The following parameters are used to control the details during compilation. Please choose according to your actual needs.

    Parameter
    		 Default Values
    		 Usage
    MNN_VULKAN OFF Compile GPU Optimization Library: MNN-VULKAN
    MNN_ARM82 OFF Compile MNN-ARM82 (CPU optimization Backend of the new architecture)
    MNN_OPENCL OFF Compile GPU Optimization Library: MNN-OpenCL
    MNN_OPENGL OFF
    MNN_USE_SYSTEM_LIB OFF When compiling OpenCL/Vulkan, it depends on the OpenCL / Vulkan library of the system or. OFF indicates looking for system drivers at runtime
    MNN_BUILD_HARD OFF Control-mfloat-abi parameter under ARM platform is hard or softfp
    MNN_BUILD_SHARED_LIBS ON Controls whether MNN-related libraries are compiled into dynamic libraries instead of static libraries.
    MNN_FORBID_MULTI_THREAD OFF Disable multithreading
    MNN_BUILD_TOOLS ON Compile test tools
    MNN_BUILD_DEMO OFF Compile Demo app
    MNN_BUILD_TRAIN OFF Compile training related
    MNN_BUILD_QUANTOOLS OFF Compile quantization tool
    MNN_BUILD_CONVERTER OFF Compile model converter
    MNN_EVALUATION OFF Compile evaluation tool
    MNN_DEBUG_MEMORY OFF Debug memory
    MNN_DEBUG_TENSOR_SIZE OFF Debug Tensor size
    MNN_SUPPORT_TFLITE_QUAN OFF Supports TFLite quantization Op
    MNN_PORTABLE_BUILD OFF Static links third-party dependencies to create executable files that do not depend on specific paths
    MNN_SEP_BUILD ON Compile each backend and MNN-express separately. Valid only when MNN_BUILD_SHARED_LIBS = ON
    MNN_AAPL_FMWK OFF Compile and package MNN as a Framework instead of a static/dynamic library on the Apple platform