glBufferSubData VBO

src/video_core/command_processor.cpp

@@ -352,10 +352,6 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
         shader_engine->SetupBatch(g_state.vs, regs.vs.main_offset);
 
         const bool use_gs = regs.pipeline.use_gs == PipelineRegs::UseGS::Yes;
-        g_state.geometry_pipeline.Reconfigure();
-        g_state.geometry_pipeline.Setup(shader_engine);
-        if (g_state.geometry_pipeline.NeedIndexInput())
-            ASSERT(is_indexed);
 
         auto VSUnitLoop = [&](u32 thread_id, auto num_threads) {
             constexpr bool single_thread = std::is_same_v<std::integral_constant<u32, 1>, decltype(num_threads)>;
@@ -436,6 +432,11 @@ static void WritePicaReg(u32 id, u32 value, u32 mask) {
             }
         }
 
+        g_state.geometry_pipeline.Reconfigure();
+        g_state.geometry_pipeline.Setup(shader_engine);
+        if (g_state.geometry_pipeline.NeedIndexInput())
+            ASSERT(is_indexed);
+
         for (unsigned int index = 0; index < regs.pipeline.num_vertices; ++index) {
             unsigned int vertex = VertexIndex(index);
             auto& cached_vertex = vs_output[is_indexed ? vertex : index];

src/video_core/primitive_assembly.cpp

@@ -15,7 +15,7 @@ PrimitiveAssembler<VertexType>::PrimitiveAssembler(PipelineRegs::TriangleTopolog
 
 template <typename VertexType>
 void PrimitiveAssembler<VertexType>::SubmitVertex(const VertexType& vtx,
-                                                  TriangleHandler triangle_handler) {
+                                                  const TriangleHandler& triangle_handler) {
     switch (topology) {
     case PipelineRegs::TriangleTopology::List:
     case PipelineRegs::TriangleTopology::Shader:

src/video_core/primitive_assembly.h

@@ -27,7 +27,7 @@ struct PrimitiveAssembler {
      * NOTE: We could specify the triangle handler in the constructor, but this way we can
      * keep event and handler code next to each other.
      */
-    void SubmitVertex(const VertexType& vtx, TriangleHandler triangle_handler);
+    void SubmitVertex(const VertexType& vtx, const TriangleHandler& triangle_handler);
 
     /**
      * Invert the vertex order of the next triangle. Called by geometry shader emitter.

src/video_core/renderer_opengl/gl_rasterizer.cpp

@@ -27,7 +27,7 @@ MICROPROFILE_DEFINE(OpenGL_Drawing, "OpenGL", "Drawing", MP_RGB(128, 128, 192));
 MICROPROFILE_DEFINE(OpenGL_Blits, "OpenGL", "Blits", MP_RGB(100, 100, 255));
 MICROPROFILE_DEFINE(OpenGL_CacheManagement, "OpenGL", "Cache Mgmt", MP_RGB(100, 255, 100));
 
-RasterizerOpenGL::RasterizerOpenGL() : shader_dirty(true) {
+RasterizerOpenGL::RasterizerOpenGL() : shader_dirty(true), vertex_buffer_size(0) {
     // Clipping plane 0 is always enabled for PICA fixed clip plane z <= 0
     state.clip_distance[0] = true;
 
@@ -236,24 +236,24 @@ void RasterizerOpenGL::DrawTriangles() {
     state.Apply();
 
     glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D,
-                           color_surface != nullptr ? color_surface->texture.handle : 0, 0);
+        color_surface != nullptr ? color_surface->texture.handle : 0, 0);
     if (depth_surface != nullptr) {
         if (regs.framebuffer.framebuffer.depth_format ==
             Pica::FramebufferRegs::DepthFormat::D24S8) {
             // attach both depth and stencil
             glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D,
-                                   depth_surface->texture.handle, 0);
+                depth_surface->texture.handle, 0);
         } else {
             // attach depth
             glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_TEXTURE_2D,
-                                   depth_surface->texture.handle, 0);
+                depth_surface->texture.handle, 0);
             // clear stencil attachment
             glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0, 0);
         }
     } else {
         // clear both depth and stencil attachment
         glFramebufferTexture2D(GL_DRAW_FRAMEBUFFER, GL_DEPTH_STENCIL_ATTACHMENT, GL_TEXTURE_2D, 0,
-                               0);
+            0);
     }
 
     // Sync the viewport
@@ -263,6 +263,11 @@ void RasterizerOpenGL::DrawTriangles() {
     GLsizei viewport_height =
         (GLsizei)Pica::float24::FromRaw(regs.rasterizer.viewport_size_y).ToFloat32() * 2;
 
+    const float res_scale_width = color_surface != nullptr ? color_surface->res_scale_width :
+                                  (depth_surface == nullptr ? 1.0f : depth_surface->res_scale_width);
+    const float res_scale_height = color_surface != nullptr ? color_surface->res_scale_height :
+                                  (depth_surface == nullptr ? 1.0f : depth_surface->res_scale_height);
+
     glViewport(
         (GLint)(rect.left + regs.rasterizer.viewport_corner.x * color_surface->res_scale_width),
         (GLint)(rect.bottom + regs.rasterizer.viewport_corner.y * color_surface->res_scale_height),
@@ -374,16 +379,21 @@ void RasterizerOpenGL::DrawTriangles() {
     // Sync the uniform data
     if (uniform_block_data.dirty) {
         glBufferData(GL_UNIFORM_BUFFER, sizeof(UniformData), &uniform_block_data.data,
-                     GL_STATIC_DRAW);
+            GL_STATIC_DRAW);
         uniform_block_data.dirty = false;
     }
 
     state.Apply();
 
     // Draw the vertex batch
-    glBufferData(GL_ARRAY_BUFFER, vertex_batch.size() * sizeof(HardwareVertex), vertex_batch.data(),
+    GLsizeiptr target_size = vertex_batch.size() * sizeof(HardwareVertex);
-                 GL_STREAM_DRAW);
+    if (vertex_buffer_size < target_size) {
-    glDrawArrays(GL_TRIANGLES, 0, (GLsizei)vertex_batch.size());
+        vertex_buffer_size = target_size * 2;
+        glBufferData(GL_ARRAY_BUFFER, vertex_buffer_size, nullptr, GL_STREAM_DRAW);
+    }
+    glBufferSubData(GL_ARRAY_BUFFER, 0, target_size, vertex_batch.data());
+    glDrawArrays(GL_TRIANGLES, 0, static_cast<GLsizei>(vertex_batch.size()));
+    vertex_batch.clear();
 
     // Mark framebuffer surfaces as dirty
     // TODO: Restrict invalidation area to the viewport
@@ -396,8 +406,6 @@ void RasterizerOpenGL::DrawTriangles() {
         res_cache.FlushRegion(depth_surface->addr, depth_surface->size, depth_surface, true);
     }
 
-    vertex_batch.clear();
-
     // Unbind textures for potential future use as framebuffer attachments
     for (unsigned texture_index = 0; texture_index < pica_textures.size(); ++texture_index) {
         state.texture_units[texture_index].texture_2d = 0;

src/video_core/renderer_opengl/gl_rasterizer.h

@@ -283,6 +283,7 @@ private:
     std::array<SamplerInfo, 3> texture_samplers;
     OGLVertexArray vertex_array;
     OGLBuffer vertex_buffer;
+    GLsizeiptr vertex_buffer_size;
     OGLBuffer uniform_buffer;
     OGLFramebuffer framebuffer;