Restore the whole mip chain of textures narrowed by a GL render pass

未关闭
#142 0 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
3-5 天
新手友好度
52/100
Issue 类型
缺陷
描述清晰度
基本清楚
活跃度
活跃
技术栈
cpp

调研方向

Start by reading source/renderer/ore/gl/ore_render_pass_gl.cpp around applySampledLevels() and finish(), then inspect ore_texture_gl.cpp and the matching headers. Trace how sampled mip ranges are applied and restored, and check the renderer’s existing OpenGL tests. Done means narrowed textures regain their full mip chain when the render pass finishes, with regression coverage for that behavior.

由索引模型根据 Issue 内容生成。

描述

Restore the whole mip chain of textures narrowed by a GL render pass.

- source/renderer/ore/gl/ore_texture_gl.hpp: Ore GL: TextureGL::restoreSampledLevels hands a narrowed texture its whole mip chain back
- source/renderer/ore/gl/ore_texture_gl.cpp: Ore GL: TextureGL::restoreSampledLevels implementation
- source/renderer/ore/gl/ore_render_pass_gl.hpp: Ore GL: a render pass remembers the textures it narrowed to a mip range
- source/renderer/ore/gl/ore_render_pass_gl.cpp: Ore GL applySampledLevels: record textures narrowed to a mip range
- source/renderer/ore/gl/ore_render_pass_gl.cpp: Ore GL finish: restore the whole mip chain of narrowed textures so the GL object leaves the pass unnarrowed

diff --git a/source/renderer/ore/gl/ore_render_pass_gl.cpp b/source/renderer/ore/gl/ore_render_pass_gl.cpp
index e07f50f468d37e9ea61544032056ee17d5e9fe8e..4871dbe3f4defe12c109cb2e271436bed70283e9 100644
--- a/source/renderer/ore/gl/ore_render_pass_gl.cpp
+++ b/source/renderer/ore/gl/ore_render_pass_gl.cpp
@@ -615,7 +615,18 @@ void RenderPassGL::applySampledLevels()
     }
   }
   for (const auto& l : m_sampledLevels)
+  {
     l.texture->setSampledLevels(l.unit, l.base, l.max);
+    const bool narrowed =
+        l.base != 0 || l.max + 1 < std::max(l.texture->numMipmaps(), 1u);
+    if (narrowed &&
+        std::none_of(m_narrowedTextures.begin(),
+                     m_narrowedTextures.end(),
+                     [&](const auto& t) { return t.get() == l.texture; }))
+    {
+      m_narrowedTextures.push_back(ref_rcp(l.texture));
+    }
+  }
 }
 
 void RenderPassGL::setViewport(
@@ -741,6 +752,10 @@ void RenderPassGL::finish()
   m_currentPipeline = nullptr;
   releaseBoundResources();
 
+  for (const auto& texture : m_narrowedTextures)
+    texture->restoreSampledLevels();
+  m_narrowedTextures.clear();
+
   glDisable(GL_SCISSOR_TEST);
   glDisable(GL_BLEND);
   glDisable(GL_DEPTH_TEST);
diff --git a/source/renderer/ore/gl/ore_render_pass_gl.hpp b/source/renderer/ore/gl/ore_render_pass_gl.hpp
index 544c779effc3228e480ef4cdc38d4bc9b2c1a942..773bcaf1daf06ab3219013c8d7e3694adf5913b9 100644
--- a/source/renderer/ore/gl/ore_render_pass_gl.hpp
+++ b/source/renderer/ore/gl/ore_render_pass_gl.hpp
@@ -83,6 +83,8 @@ private:
     uint32_t max;
   };
   std::vector<SampledLevels> m_sampledLevels;
+  // Textures this pass narrowed, restored to their whole chain at finish().
+  std::vector<rcp<TextureGL>> m_narrowedTextures;
 
   // Buffers and textures at the pipeline's units, which differ from the
   // layout's when the layout came from another shader.
diff --git a/source/renderer/ore/gl/ore_texture_gl.cpp b/source/renderer/ore/gl/ore_texture_gl.cpp
index 82d21e1049c6d0faf222e4b410f8e1648672a970..c2d4a0dbf37126e6e0e58b1a4aa70abea61cbf9c 100644
--- a/source/renderer/ore/gl/ore_texture_gl.cpp
+++ b/source/renderer/ore/gl/ore_texture_gl.cpp
@@ -340,6 +340,17 @@ void TextureGL::setSampledLevels(uint32_t unit, uint32_t base, uint32_t max)
   m_glMaxLevel = max;
 }
 
+void TextureGL::restoreSampledLevels()
+{
+  const uint32_t max = std::max(numMipmaps(), 1u) - 1;
+  if (m_glBaseLevel == 0 && m_glMaxLevel == max)
+    return;
+  glActiveTexture(GL_TEXTURE0);
+  glBindTexture(m_glTarget, m_glTexture);
+  setSampledLevels(0, 0, max);
+  glBindTexture(m_glTarget, 0);
+}
+
 TextureGL::~TextureGL()
 {
   if (m_glRenderbuffer != 0 && m_glOwnsTexture)
diff --git a/source/renderer/ore/gl/ore_texture_gl.hpp b/source/renderer/ore/gl/ore_texture_gl.hpp
index 62ea3852b16849cf3438a8dc6a5488b5199fff06..0466a32b220df7f5ca59022cbe084736536b038c 100644
--- a/source/renderer/ore/gl/ore_texture_gl.hpp
+++ b/source/renderer/ore/gl/ore_texture_gl.hpp
@@ -15,6 +15,9 @@ public:
   // GLES3 has no texture views, so a sampled view's mip range is texture
   // state. Expects the texture bound at `unit`.
   void setSampledLevels(uint32_t unit, uint32_t base, uint32_t max);
+  // Hands the whole mip chain back, so GL users outside ore never inherit a
+  // narrowed view's range. Binds and unbinds the texture on unit 0.
+  void restoreSampledLevels();
 
 private:
   friend class ContextGL;
主要语言
C++
星标
1.2k
派生
133
PR 合并指标
30 天内没有已合并 PR

环境准备

  • 提供 Dockerfile 或 Docker Compose 文件
  • 没有 Pull Request 模板
  • 没有贡献指南

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

rive-app/rive-runtime 的其他 Issue

查看 rive-app/rive-runtime 的全部 Issue

相似的 Issue

更多 C++ Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。