += fmt::format(
@@ -544,9 +607,10 @@ fn fs_main(in: VertexOutput) -> @location(0) vec4<f32> {{
 
   wgpu::ShaderModuleWGSLDescriptor wgslDescriptor{};
   wgslDescriptor.source = shaderSource.c_str();
+  const auto label = fmt::format(FMT_STRING("GX Shader {:x}"), hash);
   const auto shaderDescriptor = wgpu::ShaderModuleDescriptor{
       .nextInChain = &wgslDescriptor,
-      .label = "GX Shader",
+      .label = label.c_str(),
   };
   auto shader = gpu::g_device.CreateShaderModule(&shaderDescriptor);
 
diff --git a/aurora/lib/gfx/model/shader.cpp b/aurora/lib/gfx/model/shader.cpp
index 26a3f3973..398e034ab 100644
--- a/aurora/lib/gfx/model/shader.cpp
+++ b/aurora/lib/gfx/model/shader.cpp
@@ -10,14 +10,14 @@ enum class VtxDescAttr : u8 {
   Position = 0,
   Normal = 2,
   Color0 = 4,
-  Color1 = 8,
-  Tex0 = 10,
-  Tex1 = 12,
-  Tex2 = 14,
-  Tex3 = 16,
-  Tex4 = 18,
-  Tex5 = 20,
-  Tex6 = 22,
+  Color1 = 6,
+  Tex0 = 8,
+  Tex1 = 10,
+  Tex2 = 12,
+  Tex3 = 14,
+  Tex4 = 16,
+  Tex5 = 18,
+  Tex6 = 20,
   PnMatIdx = 24,
   Tex0MatIdx = 25,
   Tex1MatIdx = 26,
@@ -51,13 +51,13 @@ static logvisor::Module Log("aurora::gfx::model");
 
 static const std::vector<zeus::CVector3f>* vtxData;
 static const std::vector<zeus::CVector3f>* nrmData;
-static const std::vector<zeus::CVector2f>* tex0TcData;
-static const std::vector<zeus::CVector2f>* tcData;
+static const std::vector<Vec2<float>>* tex0TcData;
+static const std::vector<Vec2<float>>* tcData;
 
-void set_vertex_buffer(const std::vector<zeus::CVector3f>& data) noexcept { vtxData = &data; }
-void set_normal_buffer(const std::vector<zeus::CVector3f>& norm) noexcept { nrmData = &norm; }
-void set_tex0_tc_buffer(const std::vector<zeus::CVector2f>& tcs) noexcept { tex0TcData = &tcs; }
-void set_tc_buffer(const std::vector<zeus::CVector2f>& tcs) noexcept { tcData = &tcs; }
+void set_vertex_buffer(const std::vector<zeus::CVector3f>* data) noexcept { vtxData = data; }
+void set_normal_buffer(const std::vector<zeus::CVector3f>* norm) noexcept { nrmData = norm; }
+void set_tex0_tc_buffer(const std::vector<Vec2<float>>* tcs) noexcept { tex0TcData = tcs; }
+void set_tc_buffer(const std::vector<Vec2<float>>* tcs) noexcept { tcData = tcs; }
 
 enum class VertexFormat : u8 {
   F32F32,
@@ -170,12 +170,12 @@ void queue_surface(const u8* dlStart, u32 dlSize) noexcept {
   const auto idxRange = push_indices(ArrayRef{indices});
   const auto sVtxRange = push_storage(reinterpret_cast<const uint8_t*>(vtxData->data()), vtxData->size() * 16);
   const auto sNrmRange = push_storage(reinterpret_cast<const uint8_t*>(nrmData->data()), nrmData->size() * 16);
-  const auto sTcRange = push_storage(reinterpret_cast<const uint8_t*>(tcData->data()), tcData->size() * 16);
+  const auto sTcRange = push_storage(reinterpret_cast<const uint8_t*>(tcData->data()), tcData->size() * 8);
   Range sPackedTcRange;
   if (tcData == tex0TcData) {
     sPackedTcRange = sTcRange;
   } else {
-    sPackedTcRange = push_storage(reinterpret_cast<const uint8_t*>(tex0TcData->data()), tex0TcData->size() * 16);
+    sPackedTcRange = push_storage(reinterpret_cast<const uint8_t*>(tex0TcData->data()), tex0TcData->size() * 8);
   }
 
   model::PipelineConfig config{};
diff --git a/aurora/lib/gfx/texture.cpp b/aurora/lib/gfx/texture.cpp
index 05b14c393..bc583448b 100644
--- a/aurora/lib/gfx/texture.cpp
+++ b/aurora/lib/gfx/texture.cpp
@@ -130,27 +130,38 @@ void write_texture(const TextureHandle& handle, ArrayRef<uint8_t> data) noexcept
     }
   }
 
-  const auto dstView = wgpu::ImageCopyTexture{
-      .texture = ref.texture,
-  };
-  if (ref.size.depthOrArrayLayers != 1) {
-    Log.report(logvisor::Fatal, FMT_STRING("write_texture: unsupported depth {}"), ref.size.depthOrArrayLayers);
-    unreachable();
+  uint32_t offset = 0;
+  for (uint32_t mip = 0; mip < ref.mipCount; ++mip) {
+    const auto mipSize = wgpu::Extent3D{
+        .width = std::max(ref.size.width >> mip, 1u),
+        .height = std::max(ref.size.height >> mip, 1u),
+        .depthOrArrayLayers = ref.size.depthOrArrayLayers,
+    };
+    const auto info = format_info(ref.format);
+    const auto physicalSize = physical_size(mipSize, info);
+    const uint32_t widthBlocks = physicalSize.width / info.blockWidth;
+    const uint32_t heightBlocks = physicalSize.height / info.blockHeight;
+    const uint32_t bytesPerRow = widthBlocks * info.blockSize;
+    const uint32_t dataSize = bytesPerRow * heightBlocks * mipSize.depthOrArrayLayers;
+    if (offset + dataSize > data.size()) {
+      Log.report(logvisor::Fatal, FMT_STRING("write_texture: expected at least {} bytes, got {}"), offset + dataSize,
+                 data.size());
+      unreachable();
+    }
+    const auto dstView = wgpu::ImageCopyTexture{
+        .texture = ref.texture,
+        .mipLevel = mip,
+    };
+    const auto dataLayout = wgpu::TextureDataLayout{
+        .bytesPerRow = bytesPerRow,
+        .rowsPerImage = heightBlocks,
+    };
+    g_queue.WriteTexture(&dstView, data.data() + offset, dataSize, &dataLayout, &physicalSize);
+    offset += dataSize;
   }
-  const auto info = format_info(ref.format);
-  const auto physicalSize = physical_size(ref.size, info);
-  const uint32_t widthBlocks = physicalSize.width / info.blockWidth;
-  const uint32_t heightBlocks = physicalSize.height / info.blockHeight;
-  const uint32_t bytesPerRow = widthBlocks * info.blockSize;
-  const uint32_t dataSize = bytesPerRow * heightBlocks * ref.size.depthOrArrayLayers;
-  if (dataSize > data.size()) {
-    Log.report(logvisor::Fatal, FMT_STRING("write_texture: expected at least {} bytes, got {}"), dataSize, data.size());
-    unreachable();
+  if (offset < data.size()) {
+    Log.report(logvisor::Warning, FMT_STRING("write_texture: texture used {} bytes, but given {} bytes"), offset,
+               data.size());
   }
-  const auto dataLayout = wgpu::TextureDataLayout{
-      .bytesPerRow = bytesPerRow,
-      .rowsPerImage = heightBlocks,
-  };
-  g_queue.WriteTexture(&dstView, data.data(), dataSize, &dataLayout, &ref.size);
 }
 } // namespace aurora::gfx
diff --git a/aurora/lib/gpu.cpp b/aurora/lib/gpu.cpp
index fb903395c..b5453ccaa 100644
--- a/aurora/lib/gpu.cpp
+++ b/aurora/lib/gpu.cpp
@@ -102,7 +102,7 @@ static TextureWithSampler create_depth_texture() {
 }
 
 static void error_callback(WGPUErrorType type, char const* message, void* userdata) {
-  Log.report(logvisor::Error, FMT_STRING("Dawn error {}: {}"),
+  Log.report(logvisor::Fatal, FMT_STRING("Dawn error {}: {}"),
              magic_enum::enum_name(static_cast<wgpu::ErrorType>(type)), message);
 }