diff --git a/docs/common/ai/cubie/_g2d-usage-guide.mdx b/docs/common/ai/cubie/_g2d-usage-guide.mdx
new file mode 100644
index 000000000..a7817aea2
--- /dev/null
+++ b/docs/common/ai/cubie/_g2d-usage-guide.mdx
@@ -0,0 +1,236 @@
+## 概述
+
+G2D 是 Allwinner SoC 集成的 2D 图形硬件加速器，负责图像旋转、缩放、格式转换、颜色填充等操作。
+
+典型应用场景：
+
+- 视频编解码前后的图像预处理（缩放、色彩空间转换）
+- 相机实时预览的图像处理
+- 显示输出前的格式转换（RGB ↔ YUV）
+- 批量图像处理流水线
+
+**本项目环境：**
+
+| 项目 | 值 |
+| ------------ | ---------------------------------- |
+| SoC | Allwinner A733 |
+| Linux | 5.15.147-100-a733 |
+| G2D 驱动版本 | 1.0.0 |
+| 驱动模块 | `g2d_sunxi` |
+| 设备节点 | `/dev/g2d`、`/dev/dma_heap/system` |
+
+## 环境准备
+
+### 验证驱动状态
+
+
+
+```bash
+# 检查驱动模块是否已加载
+lsmod | grep g2d
+# 输出类似: g2d_sunxi 90112 0
+
+# 查看驱动版本
+cat /sys/module/g2d_sunxi/version
+# 输出: 1.0.0
+```
+
+
+
+若未加载，手动加载：
+
+
+
+```bash
+sudo modprobe g2d_sunxi
+```
+
+
+
+### 设备节点权限
+
+当前系统已配置为无需 root 运行：
+
+```text
+/dev/g2d (0666)
+/dev/dma_heap/system (0666)
+```
+
+如需重新配置 udev 规则：
+
+
+
+```bash
+sudo sh -c 'echo "KERNEL==\"system\", SUBSYSTEM==\"dma_heap\", MODE=\"0666\"" > /etc/udev/rules.d/99-dma-heap.rules'
+sudo udevadm control --reload-rules
+sudo udevadm trigger
+```
+
+
+
+### 头文件
+
+G2D API 头文件位于：
+
+```text
+/usr/include/bsp/linux/sunxi-g2d.h
+```
+
+## 快速启动
+
+### 核心概念
+
+使用 G2D 的标准流程：
+
+```text
+1. 分配 DMA buffer（图像数据缓冲区）
+2. 填充源图像数据
+3. 配置 g2d_blit_h 结构体
+4. 调用 ioctl(G2D_CMD_BITBLT_H, ...)
+5. 从目标 DMA buffer 读取结果
+6. 释放资源
+```
+
+**关键：G2D 操作的是 DMA buffer，不是普通内存。**
+
+DMA buffer 由 `/dev/dma_heap/system` 分配，物理地址连续，硬件可直接访问。
+
+### 示例（旋转 90°）
+
+
+
+```c
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+
+#define W 1920
+#define H 1080
+
+// 分配 DMA buffer，返回 fd 和虚拟地址
+static int alloc_dmabuf(int *fd, void **vaddr, size_t size)
+{
+ struct dma_heap_allocation_data alloc_data = {
+ .len = size, .fd_flags = O_RDWR | O_CLOEXEC, .heap_flags = 0,
+ };
+ int heap_fd = open("/dev/dma_heap/system", O_RDONLY);
+ if (heap_fd < 0) return -1;
+ if (ioctl(heap_fd, DMA_HEAP_IOCTL_ALLOC, &alloc_data) < 0) {
+ close(heap_fd); return -1;
+ }
+ close(heap_fd);
+ *fd = alloc_data.fd;
+ *vaddr = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED, *fd, 0);
+ if (*vaddr == MAP_FAILED) { close(*fd); return -1; }
+ return 0;
+}
+
+int main(void)
+{
+ int g2d_fd, src_fd, dst_fd;
+ void *src_v, *dst_v;
+ g2d_blt_h blit;
+
+ // 1. 分配两个 DMA buffer（源和目标）
+ alloc_dmabuf(&src_fd, &src_v, W * H * 4);
+ alloc_dmabuf(&dst_fd, &dst_v, W * H * 4);
+
+ // 2. 填充源图像数据（渐变示例）
+ fill_pattern(src_v, W, H);
+
+ // 3. 打开 G2D 设备
+ g2d_fd = open("/dev/g2d", O_RDWR);
+
+ // 4. 配置操作参数
+ memset(&blit, 0, sizeof(blit));
+ blit.flag_h = G2D_ROT_90; // 旋转 90°
+
+ blit.src_image_h.fd = src_fd;
+ blit.src_image_h.format = G2D_FORMAT_ARGB8888;
+ blit.src_image_h.width = W;
+ blit.src_image_h.height = H;
+
+ blit.dst_image_h.fd = dst_fd;
+ blit.dst_image_h.format = G2D_FORMAT_ARGB8888;
+ blit.dst_image_h.width = H; // 旋转后宽高互换
+ blit.dst_image_h.height = W;
+
+ // 5. 执行硬件加速操作
+ ioctl(g2d_fd, G2D_CMD_BITBLT_H, (unsigned long)(&blit));
+
+ // 6. 结果已在 dst_v 中，验证或提交给后续流程
+
+ // 7. 释放资源
+ close(g2d_fd);
+ munmap(src_v, W * H * 4); close(src_fd);
+ munmap(dst_v, W * H * 4); close(dst_fd);
+ return 0;
+}
+```
+
+
+
+编译
+
+
+
+```bash
+sudo apt install gcc g++ cmake
+gcc -o g2d_rotation g2d_rotation.c
+```
+
+
+
+## 示例运行结果
+
+| 示例 | 操作 | 结果 |
+| ----------------- | --------------------------- | -------------------------- |
+| `g2d_rotation` | ARGB8888 1920×1080 旋转 90° | **8.24 ms**，251.8 MP/sec |
+| `g2d_format_conv` | ARGB8888 → YUV420 1920×1080 | **9.56 ms**，216.8 MP/sec |
+| `g2d_scaler` | 4096×4096 → 1920×1080 缩放 | **68.60 ms**，244.6 MP/sec |
+| `g2d_color_fill` | 1920×1080 纯色填充 | **8.53 ms**，243.2 MP/sec |
+
+注：`g2d_scaler` 源图分辨率上限为 4096×4096（8192×8192 会触发 `EPERM`）。
+
+## API 参考
+
+完整 API 定义和所有支持格式见头文件：
+
+```text
+/usr/include/bsp/linux/sunxi-g2d.h
+```
+
+常用 ioctl 命令：
+
+| 命令 | 用途 |
+| -------------------- | ------------------------------------ |
+| `G2D_CMD_BITBLT_H` | 单图像位块传输（旋转/缩放/格式转换） |
+| `G2D_CMD_FILLRECT_H` | 颜色填充矩形 |
+| `G2D_CMD_STRETCHBLT` | 伸缩位块传输 |
+| `G2D_CMD_BLD_H` | Porter-Duff 混合操作 |
+| `G2D_CMD_MIXER_TASK` | 批量任务（一次提交多个操作） |
+
+支持格式（部分）：
+
+| 格式 | 说明 |
+| ------------------------------- | --------------- |
+| `G2D_FORMAT_ARGB8888` | 32bpp Alpha-RGB |
+| `G2D_FORMAT_RGB888` | 24bpp RGB |
+| `G2D_FORMAT_RGB565` | 16bpp |
+| `G2D_FORMAT_YUV420UVC_U1V1U0V0` | NV12 标准格式 |
+
+旋转标志：`G2D_ROT_0` / `G2D_ROT_90` / `G2D_ROT_180` / `G2D_ROT_270` / `G2D_ROT_H`（水平翻转）/ `G2D_ROT_V`（垂直翻转）
+
+## 注意事项
+
+- **DMA buffer 必须物理连续**，普通 `malloc` 不能用于 G2D
+- **IOMMU 负责地址翻译**，G2D 访问的是 IOMMU 映射后的物理地址，无需关心具体地址值
+- **DMA buffer 通过 fd 传递**，用户空间用 `mmap` 后的虚拟地址读写
+- **目标分辨率应与源配合**，旋转 90° 时宽高互换
+- **格式转换**（如 RGB → YUV）由 G2D 硬件完成，源和目标 format 字段分别设置即可
+- **调试时**，失败返回 `-1`，`errno` 记录具体原因，用 `perror()` 打印
diff --git a/docs/common/ai/rockchip/_rga_usage_guide.mdx b/docs/common/ai/rockchip/_rga_usage_guide.mdx
new file mode 100644
index 000000000..fc2081e00
--- /dev/null
+++ b/docs/common/ai/rockchip/_rga_usage_guide.mdx
@@ -0,0 +1,416 @@
+## RGA 简介
+
+RGA (Raster Graphic Acceleration) 是 Rockchip 提供的 2D 图像处理硬件加速器，专门用于加速图像的缩放、颜色转换、格式转换等操作。在 RK3588 等芯片上可用。
+
+**典型优势：**
+
+- 硬件加速，CPU 零开销
+- 高效的图像缩放 (resize)
+- 色彩空间转换 (BGR↔RGB, RGB↔YUV)
+- 支持 DMA 零拷贝
+
+**适用场景：**
+
+- YOLO 等目标检测的图像预处理
+- 视频/图像的格式转换
+- 相机 pipeline 的图像处理
+
+## RGA 快速上手
+
+### 克隆仓库
+
+
+
+```bash
+git clone https://github.com/airockchip/librga.git
+```
+
+
+
+### 目录结构
+
+
+
+```bash
+./
+├── CHANGELOG.md
+├── COPYING
+├── docs
+│ ├── RGA_FAQ.assets
+│ ├── Rockchip_Developer_Guide_RGA_CN.md
+│ ├── Rockchip_Developer_Guide_RGA_EN.md
+│ ├── Rockchip_FAQ_RGA_CN.md
+│ └── Rockchip_FAQ_RGA_EN.md
+├── include
+│ ├── drmrga.h
+│ ├── GrallocOps.h
+│ ├── im2d_buffer.h
+│ ├── im2d_common.h
+│ ├── im2d_expand.h
+│ ├── im2d.h
+│ ├── im2d.hpp
+│ ├── im2d_mpi.h
+│ ├── im2d_single.h
+│ ├── im2d_task.h
+│ ├── im2d_type.h
+│ ├── im2d_version.h
+│ ├── RgaApi.h
+│ ├── rga.h
+│ ├── RgaMutex.h
+│ ├── RgaSingleton.h
+│ ├── RgaUtils.h
+│ └── RockchipRga.h
+├── libs
+│ ├── AndroidNdk
+│ └── Linux
+├── README.md
+├── samples
+│ ├── allocator_demo
+│ ├── alpha_demo
+│ ├── async_demo
+│ ├── build
+│ ├── cmake-android.sh
+│ ├── cmake-linux.sh
+│ ├── CMakeLists.txt
+│ ├── config_demo
+│ ├── copy_demo
+│ ├── crop_demo
+│ ├── cvtcolor_demo
+│ ├── fill_demo
+│ ├── gauss_demo
+│ ├── im2d_api_demo
+│ ├── mosaic_demo
+│ ├── padding_demo
+│ ├── palette_demo
+│ ├── README.md
+│ ├── resize_demo
+│ ├── rop_demo
+│ ├── sample_file
+│ ├── SConscript
+│ ├── toolchain_local.cmake
+│ ├── transform_demo
+│ └── utils
+├── toolchains
+│ ├── toolchain_android_ndk.cmake
+│ ├── toolchain_linux_1106.cmake
+│ ├── toolchain_linux.cmake
+│ └── toolchain_rt_thread.cmake
+└── tools
+ └── bin
+```
+
+
+
+### 运行 Demo
+
+
+
+```bash
+cd librga/tools/bin/Linux/gcc-aarch64
+./rgaImDemo -h
+```
+
+
+
+
+
+```bash
+./rgaImDemo --copy
+```
+
+
+
+
+
+```bash
+./rgaImDemo --resize up
+```
+
+
+
+### 头文件与库文件
+
+**头文件引用：**
+
+- RGA 头文件位置: `librga/include/`
+- C++ 调用 im2d API：`im2d.hpp`
+- C 调用 im2d API：`im2d.h`
+
+**库文件：**
+
+- RGA 库文件位置： `librga/libs/Linux/gcc-aarch64/`
+- `librga.so`
+- `librga.a`
+
+## 核心概念
+
+### Buffer 管理
+
+RGA 操作的是 buffer，需要先导入再用。
+
+
+
+```cpp
+#include
+#include
+
+// 1. 分配用户态内存
+uint8_t* src_buf = malloc(width * height * 3);
+
+// 2. 导入到 RGA (虚拟地址方式)
+rga_buffer_handle_t src_handle = importbuffer_virtualaddr(src_buf, buf_size);
+
+// 3. 包装成 rga_buffer_t
+rga_buffer_t src_img = wrapbuffer_handle(src_handle, width, height, RK_FORMAT_BGR_888);
+
+// 4. 释放
+releasebuffer_handle(src_handle);
+free(src_buf);
+```
+
+
+
+### 常用格式
+
+| 格式 | 说明 |
+| ------------------------ | ----------- |
+| `RK_FORMAT_BGR_888` | 3 通道 BGR |
+| `RK_FORMAT_RGB_888` | 3 通道 RGB |
+| `RK_FORMAT_RGBA_8888` | 4 通道 RGBA |
+| `RK_FORMAT_YCbCr_420_SP` | NV12 |
+
+### API 概览
+
+| 操作 | API | 说明 |
+| -------- | -------------- | -------------------- |
+| 缩放 | `imresize()` | 图像缩放 |
+| 颜色转换 | `imcvtcolor()` | BGR↔RGB, RGB↔YUV |
+| 填充边框 | `imfill()` | 矩形区域填充 |
+| 综合处理 | `improcess()` | crop+resize+格式转换 |
+| 内存拷贝 | `imcopy()` | buffer 拷贝 |
+
+## 代码示例
+
+### 简单示例：图像缩放
+
+
+
+```cpp
+#include
+#include
+
+void resize_example() {
+ int src_w = 1920, src_h = 1080;
+ int dst_w = 640, dst_h = 640;
+
+ // 分配内存
+ uint8_t* src = malloc(src_w * src_h * 3);
+ uint8_t* dst = malloc(dst_w * dst_h * 3);
+
+ // 导入 buffer
+ rga_buffer_handle_t src_handle = importbuffer_virtualaddr(src, src_w * src_h * 3);
+ rga_buffer_handle_t dst_handle = importbuffer_virtualaddr(dst, dst_w * dst_h * 3);
+
+ // 包装
+ rga_buffer_t src_img = wrapbuffer_handle(src_handle, src_w, src_h, RK_FORMAT_BGR_888);
+ rga_buffer_t dst_img = wrapbuffer_handle(dst_handle, dst_w, dst_h, RK_FORMAT_BGR_888);
+
+ // 缩放
+ IM_STATUS ret = imresize(src_img, dst_img);
+
+ // 释放
+ releasebuffer_handle(src_handle);
+ releasebuffer_handle(dst_handle);
+ free(src);
+ free(dst);
+}
+```
+
+
+
+### 颜色转换
+
+
+
+```cpp
+// BGR -> RGB
+imcvtcolor(src_img, dst_img, RK_FORMAT_BGR_888, RK_FORMAT_RGB_888);
+
+// BGR -> YUV (NV12)
+imcvtcolor(src_img, dst_img, RK_FORMAT_BGR_888, RK_FORMAT_YCbCr_420_SP);
+```
+
+
+
+### 填充背景
+
+
+
+```cpp
+// 填充灰色矩形区域
+rga_buffer_t canvas = wrapbuffer_handle(handle, width, height, RK_FORMAT_BGR_888);
+
+// 灰色: B=114, G=114, R=114 (按 BGR 顺序)
+int gray = (114 << 16) | (114 << 8) | 114;
+
+// 填充整个图像为灰色
+imfill(canvas, {0, 0, width, height}, gray);
+```
+
+
+
+### 综合处理 improcess()
+
+`improcess()` 是最强大的 API，可以同时完成：
+
+- 指定源区域 (crop)
+- 缩放到目标区域
+- 颜色格式转换
+
+
+
+```cpp
+// 语法
+improcess(src_img, dst_img, {}, srect, drect, {}, -1, nullptr, nullptr, IM_SYNC);
+
+// 参数说明:
+// src_img, dst_img: 源和目标 buffer
+// {}: 第三个参数是 pat (pattern)，一般不用
+// srect: 源区域 {x, y, width, height}
+// drect: 目标区域 {x, y, width, height}
+// {}: prect，一般不用
+// -1: acquire_fence_fd
+// nullptr: release_fence_fd
+// nullptr: opt_ptr
+// IM_SYNC: 同步执行
+```
+
+
+
+## YOLO 预处理中的使用
+
+### 问题分析
+
+YOLO 预处理需要：
+
+1. 保持纵横比的缩放 (letterbox)
+2. 灰色填充边缘
+3. BGR → RGB 转换
+
+### 完整实现
+
+
+
+```cpp
+int preprocess_with_rga(const cv::Mat& img0, int orig_w, int orig_h,
+ int target_size, float ratio, int dw, int dh,
+ std::vector& input_data, PerfStats& perf) {
+ Timer timer;
+ double t_padding = 0, t_resize = 0;
+
+ int new_unpad_w = static_cast(orig_w * ratio);
+ int new_unpad_h = static_cast(orig_h * ratio);
+
+ // RGA 要求 stride 和 width 像素数都按 16 对齐
+ // BGR_888: stride_bytes 必须是 16 的倍数，且 width_px 也必须是 16 的倍数
+ int src_width_px = ((orig_w + 15) / 16) * 16;
+ int dst_width_px = ((target_size + 15) / 16) * 16;
+ int src_stride = src_width_px * 3;
+ int dst_stride = dst_width_px * 3;
+
+ int src_buf_size = src_stride * orig_h;
+ int dst_buf_size = dst_stride * target_size;
+
+ std::vector src_buf(src_buf_size);
+ std::vector dst_buf(dst_buf_size);
+
+ // 复制原图数据 (逐行拷贝，保留行间距)
+ for (int h = 0; h < orig_h; ++h) {
+ memcpy(src_buf.data() + h * src_stride, img0.data + h * orig_w * 3, orig_w * 3);
+ }
+
+ // Step 1: RGA 填充灰色背景
+ {
+ rga_buffer_handle_t canvas_handle = importbuffer_virtualaddr(dst_buf.data(), dst_width_px, target_size, RK_FORMAT_BGR_888);
+ if (canvas_handle == 0) {
+ std::cerr << "Warning: RGA importbuffer failed for canvas\n";
+ return -1;
+ }
+
+ rga_buffer_t canvas = wrapbuffer_handle(canvas_handle, dst_width_px, target_size, RK_FORMAT_BGR_888);
+
+ int gray_color = (114 << 16) | (114 << 8) | 114;
+
+ timer.reset();
+ IM_STATUS ret = imfill(canvas, {0, 0, dst_width_px, target_size}, gray_color);
+ t_padding = timer.elapsed_ms();
+
+ if (ret != IM_STATUS_SUCCESS) {
+ std::cerr << "Warning: RGA imfill failed: " << imStrError(ret) << "\n";
+ releasebuffer_handle(canvas_handle);
+ return -1;
+ }
+
+ releasebuffer_handle(canvas_handle);
+ }
+
+ // Step 2: RGA improcess - 缩放原图到中间区域 + BGR to RGB
+ {
+ rga_buffer_handle_t src_handle = importbuffer_virtualaddr(src_buf.data(), src_width_px, orig_h, RK_FORMAT_BGR_888);
+ rga_buffer_handle_t dst_handle = importbuffer_virtualaddr(dst_buf.data(), dst_width_px, target_size, RK_FORMAT_RGB_888);
+
+ if (src_handle == 0 || dst_handle == 0) {
+ std::cerr << "Warning: RGA importbuffer failed\n";
+ return -1;
+ }
+
+ rga_buffer_t src_img = wrapbuffer_handle(src_handle, src_width_px, orig_h, RK_FORMAT_BGR_888);
+ rga_buffer_t dst_img = wrapbuffer_handle(dst_handle, dst_width_px, target_size, RK_FORMAT_RGB_888);
+
+ im_rect srect = {0, 0, orig_w, orig_h};
+ im_rect drect = {dw, dh, new_unpad_w, new_unpad_h};
+
+ timer.reset();
+ IM_STATUS ret = improcess(src_img, dst_img, {}, srect, drect, {}, -1, nullptr, nullptr, IM_SYNC);
+ t_resize = timer.elapsed_ms();
+
+ if (ret != IM_STATUS_SUCCESS) {
+ std::cerr << "Warning: RGA improcess failed: " << imStrError(ret) << "\n";
+ releasebuffer_handle(src_handle);
+ releasebuffer_handle(dst_handle);
+ return -1;
+ }
+
+ releasebuffer_handle(src_handle);
+ releasebuffer_handle(dst_handle);
+ }
+
+ // 复制到输出
+ input_data = std::move(dst_buf);
+
+ perf.preprocess = t_padding + t_resize;
+ return 0;
+}
+```
+
+
+
+## 性能对比
+
+```bash
+================================================
+ BENCHMARK RESULTS
+================================================
+
+Metric OpenCV RGA Speedup
+--------------------------------------------------
+Preprocess (ms) 2.01 2.42 .83x
+Inference (ms) 63.62 63.81
+TOTAL (ms) 128.30 126.93
+--------------------------------------------------
+
+Iterations: 10 per version
+Test completed.
+```
+
+可以看到在预处理非常简单的前提下，直接用 OpenCV 处理和 RGA 处理差异不大。
diff --git a/docs/cubie/a5e/app-dev/npu-dev/g2d-usage-guide.md b/docs/cubie/a5e/app-dev/npu-dev/g2d-usage-guide.md
new file mode 100644
index 000000000..3b00a98bf
--- /dev/null
+++ b/docs/cubie/a5e/app-dev/npu-dev/g2d-usage-guide.md
@@ -0,0 +1,14 @@
+---
+sidebar_position: 4.5
+
+doc_kind: wrapper
+source_of_truth: common
+imports_resolve_to:
+ - docs/common/ai/cubie/_g2d-usage-guide.mdx
+---
+
+# G2D 使用指南
+
+import G2D from '../../../../common/ai/cubie/\_g2d-usage-guide.mdx';
+
+
diff --git a/docs/cubie/a7a/app-dev/npu-dev/g2d-usage-guide.md b/docs/cubie/a7a/app-dev/npu-dev/g2d-usage-guide.md
new file mode 100644
index 000000000..3b00a98bf
--- /dev/null
+++ b/docs/cubie/a7a/app-dev/npu-dev/g2d-usage-guide.md
@@ -0,0 +1,14 @@
+---
+sidebar_position: 4.5
+
+doc_kind: wrapper
+source_of_truth: common
+imports_resolve_to:
+ - docs/common/ai/cubie/_g2d-usage-guide.mdx
+---
+
+# G2D 使用指南
+
+import G2D from '../../../../common/ai/cubie/\_g2d-usage-guide.mdx';
+
+
diff --git a/docs/cubie/a7s/app-dev/npu-dev/g2d-usage-guide.md b/docs/cubie/a7s/app-dev/npu-dev/g2d-usage-guide.md
new file mode 100644
index 000000000..3b00a98bf
--- /dev/null
+++ b/docs/cubie/a7s/app-dev/npu-dev/g2d-usage-guide.md
@@ -0,0 +1,14 @@
+---
+sidebar_position: 4.5
+
+doc_kind: wrapper
+source_of_truth: common
+imports_resolve_to:
+ - docs/common/ai/cubie/_g2d-usage-guide.mdx
+---
+
+# G2D 使用指南
+
+import G2D from '../../../../common/ai/cubie/\_g2d-usage-guide.mdx';
+
+
diff --git a/docs/cubie/a7z/app-dev/npu-dev/g2d-usage-guide.md b/docs/cubie/a7z/app-dev/npu-dev/g2d-usage-guide.md
new file mode 100644
index 000000000..3b00a98bf
--- /dev/null
+++ b/docs/cubie/a7z/app-dev/npu-dev/g2d-usage-guide.md
@@ -0,0 +1,14 @@
+---
+sidebar_position: 4.5
+
+doc_kind: wrapper
+source_of_truth: common
+imports_resolve_to:
+ - docs/common/ai/cubie/_g2d-usage-guide.mdx
+---
+
+# G2D 使用指南
+
+import G2D from '../../../../common/ai/cubie/\_g2d-usage-guide.mdx';
+
+
diff --git a/docs/rock5/rock5b/app-development/rga-usage-guide.md b/docs/rock5/rock5b/app-development/rga-usage-guide.md
new file mode 100644
index 000000000..c98ccb70e
--- /dev/null
+++ b/docs/rock5/rock5b/app-development/rga-usage-guide.md
@@ -0,0 +1,15 @@
+---
+sidebar_position: 5
+description: "RGA 2D 图像加速开发指南"
+
+doc_kind: wrapper
+source_of_truth: common
+imports_resolve_to:
+ - docs/common/ai/rockchip/_rga_usage_guide.mdx
+---
+
+import RGA from '../../../common/ai/rockchip/\_rga_usage_guide.mdx';
+
+# RGA 使用指南
+
+
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_g2d-usage-guide.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_g2d-usage-guide.mdx
new file mode 100644
index 000000000..efedbe332
--- /dev/null
+++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_g2d-usage-guide.mdx
@@ -0,0 +1,238 @@
+## Overview
+
+G2D is a 2D graphics hardware accelerator integrated into Allwinner SoCs, responsible for image rotation, scaling, format conversion, color filling and other operations.
+
+Typical application scenarios:
+
+- Image preprocessing (scaling, color space conversion) before and after video encoding/decoding
+- Camera real-time preview image processing
+- Format conversion (RGB ↔ YUV) before display output
+- Batch image processing pipeline
+
+**Project Environment:**
+
+| Item | Value |
+| ------------------ | ---------------------------------- |
+| SoC | Allwinner A733 |
+| Linux | 5.15.147-100-a733 |
+| G2D Driver Version | 1.0.0 |
+| Driver Module | `g2d_sunxi` |
+| Device Node | `/dev/g2d`, `/dev/dma_heap/system` |
+
+## Environment Setup
+
+### Verify Driver Status
+
+
+
+```bash
+# Check if driver module is loaded
+lsmod | grep g2d
+# Output example: g2d_sunxi 90112 0
+
+# Check driver version
+cat /sys/module/g2d_sunxi/version
+# Output: 1.0.0
+```
+
+
+
+If not loaded, manually load:
+
+
+
+```bash
+sudo modprobe g2d_sunxi
+```
+
+
+
+### Device Node Permissions
+
+Current system is configured to run without root:
+
+```text
+/dev/g2d (0666)
+/dev/dma_heap/system (0666)
+```
+
+To reconfigure udev rules if needed:
+
+
+
+```bash
+sudo sh -c 'echo "KERNEL==\"system\", SUBSYSTEM==\"dma_heap\", MODE=\"0666\"" > /etc/udev/rules.d/99-dma-heap.rules'
+sudo udevadm control --reload-rules
+sudo udevadm trigger
+```
+
+
+
+### Header Files
+
+G2D API header file location:
+
+```text
+/usr/include/bsp/linux/sunxi-g2d.h
+```
+
+## Quick Start
+
+### Core Concepts
+
+Standard workflow for using G2D:
+
+```text
+1. Allocate DMA buffer (image data buffer)
+2. Fill source image data
+3. Configure g2d_blit_h structure
+4. Call ioctl(G2D_CMD_BITBLT_H, ...)
+5. Read result from target DMA buffer
+6. Release resources
+```
+
+**Key: G2D operates on DMA buffer, not regular memory.**
+
+DMA buffer is allocated by `/dev/dma_heap/system`, physically contiguous, hardware can access directly.
+
+### Example (Rotate 90°)
+
+Complete runnable code see `g2d_rotation_or_mirror.c`.
+
+
+
+```c
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+#include
+
+#define W 1920
+#define H 1080
+
+// Allocate DMA buffer, return fd and virtual address
+static int alloc_dmabuf(int *fd, void **vaddr, size_t size)
+{
+ struct dma_heap_allocation_data alloc_data = {
+ .len = size, .fd_flags = O_RDWR | O_CLOEXEC, .heap_flags = 0,
+ };
+ int heap_fd = open("/dev/dma_heap/system", O_RDONLY);
+ if (heap_fd < 0) return -1;
+ if (ioctl(heap_fd, DMA_HEAP_IOCTL_ALLOC, &alloc_data) < 0) {
+ close(heap_fd); return -1;
+ }
+ close(heap_fd);
+ *fd = alloc_data.fd;
+ *vaddr = mmap(NULL, size, PROT_READ | PROT_WRITE, MAP_SHARED, *fd, 0);
+ if (*vaddr == MAP_FAILED) { close(*fd); return -1; }
+ return 0;
+}
+
+int main(void)
+{
+ int g2d_fd, src_fd, dst_fd;
+ void *src_v, *dst_v;
+ g2d_blt_h blit;
+
+ // 1. Allocate two DMA buffers (source and destination)
+ alloc_dmabuf(&src_fd, &src_v, W * H * 4);
+ alloc_dmabuf(&dst_fd, &dst_v, W * H * 4);
+
+ // 2. Fill source image data (gradient example)
+ fill_pattern(src_v, W, H);
+
+ // 3. Open G2D device
+ g2d_fd = open("/dev/g2d", O_RDWR);
+
+ // 4. Configure operation parameters
+ memset(&blit, 0, sizeof(blit));
+ blit.flag_h = G2D_ROT_90; // Rotate 90°
+
+ blit.src_image_h.fd = src_fd;
+ blit.src_image_h.format = G2D_FORMAT_ARGB8888;
+ blit.src_image_h.width = W;
+ blit.src_image_h.height = H;
+
+ blit.dst_image_h.fd = dst_fd;
+ blit.dst_image_h.format = G2D_FORMAT_ARGB8888;
+ blit.dst_image_h.width = H; // Width and height swap after rotation
+ blit.dst_image_h.height = W;
+
+ // 5. Execute hardware accelerated operation
+ ioctl(g2d_fd, G2D_CMD_BITBLT_H, (unsigned long)(&blit));
+
+ // 6. Result is already in dst_v, verify or submit to subsequent process
+
+ // 7. Release resources
+ close(g2d_fd);
+ munmap(src_v, W * H * 4); close(src_fd);
+ munmap(dst_v, W * H * 4); close(dst_fd);
+ return 0;
+}
+```
+
+
+
+Compilation
+
+
+
+```bash
+sudo apt install gcc g++ cmake
+gcc -o g2d_rotation g2d_rotation.c
+```
+
+
+
+## Example Results
+
+| Example | Operation | Result |
+| ----------------- | ----------------------------- | -------------------------- |
+| `g2d_rotation` | ARGB8888 1920×1080 rotate 90° | **8.24 ms**, 251.8 MP/sec |
+| `g2d_format_conv` | ARGB8888 → YUV420 1920×1080 | **9.56 ms**, 216.8 MP/sec |
+| `g2d_scaler` | 4096×4096 → 1920×1080 scale | **68.60 ms**, 244.6 MP/sec |
+| `g2d_color_fill` | 1920×1080 solid color fill | **8.53 ms**, 243.2 MP/sec |
+
+Note: `g2d_scaler` source resolution upper limit is 4096×4096 (8192×8192 triggers `EPERM`).
+
+## API Reference
+
+Complete API definition and all supported formats see header file:
+
+```text
+/usr/include/bsp/linux/sunxi-g2d.h
+```
+
+Common ioctl commands:
+
+| Command | Purpose |
+| -------------------- | ------------------------------------------------------------- |
+| `G2D_CMD_BITBLT_H` | Single image bit block transfer (rotate/scale/format convert) |
+| `G2D_CMD_FILLRECT_H` | Color fill rectangle |
+| `G2D_CMD_STRETCHBLT` | Stretch bit block transfer |
+| `G2D_CMD_BLD_H` | Porter-Duff blend operation |
+| `G2D_CMD_MIXER_TASK` | Batch tasks (submit multiple operations at once) |
+
+Supported formats (partial):
+
+| Format | Description |
+| ------------------------------- | -------------------- |
+| `G2D_FORMAT_ARGB8888` | 32bpp Alpha-RGB |
+| `G2D_FORMAT_RGB888` | 24bpp RGB |
+| `G2D_FORMAT_RGB565` | 16bpp |
+| `G2D_FORMAT_YUV420UVC_U1V1U0V0` | NV12 standard format |
+
+Rotation flags: `G2D_ROT_0` / `G2D_ROT_90` / `G2D_ROT_180` / `G2D_ROT_270` / `G2D_ROT_H` (horizontal flip) / `G2D_ROT_V` (vertical flip)
+
+## Notes
+
+- **DMA buffer must be physically contiguous**, regular `malloc` cannot be used for G2D
+- **IOMMU is responsible for address translation**, G2D accesses IOMMU-mapped physical addresses, no need to care about specific address values
+- **DMA buffer passed through fd**, user space uses `mmap`-ed virtual address for read/write
+- **Target resolution should match source**, width and height swap when rotating 90°
+- **Format conversion** (e.g. RGB → YUV) is done by G2D hardware, just set source and destination format fields separately
+- **For debugging**, failure returns `-1`, `errno` records specific reason, use `perror()` to print
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/rockchip/_rga_usage_guide.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/rockchip/_rga_usage_guide.mdx
new file mode 100644
index 000000000..a9f1460da
--- /dev/null
+++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/rockchip/_rga_usage_guide.mdx
@@ -0,0 +1,416 @@
+## RGA Overview
+
+RGA (Raster Graphic Acceleration) is a 2D image processing hardware accelerator provided by Rockchip, specifically designed for accelerating image scaling, color conversion, format conversion and other operations. Available on RK3588 and other chips.
+
+**Typical Advantages:**
+
+- Hardware acceleration, zero CPU overhead
+- Efficient image scaling (resize)
+- Color space conversion (BGR↔RGB, RGB↔YUV)
+- Supports DMA zero-copy
+
+**Application Scenarios:**
+
+- Image preprocessing for object detection like YOLO
+- Video/image format conversion
+- Camera pipeline image processing
+
+## RGA Quick Start
+
+### Clone Repository
+
+
+
+```bash
+git clone https://github.com/airockchip/librga.git
+```
+
+
+
+### Directory Structure
+
+
+
+```bash
+./
+├── CHANGELOG.md
+├── COPYING
+├── docs
+│ ├── RGA_FAQ.assets
+│ ├── Rockchip_Developer_Guide_RGA_CN.md
+│ ├── Rockchip_Developer_Guide_RGA_EN.md
+│ ├── Rockchip_FAQ_RGA_CN.md
+│ └── Rockchip_FAQ_RGA_EN.md
+├── include
+│ ├── drmrga.h
+│ ├── GrallocOps.h
+│ ├── im2d_buffer.h
+│ ├── im2d_common.h
+│ ├── im2d_expand.h
+│ ├── im2d.h
+│ ├── im2d.hpp
+│ ├── im2d_mpi.h
+│ ├── im2d_single.h
+│ ├── im2d_task.h
+│ ├── im2d_type.h
+│ ├── im2d_version.h
+│ ├── RgaApi.h
+│ ├── rga.h
+│ ├── RgaMutex.h
+│ ├── RgaSingleton.h
+│ ├── RgaUtils.h
+│ └── RockchipRga.h
+├── libs
+│ ├── AndroidNdk
+│ └── Linux
+├── README.md
+├── samples
+│ ├── allocator_demo
+│ ├── alpha_demo
+│ ├── async_demo
+│ ├── build
+│ ├── cmake-android.sh
+│ ├── cmake-linux.sh
+│ ├── CMakeLists.txt
+│ ├── config_demo
+│ ├── copy_demo
+│ ├── crop_demo
+│ ├── cvtcolor_demo
+│ ├── fill_demo
+│ ├── gauss_demo
+│ ├── im2d_api_demo
+│ ├── mosaic_demo
+│ ├── padding_demo
+│ ├── palette_demo
+│ ├── README.md
+│ ├── resize_demo
+│ ├── rop_demo
+│ ├── sample_file
+│ ├── SConscript
+│ ├── toolchain_local.cmake
+│ ├── transform_demo
+│ └── utils
+├── toolchains
+│ ├── toolchain_android_ndk.cmake
+│ ├── toolchain_linux_1106.cmake
+│ ├── toolchain_linux.cmake
+│ └── toolchain_rt_thread.cmake
+└── tools
+ └── bin
+```
+
+
+
+### Run Demo
+
+
+
+```bash
+cd librga/tools/bin/Linux/gcc-aarch64
+./rgaImDemo -h
+```
+
+
+
+
+
+```bash
+./rgaImDemo --copy
+```
+
+
+
+
+
+```bash
+./rgaImDemo --resize up
+```
+
+
+
+### Header Files and Libraries
+
+**Header File References:**
+
+- RGA header location: `librga/include/`
+- C++ call im2d API: `im2d.hpp`
+- C call im2d API: `im2d.h`
+
+**Library Files:**
+
+- RGA library location: `librga/libs/Linux/gcc-aarch64/`
+- `librga.so`
+- `librga.a`
+
+## Core Concepts
+
+### Buffer Management
+
+RGA operates on buffers, which need to be imported first.
+
+
+
+```cpp
+#include
+#include
+
+// 1. Allocate user-space memory
+uint8_t* src_buf = malloc(width * height * 3);
+
+// 2. Import to RGA (virtual address method)
+rga_buffer_handle_t src_handle = importbuffer_virtualaddr(src_buf, buf_size);
+
+// 3. Wrap as rga_buffer_t
+rga_buffer_t src_img = wrapbuffer_handle(src_handle, width, height, RK_FORMAT_BGR_888);
+
+// 4. Release
+releasebuffer_handle(src_handle);
+free(src_buf);
+```
+
+
+
+### Common Formats
+
+| Format | Description |
+| ------------------------ | -------------- |
+| `RK_FORMAT_BGR_888` | 3-channel BGR |
+| `RK_FORMAT_RGB_888` | 3-channel RGB |
+| `RK_FORMAT_RGBA_8888` | 4-channel RGBA |
+| `RK_FORMAT_YCbCr_420_SP` | NV12 |
+
+### API Overview
+
+| Operation | API | Description |
+| ------------------- | -------------- | ----------------------------- |
+| Scale | `imresize()` | Image scaling |
+| Color conversion | `imcvtcolor()` | BGR↔RGB, RGB↔YUV |
+| Fill border | `imfill()` | Rectangle area fill |
+| Combined processing | `improcess()` | crop+resize+format conversion |
+| Memory copy | `imcopy()` | Buffer copy |
+
+## Code Examples
+
+### Simple Example: Image Scaling
+
+
+
+```cpp
+#include
+#include
+
+void resize_example() {
+ int src_w = 1920, src_h = 1080;
+ int dst_w = 640, dst_h = 640;
+
+ // Allocate memory
+ uint8_t* src = malloc(src_w * src_h * 3);
+ uint8_t* dst = malloc(dst_w * dst_h * 3);
+
+ // Import buffer
+ rga_buffer_handle_t src_handle = importbuffer_virtualaddr(src, src_w * src_h * 3);
+ rga_buffer_handle_t dst_handle = importbuffer_virtualaddr(dst, dst_w * dst_h * 3);
+
+ // Wrap
+ rga_buffer_t src_img = wrapbuffer_handle(src_handle, src_w, src_h, RK_FORMAT_BGR_888);
+ rga_buffer_t dst_img = wrapbuffer_handle(dst_handle, dst_w, dst_h, RK_FORMAT_BGR_888);
+
+ // Scale
+ IM_STATUS ret = imresize(src_img, dst_img);
+
+ // Release
+ releasebuffer_handle(src_handle);
+ releasebuffer_handle(dst_handle);
+ free(src);
+ free(dst);
+}
+```
+
+
+
+### Color Conversion
+
+
+
+```cpp
+// BGR -> RGB
+imcvtcolor(src_img, dst_img, RK_FORMAT_BGR_888, RK_FORMAT_RGB_888);
+
+// BGR -> YUV (NV12)
+imcvtcolor(src_img, dst_img, RK_FORMAT_BGR_888, RK_FORMAT_YCbCr_420_SP);
+```
+
+
+
+### Fill Background
+
+
+
+```cpp
+// Fill gray rectangle area
+rga_buffer_t canvas = wrapbuffer_handle(handle, width, height, RK_FORMAT_BGR_888);
+
+// Gray: B=114, G=114, R=114 (BGR order)
+int gray = (114 << 16) | (114 << 8) | 114;
+
+// Fill entire image with gray
+imfill(canvas, {0, 0, width, height}, gray);
+```
+
+
+
+### Combined Processing improcess()
+
+`improcess()` is the most powerful API, can simultaneously:
+
+- Specify source region (crop)
+- Scale to target region
+- Color format conversion
+
+
+
+```cpp
+// Syntax
+improcess(src_img, dst_img, {}, srect, drect, {}, -1, nullptr, nullptr, IM_SYNC);
+
+// Parameter description:
+// src_img, dst_img: source and destination buffer
+// {}: third parameter is pat (pattern), generally not used
+// srect: source region {x, y, width, height}
+// drect: destination region {x, y, width, height}
+// {}: prect, generally not used
+// -1: acquire_fence_fd
+// nullptr: release_fence_fd
+// nullptr: opt_ptr
+// IM_SYNC: synchronous execution
+```
+
+
+
+## Usage in YOLO Preprocessing
+
+### Problem Analysis
+
+YOLO preprocessing requires:
+
+1. Aspect ratio preserved scaling (letterbox)
+2. Gray padding edges
+3. BGR → RGB conversion
+
+### Complete Implementation
+
+
+
+```cpp
+int preprocess_with_rga(const cv::Mat& img0, int orig_w, int orig_h,
+ int target_size, float ratio, int dw, int dh,
+ std::vector& input_data, PerfStats& perf) {
+ Timer timer;
+ double t_padding = 0, t_resize = 0;
+
+ int new_unpad_w = static_cast(orig_w * ratio);
+ int new_unpad_h = static_cast(orig_h * ratio);
+
+ // RGA requires both stride and width pixels to be 16-aligned
+ // BGR_888: stride_bytes must be multiple of 16, and width_px must also be multiple of 16
+ int src_width_px = ((orig_w + 15) / 16) * 16;
+ int dst_width_px = ((target_size + 15) / 16) * 16;
+ int src_stride = src_width_px * 3;
+ int dst_stride = dst_width_px * 3;
+
+ int src_buf_size = src_stride * orig_h;
+ int dst_buf_size = dst_stride * target_size;
+
+ std::vector src_buf(src_buf_size);
+ std::vector dst_buf(dst_buf_size);
+
+ // Copy original image data (line by line, preserve row spacing)
+ for (int h = 0; h < orig_h; ++h) {
+ memcpy(src_buf.data() + h * src_stride, img0.data + h * orig_w * 3, orig_w * 3);
+ }
+
+ // Step 1: RGA fill gray background
+ {
+ rga_buffer_handle_t canvas_handle = importbuffer_virtualaddr(dst_buf.data(), dst_width_px, target_size, RK_FORMAT_BGR_888);
+ if (canvas_handle == 0) {
+ std::cerr << "Warning: RGA importbuffer failed for canvas\n";
+ return -1;
+ }
+
+ rga_buffer_t canvas = wrapbuffer_handle(canvas_handle, dst_width_px, target_size, RK_FORMAT_BGR_888);
+
+ int gray_color = (114 << 16) | (114 << 8) | 114;
+
+ timer.reset();
+ IM_STATUS ret = imfill(canvas, {0, 0, dst_width_px, target_size}, gray_color);
+ t_padding = timer.elapsed_ms();
+
+ if (ret != IM_STATUS_SUCCESS) {
+ std::cerr << "Warning: RGA imfill failed: " << imStrError(ret) << "\n";
+ releasebuffer_handle(canvas_handle);
+ return -1;
+ }
+
+ releasebuffer_handle(canvas_handle);
+ }
+
+ // Step 2: RGA improcess - scale original to center region + BGR to RGB
+ {
+ rga_buffer_handle_t src_handle = importbuffer_virtualaddr(src_buf.data(), src_width_px, orig_h, RK_FORMAT_BGR_888);
+ rga_buffer_handle_t dst_handle = importbuffer_virtualaddr(dst_buf.data(), dst_width_px, target_size, RK_FORMAT_RGB_888);
+
+ if (src_handle == 0 || dst_handle == 0) {
+ std::cerr << "Warning: RGA importbuffer failed\n";
+ return -1;
+ }
+
+ rga_buffer_t src_img = wrapbuffer_handle(src_handle, src_width_px, orig_h, RK_FORMAT_BGR_888);
+ rga_buffer_t dst_img = wrapbuffer_handle(dst_handle, dst_width_px, target_size, RK_FORMAT_RGB_888);
+
+ im_rect srect = {0, 0, orig_w, orig_h};
+ im_rect drect = {dw, dh, new_unpad_w, new_unpad_h};
+
+ timer.reset();
+ IM_STATUS ret = improcess(src_img, dst_img, {}, srect, drect, {}, -1, nullptr, nullptr, IM_SYNC);
+ t_resize = timer.elapsed_ms();
+
+ if (ret != IM_STATUS_SUCCESS) {
+ std::cerr << "Warning: RGA improcess failed: " << imStrError(ret) << "\n";
+ releasebuffer_handle(src_handle);
+ releasebuffer_handle(dst_handle);
+ return -1;
+ }
+
+ releasebuffer_handle(src_handle);
+ releasebuffer_handle(dst_handle);
+ }
+
+ // Copy to output
+ input_data = std::move(dst_buf);
+
+ perf.preprocess = t_padding + t_resize;
+ return 0;
+}
+```
+
+
+
+## Performance Comparison
+
+```bash
+================================================
+ BENCHMARK RESULTS
+================================================
+
+Metric OpenCV RGA Speedup
+--------------------------------------------------
+Preprocess (ms) 2.01 2.42 .83x
+Inference (ms) 63.62 63.81
+TOTAL (ms) 128.30 126.93
+--------------------------------------------------
+
+Iterations: 10 per version
+Test completed.
+```
+
+As you can see, with very simple preprocessing, OpenCV and RGA have minimal difference.
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/g2d-usage-guide.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/g2d-usage-guide.md
new file mode 100644
index 000000000..7a0e4ce9f
--- /dev/null
+++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/g2d-usage-guide.md
@@ -0,0 +1,14 @@
+---
+sidebar_position: 4.5
+
+doc_kind: wrapper
+source_of_truth: common
+imports_resolve_to:
+ - docs/common/ai/cubie/_g2d-usage-guide.mdx
+---
+
+# G2D Usage Guide
+
+import G2D from '../../../../common/ai/cubie/\_g2d-usage-guide.mdx';
+
+
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/g2d-usage-guide.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/g2d-usage-guide.md
new file mode 100644
index 000000000..7a0e4ce9f
--- /dev/null
+++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/g2d-usage-guide.md
@@ -0,0 +1,14 @@
+---
+sidebar_position: 4.5
+
+doc_kind: wrapper
+source_of_truth: common
+imports_resolve_to:
+ - docs/common/ai/cubie/_g2d-usage-guide.mdx
+---
+
+# G2D Usage Guide
+
+import G2D from '../../../../common/ai/cubie/\_g2d-usage-guide.mdx';
+
+
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/g2d-usage-guide.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/g2d-usage-guide.md
new file mode 100644
index 000000000..7a0e4ce9f
--- /dev/null
+++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/g2d-usage-guide.md
@@ -0,0 +1,14 @@
+---
+sidebar_position: 4.5
+
+doc_kind: wrapper
+source_of_truth: common
+imports_resolve_to:
+ - docs/common/ai/cubie/_g2d-usage-guide.mdx
+---
+
+# G2D Usage Guide
+
+import G2D from '../../../../common/ai/cubie/\_g2d-usage-guide.mdx';
+
+
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/g2d-usage-guide.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/g2d-usage-guide.md
new file mode 100644
index 000000000..7a0e4ce9f
--- /dev/null
+++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/g2d-usage-guide.md
@@ -0,0 +1,14 @@
+---
+sidebar_position: 4.5
+
+doc_kind: wrapper
+source_of_truth: common
+imports_resolve_to:
+ - docs/common/ai/cubie/_g2d-usage-guide.mdx
+---
+
+# G2D Usage Guide
+
+import G2D from '../../../../common/ai/cubie/\_g2d-usage-guide.mdx';
+
+
diff --git a/i18n/en/docusaurus-plugin-content-docs/current/rock5/rock5b/app-development/rga-usage-guide.md b/i18n/en/docusaurus-plugin-content-docs/current/rock5/rock5b/app-development/rga-usage-guide.md
new file mode 100644
index 000000000..5564e8b4a
--- /dev/null
+++ b/i18n/en/docusaurus-plugin-content-docs/current/rock5/rock5b/app-development/rga-usage-guide.md
@@ -0,0 +1,15 @@
+---
+sidebar_position: 5
+description: "RGA 2D Image Acceleration Development Guide"
+
+doc_kind: wrapper
+source_of_truth: common
+imports_resolve_to:
+ - docs/common/ai/rockchip/_rga_usage_guide.mdx
+---
+
+import RGA from '../../../common/ai/rockchip/\_rga_usage_guide.mdx';
+
+# RGA Usage Guide
+
+