"Undefined symbols for architecture arm64" on Apple M1 Max when compiling Inference library

[AzureUni]

问题描述 Issue Description

参考文档/This doc was referred：https://paddle-inference.readthedocs.io/en/latest/guides/install/compile/source_compile_under_MacOS.html

制作Makefile命令/Command to make Makefile：cmake .. -DPY_VERSION=3.9 -DPYTHON_INCLUDE_DIR=${PYTHON_INCLUDE_DIRS}
-DPYTHON_LIBRARY=${PYTHON_LIBRARY} -DWITH_GPU=OFF -DWITH_TESTING=OFF
-DWITH_AVX=OFF -DWITH_ARM=ON -DCMAKE_BUILD_TYPE=Release -DON_INFER=ON

编译命令/Command to compile：make inference_lib_dist TARGET=ARMV8 -j8

报错信息
Error Message：

Undefined symbols for architecture arm64:
"void phi::Copyphi::CPUContext(phi::CPUContext const&, phi::DenseTensor const&, phi::Place, bool, phi::DenseTensor*)", referenced from:
void phi::RnnFunc<phi::LSTMCell, phi::Layer, phi::SingleLayer, phi::BidirLayer, float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const*, std::__1::vector<phi::DenseTensor const*, std::__1::allocator<phi::DenseTensor const*> > const&, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, int, int, int, int, bool, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, float, bool, int, phi::DenseTensor*) in libphi_static_1.a(rnn_kernel.cc.o)
void phi::RnnFunc<phi::SimpleRNNCell<float, phi::funcs::ReluCPUFunctor, (phi::funcs::detail::ActivationType)2>, phi::Layer, phi::SingleLayer, phi::BidirLayer, float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const*, std::__1::vector<phi::DenseTensor const*, std::__1::allocator<phi::DenseTensor const*> > const&, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, int, int, int, int, bool, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, float, bool, int, phi::DenseTensor*) in libphi_static_1.a(rnn_kernel.cc.o)
void phi::RnnFunc<phi::SimpleRNNCell<float, phi::funcs::TanhFunctor, (phi::funcs::detail::ActivationType)4>, phi::Layer, phi::SingleLayer, phi::BidirLayer, float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const*, std::__1::vector<phi::DenseTensor const*, std::__1::allocator<phi::DenseTensor const*> > const&, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, int, int, int, int, bool, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, float, bool, int, phi::DenseTensor*) in libphi_static_1.a(rnn_kernel.cc.o)
void phi::RnnFunc<phi::GRUCell, phi::Layer, phi::SingleLayer, phi::BidirLayer, float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const*, std::__1::vector<phi::DenseTensor const*, std::__1::allocator<phi::DenseTensor const*> > const&, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, int, int, int, int, bool, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, float, bool, int, phi::DenseTensor*) in libphi_static_1.a(rnn_kernel.cc.o)
void phi::RnnFunc<phi::LSTMCell, phi::Layer, phi::SingleLayer, phi::BidirLayer, double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const*, std::__1::vector<phi::DenseTensor const*, std::__1::allocator<phi::DenseTensor const*> > const&, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, int, int, int, int, bool, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, float, bool, int, phi::DenseTensor*) in libphi_static_1.a(rnn_kernel.cc.o)
void phi::RnnFunc<phi::SimpleRNNCell<double, phi::funcs::ReluCPUFunctor, (phi::funcs::detail::ActivationType)2>, phi::Layer, phi::SingleLayer, phi::BidirLayer, double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const*, std::__1::vector<phi::DenseTensor const*, std::__1::allocator<phi::DenseTensor const*> > const&, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, int, int, int, int, bool, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, float, bool, int, phi::DenseTensor*) in libphi_static_1.a(rnn_kernel.cc.o)
void phi::RnnFunc<phi::SimpleRNNCell<double, phi::funcs::TanhFunctor, (phi::funcs::detail::ActivationType)4>, phi::Layer, phi::SingleLayer, phi::BidirLayer, double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const*, std::__1::vector<phi::DenseTensor const*, std::__1::allocator<phi::DenseTensor const*> > const&, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor const*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, int, int, int, int, bool, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, float, bool, int, phi::DenseTensor*) in libphi_static_1.a(rnn_kernel.cc.o)
...
"void phi::Copyphi::CPUContext(phi::CPUContext const&, phi::SparseCooTensor const&, phi::Place, bool, phi::SparseCooTensor*)", referenced from:
void phi::sparse::ElementWiseAddCooGradCPUKernel<float, int, phi::CPUContext>(phi::CPUContext const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor*, phi::SparseCooTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
void phi::sparse::ElementWiseAddCooGradCPUKernel<float, long long, phi::CPUContext>(phi::CPUContext const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor*, phi::SparseCooTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
void phi::sparse::ElementWiseAddCooGradCPUKernel<float, signed char, phi::CPUContext>(phi::CPUContext const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor*, phi::SparseCooTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
void phi::sparse::ElementWiseAddCooGradCPUKernel<float, unsigned char, phi::CPUContext>(phi::CPUContext const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor*, phi::SparseCooTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
void phi::sparse::ElementWiseAddCooGradCPUKernel<float, short, phi::CPUContext>(phi::CPUContext const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor*, phi::SparseCooTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
void phi::sparse::ElementWiseAddCooGradCPUKernel<double, int, phi::CPUContext>(phi::CPUContext const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor*, phi::SparseCooTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
void phi::sparse::ElementWiseAddCooGradCPUKernel<double, long long, phi::CPUContext>(phi::CPUContext const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor const&, phi::SparseCooTensor*, phi::SparseCooTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
...
"void phi::Copyphi::CPUContext(phi::CPUContext const&, phi::SparseCsrTensor const&, phi::Place, bool, phi::SparseCsrTensor*)", referenced from:
void phi::sparse::ElementWiseAddCsrGradCPUKernel<float, int, phi::CPUContext>(phi::CPUContext const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor*, phi::SparseCsrTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
void phi::sparse::ElementWiseAddCsrGradCPUKernel<float, long long, phi::CPUContext>(phi::CPUContext const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor*, phi::SparseCsrTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
void phi::sparse::ElementWiseAddCsrGradCPUKernel<float, signed char, phi::CPUContext>(phi::CPUContext const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor*, phi::SparseCsrTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
void phi::sparse::ElementWiseAddCsrGradCPUKernel<float, unsigned char, phi::CPUContext>(phi::CPUContext const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor*, phi::SparseCsrTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
void phi::sparse::ElementWiseAddCsrGradCPUKernel<float, short, phi::CPUContext>(phi::CPUContext const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor*, phi::SparseCsrTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
void phi::sparse::ElementWiseAddCsrGradCPUKernel<double, int, phi::CPUContext>(phi::CPUContext const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor*, phi::SparseCsrTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
void phi::sparse::ElementWiseAddCsrGradCPUKernel<double, long long, phi::CPUContext>(phi::CPUContext const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor const&, phi::SparseCsrTensor*, phi::SparseCsrTensor*) in libphi_static_1.a(elementwise_grad_kernel.cc.o)
...
"phi::funcs::SegmentPoolFunctor<phi::CPUContext, double, int>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, phi::DenseTensor*, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentKernelLaunchHelper<phi::CPUContext, double, int>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(segment_pool_kernel.cc.o)
"phi::funcs::SegmentPoolFunctor<phi::CPUContext, double, long long>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, phi::DenseTensor*, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentKernelLaunchHelper<phi::CPUContext, double, long long>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(segment_pool_kernel.cc.o)
"phi::funcs::SegmentPoolFunctor<phi::CPUContext, float, int>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, phi::DenseTensor*, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentKernelLaunchHelper<phi::CPUContext, float, int>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(segment_pool_kernel.cc.o)
"phi::funcs::SegmentPoolFunctor<phi::CPUContext, float, long long>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, phi::DenseTensor*, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentKernelLaunchHelper<phi::CPUContext, float, long long>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(segment_pool_kernel.cc.o)
"phi::funcs::SegmentPoolFunctor<phi::CPUContext, int, int>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, phi::DenseTensor*, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentKernelLaunchHelper<phi::CPUContext, int, int>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(segment_pool_kernel.cc.o)
"phi::funcs::SegmentPoolFunctor<phi::CPUContext, int, long long>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, phi::DenseTensor*, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentKernelLaunchHelper<phi::CPUContext, int, long long>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(segment_pool_kernel.cc.o)
"phi::funcs::SegmentPoolFunctor<phi::CPUContext, long long, int>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, phi::DenseTensor*, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentKernelLaunchHelper<phi::CPUContext, long long, int>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(segment_pool_kernel.cc.o)
"phi::funcs::SegmentPoolFunctor<phi::CPUContext, long long, long long>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, phi::DenseTensor*, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentKernelLaunchHelper<phi::CPUContext, long long, long long>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(segment_pool_kernel.cc.o)
"phi::funcs::MatrixReduceSumFunctor<double, phi::CPUContext>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor*)", referenced from:
void phi::CholeskySolveGradKernel<double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, bool, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(cholesky_solve_grad_kernel.cc.o)
void phi::TriangularSolveGradKernel<double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, bool, bool, bool, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(triangular_solve_grad_kernel.cc.o)
"phi::funcs::MatrixReduceSumFunctor<float, phi::CPUContext>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor*)", referenced from:
void phi::CholeskySolveGradKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, bool, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(cholesky_solve_grad_kernel.cc.o)
void phi::TriangularSolveGradKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, bool, bool, bool, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(triangular_solve_grad_kernel.cc.o)
"phi::funcs::SegmentPoolGradFunctor<phi::CPUContext, double, int>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, paddle::optionalphi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentPoolGradKernel<double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, paddle::optionalphi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*) in libphi_static_1.a(segment_pool_grad_kernel.cc.o)
"phi::funcs::SegmentPoolGradFunctor<phi::CPUContext, double, long long>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, paddle::optionalphi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentPoolGradKernel<double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, paddle::optionalphi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*) in libphi_static_1.a(segment_pool_grad_kernel.cc.o)
"phi::funcs::SegmentPoolGradFunctor<phi::CPUContext, float, int>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, paddle::optionalphi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentPoolGradKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, paddle::optionalphi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*) in libphi_static_1.a(segment_pool_grad_kernel.cc.o)
"phi::funcs::SegmentPoolGradFunctor<phi::CPUContext, float, long long>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, paddle::optionalphi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentPoolGradKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, paddle::optionalphi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*) in libphi_static_1.a(segment_pool_grad_kernel.cc.o)
1 warning generated.
[ 97%] Linking CXX static library libir_analysis_pass.a
[ 97%] Built target ir_analysis_pass
"phi::funcs::SegmentPoolGradFunctor<phi::CPUContext, int, int>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, paddle::optionalphi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
1 warning generated.
void phi::SegmentPoolGradKernel<int, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, paddle::optionalphi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*) in libphi_static_1.a(segment_pool_grad_kernel.cc.o)
[ 97%] Linking CXX static library libir_params_sync_among_devices_pass.a
[ 97%] Built target ir_params_sync_among_devices_pass
"phi::funcs::SegmentPoolGradFunctor<phi::CPUContext, int, long long>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, paddle::optionalphi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentPoolGradKernel<int, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, paddle::optionalphi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*) in libphi_static_1.a(segment_pool_grad_kernel.cc.o)
"phi::funcs::SegmentPoolGradFunctor<phi::CPUContext, long long, int>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, paddle::optionalphi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentPoolGradKernel<long long, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, paddle::optionalphi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*) in libphi_static_1.a(segment_pool_grad_kernel.cc.o)
"phi::funcs::SegmentPoolGradFunctor<phi::CPUContext, long long, long long>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, paddle::optionalphi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator >)", referenced from:
void phi::SegmentPoolGradKernel<long long, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, paddle::optionalphi::DenseTensor const&, phi::DenseTensor const&, std::__1::basic_string<char, std::__1::char_traits, std::__1::allocator > const&, phi::DenseTensor*) in libphi_static_1.a(segment_pool_grad_kernel.cc.o)
"void phi::funcs::ModulatedDeformableIm2col<double, phi::CPUContext>(phi::CPUContext const&, double const*, double const*, double const*, std::__1::vector<long long, std::__1::allocator > const&, std::__1::vector<long long, std::__1::allocator > const&, std::__1::vector<long long, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, int, double*)", referenced from:
void phi::DeformableConvKernel<double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, paddle::optionalphi::DenseTensor const&, std::__1::vector<int, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, int, int, int, phi::DenseTensor*) in libphi_static_1.a(deformable_conv_kernel.cc.o)
void phi::DeformableConvGradKernel<double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, paddle::optionalphi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<int, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, int, int, int, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(deformable_conv_grad_kernel.cc.o)
"void phi::funcs::ModulatedDeformableIm2col<float, phi::CPUContext>(phi::CPUContext const&, float const*, float const*, float const*, std::__1::vector<long long, std::__1::allocator > const&, std::__1::vector<long long, std::__1::allocator > const&, std::__1::vector<long long, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, int, float*)", referenced from:
void phi::DeformableConvKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, paddle::optionalphi::DenseTensor const&, std::__1::vector<int, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, int, int, int, phi::DenseTensor*) in libphi_static_1.a(deformable_conv_kernel.cc.o)
void phi::DeformableConvGradKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, paddle::optionalphi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<int, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, std::__1::vector<int, std::__1::allocator > const&, int, int, int, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*, phi::DenseTensor*) in libphi_static_1.a(deformable_conv_grad_kernel.cc.o)
ld: symbol(s) not found for architecture arm64
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make[3]: *** [paddle/fluid/eager/auto_code_generator/eager_generator] Error 1
make[2]: *** [paddle/fluid/eager/auto_code_generator/CMakeFiles/eager_generator.dir/all] Error 2
make[2]: *** Waiting for unfinished jobs....
1 warning generated.

之所以未按照文档指示使用-DWITH_INFER=ON是因为develop和release/2.3两个分支的此选项均无效，CMakeList.txt中没有相应的配置，实测编译完成后也不会出现c/c++预测库。所以我选择使用-DON_INFER=ON配合make的inference_lib_dist参数进行编译。
The reason why I did not stick to the doc to use -DWITH_INFER=ON but rather chose to use -DON_INFER=ON instead, is that -DWITH_INFER=ON was not working in either develop or release/2.3 branch. So I choose to use -DON_INFER=ON with "inference_lib_dist" parameter.

之所以选用develop分支是因为release/2.3分支在使用-DON_INFER=ON时编译会报illegal hardware instruction问题，此问题已经在其他issue中提出但是没有见到后续回复。develop分支编译没有遇到此问题。
The reason why I chose the develop branch is that I have encountered with "illegal hardware instruction" issue in release/2.3 branch with -DON_INFER=ON. This issue has been raised earlier by someone else and I do not see a solution as of now. There is no such issue in the current develop branch.

版本&环境信息 Version & Environment Information

---

Paddle version: None
Paddle With CUDA: None

OS: macOs 12.4
Python version: 3.9.12

CUDA version: None
cuDNN version: None.None.None
Nvidia driver version: None

---

Branch：develop
Commit：fix logging debug level (#44684) 8aa286d
CPU：Apple M1 Max
GPU：Apple M1 Max

[jzhang533]

try import this PR: #42325

[AzureUni]

try import this PR: #42325

This helps on building the paddle_python target with no problem. But when it comes to c/c++ inference library it still fails.

After the success of the python build, the errors are:

[100%] Built target paddle_python
  NOTE: a missing vtable usually means the first non-inline virtual member function has no definition.
  "vtable for CryptoPP::SHA1", referenced from:
      paddle::framework::GetSha1(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >) in libpaddle_inference.a(program_desc.cc.o)
      CryptoPP::SHA1::SHA1(CryptoPP::SHA1 const&) in libpaddle_inference.a(program_desc.cc.o)
  NOTE: a missing vtable usually means the first non-inline virtual member function has no definition.
  "vtable for CryptoPP::Filter", referenced from:
      paddle::framework::HexEncoding(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >) in libpaddle_inference.a(program_desc.cc.o)
      CryptoPP::StringSource::~StringSource() in libpaddle_inference.a(program_desc.cc.o)
      CryptoPP::HexEncoder::HexEncoder(CryptoPP::BufferedTransformation*, bool, int, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&) in libpaddle_inference.a(program_desc.cc.o)
      CryptoPP::Grouper::Grouper(CryptoPP::BufferedTransformation*) in libpaddle_inference.a(program_desc.cc.o)
      CryptoPP::StringSource::StringSource(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, bool, CryptoPP::BufferedTransformation*) in libpaddle_inference.a(program_desc.cc.o)
      CryptoPP::StringSource::~StringSource() in libpaddle_inference.a(program_desc.cc.o)
      non-virtual thunk to CryptoPP::StringSource::~StringSource() in libpaddle_inference.a(program_desc.cc.o)
      ...
  NOTE: a missing vtable usually means the first non-inline virtual member function has no definition.
  "vtable for CryptoPP::Source", referenced from:
      CryptoPP::StringSource::StringSource(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, bool, CryptoPP::BufferedTransformation*) in libpaddle_inference.a(program_desc.cc.o)
      CryptoPP::StringSource::StringSource(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, bool, CryptoPP::BufferedTransformation*) in libpaddle_inference.a(aes_cipher.cc.o)
  NOTE: a missing vtable usually means the first non-inline virtual member function has no definition.
  "vtable for CryptoPP::Grouper", referenced from:
      CryptoPP::Grouper::Grouper(CryptoPP::BufferedTransformation*) in libpaddle_inference.a(program_desc.cc.o)
  NOTE: a missing vtable usually means the first non-inline virtual member function has no definition.
  "vtable for CryptoPP::GCM_Base::GCTR", referenced from:
      CryptoPP::GCM_Base::GCM_Base() in libpaddle_inference.a(aes_cipher.cc.o)
  NOTE: a missing vtable usually means the first non-inline virtual member function has no definition.
  "vtable for CryptoPP::GCM_Base", referenced from:
      paddle::framework::AESCipher::BuildAuthEncCipher(bool*, CryptoPP::member_ptr<CryptoPP::AuthenticatedSymmetricCipher>*, CryptoPP::member_ptr<CryptoPP::AuthenticatedEncryptionFilter>*) in libpaddle_inference.a(aes_cipher.cc.o)
      paddle::framework::AESCipher::BuildAuthDecCipher(bool*, CryptoPP::member_ptr<CryptoPP::AuthenticatedSymmetricCipher>*, CryptoPP::member_ptr<CryptoPP::AuthenticatedDecryptionFilter>*) in libpaddle_inference.a(aes_cipher.cc.o)
      CryptoPP::GCM_Base::GCM_Base() in libpaddle_inference.a(aes_cipher.cc.o)
      CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, true>::~GCM_Final() in libpaddle_inference.a(aes_cipher.cc.o)
      CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, false>::~GCM_Final() in libpaddle_inference.a(aes_cipher.cc.o)
  NOTE: a missing vtable usually means the first non-inline virtual member function has no definition.
  "vtable for CryptoPP::Rijndael::Base", referenced from:
      CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, true>::~GCM_Final() in libpaddle_inference.a(aes_cipher.cc.o)
      CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>::~BlockCipherFinal() in libpaddle_inference.a(aes_cipher.cc.o)
      CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>::~BlockCipherFinal() in libpaddle_inference.a(aes_cipher.cc.o)
      non-virtual thunk to CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>::~BlockCipherFinal() in libpaddle_inference.a(aes_cipher.cc.o)
      non-virtual thunk to CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>::~BlockCipherFinal() in libpaddle_inference.a(aes_cipher.cc.o)
      CryptoPP::Rijndael::Base::Base(CryptoPP::Rijndael::Base const&) in libpaddle_inference.a(aes_cipher.cc.o)
      CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, false>::~GCM_Final() in libpaddle_inference.a(aes_cipher.cc.o)
      ...
  NOTE: a missing vtable usually means the first non-inline virtual member function has no definition.
  "non-virtual thunk to CryptoPP::AuthenticatedSymmetricCipherBase::ProcessData(unsigned char*, unsigned char const*, unsigned long)", referenced from:
      vtable for CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, true> in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, false> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::GCM_Base::OptimalDataAlignment() const", referenced from:
      vtable for CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, true> in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, false> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::CTR_ModePolicy::SeekToIteration(unsigned long)", referenced from:
      vtable for CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::ConcretePolicyHolder<CryptoPP::Empty, CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >, CryptoPP::AdditiveCipherAbstractPolicy> > in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::CTR_ModePolicy::OperateKeystream(CryptoPP::KeystreamOperation, unsigned char*, unsigned char const*, unsigned long)", referenced from:
      vtable for CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::ConcretePolicyHolder<CryptoPP::Empty, CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >, CryptoPP::AdditiveCipherAbstractPolicy> > in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::CTR_ModePolicy::CipherResynchronize(unsigned char*, unsigned char const*, unsigned long)", referenced from:
      vtable for CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::ConcretePolicyHolder<CryptoPP::Empty, CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >, CryptoPP::AdditiveCipherAbstractPolicy> > in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::ECB_OneWay::ProcessData(unsigned char*, unsigned char const*, unsigned long)", referenced from:
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::ECB_OneWay> in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)1, CryptoPP::Rijndael::Dec>, CryptoPP::ECB_OneWay> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::CBC_Decryption::ProcessData(unsigned char*, unsigned char const*, unsigned long)", referenced from:
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)1, CryptoPP::Rijndael::Dec>, CryptoPP::CBC_Decryption> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::CBC_Encryption::ProcessData(unsigned char*, unsigned char const*, unsigned long)", referenced from:
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::CBC_Encryption> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >::ProcessData(unsigned char*, unsigned char const*, unsigned long)", referenced from:
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::ConcretePolicyHolder<CryptoPP::Empty, CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >, CryptoPP::AdditiveCipherAbstractPolicy> > in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >::Seek(unsigned long)", referenced from:
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::ConcretePolicyHolder<CryptoPP::Empty, CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >, CryptoPP::AdditiveCipherAbstractPolicy> > in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::BufferedTransformation::GetWaitObjects(CryptoPP::WaitObjectContainer&, CryptoPP::CallStack const&)", referenced from:
      vtable for CryptoPP::StringSource in libpaddle_inference.a(program_desc.cc.o)
      vtable for CryptoPP::SourceTemplate<CryptoPP::StringStore> in libpaddle_inference.a(program_desc.cc.o)
      vtable for CryptoPP::StringSource in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::SourceTemplate<CryptoPP::StringStore> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::AuthenticatedSymmetricCipherBase::TruncatedFinal(unsigned char*, unsigned long)", referenced from:
      vtable for CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, true> in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, false> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::AuthenticatedSymmetricCipherBase::Update(unsigned char const*, unsigned long)", referenced from:
      vtable for CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, true> in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, false> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >::IsRandomAccess() const", referenced from:
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::ConcretePolicyHolder<CryptoPP::Empty, CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >, CryptoPP::AdditiveCipherAbstractPolicy> > in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >::IsSelfInverting() const", referenced from:
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::ConcretePolicyHolder<CryptoPP::Empty, CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >, CryptoPP::AdditiveCipherAbstractPolicy> > in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >::OptimalBlockSize() const", referenced from:
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::ConcretePolicyHolder<CryptoPP::Empty, CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >, CryptoPP::AdditiveCipherAbstractPolicy> > in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >::OptimalDataAlignment() const", referenced from:
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::ConcretePolicyHolder<CryptoPP::Empty, CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >, CryptoPP::AdditiveCipherAbstractPolicy> > in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >::IsForwardTransformation() const", referenced from:
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::ConcretePolicyHolder<CryptoPP::Empty, CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >, CryptoPP::AdditiveCipherAbstractPolicy> > in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::BufferedTransformation::GetMaxWaitObjectCount() const", referenced from:
      vtable for CryptoPP::StringSource in libpaddle_inference.a(program_desc.cc.o)
      vtable for CryptoPP::SourceTemplate<CryptoPP::StringStore> in libpaddle_inference.a(program_desc.cc.o)
      vtable for CryptoPP::StringSource in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::SourceTemplate<CryptoPP::StringStore> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::GCM_Base::OptimalDataAlignment() const", referenced from:
      vtable for CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, true> in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::GCM_Final<CryptoPP::Rijndael, (CryptoPP::GCM_TablesOption)0, false> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::Rijndael::Dec::ProcessAndXorBlock(unsigned char const*, unsigned char const*, unsigned char*) const", referenced from:
      vtable for CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)1, CryptoPP::Rijndael::Dec> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::Rijndael::Dec::AdvancedProcessBlocks(unsigned char const*, unsigned char const*, unsigned char*, unsigned long, unsigned int) const", referenced from:
      vtable for CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)1, CryptoPP::Rijndael::Dec> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::Rijndael::Enc::ProcessAndXorBlock(unsigned char const*, unsigned char const*, unsigned char*) const", referenced from:
      vtable for CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::Rijndael::Enc::AdvancedProcessBlocks(unsigned char const*, unsigned char const*, unsigned char*, unsigned long, unsigned int) const", referenced from:
      vtable for CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::Rijndael::Base::AlgorithmProvider() const", referenced from:
      vtable for CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc> in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)1, CryptoPP::Rijndael::Dec> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::Rijndael::Base::OptimalDataAlignment() const", referenced from:
      vtable for CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc> in libpaddle_inference.a(aes_cipher.cc.o)
      vtable for CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)1, CryptoPP::Rijndael::Dec> in libpaddle_inference.a(aes_cipher.cc.o)
  "non-virtual thunk to CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >::GenerateBlock(unsigned char*, unsigned long)", referenced from:
      vtable for CryptoPP::CipherModeFinalTemplate_CipherHolder<CryptoPP::BlockCipherFinal<(CryptoPP::CipherDir)0, CryptoPP::Rijndael::Enc>, CryptoPP::ConcretePolicyHolder<CryptoPP::Empty, CryptoPP::AdditiveCipherTemplate<CryptoPP::AbstractPolicyHolder<CryptoPP::AdditiveCipherAbstractPolicy, CryptoPP::CTR_ModePolicy> >, CryptoPP::AdditiveCipherAbstractPolicy> > in libpaddle_inference.a(aes_cipher.cc.o)
  "_cblas_caxpy", referenced from:
      std::__1::enable_if<!(std::is_same<phi::dtype::complex<float>, phi::dtype::bfloat16>::value), void>::type paddle::operators::math::scatter::add_sparse_inputs<phi::dtype::complex<float>, phi::CPUContext>(std::__1::vector<phi::SelectedRows const*, std::__1::allocator<phi::SelectedRows const*> > const&, std::__1::unordered_map<long long, unsigned long, std::__1::hash<long long>, std::__1::equal_to<long long>, std::__1::allocator<std::__1::pair<long long const, unsigned long> > > const&, long long, phi::CPUContext const&, phi::dtype::complex<float>*) in libpaddle_inference.a(selected_rows_functor.cc.o)
  "_cblas_cgemm", referenced from:
      void phi::MatMulFunction<phi::CPUContext, phi::dtype::complex<float> >(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, phi::DenseTensor*, bool, bool, bool) in libpaddle_inference.a(matmul_kernel.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::BatchedGEMM<phi::dtype::complex<float> >(CBLAS_TRANSPOSE, CBLAS_TRANSPOSE, int, int, int, phi::dtype::complex<float>, phi::dtype::complex<float> const*, phi::dtype::complex<float> const*, phi::dtype::complex<float>, phi::dtype::complex<float>*, int, long long, long long) const in libpaddle_inference.a(matmul_kernel.cc.o)
      void phi::MatMulFunction<phi::CPUContext, phi::dtype::complex<float> >(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, phi::DenseTensor*, bool, bool, bool) in libpaddle_inference.a(matmul_grad_kernel.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::MatMul<phi::dtype::complex<float> >(phi::dtype::complex<float> const*, phi::funcs::MatDescriptor const&, phi::dtype::complex<float> const*, phi::funcs::MatDescriptor const&, phi::dtype::complex<float>, phi::dtype::complex<float>*, phi::dtype::complex<float>) const in libpaddle_inference.a(matmul_grad_kernel.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::BatchedGEMM<phi::dtype::complex<float> >(CBLAS_TRANSPOSE, CBLAS_TRANSPOSE, int, int, int, phi::dtype::complex<float>, phi::dtype::complex<float> const*, phi::dtype::complex<float> const*, phi::dtype::complex<float>, phi::dtype::complex<float>*, int, long long, long long) const in libpaddle_inference.a(matmul_grad_kernel.cc.o)
  "_cblas_cgemv", referenced from:
      void phi::MatMulFunction<phi::CPUContext, phi::dtype::complex<float> >(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, phi::DenseTensor*, bool, bool, bool) in libpaddle_inference.a(matmul_kernel.cc.o)
      void phi::MatMulFunction<phi::CPUContext, phi::dtype::complex<float> >(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, phi::DenseTensor*, bool, bool, bool) in libpaddle_inference.a(matmul_grad_kernel.cc.o)
  "_cblas_daxpy", referenced from:
      paddle::operators::LRNFunctor<phi::CPUContext, double>::operator()(paddle::framework::ExecutionContext const&, phi::DenseTensor const&, phi::DenseTensor*, phi::DenseTensor*, int, int, int, int, int, double, double, double, paddle::experimental::DataLayout) in libpaddle_inference.a(lrn_op.cc.o)
      paddle::operators::CenterLossKernel<phi::CPUContext, double>::Compute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(center_loss_op.cc.o)
      paddle::operators::math::ContextProjectGradFunctor<phi::CPUContext, double>::operator()(phi::CPUContext const&, phi::DenseTensor const&, bool, int, int, int, int, int, bool, bool, phi::DenseTensor*, phi::DenseTensor*) in libpaddle_inference.a(sequence_conv_op.cc.o)
      paddle::operators::AttentionLSTMKernel<double>::Compute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(attention_lstm_op.cc.o)
      paddle::operators::FusedEmbeddingFCLSTMKernel<double>::SeqCompute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(fused_embedding_fc_lstm_op.cc.o)
      paddle::operators::FusedEmbeddingFCLSTMKernel<double>::BatchCompute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(fused_embedding_fc_lstm_op.cc.o)
      phi::SameDimsAddFunctor<phi::CPUContext, double, void>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*) in libpaddle_inference.a(elementwise_add_kernel.cc.o)
      ...
  "_cblas_dcopy", referenced from:
      paddle::operators::CenterLossKernel<phi::CPUContext, double>::Compute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(center_loss_op.cc.o)
      paddle::operators::FuisonLSTMKernel<double>::BatchCompute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(fusion_lstm_op.cc.o)
      paddle::operators::LookupTableKernel<double>::Compute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(lookup_table_op.cc.o)
      paddle::operators::AttentionLSTMKernel<double>::Compute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(attention_lstm_op.cc.o)
      paddle::operators::FusedEmbeddingFCLSTMKernel<double>::SeqCompute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(fused_embedding_fc_lstm_op.cc.o)
      paddle::operators::FusedEmbeddingFCLSTMKernel<double>::BatchCompute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(fused_embedding_fc_lstm_op.cc.o)
      phi::SameDimsAddFunctor<phi::CPUContext, double, void>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*) in libpaddle_inference.a(elementwise_add_kernel.cc.o)
      ...
  "_cblas_dgemm", referenced from:
      void phi::funcs::Blas<phi::CPUContext>::MatMul<double>(double const*, phi::funcs::MatDescriptor const&, double const*, phi::funcs::MatDescriptor const&, double, double*, double) const in libpaddle_inference.a(fsp_op.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::BatchedGEMM<double>(CBLAS_TRANSPOSE, CBLAS_TRANSPOSE, int, int, int, double, double const*, double const*, double, double*, int, long long, long long) const in libpaddle_inference.a(fsp_op.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::MatMul<double>(phi::DenseTensor const&, bool, phi::DenseTensor const&, bool, double, phi::DenseTensor*, double) const in libpaddle_inference.a(lstm_op.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::MatMul<double>(phi::DenseTensor const&, bool, phi::DenseTensor const&, bool, double, phi::DenseTensor*, double) const in libpaddle_inference.a(lstmp_op.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::MatMul<double>(double const*, phi::funcs::MatDescriptor const&, double const*, phi::funcs::MatDescriptor const&, double, double*, double) const in libpaddle_inference.a(matmul_op.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::BatchedGEMM<double>(CBLAS_TRANSPOSE, CBLAS_TRANSPOSE, int, int, int, double, double const*, double const*, double, double*, int, long long, long long) const in libpaddle_inference.a(matmul_op.cc.o)
      paddle::operators::GRUUnitKernel<phi::CPUContext, double>::Compute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(gru_unit_op.cc.o)
      ...
  "_cblas_dgemv", referenced from:
      void phi::MvKernel<double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*) in libpaddle_inference.a(mv_kernel.cc.o)
      phi::KernelImpl<void (*)(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*), &(void phi::MvKernel<double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*))>::VariadicCompute(phi::DeviceContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*) in libpaddle_inference.a(mv_kernel.cc.o)
      void phi::KernelImpl<void (*)(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*), &(void phi::MvKernel<double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*))>::KernelCallHelper<phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, phi::TypeTag<int> >::Compute<0, 0, 0, 0>(phi::KernelContext*) in libpaddle_inference.a(mv_kernel.cc.o)
      void phi::MatMulFunction<phi::CPUContext, double>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, phi::DenseTensor*, bool, bool, bool) in libpaddle_inference.a(matmul_kernel.cc.o)
      void phi::MvGradKernel<double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, phi::DenseTensor*) in libpaddle_inference.a(mv_grad_kernel.cc.o)
      void phi::MatMulFunction<phi::CPUContext, double>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, phi::DenseTensor*, bool, bool, bool) in libpaddle_inference.a(matmul_grad_kernel.cc.o)
  "_cblas_dtrsm", referenced from:
      void phi::CholeskyGradKernel<double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, bool, phi::DenseTensor*) in libpaddle_inference.a(cholesky_grad_kernel.cc.o)
      void phi::TriangularSolveKernel<double, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, bool, bool, bool, phi::DenseTensor*) in libpaddle_inference.a(triangular_solve_kernel.cc.o)
  "_cblas_saxpy", referenced from:
      paddle::operators::LRNFunctor<phi::CPUContext, float>::operator()(paddle::framework::ExecutionContext const&, phi::DenseTensor const&, phi::DenseTensor*, phi::DenseTensor*, int, int, int, int, int, float, float, float, paddle::experimental::DataLayout) in libpaddle_inference.a(lrn_op.cc.o)
      paddle::operators::CenterLossKernel<phi::CPUContext, float>::Compute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(center_loss_op.cc.o)
      paddle::operators::math::ContextProjectGradFunctor<phi::CPUContext, float>::operator()(phi::CPUContext const&, phi::DenseTensor const&, bool, int, int, int, int, int, bool, bool, phi::DenseTensor*, phi::DenseTensor*) in libpaddle_inference.a(sequence_conv_op.cc.o)
      paddle::operators::AttentionLSTMKernel<float>::Compute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(attention_lstm_op.cc.o)
      paddle::operators::FusedEmbeddingFCLSTMKernel<float>::SeqCompute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(fused_embedding_fc_lstm_op.cc.o)
      paddle::operators::FusedEmbeddingFCLSTMKernel<float>::BatchCompute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(fused_embedding_fc_lstm_op.cc.o)
      phi::SameDimsAddFunctor<phi::CPUContext, float, void>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*) in libpaddle_inference.a(elementwise_add_kernel.cc.o)
      ...
  "_cblas_scopy", referenced from:
      paddle::operators::CenterLossKernel<phi::CPUContext, float>::Compute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(center_loss_op.cc.o)
      paddle::operators::FuisonLSTMKernel<float>::BatchCompute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(fusion_lstm_op.cc.o)
      paddle::operators::LookupTableKernel<float>::Compute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(lookup_table_op.cc.o)
      paddle::operators::AttentionLSTMKernel<float>::Compute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(attention_lstm_op.cc.o)
      paddle::operators::FusedEmbeddingFCLSTMKernel<float>::SeqCompute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(fused_embedding_fc_lstm_op.cc.o)
      paddle::operators::FusedEmbeddingFCLSTMKernel<float>::BatchCompute(paddle::framework::ExecutionContext const&) const in libpaddle_inference.a(fused_embedding_fc_lstm_op.cc.o)
      phi::SameDimsAddFunctor<phi::CPUContext, float, void>::operator()(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*) in libpaddle_inference.a(elementwise_add_kernel.cc.o)
      ...
  "_cblas_sgemm", referenced from:
      paddle::framework::ir::BuildFusion(paddle::framework::ir::Graph*, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, paddle::framework::Scope*, bool)::$_1::operator()(paddle::framework::ir::Node*, paddle::framework::ir::Node*, paddle::framework::ir::Node*, paddle::framework::ir::Node*, paddle::framework::ir::Node*, paddle::framework::ir::Node*, paddle::framework::ir::Node*, paddle::framework::ir::Node*, paddle::framework::ir::Node*, paddle::framework::ir::Node*, paddle::framework::ir::Node*) const in libpaddle_inference.a(embedding_fc_lstm_fuse_pass.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::MatMul<float>(float const*, phi::funcs::MatDescriptor const&, float const*, phi::funcs::MatDescriptor const&, float, float*, float) const in libpaddle_inference.a(fsp_op.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::BatchedGEMM<float>(CBLAS_TRANSPOSE, CBLAS_TRANSPOSE, int, int, int, float, float const*, float const*, float, float*, int, long long, long long) const in libpaddle_inference.a(fsp_op.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::MatMul<float>(phi::DenseTensor const&, bool, phi::DenseTensor const&, bool, float, phi::DenseTensor*, float) const in libpaddle_inference.a(lstm_op.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::MatMul<float>(phi::DenseTensor const&, bool, phi::DenseTensor const&, bool, float, phi::DenseTensor*, float) const in libpaddle_inference.a(lstmp_op.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::MatMul<float>(float const*, phi::funcs::MatDescriptor const&, float const*, phi::funcs::MatDescriptor const&, float, float*, float) const in libpaddle_inference.a(matmul_op.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::BatchedGEMM<float>(CBLAS_TRANSPOSE, CBLAS_TRANSPOSE, int, int, int, float, float const*, float const*, float, float*, int, long long, long long) const in libpaddle_inference.a(matmul_op.cc.o)
      ...
  "_cblas_sgemv", referenced from:
      void phi::MvKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*) in libpaddle_inference.a(mv_kernel.cc.o)
      phi::KernelImpl<void (*)(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*), &(void phi::MvKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*))>::VariadicCompute(phi::DeviceContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*) in libpaddle_inference.a(mv_kernel.cc.o)
      void phi::KernelImpl<void (*)(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*), &(void phi::MvKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*))>::KernelCallHelper<phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, phi::TypeTag<int> >::Compute<0, 0, 0, 0>(phi::KernelContext*) in libpaddle_inference.a(mv_kernel.cc.o)
      void phi::MatMulFunction<phi::CPUContext, float>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, phi::DenseTensor*, bool, bool, bool) in libpaddle_inference.a(matmul_kernel.cc.o)
      void phi::MvGradKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor const&, phi::DenseTensor*, phi::DenseTensor*) in libpaddle_inference.a(mv_grad_kernel.cc.o)
      void phi::MatMulFunction<phi::CPUContext, float>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, phi::DenseTensor*, bool, bool, bool) in libpaddle_inference.a(matmul_grad_kernel.cc.o)
  "_cblas_strsm", referenced from:
      void phi::CholeskyGradKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, bool, phi::DenseTensor*) in libpaddle_inference.a(cholesky_grad_kernel.cc.o)
      void phi::TriangularSolveKernel<float, phi::CPUContext>(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, bool, bool, bool, phi::DenseTensor*) in libpaddle_inference.a(triangular_solve_kernel.cc.o)
  "_cblas_zaxpy", referenced from:
      std::__1::enable_if<!(std::is_same<phi::dtype::complex<double>, phi::dtype::bfloat16>::value), void>::type paddle::operators::math::scatter::add_sparse_inputs<phi::dtype::complex<double>, phi::CPUContext>(std::__1::vector<phi::SelectedRows const*, std::__1::allocator<phi::SelectedRows const*> > const&, std::__1::unordered_map<long long, unsigned long, std::__1::hash<long long>, std::__1::equal_to<long long>, std::__1::allocator<std::__1::pair<long long const, unsigned long> > > const&, long long, phi::CPUContext const&, phi::dtype::complex<double>*) in libpaddle_inference.a(selected_rows_functor.cc.o)
  "_cblas_zgemm", referenced from:
      void phi::MatMulFunction<phi::CPUContext, phi::dtype::complex<double> >(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, phi::DenseTensor*, bool, bool, bool) in libpaddle_inference.a(matmul_kernel.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::BatchedGEMM<phi::dtype::complex<double> >(CBLAS_TRANSPOSE, CBLAS_TRANSPOSE, int, int, int, phi::dtype::complex<double>, phi::dtype::complex<double> const*, phi::dtype::complex<double> const*, phi::dtype::complex<double>, phi::dtype::complex<double>*, int, long long, long long) const in libpaddle_inference.a(matmul_kernel.cc.o)
      void phi::MatMulFunction<phi::CPUContext, phi::dtype::complex<double> >(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, phi::DenseTensor*, bool, bool, bool) in libpaddle_inference.a(matmul_grad_kernel.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::MatMul<phi::dtype::complex<double> >(phi::dtype::complex<double> const*, phi::funcs::MatDescriptor const&, phi::dtype::complex<double> const*, phi::funcs::MatDescriptor const&, phi::dtype::complex<double>, phi::dtype::complex<double>*, phi::dtype::complex<double>) const in libpaddle_inference.a(matmul_grad_kernel.cc.o)
      void phi::funcs::Blas<phi::CPUContext>::BatchedGEMM<phi::dtype::complex<double> >(CBLAS_TRANSPOSE, CBLAS_TRANSPOSE, int, int, int, phi::dtype::complex<double>, phi::dtype::complex<double> const*, phi::dtype::complex<double> const*, phi::dtype::complex<double>, phi::dtype::complex<double>*, int, long long, long long) const in libpaddle_inference.a(matmul_grad_kernel.cc.o)
  "_cblas_zgemv", referenced from:
      void phi::MatMulFunction<phi::CPUContext, phi::dtype::complex<double> >(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, phi::DenseTensor*, bool, bool, bool) in libpaddle_inference.a(matmul_kernel.cc.o)
      void phi::MatMulFunction<phi::CPUContext, phi::dtype::complex<double> >(phi::CPUContext const&, phi::DenseTensor const&, phi::DenseTensor const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, std::__1::vector<long long, std::__1::allocator<long long> > const&, phi::DenseTensor*, bool, bool, bool) in libpaddle_inference.a(matmul_grad_kernel.cc.o)
  "_openblas_set_num_threads", referenced from:
      paddle::platform::SetNumThreads(int) in libpaddle_inference.a(cpu_helper.cc.o)
  "_utf8proc_NFD", referenced from:
      paddle::framework::NFD(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> >*) in libpaddle_inference.a(string_array.cc.o)
  "_utf8proc_category", referenced from:
      paddle::operators::BasicTokenizer::Tokenize(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, std::__1::vector<std::__1::basic_string<wchar_t, std::__1::char_traits<wchar_t>, std::__1::allocator<wchar_t> >, std::__1::allocator<std::__1::basic_string<wchar_t, std::__1::char_traits<wchar_t>, std::__1::allocator<wchar_t> > > >*) const in libpaddle_inference.a(faster_tokenizer_op.cc.o)
      paddle::operators::IsPunctuation(wchar_t const&) in libpaddle_inference.a(faster_tokenizer_op.cc.o)
  "_utf8proc_islower", referenced from:
      phi::strings::GetCharcasesMap() in libpaddle_inference.a(unicode.cc.o)
  "_utf8proc_isupper", referenced from:
      phi::strings::GetCharcasesMap() in libpaddle_inference.a(unicode.cc.o)
  "_utf8proc_tolower", referenced from:
      paddle::operators::BasicTokenizer::do_lower_case(wchar_t) const in libpaddle_inference.a(faster_tokenizer_op.cc.o)
      paddle::operators::BasicTokenizer::Tokenize(std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&, std::__1::vector<std::__1::basic_string<wchar_t, std::__1::char_traits<wchar_t>, std::__1::allocator<wchar_t> >, std::__1::allocator<std::__1::basic_string<wchar_t, std::__1::char_traits<wchar_t>, std::__1::allocator<wchar_t> > > >*) const in libpaddle_inference.a(faster_tokenizer_op.cc.o)
      phi::strings::GetCharcasesMap() in libpaddle_inference.a(unicode.cc.o)
  "_utf8proc_toupper", referenced from:
      phi::strings::GetCharcasesMap() in libpaddle_inference.a(unicode.cc.o)
ld: symbol(s) not found for architecture arm64
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make[2]: *** [paddle/fluid/inference/capi_exp/libpaddle_inference_c.dylib] Error 1
make[1]: *** [paddle/fluid/inference/capi_exp/CMakeFiles/paddle_inference_c_shared.dir/all] Error 2
make: *** [all] Error 2

The command I used to build is:

cmake .. -DPY_VERSION=3.9 -DPYTHON_INCLUDE_DIR=${PYTHON_INCLUDE_DIRS} \
-DPYTHON_LIBRARY=${PYTHON_LIBRARY} -DWITH_GPU=OFF -DWITH_TESTING=OFF \
-DWITH_AVX=OFF -DWITH_ARM=ON -DCMAKE_BUILD_TYPE=Release -DON_INFER=ON
make TARGET=ARMV8 -j8

Branch: develop
Commit: 9e5f3a3

[amitchaudhary140]

I am also facing the same issue. Pls help.

[ariefwijaya]

any update on this?

[paddle-bot]

Since you haven't replied for more than a year, we have closed this issue/pr.
If the problem is not solved or there is a follow-up one, please reopen it at any time and we will continue to follow up.
由于您超过一年未回复，我们将关闭这个issue/pr。
若问题未解决或有后续问题，请随时重新打开，我们会继续跟进。