HDR功能初步实现#1300
Open
steven42121 wants to merge 2 commits into
Open
Conversation
![sourcery-ai[bot]](https://avatars.githubusercontent.com/in/48477?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Hey - 我发现了 1 个问题，并留下了一些总体反馈：
- HDR 色调映射工具（HdrSceneStats、曝光计算、tone_map_filmic、linear_to_srgb 等）在 FramePoolScreencap.cpp 和 ControllerAgent.cpp 中基本是重复的；建议把它们抽取到一个共享的辅助模块中，以避免两边逻辑偏移，并让将来的调优更轻松。
- 在 FramePoolScreencap::screencap() 中，switch 的 default 分支记录的是
texture_desc_.Format，而不是实际查询到的current_desc.Format；在重新创建纹理之后，前者可能已经过期。在日志中使用当前的描述符会让调试不支持的格式时更清晰。 - 多处使用了函数静态变量
std::chrono::steady_clock::time_point（例如 ControllerAgent 中的 HDR 日志节流以及 HDR 显示补偿路径）却没有做同步；如果截图处理可能在多个线程中进行，这会引入数据竞争。建议用互斥量进行保护，或者把节流状态移动到线程隔离的上下文中。
给 AI Agents 的提示
Please address the comments from this code review:
## Overall Comments
- The HDR tone mapping utilities (HdrSceneStats, exposure calculation, tone_map_filmic, linear_to_srgb, etc.) are essentially duplicated between FramePoolScreencap.cpp and ControllerAgent.cpp; consider extracting these into a shared helper to avoid drift and make future tuning easier.
- In FramePoolScreencap::screencap() the default case of the switch logs `texture_desc_.Format` instead of the actually queried `current_desc.Format`, which can be stale after recreating textures; using the current descriptor in the log will make debugging unsupported formats much clearer.
- Several places use function‑static `std::chrono::steady_clock::time_point` (e.g., HDR logging throttles in ControllerAgent and the HDR display compensation paths) without synchronization; if screenshots can be processed from multiple threads this introduces data races, so consider guarding these with a mutex or moving the throttling state to a thread‑confined context.
## Individual Comments
### Comment 1
<location path="source/MaaWin32ControlUnit/Screencap/FramePoolScreencap.cpp" line_range="667-668" />
<code_context>
+ case ScreencapMethod::ScreenDC:
+ return "ScreenDC";
+ case ScreencapMethod::UnknownYet:
+ default:
+ return "UnknownYet";
+ }
</code_context>
<issue_to_address>
**suggestion:** Log uses `texture_desc_` instead of the current frame’s `current_desc` when reporting an unsupported DXGI format.
This branch switches on `current_desc.Format`, but the log prints `texture_desc_.Format`. If the staging texture isn’t created yet or uses a different format, the log will show the wrong value and hinder debugging. Please log `current_desc.Format` here (or both, if you want to expose mismatches).
Suggested implementation:
```cpp
LOG_ERROR(
"Unsupported DXGI format: current_desc.Format={}, texture_desc_.Format={}",
static_cast<int>(current_desc.Format),
static_cast<int>(texture_desc_.Format));
```
If the logging macro or function name differs (e.g. `spdlog::error`, `LOGE`, `Logger::Error`, etc.), adjust the above replacement to match the existing logging style while keeping the two-format logging and the `static_cast<int>` conversions (or equivalent) to format the enum values.
</issue_to_address>请帮我变得更有用！请对每条评论点 👍 或 👎，我会根据你的反馈改进后续评审。
Original comment in English
Hey - I've found 1 issue, and left some high level feedback:
- The HDR tone mapping utilities (HdrSceneStats, exposure calculation, tone_map_filmic, linear_to_srgb, etc.) are essentially duplicated between FramePoolScreencap.cpp and ControllerAgent.cpp; consider extracting these into a shared helper to avoid drift and make future tuning easier.
- In FramePoolScreencap::screencap() the default case of the switch logs
texture_desc_.Formatinstead of the actually queriedcurrent_desc.Format, which can be stale after recreating textures; using the current descriptor in the log will make debugging unsupported formats much clearer. - Several places use function‑static
std::chrono::steady_clock::time_point(e.g., HDR logging throttles in ControllerAgent and the HDR display compensation paths) without synchronization; if screenshots can be processed from multiple threads this introduces data races, so consider guarding these with a mutex or moving the throttling state to a thread‑confined context.
Prompt for AI Agents
Please address the comments from this code review:
## Overall Comments
- The HDR tone mapping utilities (HdrSceneStats, exposure calculation, tone_map_filmic, linear_to_srgb, etc.) are essentially duplicated between FramePoolScreencap.cpp and ControllerAgent.cpp; consider extracting these into a shared helper to avoid drift and make future tuning easier.
- In FramePoolScreencap::screencap() the default case of the switch logs `texture_desc_.Format` instead of the actually queried `current_desc.Format`, which can be stale after recreating textures; using the current descriptor in the log will make debugging unsupported formats much clearer.
- Several places use function‑static `std::chrono::steady_clock::time_point` (e.g., HDR logging throttles in ControllerAgent and the HDR display compensation paths) without synchronization; if screenshots can be processed from multiple threads this introduces data races, so consider guarding these with a mutex or moving the throttling state to a thread‑confined context.
## Individual Comments
### Comment 1
<location path="source/MaaWin32ControlUnit/Screencap/FramePoolScreencap.cpp" line_range="667-668" />
<code_context>
+ case ScreencapMethod::ScreenDC:
+ return "ScreenDC";
+ case ScreencapMethod::UnknownYet:
+ default:
+ return "UnknownYet";
+ }
</code_context>
<issue_to_address>
**suggestion:** Log uses `texture_desc_` instead of the current frame’s `current_desc` when reporting an unsupported DXGI format.
This branch switches on `current_desc.Format`, but the log prints `texture_desc_.Format`. If the staging texture isn’t created yet or uses a different format, the log will show the wrong value and hinder debugging. Please log `current_desc.Format` here (or both, if you want to expose mismatches).
Suggested implementation:
```cpp
LOG_ERROR(
"Unsupported DXGI format: current_desc.Format={}, texture_desc_.Format={}",
static_cast<int>(current_desc.Format),
static_cast<int>(texture_desc_.Format));
```
If the logging macro or function name differs (e.g. `spdlog::error`, `LOGE`, `Logger::Error`, etc.), adjust the above replacement to match the existing logging style while keeping the two-format logging and the `static_cast<int>` conversions (or equivalent) to format the enum values.
</issue_to_address>Help me be more useful! Please click 👍 or 👎 on each comment and I'll use the feedback to improve your reviews.
Comment on lines
+667
to
+668
| default: | ||
| LogError << "Unsupported frame format" << VAR(texture_desc_.Format); |
Contributor
There was a problem hiding this comment.
suggestion: 日志在报告不支持的 DXGI 格式时，使用了 texture_desc_，而不是当前帧的 current_desc。
这个分支是根据 current_desc.Format 进行 switch 的，但日志打印的是 texture_desc_.Format。如果暂存纹理尚未创建或使用了不同的格式，日志中显示的值就会不正确，从而妨碍调试。请在这里记录 current_desc.Format（或者两者都记录，如果你希望暴露格式不匹配的问题）。
建议实现：
LOG_ERROR(
"Unsupported DXGI format: current_desc.Format={}, texture_desc_.Format={}",
static_cast<int>(current_desc.Format),
static_cast<int>(texture_desc_.Format));
如果日志宏或函数名不同（例如 spdlog::error、LOGE、Logger::Error 等），请根据现有的日志风格调整上述替换代码，同时保留对两种格式的记录以及用于格式化枚举值的 static_cast<int> 转换（或等价方式）。
Original comment in English
suggestion: Log uses texture_desc_ instead of the current frame’s current_desc when reporting an unsupported DXGI format.
This branch switches on current_desc.Format, but the log prints texture_desc_.Format. If the staging texture isn’t created yet or uses a different format, the log will show the wrong value and hinder debugging. Please log current_desc.Format here (or both, if you want to expose mismatches).
Suggested implementation:
LOG_ERROR(
"Unsupported DXGI format: current_desc.Format={}, texture_desc_.Format={}",
static_cast<int>(current_desc.Format),
static_cast<int>(texture_desc_.Format));
If the logging macro or function name differs (e.g. spdlog::error, LOGE, Logger::Error, etc.), adjust the above replacement to match the existing logging style while keeping the two-format logging and the static_cast<int> conversions (or equivalent) to format the enum values.
Contributor
There was a problem hiding this comment.
Pull request overview
该 PR 为 Win32 截图链路加入 HDR 相关能力：在检测到显示器开启 HDR 时，对传统 GDI/PrintWindow/ScreenDC 截图做 SDR 白点补偿；同时扩展 FramePool (WGC) 支持 HDR 浮点帧并提供 GPU 端预处理/色调映射，最终将 HDR/补偿状态上报到 Controller 层并输出调试探针图。
Changes:
- 新增
HdrDisplayUtils：查询 Windows 高级色彩/HDR 状态与 SDR 白点，并提供 HDR 显示补偿算法（CPU）。 - 传统截图方法（GDI/PrintWindow/ScreenDC）在 HDR 显示开启时对截图做补偿，并记录
ScreencapInfo。 - FramePool 截图支持 HDR 像素格式探测与 GPU 预处理管线；Win32ControlUnitMgr/ControllerAgent 增加 HDR 状态上报、通知与探针图保存。
Reviewed changes
Copilot reviewed 20 out of 20 changed files in this pull request and generated 5 comments.
Show a summary per file
| File | Description |
|---|---|
| source/MaaWin32ControlUnit/Screencap/ScreenDCScreencap.h | 增加 last_screencap_info() 与状态缓存字段。 |
| source/MaaWin32ControlUnit/Screencap/ScreenDCScreencap.cpp | HDR 显示开启时对 ScreenDC 截图做补偿并限频日志。 |
| source/MaaWin32ControlUnit/Screencap/PrintWindowWithPseudoMinimizeScreencap.h | 透传 last_screencap_info()。 |
| source/MaaWin32ControlUnit/Screencap/PrintWindowScreencap.h | 增加 last_screencap_info() 与状态缓存字段。 |
| source/MaaWin32ControlUnit/Screencap/PrintWindowScreencap.cpp | HDR 显示开启时对 PrintWindow 截图做补偿并限频日志。 |
| source/MaaWin32ControlUnit/Screencap/HwndUtils.hpp | 全屏/无边框时使用 rcMonitor 以允许覆盖任务栏区域。 |
| source/MaaWin32ControlUnit/Screencap/HdrDisplayUtils.hpp | 新增 HDR 查询/补偿接口与 HdrDisplayState 定义。 |
| source/MaaWin32ControlUnit/Screencap/HdrDisplayUtils.cpp | 实现 per-monitor HDR 状态查询、SDR 白点读取与补偿算法。 |
| source/MaaWin32ControlUnit/Screencap/GdiScreencap.h | 增加 last_screencap_info() 与状态缓存字段。 |
| source/MaaWin32ControlUnit/Screencap/GdiScreencap.cpp | HDR 显示开启时对 GDI 截图做补偿并限频日志。 |
| source/MaaWin32ControlUnit/Screencap/FramePoolWithPseudoMinimizeScreencap.h | 透传 last_screencap_info()。 |
| source/MaaWin32ControlUnit/Screencap/FramePoolScreencap.h | 增加 HDR GPU 处理相关成员与 last_screencap_info()。 |
| source/MaaWin32ControlUnit/Screencap/FramePoolScreencap.cpp | HDR/SDR 帧池探测、GPU 预处理/色调映射、缓存回退与包含从属窗口等增强。 |
| source/MaaWin32ControlUnit/Screencap/DesktopDupWindowScreencap.h | 增加 ScreenDC 回退与 last_screencap_info()。 |
| source/MaaWin32ControlUnit/Screencap/DesktopDupWindowScreencap.cpp | 多显示器跨度/越界等情况下回退到 ScreenDC。 |
| source/MaaWin32ControlUnit/Manager/Win32ControlUnitMgr.h | 记录活动截图方法与最近截图信息并提供方法名。 |
| source/MaaWin32ControlUnit/Manager/Win32ControlUnitMgr.cpp | 截图失败时重初始化截图方式；get_info() 上报 HDR/方法信息。 |
| source/MaaWin32ControlUnit/Base/UnitBase.h | 新增 ScreencapInfo 与 ScreencapBase::last_screencap_info() 默认实现。 |
| source/MaaFramework/Controller/ControllerAgent.h | 增加 HDR 模式通知/探针字段与若干互斥锁。 |
| source/MaaFramework/Controller/ControllerAgent.cpp | HDR 帧色调映射与探针图保存；改进通知构建与若干线程安全/边界处理。 |
| rgb *= scale; | ||
| } | ||
|
|
||
| return float4(linear_to_srgb(rgb.r), linear_to_srgb(rgb.g), linear_to_srgb(rgb.b), 1.0); |
There was a problem hiding this comment.
The tonemap shader writes float4(R,G,B,A) but the render target is created as DXGI_FORMAT_B8G8R8A8_UNORM. For BGRA render targets, the first component maps to B, so this will swap red/blue in GPU HDR output (and bgra_to_bgr() will preserve the swap). Output BGRA order (B,G,R,A) here, or change the output texture format to an RGBA render target format.
Suggested change
| return float4(linear_to_srgb(rgb.r), linear_to_srgb(rgb.g), linear_to_srgb(rgb.b), 1.0); | |
| return float4(linear_to_srgb(rgb.b), linear_to_srgb(rgb.g), linear_to_srgb(rgb.r), 1.0); |
Comment on lines
+31
to
+33
| constexpr LONG kDisplayConfigErrorInsufficientBuffer = 122; | ||
| constexpr int kMaxDisplayConfigRetries = 3; | ||
|
|
There was a problem hiding this comment.
kDisplayConfigErrorInsufficientBuffer and kMaxDisplayConfigRetries are declared here but never used in this file, which can trigger unused-variable warnings. Remove them or use the corresponding constants from the HDR query helper if needed.
Suggested change
| constexpr LONG kDisplayConfigErrorInsufficientBuffer = 122; | |
| constexpr int kMaxDisplayConfigRetries = 3; |
| return cached_image_; | ||
| } | ||
| default: | ||
| LogError << "Unsupported frame format" << VAR(texture_desc_.Format); |
There was a problem hiding this comment.
This error log reports texture_desc_.Format, but the switch is on current_desc.Format. If the staging texture desc differs (or texture_desc_ was reset), the log can be misleading—log current_desc.Format instead.
Suggested change
| LogError << "Unsupported frame format" << VAR(texture_desc_.Format); | |
| LogError << "Unsupported frame format" << VAR(current_desc.Format); |
| readable_texture_ = nullptr; | ||
| texture_desc_ = { 0 }; | ||
| if (!init_texture(texture)) { | ||
| LogError << "falied to init_texture"; |
There was a problem hiding this comment.
Typo in log message: "falied" should be "failed".
Suggested change
| LogError << "falied to init_texture"; | |
| LogError << "failed to init_texture"; |
Comment on lines
+61
to
+72
| bool device_name_equals(const wchar_t (&lhs)[CCHDEVICENAME], const DeviceNameBuffer& rhs) | ||
| { | ||
| for (size_t i = 0; i < CCHDEVICENAME; ++i) { | ||
| if (lhs[i] != rhs[i]) { | ||
| return false; | ||
| } | ||
| if (lhs[i] == L'\0') { | ||
| return true; | ||
| } | ||
| } | ||
| return true; | ||
| } |
There was a problem hiding this comment.
This overload of device_name_equals() is not referenced in this translation unit. Consider removing it to avoid unused-function warnings and reduce duplicate logic.
zmdyy0318
reviewed
Apr 28, 2026
| processed = std::move(hdr_bgr8); | ||
| } | ||
| else { | ||
| cv::resize(hdr_bgr8, processed, { target_width, target_height }, 0, 0, resize_method); |
zmdyy0318
reviewed
Apr 28, 2026
| resized = raw; | ||
| } | ||
| else { | ||
| cv::resize(raw, resized, { target_width, target_height }, 0, 0, resize_method); |
zmdyy0318
reviewed
Apr 28, 2026
| } | ||
|
|
||
| if (resized.type() == CV_8UC3) { | ||
| processed = resized.clone(); |
zmdyy0318
reviewed
Apr 28, 2026
| processed = resized.clone(); | ||
| } | ||
| else if (resized.type() == CV_8UC4) { | ||
| cv::cvtColor(resized, processed, cv::COLOR_BGRA2BGR); |
zmdyy0318
reviewed
Apr 28, 2026
|
|
||
| cv::Mat rgba32f; | ||
| if (hdr_rgba.type() == CV_32FC4) { | ||
| rgba32f = hdr_rgba; |
zmdyy0318
reviewed
Apr 28, 2026
| rgba32f = hdr_rgba; | ||
| } | ||
| else { | ||
| hdr_rgba.convertTo(rgba32f, CV_32FC4); |
zmdyy0318
reviewed
Apr 28, 2026
| stats.max_linear = std::numeric_limits<float>::lowest(); | ||
|
|
||
| std::vector<float> luminances; | ||
| luminances.reserve(static_cast<size_t>((rgba32f.rows / stride + 1) * (rgba32f.cols / stride + 1))); |
zmdyy0318
reviewed
Apr 28, 2026
| double log_sum = 0.0; | ||
| size_t sample_count = 0; | ||
|
|
||
| for (int y = 0; y < rgba32f.rows; y += stride) { |
zmdyy0318
reviewed
Apr 28, 2026
| } | ||
| } | ||
|
|
||
| cv::Mat bgr8(rgba32f.rows, rgba32f.cols, CV_8UC3); |
zmdyy0318
reviewed
Apr 28, 2026
| cv::Mat bgr8(rgba32f.rows, rgba32f.cols, CV_8UC3); | ||
|
|
||
| cv::parallel_for_(cv::Range(0, rgba32f.rows), [&](const cv::Range& range) { | ||
| for (int y = range.start; y < range.end; ++y) { |
zmdyy0318
reviewed
Apr 28, 2026
| rgba32f = hdr_rgba; | ||
| } | ||
| else { | ||
| hdr_rgba.convertTo(rgba32f, CV_32FC4); |
zmdyy0318
reviewed
Apr 28, 2026
| stats.max_linear = std::numeric_limits<float>::lowest(); | ||
|
|
||
| std::vector<float> luminances; | ||
| luminances.reserve(static_cast<size_t>(rgba32f.rows * rgba32f.cols)); |
zmdyy0318
reviewed
Apr 28, 2026
| size_t sample_count = 0; | ||
|
|
||
| for (int y = 0; y < rgba32f.rows; ++y) { | ||
| const auto* row = rgba32f.ptr<cv::Vec4f>(y); |
zmdyy0318
reviewed
Apr 28, 2026
| } | ||
| } | ||
|
|
||
| d3d_context_->CopyResource(readable_texture_.get(), texture.get()); |
zmdyy0318
reviewed
Apr 28, 2026
| d3d_context_->CopyResource(readable_texture_.get(), texture.get()); | ||
|
|
||
| D3D11_MAPPED_SUBRESOURCE mapped { 0 }; | ||
| ret = d3d_context_->Map(readable_texture_.get(), 0, D3D11_MAP_READ, 0, &mapped); |
zmdyy0318
reviewed
Apr 28, 2026
| case DXGI_FORMAT_B8G8R8A8_UNORM: { | ||
| cv::Mat raw(raw_height, raw_width, CV_8UC4, mapped.pData, mapped.RowPitch); | ||
| last_screencap_info_ = {}; | ||
| cached_image_ = bgra_to_bgr(raw(client_roi)); |
zmdyy0318
reviewed
Apr 28, 2026
| .display_hdr_active = true, | ||
| .display_hdr_compensated = false, | ||
| }; | ||
| cached_image_ = raw(client_roi).clone(); |
Author
|
我测了反正自动战斗可以用 |
Author
|
没办法全部做进显存里面的 |
zmdyy0318
reviewed
Apr 28, 2026
| d3d_context_->CopyResource(hdr_stats_staging_.get(), hdr_stats_texture_.get()); | ||
|
|
||
| D3D11_MAPPED_SUBRESOURCE mapped_stats {}; | ||
| HRESULT ret = d3d_context_->Map(hdr_stats_staging_.get(), 0, D3D11_MAP_READ, 0, &mapped_stats); |
zmdyy0318
reviewed
Apr 28, 2026
| } | ||
| OnScopeLeave([&]() { d3d_context_->Unmap(hdr_stats_staging_.get(), 0); }); | ||
|
|
||
| cv::Mat hdr_stats_mat(hdr_gpu_stats_size_.height, hdr_gpu_stats_size_.width, CV_16FC4, mapped_stats.pData, mapped_stats.RowPitch); |
Author
|
没事不行就关了 |
zmdyy0318
reviewed
Apr 28, 2026
| d3d_context_->OMSetRenderTargets(1, stats_rtv, nullptr); | ||
| d3d_context_->PSSetShader(hdr_copy_pixel_shader_.get(), nullptr, 0); | ||
| d3d_context_->Draw(3, 0); | ||
| d3d_context_->CopyResource(hdr_stats_staging_.get(), hdr_stats_texture_.get()); |
zmdyy0318
reviewed
Apr 28, 2026
| ID3D11ShaderResourceView* null_shader_resources[] = { nullptr }; | ||
| d3d_context_->PSSetShaderResources(0, 1, null_shader_resources); | ||
|
|
||
| d3d_context_->CopyResource(hdr_output_staging_.get(), hdr_output_texture_.get()); |
zmdyy0318
reviewed
Apr 28, 2026
| d3d_context_->CopyResource(hdr_output_staging_.get(), hdr_output_texture_.get()); | ||
|
|
||
| D3D11_MAPPED_SUBRESOURCE mapped_output {}; | ||
| ret = d3d_context_->Map(hdr_output_staging_.get(), 0, D3D11_MAP_READ, 0, &mapped_output); |
Author
|
主要是想找点人试试有，看转换能不能行 |
zmdyy0318
reviewed
Apr 28, 2026
| OnScopeLeave([&]() { d3d_context_->Unmap(hdr_output_staging_.get(), 0); }); | ||
|
|
||
| cv::Mat gpu_bgra(hdr_gpu_output_size_.height, hdr_gpu_output_size_.width, CV_8UC4, mapped_output.pData, mapped_output.RowPitch); | ||
| output = bgra_to_bgr(gpu_bgra); |
Author
|
因为每个人颜色管理不一样的 |
zmdyy0318
reviewed
Apr 28, 2026
| return false; | ||
| } | ||
|
|
||
| d3d_context_->CopyResource(hdr_shader_input_texture_.get(), texture.get()); |
zmdyy0318
reviewed
Apr 28, 2026
| d3d_context_->RSSetViewports(1, &output_viewport); | ||
| d3d_context_->OMSetRenderTargets(1, output_rtv, nullptr); | ||
| d3d_context_->PSSetShader(hdr_tonemap_pixel_shader_.get(), nullptr, 0); | ||
| d3d_context_->Draw(3, 0); |
Contributor
|
在视频转换项目里，做多次内存拷贝、显存拷贝，大不了时间长一点无所谓，时长x2 x3用户感知也不明显，多等一会就是了。 |
Author
|
其实没有这个的原生Maaend 147kf的CPU占用也有4%（ |
Member
|
🤖 附带的线程安全修复、swipe 插值、null guard 这些都不错，但 HDR 核心路径有几个问题：
如果核心问题（1）能有数据佐证的话，后面几个改一下我们可以继续谈 |
Member
|
🤖 MistEO 那边有个 #1105 在做类似的事，等他回来一起看看方向 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
是Add，不影响现在非HDR的功能实现。
采用了转换算法，延迟和CPU占用控制在合理范围啦！
实测整个Maaend 5%CPU占用(147kf)
延迟6~20ms吧
Summary by Sourcery
为 Win32 控制单元和控制器新增支持 HDR 的截图管线，包括色调映射、HDR 显示补偿以及按显示器检测 HDR，同时保持现有 SDR 路径不变。
New Features:
Bug Fixes:
Enhancements:
Original summary in English
Summary by Sourcery
Add HDR-aware screenshot pipeline for Win32 control unit and controller, including tone mapping, HDR display compensation, and per-monitor HDR detection while keeping existing SDR paths intact.
New Features:
Bug Fixes:
Enhancements: