Hello Torch Team & Torch Enthusiasts,
I am compiling torch with given scripts/build_local.sh in Linux 64 bit.
I am succeed compiled torch statically but CUDA not.
So I repeated the steps to see what the options.
I found this:
CUDA Static still OFF but Torch Static On
How to set CUDA static link to ON?
This is my CMake Args I made in build_local.sh :
CMAKE_ARGS+=("-DCMAKE_PREFIX_PATH=$(python -c 'from distutils.sysconfig import get_python_lib; print(get_python_lib())')")
CMAKE_ARGS+=("-DPYTHON_EXECUTABLE=$(python -c 'import sys; print(sys.executable)')")
CMAKE_ARGS+=("-DBUILD_CUSTOM_PROTOBUF=OFF")
CMAKE_ARGS+=("-DBUILD_SHARED_LIBS=OFF")
CMAKE_ARGS+=("-DCUDA_TOOLKIT_ROOT_DIR=/opt/cuda-10.1")
CMAKE_ARGS+=("-DCUDA_INCLUDE_DIRS=/opt/cuda-10.1/bin/include")
CMAKE_ARGS+=("-DCUDA_CUDART_LIBRARY=/opt/cuda-10.1/bin/lib64/libcudart_static.a")
CMAKE_ARGS+=("-DUSE_CUDA=ON")
CMAKE_ARGS+=("-DCMAKE_C_COMPILER=/usr/bin/gcc-8")
CMAKE_ARGS+=("-DCMAKE_CXX_COMPILER=/usr/bin/g++-8")
Error when compiling (I assume because cuda also included dynamic libs):
[363/366] Linking CXX shared library lib/libtorch_python.so
FAILED: lib/libtorch_python.so
: && /usr/bin/g++-8 -fPIC -Wno-deprecated -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -DUSE_VULKAN_WRAPPER -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Wno-stringop-overflow -DHAVE_AVX_CPU_DEFINITION -DHAVE_AVX2_CPU_DEFINITION -O3 -DNDEBUG -DNDEBUG -Wl,--no-as-needed -rdynamic -shared -Wl,-soname,libtorch_python.so -o lib/libtorch_python.so caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/generated/python_functions.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/generated/python_variable_methods.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/generated/python_torch_functions.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/generated/python_nn_functions.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/generated/python_fft_functions.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/generated/python_linalg_functions.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/CudaIPCTypes.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/DataLoader.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/Device.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/Dtype.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/DynamicTypes.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/Exceptions.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/Generator.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/Layout.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/MemoryFormat.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/QScheme.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/Module.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/PtrWrapper.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/python_dimname.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/Size.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/Storage.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/TypeInfo.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/api/src/python/init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/functions/init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/python_anomaly_mode.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/python_cpp_function.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/python_engine.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/python_function.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/python_hook.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/python_legacy_variable.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/python_variable.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/autograd/python_variable_indexing.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/backends/backend_init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/backends/backend_resolver.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/python/init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/cast_all_constant_to_floating.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/eval_peephole.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/constant_fold.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/eliminate_unused_items.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/fixup_onnx_controlflow.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/function_substitution.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/helper.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/peephole.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/preprocess_for_onnx.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/prepare_division_for_onnx.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/scalar_type_analysis.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/unpack_quantized_weights.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/passes/onnx/prepare_inplace_ops_for_onnx.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/python/python_arg_flatten.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/python/python_custom_class.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/python/python_interpreter.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/python/python_ir.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/python/python_tracer.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/python/script_init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/frontend/concrete_module_type.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/python/python_sugared_value.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/python/python_tree_views.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/multiprocessing/init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/onnx/init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/serialization.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/tensor/python_tensor.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/throughput_benchmark.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/cuda_lazy_init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/invalid_arguments.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/object_ptr.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/python_arg_parser.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/python_dispatch.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/structseq.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/tensor_apply.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/tensor_dtypes.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/tensor_layouts.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/tensor_memoryformats.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/tensor_qschemes.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/tensor_list.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/tensor_new.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/tensor_numpy.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/tensor_types.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/utils/disable_torch_function.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/cuda/Event.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/cuda/Module.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/cuda/python_comm.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/cuda/Storage.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/cuda/Stream.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/cuda/serialization.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/cuda/shared/cudart.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/cuda/shared/nvtx.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/cuda/utils.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/autograd/init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/c10d/comm.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/c10d/init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/c10d/reducer.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/rpc/init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/rpc/process_group_agent.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/rpc/py_rref.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/rpc/python_functions.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/rpc/python_rpc_handler.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/rpc/request_callback_impl.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/rpc/tensorpipe_agent.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/rpc/testing/faulty_process_group_agent.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/rpc/testing/init.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/rpc/unpickled_python_call.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/distributed/rpc/unpickled_python_remote_call.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/jit/runtime/register_distributed_ops.cpp.o caffe2/torch/CMakeFiles/torch_python.dir/csrc/cuda/python_nccl.cpp.o -Wl,-rpath,/run/media/aviezab/Binary/SourceCode/cpp/pytorch/build/lib:/opt/cuda-10.1/lib:/opt/cuda-10.1/lib64:/usr/lib/openmpi: lib/libshm.so /opt/cuda-10.1/lib/libnvToolsExt.so lib/libc10d.a lib/libtensorpipe.a -Wl,--whole-archive,"/run/media/aviezab/Binary/SourceCode/cpp/pytorch/build/lib/libtorch.a" -Wl,--no-whole-archive lib/libtorch.a lib/libtensorpipe.a lib/libtensorpipe_uv.a -Wl,--whole-archive,"/run/media/aviezab/Binary/SourceCode/cpp/pytorch/build/lib/libtorch_cpu.a" -Wl,--no-whole-archive /usr/lib/gcc/x86_64-pc-linux-gnu/8.4.0/libgomp.so /usr/lib/libpthread.so /usr/lib/gcc/x86_64-pc-linux-gnu/8.4.0/libgomp.so /usr/lib/libpthread.so -Wl,--whole-archive,"/run/media/aviezab/Binary/SourceCode/cpp/pytorch/build/lib/libcaffe2_protos.a" -Wl,--no-whole-archive lib/libqnnpack.a lib/libpytorch_qnnpack.a lib/libnnpack.a lib/libXNNPACK.a lib/libpthreadpool.a -lm lib/libfbgemm.a lib/libcpuinfo.a lib/libclog.a lib/libasmjit.a lib/libfoxi_loader.a -lgcc_s -lgcc /usr/lib/libblas.so /usr/lib/liblapack.so -lm lib/libdnnl.a -ldl sleef/lib/libsleef.a -Wl,--whole-archive,"/run/media/aviezab/Binary/SourceCode/cpp/pytorch/build/lib/libonnx.a" -Wl,--no-whole-archive lib/libonnx_proto.a /usr/local/lib/libprotobuf.a -pthread -Wl,--whole-archive,"/run/media/aviezab/Binary/SourceCode/cpp/pytorch/build/lib/libCaffe2_perfkernels_avx.a" -Wl,--no-whole-archive -Wl,--whole-archive,"/run/media/aviezab/Binary/SourceCode/cpp/pytorch/build/lib/libCaffe2_perfkernels_avx2.a" -Wl,--no-whole-archive -Wl,--whole-archive,"/run/media/aviezab/Binary/SourceCode/cpp/pytorch/build/lib/libCaffe2_perfkernels_avx512.a" -Wl,--no-whole-archive -Wl,--whole-archive,"/run/media/aviezab/Binary/SourceCode/cpp/pytorch/build/lib/libtorch_cuda.a" -Wl,--no-whole-archive lib/libc10_cuda.a lib/libc10.a /opt/cuda-10.1/lib/libnvToolsExt.so /opt/cuda-10.1/lib64/libcusparse.so /opt/cuda-10.1/lib64/libcurand.so /opt/cuda-10.1/lib64/libcufft.so /opt/cuda-10.1/lib64/libcurand.so /opt/cuda-10.1/lib64/libcublas.so nccl/lib/libnccl_static.a -lrt lib/libgloo_cuda.a lib/libgloo.a /usr/lib/openmpi/libmpi_cxx.so /usr/lib/openmpi/libmpi.so -lpthread /opt/cuda-10.1/lib64/libcudart.so && :
/usr/bin/ld: lib/libtensorpipe.a(context.cc.o): relocation R_X86_64_PC32 against symbol `_ZGVZ24TensorPipeVerbosityLevelvE5level' can not be used when making a shared object; recompile with -fPIC
/usr/bin/ld: final link failed: bad value
collect2: error: ld returned 1 exit status
[365/366] Building NVCC (Device) object modules/detectron/CMakeFiles/caffe2_detectron_ops_gpu.dir/caffe2_detectron_ops_gpu_generated_upsample_nearest_op.cu.o
ninja: build stopped: subcommand failed.
Am I supposed to pass nvcc build options too?
Thank You for your kindness.