【编译MNN】编译完成后，已在虚拟Python环境中安装PyMNN，导入MNN报错

[BerineYang]

平台(如果交叉编译请再附上交叉编译目标平台):

RK3588，本地编译
gcc版本11.4.0
g++版本
Ubuntu 22.04

Github版本:

git clone https://github.com/alibaba/MNN.git

编译方式:

cmake..
make -j4
python build_deps.py opencl
python setup.py install --version 2.9.6 --deps opencl

编译日志:

编译过程仅开启opencl
cmake配置如下
option(MNN_USE_SYSTEM_LIB "For opencl and vulkan, use system lib or use dlopen" OFF)
option(MNN_BUILD_HARD "Build -mfloat-abi=hard or not" OFF)
option(MNN_BUILD_SHARED_LIBS "MNN build shared or static lib" ON)
option(MNN_WIN_RUNTIME_MT "MNN use /MT on Windows dll" OFF)
option(MNN_FORBID_MULTI_THREAD "Disable Multi Thread" OFF)
option(MNN_OPENMP "Use OpenMP's thread pool implementation. Does not work on iOS or Mac OS" OFF)
option(MNN_USE_THREAD_POOL "Use MNN's own thread pool implementation" ON)
option(MNN_BUILD_TRAIN "Build MNN's training framework" OFF)
option(MNN_BUILD_DEMO "Build demo/exec or not" OFF)
option(MNN_BUILD_TOOLS "Build tools/cpp or not" ON)
option(MNN_BUILD_QUANTOOLS "Build Quantized Tools or not" ON)
option(MNN_EVALUATION "Build Evaluation Tools or not" ON)
option(MNN_BUILD_CONVERTER "Build Converter" ON)
option(MNN_SUPPORT_DEPRECATED_OP "Enable MNN's tflite quantized op" ON)
option(MNN_DEBUG_MEMORY "MNN Debug Memory Access" OFF)
option(MNN_DEBUG_TENSOR_SIZE "Enable Tensor Size" OFF)
option(MNN_GPU_TRACE "Enable MNN Gpu Debug" OFF)
option(MNN_SUPPORT_RENDER "Enable MNN Render Ops" OFF)
option(MNN_SUPPORT_TRANSFORMER_FUSE "Enable MNN transformer Fuse Ops" OFF)
option(MNN_PORTABLE_BUILD "Link the static version of third party libraries where possible to improve the portability of built executables" OFF)
option(MNN_SEP_BUILD "Build MNN Backends and expression separately. Only works with MNN_BUILD_SHARED_LIBS=ON" ON)
option(NATIVE_LIBRARY_OUTPUT "Native Library Path" OFF)
option(NATIVE_INCLUDE_OUTPUT "Native Include Path" OFF)
option(MNN_AAPL_FMWK "Build MNN.framework instead of traditional .a/.dylib" OFF)
option(MNN_WITH_PLUGIN "Build with plugin op support." OFF)
option(MNN_BUILD_MINI "Build MNN-MINI that just supports fixed shape models." OFF)
option(MNN_USE_SSE "Use SSE optimization for x86 if possiable" ON)
option(MNN_BUILD_CODEGEN "Build with codegen" OFF)
option(MNN_ENABLE_COVERAGE "Build with coverage enable" OFF)
option(MNN_BUILD_PROTOBUFFER "Build with protobuffer in MNN" ON)
option(MNN_BUILD_OPENCV "Build OpenCV api in MNN." OFF)
option(MNN_BUILD_LLM "Build llm library based MNN." OFF)
option(MNN_BUILD_DIFFUSION "Build diffusion demo based MNN." OFF)
option(MNN_INTERNAL "Build with MNN internal features, such as model authentication, metrics logging" OFF)
option(MNN_JNI "Build MNN Jni for java to use" OFF)
option(MNN_SUPPORT_BF16 "Enable MNN's bf16 op" OFF)
option(MNN_LOW_MEMORY "Build MNN support low memory for weight quant model." OFF)
option(MNN_CPU_WEIGHT_DEQUANT_GEMM "Build MNN CPU weight dequant related gemm kernels." OFF)
在python中导入时，发现如下问题
Traceback (most recent call last):
File "", line 1, in
File "/home/orangepi/archiconda3/envs/mnn/lib/python3.6/site-packages/MNN_OPENCL-2.9.6-py3.6-linux-aarch64.egg/MNN/init.py", line 4, in
from _mnncengine import *
ImportError: /home/orangepi/archiconda3/envs/mnn/lib/python3.6/site-packages/MNN_OPENCL-2.9.6-py3.6-linux-aarch64.egg/_mnncengine.cpython-36m-aarch64-linux-gnu.so: undefined symbol: _ZSt28__throw_bad_array_new_lengthv
还恳请工程师能够帮忙解决，谢谢！

[jxt1234]

像是编译时使用的 libc++ 和运行时的 libc++ 不一致

[BerineYang]

是的，我编译用的是本地环境，运行的时候用的是虚拟环境，想问问有什么办法？

[BerineYang]

你好，我的libstdc++已经正常外部链接进来了，但是我的MNN识别不到Mali-G610的显卡，想问问是什么原因，clinfo是正常的

[jxt1234]

是加载 opencl 的 so 失败?

[BerineYang]

我看了下opencl的库应该是正常的，在import MNN的事时候是否会提示可以使用GPU？

[jxt1234]

编译时打开 MNN_OPENCL 了么?

[BerineYang]

都打开了