在windows下的linux子系统wsl Ubuntu下驱动mediapipe gpu支持，问题：gpu比cpu慢；内存泄漏 #6180
Description
Have I written custom code (as opposed to using a stock example script provided in MediaPipe)
None
OS Platform and Distribution
Ubuntu 24.04LTS
Mobile device if the issue happens on mobile device
No response
Browser and version if the issue happens on browser
No response
Programming Language and version
C++，python
MediaPipe version
0.10.21
Bazel version
7.3.2
Solution
hand_tracking
Android Studio, NDK, SDK versions (if issue is related to building in Android environment)
No response
Xcode & Tulsi version (if issue is related to building for iOS)
No response
Describe the actual behavior
我在windows上的linux子系统wsl2（Ubuntu）中下载安装了CUDA、CUdnn、Tensorflow、opengl等组件，希望可以运行mediapipe的gpu项目
Describe the expected behaviour
我希望可以在wsl中成功运行mediapipe的gpu项目
Standalone code/steps you may have used to try to get what you need
无
Other info / Complete Logs
首先我在wsl中进行了相关的环境配置，这点我不细说，我只阐述我安装的版本：CUDA-12.5、Cudnn-9.3、bazel-7.3.2、tensorflow-2.20、python3.11、nccl、nvcc、mediapipe-10.21等。
在开始时，我发现通过opengl es 支持mediapipe的gpu实际上比在cpu运行的还慢，gpu内存大概只占用了35%-45%，这应该是因为opengl需要多次地上传内存到Linux和下载回windows，以及反复的虚拟内存转换。
然后，我尝试使用CUDA直接进行内存访问，我按照官方文档的指示，编译命令如下： bazel run -c opt --config=cuda --spawn_strategy=local \
--define no_aws_support=true --copt -DMESA_EGL_NO_X11_HEADERS \
mediapipe/examples/desktop/hand_tracking:hand_tracking_gpu 但报错：double free or corruption (out)，内存泄漏。
我不知道这该如何解决。以下是我相关环境的验证和配置信息：(mediapipe_gpu) xqy@DESKTOP-A2HORL1:~/mediapipe$ ls /usr/local/cuda-12.5/extras/CUPTI
doc samples
(mediapipe_gpu) xqy@DESKTOP-A2HORL1:~/mediapipe$ ls /usr/local/cuda-12.5
DOCS README compute-sanitizer extras include libnvvp nvml share targets version.json
EULA.txt bin doc gds lib64 nsightee_plugins nvvm src tools
(mediapipe_gpu) xqy@DESKTOP-A2HORL1:~/mediapipe$ nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2024 NVIDIA Corporation
Built on Thu_Jun__6_02:18:23_PDT_2024
Cuda compilation tools, release 12.5, V12.5.82
Build cuda_12.5.r12.5/compiler.34385749_0
(mediapipe_gpu) xqy@DESKTOP-A2HORL1:~/mediapipe$ ls /usr/lib/x86_64-linux-gnu/ | grep libcudnn.so
libcudnn.so
libcudnn.so.9
libcudnn.so.9.3.0
(mediapipe_gpu) xqy@DESKTOP-A2HORL1:~/mediapipe$ glxinfo | grep -i opengl
OpenGL vendor string: Microsoft Corporation
OpenGL renderer string: D3D12 (NVIDIA GeForce RTX 3060)
OpenGL core profile version string: 4.6 (Core Profile) Mesa 25.0.7-0ubuntu0.24.04.2
OpenGL core profile shading language version string: 4.60
OpenGL core profile context flags: (none)
OpenGL core profile profile mask: core profile
OpenGL core profile extensions:
OpenGL version string: 4.6 (Compatibility Profile) Mesa 25.0.7-0ubuntu0.24.04.2
OpenGL shading language version string: 4.60
OpenGL context flags: (none)
OpenGL profile mask: compatibility profile
OpenGL extensions:
OpenGL ES profile version string: OpenGL ES 3.1 Mesa 25.0.7-0ubuntu0.24.04.2
OpenGL ES profile shading language version string: OpenGL ES GLSL ES 3.10
OpenGL ES profile extensions:
(mediapipe_gpu) xqy@DESKTOP-A2HORL1:~/mediapipe$ nvidia-smi
Fri Dec 12 11:42:05 2025
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 580.95.05 Driver Version: 581.80 CUDA Version: 13.0 |
+-----------------------------------------+------------------------+----------------------+
| GPU Name Persistence-M | Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap | Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|=========================================+========================+======================|
| 0 NVIDIA GeForce RTX 3060 On | 00000000:01:00.0 On | N/A |
| 52% 38C P8 18W / 170W | 1350MiB / 12288MiB | 12% Default |
| | | N/A |
+-----------------------------------------+------------------------+----------------------+
+-----------------------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=========================================================================================|
| 0 N/A N/A 460 G /Xwayland N/A |
+-----------------------------------------------------------------------------------------+
现在我正在考虑用Vulkan来替换Opengl，或者使用Nsight Systems来检查数据流在内存中的走向