Multiple viewports are only valid while using geometry shaders so avoid uploading states that can be potentially invalid.

fix Clear viewport area by calculating overlapping area when viewport and scissor are enabled at the same time.
add warning messages for extension that are not required but are needed for a correct emulation.

src/video_core/engines/maxwell_3d.h

@@ -624,7 +624,16 @@ public:
                     }
                 } zeta;
 
-                INSERT_PADDING_WORDS(0x5B);
+                INSERT_PADDING_WORDS(0x41);
+
+                union {
+                    BitField<0, 4, u32> stencil;
+                    BitField<4, 4, u32> unknown;
+                    BitField<8, 4, u32> scissor;
+                    BitField<12, 4, u32> viewport;
+                } clear_flags;
+
+                INSERT_PADDING_WORDS(0x19);
 
                 std::array<VertexAttribute, NumVertexAttributes> vertex_attrib_format;
 
@@ -719,7 +728,20 @@ public:
 
                 u32 vb_element_base;
 
-                INSERT_PADDING_WORDS(0x38);
+                INSERT_PADDING_WORDS(0x36);
+
+                union {
+                    BitField<0, 1, u32> c0;
+                    BitField<1, 1, u32> c1;
+                    BitField<2, 1, u32> c2;
+                    BitField<3, 1, u32> c3;
+                    BitField<4, 1, u32> c4;
+                    BitField<5, 1, u32> c5;
+                    BitField<6, 1, u32> c6;
+                    BitField<7, 1, u32> c7;
+                } clip_distance_enabled;
+
+                INSERT_PADDING_WORDS(0x1);
 
                 float point_size;
 
@@ -863,7 +885,32 @@ public:
 
                 Cull cull;
 
-                INSERT_PADDING_WORDS(0x28);
+                u32 pixel_center_integer;
+
+                INSERT_PADDING_WORDS(0x1);
+
+                u32 viewport_transform_enabled;
+
+                INSERT_PADDING_WORDS(0x3);
+
+                union {
+                    BitField<0, 1, u32> depth_range_0_1;
+                    BitField<3, 1, u32> depth_clamp_near;
+                    BitField<4, 1, u32> depth_clamp_far;
+                } view_volume_clip_control;
+
+                union {
+                    BitField<0, 4, u32> c0;
+                    BitField<4, 4, u32> c1;
+                    BitField<8, 4, u32> c2;
+                    BitField<12, 4, u32> c3;
+                    BitField<16, 4, u32> c4;
+                    BitField<20, 4, u32> c5;
+                    BitField<24, 4, u32> c6;
+                    BitField<28, 4, u32> c7;
+                } clip_distance_mode;
+
+                INSERT_PADDING_WORDS(0x20);
 
                 struct {
                     u32 enable;
@@ -1127,6 +1174,7 @@ ASSERT_REG_POSITION(stencil_back_func_mask, 0x3D7);
 ASSERT_REG_POSITION(color_mask_common, 0x3E4);
 ASSERT_REG_POSITION(rt_separate_frag_data, 0x3EB);
 ASSERT_REG_POSITION(zeta, 0x3F8);
+ASSERT_REG_POSITION(clear_flags, 0x43E);
 ASSERT_REG_POSITION(vertex_attrib_format, 0x458);
 ASSERT_REG_POSITION(rt_control, 0x487);
 ASSERT_REG_POSITION(zeta_width, 0x48a);
@@ -1153,6 +1201,7 @@ ASSERT_REG_POSITION(stencil_front_mask, 0x4E7);
 ASSERT_REG_POSITION(frag_color_clamp, 0x4EA);
 ASSERT_REG_POSITION(screen_y_control, 0x4EB);
 ASSERT_REG_POSITION(vb_element_base, 0x50D);
+ASSERT_REG_POSITION(clip_distance_enabled, 0x544);
 ASSERT_REG_POSITION(point_size, 0x546);
 ASSERT_REG_POSITION(zeta_enable, 0x54E);
 ASSERT_REG_POSITION(multisample_control, 0x54F);
@@ -1171,6 +1220,9 @@ ASSERT_REG_POSITION(primitive_restart, 0x591);
 ASSERT_REG_POSITION(index_array, 0x5F2);
 ASSERT_REG_POSITION(instanced_arrays, 0x620);
 ASSERT_REG_POSITION(cull, 0x646);
+ASSERT_REG_POSITION(pixel_center_integer, 0x649);
+ASSERT_REG_POSITION(viewport_transform_enabled, 0x64B);
+ASSERT_REG_POSITION(clip_distance_mode, 0x650);
 ASSERT_REG_POSITION(logic_op, 0x671);
 ASSERT_REG_POSITION(clear_buffers, 0x674);
 ASSERT_REG_POSITION(color_mask, 0x680);

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -118,10 +118,43 @@ RasterizerOpenGL::RasterizerOpenGL(Core::Frontend::EmuWindow& window, ScreenInfo
     glGetIntegerv(GL_UNIFORM_BUFFER_OFFSET_ALIGNMENT, &uniform_buffer_alignment);
 
     LOG_CRITICAL(Render_OpenGL, "Sync fixed function OpenGL state here!");
+    CheckExtensions();
 }
 
 RasterizerOpenGL::~RasterizerOpenGL() {}
 
+void RasterizerOpenGL::CheckExtensions() {
+    if (!GLAD_GL_ARB_texture_filter_anisotropic && !GLAD_GL_EXT_texture_filter_anisotropic) {
+        LOG_WARNING(
+            Render_OpenGL,
+            "Anisotropic filter is not supported! this can cause graphical issues in some games.");
+    }
+    if (!GLAD_GL_ARB_viewport_array) {
+        LOG_WARNING(Render_OpenGL, "Viewport arrays are not supported! this can potyentially cause "
+                                   "issues in games that use geometry shaders.");
+    }
+    if (!GLAD_GL_ARB_color_buffer_float) {
+        LOG_WARNING(
+            Render_OpenGL,
+            "Color clamp control is not supported! this can cause graphical issues in some games.");
+    }
+    if (!GLAD_GL_ARB_buffer_storage) {
+        LOG_WARNING(
+            Render_OpenGL,
+            "Buffer storage control is not supported! this can cause performance degradation.");
+    }
+    if (!GLAD_GL_AMD_depth_clamp_separate) {
+        if (!GLAD_GL_ARB_depth_clamp) {
+            LOG_WARNING(
+                Render_OpenGL,
+                "Depth Clamp is not supported! this can cause graphical issues in some games.");
+        } else {
+            LOG_WARNING(Render_OpenGL, "Separate Depth Clamp is not supported! this can cause "
+                                       "graphical issues in some games.");
+        }
+    }
+}
+
 void RasterizerOpenGL::SetupVertexFormat() {
     auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
     const auto& regs = gpu.regs;
@@ -266,6 +299,11 @@ DrawParameters RasterizerOpenGL::SetupDraw() {
     return params;
 }
 
+bool RasterizerOpenGL::GeometryShaderActive() {
+    const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
+    return gpu.regs.IsShaderConfigEnabled(static_cast<size_t>(Maxwell::ShaderProgram::Geometry));
+}
+
 void RasterizerOpenGL::SetupShaders(GLenum primitive_mode) {
     MICROPROFILE_SCOPE(OpenGL_Shader);
     const auto& gpu = Core::System::GetInstance().GPU().Maxwell3D();
@@ -542,6 +580,30 @@ void RasterizerOpenGL::Clear() {
         ASSERT_MSG(regs.zeta_enable != 0, "Tried to clear stencil but buffer is not enabled!");
         use_stencil = true;
         clear_state.stencil.test_enabled = true;
+        if (regs.clear_flags.stencil) {
+            // Stencil affects the clear so  fill it with the used masks
+            clear_state.stencil.front.test_func = GL_ALWAYS;
+            clear_state.stencil.front.test_mask = regs.stencil_front_func_mask;
+            clear_state.stencil.front.action_stencil_fail = GL_KEEP;
+            clear_state.stencil.front.action_depth_fail = GL_KEEP;
+            clear_state.stencil.front.action_depth_pass = GL_KEEP;
+            clear_state.stencil.front.write_mask = regs.stencil_front_mask;
+            if (regs.stencil_two_side_enable) {
+                clear_state.stencil.back.test_func = GL_ALWAYS;
+                clear_state.stencil.back.test_mask = regs.stencil_back_func_mask;
+                clear_state.stencil.back.action_stencil_fail = GL_KEEP;
+                clear_state.stencil.back.action_depth_fail = GL_KEEP;
+                clear_state.stencil.back.action_depth_pass = GL_KEEP;
+                clear_state.stencil.back.write_mask = regs.stencil_back_mask;
+            } else {
+                clear_state.stencil.back.test_func = GL_ALWAYS;
+                clear_state.stencil.back.test_mask = 0xFFFFFFFF;
+                clear_state.stencil.back.write_mask = 0xFFFFFFFF;
+                clear_state.stencil.back.action_stencil_fail = GL_KEEP;
+                clear_state.stencil.back.action_depth_fail = GL_KEEP;
+                clear_state.stencil.back.action_depth_pass = GL_KEEP;
+            }
+        }
     }
 
     if (!use_color && !use_depth && !use_stencil) {
@@ -553,6 +615,15 @@ void RasterizerOpenGL::Clear() {
 
     ConfigureFramebuffers(clear_state, use_color, use_depth || use_stencil, false,
                           regs.clear_buffers.RT.Value());
+    if (regs.clear_flags.scissor) {
+        SyncScissorTest(clear_state);
+    }
+
+    if (regs.clear_flags.viewport) {
+        // Viewport does not affects glClearBuffer so emulate viewport using  Scissor
+        clear_state.TranslateViewportToScissor();
+    }
+
     clear_state.Apply();
 
     if (use_color) {
@@ -577,7 +648,7 @@ void RasterizerOpenGL::DrawArrays() {
     const auto& regs = gpu.regs;
 
     ScopeAcquireGLContext acquire_context{emu_window};
-
+    state.geometry_shaders.enabled = GeometryShaderActive();
     ConfigureFramebuffers(state);
     SyncColorMask();
     SyncFragmentColorClampState();
@@ -588,7 +659,7 @@ void RasterizerOpenGL::DrawArrays() {
     SyncLogicOpState();
     SyncCullMode();
     SyncPrimitiveRestart();
-    SyncScissorTest();
+    SyncScissorTest(state);
     // Alpha Testing is synced on shaders.
     SyncTransformFeedback();
     SyncPointState();
@@ -815,7 +886,7 @@ void RasterizerOpenGL::SamplerInfo::SyncWithConfig(const Tegra::Texture::TSCEntr
     }
     const u32 bias = config.mip_lod_bias.Value();
     // Sign extend the 13-bit value.
-    const u32 mask = 1U << (13 - 1);
+    constexpr u32 mask = 1U << (13 - 1);
     const float bias_lod = static_cast<s32>((bias ^ mask) - mask) / 256.f;
     if (lod_bias != bias_lod) {
         lod_bias = bias_lod;
@@ -942,16 +1013,28 @@ u32 RasterizerOpenGL::SetupTextures(Maxwell::ShaderStage stage, Shader& shader,
 
 void RasterizerOpenGL::SyncViewport(OpenGLState& current_state) {
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
-    for (std::size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumViewports; i++) {
-        const MathUtil::Rectangle<s32> viewport_rect{regs.viewport_transform[i].GetRect()};
+    const std::size_t viewport_count =
+        current_state.geometry_shaders.enabled ? Tegra::Engines::Maxwell3D::Regs::NumViewports : 1;
+    for (std::size_t i = 0; i < viewport_count; i++) {
         auto& viewport = current_state.viewports[i];
-        viewport.x = viewport_rect.left;
-        viewport.y = viewport_rect.bottom;
-        viewport.width = static_cast<GLfloat>(viewport_rect.GetWidth());
-        viewport.height = static_cast<GLfloat>(viewport_rect.GetHeight());
+        const auto& src = regs.viewports[i];
+        if (regs.viewport_transform_enabled) {
+            const MathUtil::Rectangle<s32> viewport_rect{regs.viewport_transform[i].GetRect()};
+            viewport.x = viewport_rect.left;
+            viewport.y = viewport_rect.bottom;
+            viewport.width = viewport_rect.GetWidth();
+            viewport.height = viewport_rect.GetHeight();
+        } else {
+            viewport.x = src.x;
+            viewport.y = src.y;
+            viewport.width = src.width;
+            viewport.height = src.height;
+        }
         viewport.depth_range_far = regs.viewports[i].depth_range_far;
         viewport.depth_range_near = regs.viewports[i].depth_range_near;
     }
+    current_state.depth_clamp.far_plane = regs.view_volume_clip_control.depth_clamp_far != 0;
+    current_state.depth_clamp.near_plane = regs.view_volume_clip_control.depth_clamp_near != 0;
 }
 
 void RasterizerOpenGL::SyncClipEnabled() {
@@ -1120,11 +1203,13 @@ void RasterizerOpenGL::SyncLogicOpState() {
     state.logic_op.operation = MaxwellToGL::LogicOp(regs.logic_op.operation);
 }
 
-void RasterizerOpenGL::SyncScissorTest() {
+void RasterizerOpenGL::SyncScissorTest(OpenGLState& current_state) {
     const auto& regs = Core::System::GetInstance().GPU().Maxwell3D().regs;
-    for (std::size_t i = 0; i < Tegra::Engines::Maxwell3D::Regs::NumViewports; i++) {
+    const std::size_t viewport_count =
+        current_state.geometry_shaders.enabled ? Tegra::Engines::Maxwell3D::Regs::NumViewports : 1;
+    for (std::size_t i = 0; i < viewport_count; i++) {
         const auto& src = regs.scissor_test[i];
-        auto& dst = state.viewports[i].scissor;
+        auto& dst = current_state.viewports[i].scissor;
         dst.enabled = (src.enable != 0);
         if (dst.enabled == 0) {
             return;

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -91,19 +91,20 @@ private:
         void SyncWithConfig(const Tegra::Texture::TSCEntry& info);
 
     private:
-        Tegra::Texture::TextureFilter mag_filter;
-        Tegra::Texture::TextureFilter min_filter;
-        Tegra::Texture::TextureMipmapFilter mip_filter;
-        Tegra::Texture::WrapMode wrap_u;
-        Tegra::Texture::WrapMode wrap_v;
-        Tegra::Texture::WrapMode wrap_p;
-        bool uses_depth_compare;
-        Tegra::Texture::DepthCompareFunc depth_compare_func;
-        GLvec4 border_color;
-        float min_lod;
-        float max_lod;
-        float lod_bias;
-        float max_anisotropic;
+        Tegra::Texture::TextureFilter mag_filter = Tegra::Texture::TextureFilter::Nearest;
+        Tegra::Texture::TextureFilter min_filter = Tegra::Texture::TextureFilter::Nearest;
+        Tegra::Texture::TextureMipmapFilter mip_filter = Tegra::Texture::TextureMipmapFilter::None;
+        Tegra::Texture::WrapMode wrap_u = Tegra::Texture::WrapMode::ClampToEdge;
+        Tegra::Texture::WrapMode wrap_v = Tegra::Texture::WrapMode::ClampToEdge;
+        Tegra::Texture::WrapMode wrap_p = Tegra::Texture::WrapMode::ClampToEdge;
+        bool uses_depth_compare = false;
+        Tegra::Texture::DepthCompareFunc depth_compare_func =
+            Tegra::Texture::DepthCompareFunc::Always;
+        GLvec4 border_color = {};
+        float min_lod = 0.f;
+        float max_lod = 16.0f;
+        float lod_bias = 0.0f;
+        float max_anisotropic = 1.0f;
     };
 
     /**
@@ -171,7 +172,7 @@ private:
     void SyncMultiSampleState();
 
     /// Syncs the scissor test state to match the guest state
-    void SyncScissorTest();
+    void SyncScissorTest(OpenGLState& current_state);
 
     /// Syncs the transform feedback state to match the guest state
     void SyncTransformFeedback();
@@ -185,6 +186,10 @@ private:
     /// Check asserts for alpha testing.
     void CheckAlphaTests();
 
+    /// check for extension that are not strictly required
+    /// but are needed for correct emulation
+    void CheckExtensions();
+
     bool has_ARB_direct_state_access = false;
     bool has_ARB_multi_bind = false;
     bool has_ARB_separate_shader_objects = false;
@@ -223,6 +228,7 @@ private:
     DrawParameters SetupDraw();
 
     void SetupShaders(GLenum primitive_mode);
+    bool GeometryShaderActive();
 
     enum class AccelDraw { Disabled, Arrays, Indexed };
     AccelDraw accelerate_draw = AccelDraw::Disabled;

src/video_core/renderer_opengl/gl_state.cpp

@@ -92,6 +92,8 @@ OpenGLState::OpenGLState() {
 
     point.size = 1;
     fragment_color_clamp.enabled = false;
+    depth_clamp.far_plane = false;
+    depth_clamp.near_plane = false;
 }
 
 void OpenGLState::ApplyDefaultState() {
@@ -234,6 +236,28 @@ void OpenGLState::ApplyStencilTest() const {
     }
 }
 
+void OpenGLState::TranslateViewportToScissor() {
+    auto& current = this->viewports[0];
+    if (current.scissor.enabled) {
+        GLint left = std::max(current.x, current.scissor.x);
+        GLint right =
+            std::max(current.x + current.width, current.scissor.x + current.scissor.width);
+        GLint bottom = std::max(current.y, current.scissor.y);
+        GLint top =
+            std::max(current.y + current.height, current.scissor.y + current.scissor.height);
+        current.scissor.x = std::max(left, 0);
+        current.scissor.y = std::max(bottom, 0);
+        current.scissor.width = std::max(right - left, 0);
+        current.scissor.height = std::max(top - bottom, 0);
+    } else {
+        current.scissor.enabled = true;
+        current.scissor.x = current.x;
+        current.scissor.y = current.y;
+        current.scissor.width = current.width;
+        current.scissor.height = current.height;
+    }
+}
+
 void OpenGLState::ApplyViewport() const {
     if (GLAD_GL_ARB_viewport_array && geometry_shaders.enabled) {
         for (GLuint i = 0; i < static_cast<GLuint>(Tegra::Engines::Maxwell3D::Regs::NumViewports);
@@ -242,7 +266,9 @@ void OpenGLState::ApplyViewport() const {
             const auto& updated = viewports[i];
             if (updated.x != current.x || updated.y != current.y ||
                 updated.width != current.width || updated.height != current.height) {
-                glViewportIndexedf(i, updated.x, updated.y, updated.width, updated.height);
+                glViewportIndexedf(
+                    i, static_cast<GLfloat>(updated.x), static_cast<GLfloat>(updated.y),
+                    static_cast<GLfloat>(updated.width), static_cast<GLfloat>(updated.height));
             }
             if (updated.depth_range_near != current.depth_range_near ||
                 updated.depth_range_far != current.depth_range_far) {
@@ -270,8 +296,7 @@ void OpenGLState::ApplyViewport() const {
         const auto& updated = viewports[0];
         if (updated.x != current.x || updated.y != current.y || updated.width != current.width ||
             updated.height != current.height) {
-            glViewport(static_cast<GLint>(updated.x), static_cast<GLint>(updated.y),
-                       static_cast<GLsizei>(updated.width), static_cast<GLsizei>(updated.height));
+            glViewport(updated.x, updated.y, updated.width, updated.height);
         }
         if (updated.depth_range_near != current.depth_range_near ||
             updated.depth_range_far != current.depth_range_far) {
@@ -483,6 +508,29 @@ void OpenGLState::Apply() const {
                          fragment_color_clamp.enabled ? GL_TRUE : GL_FALSE);
         }
     }
+    if (depth_clamp.far_plane != cur_state.depth_clamp.far_plane ||
+        depth_clamp.near_plane != cur_state.depth_clamp.near_plane) {
+        if (GLAD_GL_AMD_depth_clamp_separate) {
+            if (depth_clamp.far_plane) {
+                glEnable(GL_DEPTH_CLAMP_FAR_AMD);
+            } else {
+                glDisable(GL_DEPTH_CLAMP_FAR_AMD);
+            }
+            if (depth_clamp.near_plane) {
+                glEnable(GL_DEPTH_CLAMP_NEAR_AMD);
+            } else {
+                glDisable(GL_DEPTH_CLAMP_NEAR_AMD);
+            }
+        } else {
+            if (depth_clamp.far_plane || depth_clamp.near_plane) {
+                glEnable(GL_DEPTH_CLAMP);
+            } else {
+                glDisable(GL_DEPTH_CLAMP);
+            }
+            UNIMPLEMENTED_IF_MSG(depth_clamp.far_plane != depth_clamp.near_plane,
+                                 "Unimplemented Depth Clamp Separation!");
+        }
+    }
     if (multisample_control.alpha_to_coverage != cur_state.multisample_control.alpha_to_coverage) {
         if (multisample_control.alpha_to_coverage) {
             glEnable(GL_SAMPLE_ALPHA_TO_COVERAGE);

src/video_core/renderer_opengl/gl_state.h

@@ -48,6 +48,11 @@ public:
         bool enabled; // GL_CLAMP_FRAGMENT_COLOR_ARB
     } fragment_color_clamp;
 
+    struct {
+        bool far_plane;
+        bool near_plane;
+    } depth_clamp; // GL_DEPTH_CLAMP
+
     struct {
         bool enabled; // viewports arrays are only supported when geometry shaders are enabled.
     } geometry_shaders;
@@ -156,10 +161,10 @@ public:
     } draw;
 
     struct viewport {
-        GLfloat x;
-        GLfloat y;
-        GLfloat width;
-        GLfloat height;
+        GLint x;
+        GLint y;
+        GLint width;
+        GLint height;
         GLfloat depth_range_near; // GL_DEPTH_RANGE
         GLfloat depth_range_far;  // GL_DEPTH_RANGE
         struct {
@@ -206,6 +211,7 @@ public:
     OpenGLState& ResetBuffer(GLuint handle);
     OpenGLState& ResetVertexArray(GLuint handle);
     OpenGLState& ResetFramebuffer(GLuint handle);
+    void TranslateViewportToScissor();
 
 private:
     static OpenGLState cur_state;