Community
cancel
Showing results for 
Search instead for 
Did you mean: 
idata
Community Manager
410 Views

Tensorflow with Intel-Compiler

I currently try to install Tensorflow 1.5 with Intel Compiler 2017 but it results in some errors:

 

/usr/bin/ld: bazel-out/local_linux-py3-opt/bin/tensorflow/python/_objs/framework/fast_tensor_util.so/tensorflow/python/framework/fast_tensor_util.pic.o: relocation R_X86_64_PC32 against symbol `pyx_module_is_main_tensorflowpythonframeworkfast_tensor_util' can not be used when making a shared object; recompile with -fPIC

 

/usr/bin/ld: final link failed: Bad value

 

Environment:

 

CUDA_TOOLKIT_PATH=/software/CUDA/9.0.176/ \

 

CUDNN_INSTALL_PATH=/software/cuDNN/9.0/7.0 \

 

GCC_HOST_COMPILER_PATH=/software/intel/xe2017/compilers_and_libraries_2017.3.191/linux/mpi/intel64/bin/mpiicc \

 

LD_LIBRARY_PATH=/software/intel/xe2017/compilers_and_libraries_2017.3.191/linux/compiler/lib/intel64:/software/intel/xe2017/compilers_and_libraries_2017.3.191/linux/mkl/lib/intel64:/software/intel/xe2017/compilers_and_libraries_2017.3.191/linux/linux/ipp/lib/intel64:/software/intel/xe2017/compilers_and_libraries_2017.3.191/linux/tbb/lib/intel64:/software/intel/xe2017/compilers_and_libraries_2017.3.191/linux/mpi/intel64/lib:/software/python/install/Python-3.4.5/Include/:/software/python/3.4.5/ivybridge/lib:/software/intel/xe2015/composer_xe_2015.1.133/compiler/lib/intel64:/software/intel/xe2015/composer_xe_2015.1.133/ipp/lib/intel64:/software/intel/xe2015/composer_xe_2015.1.133/mkl/lib/intel64:/software/intel/xe2015/composer_xe_2015.1.133/tbb/lib/intel64:/software/cuDNN/9.0/7.0//lib64:/software/java/1.8.0_151//lib:/lib64:/usr/lib64:/usr/local/lib64:/usr/X11R6/lib64:/software/CUDA/9.0.176/lib64 \

 

PATH=/software/intel/xe2017/compilers_and_libraries_2017.3.191/linux/bin/intel64:/software/intel/xe2017/compilers_and_libraries_2017.3.191/linux/debugger/gui/intel64:/software/intel/xe2017/compilers_and_libraries_2017.3.191/linux/mpi/intel64/bin:/software/python/3.4.5/ivybridge/bin:/software/intel/xe2015/composer_xe_2015.1.133/bin/intel64:/software/intel/xe2015/composer_xe_2015.1.133/bin/intel64_mic:/software/intel/xe2015/composer_xe_2015.1.133/debugger/gui/intel64:/software/CUDA/9.0.176/bin:/software/bazel/0.7.0/ivybridge:/software/java/1.8.0_151//bin:/usr/pbs/bin:/usr/lpp/mmfs/bin/:/usr/pbs/bin/:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/opt/pbs/bin:/opt/pbs/sbin:/root/bin:/bin:/usr/X11R6/bin \

 

PWD=/proc/self/cwd \

 

PYTHON_BIN_PATH=/software/python/3.4.5/ivybridge/bin/python \

 

PYTHON_LIB_PATH=/software/tensorflow/1.5.0/python3.4.5/gpu/ \

 

TF_CUDA_CLANG=0 \

 

TF_CUDA_COMPUTE_CAPABILITIES=3.5,6.1 \

 

TF_CUDA_VERSION=9.0 \

 

TF_CUDNN_VERSION=7 \

 

TF_NEED_CUDA=1 \

 

TF_NEED_OPENCL_SYCL=0

 

Calling the Tensorflow Function:

 

external/local_config_cuda/crosstool/clang/bin/crosstool_wrapper_driver_is_not_gcc -shared -o bazel-out/local_linux-py3-opt/bin/tensorflow/python/framework/fast_tensor_util.so -Wl,-no-as-needed -B/usr/bin/ -Wl,-z,relro,-z,now -no-canonical-prefixes -pass-exit-codes '-Wl,--build-id=md5' '-Wl,--hash-style=gnu' -Wl,--gc-sections -Wl,@bazel-out/local_linux-py3-opt/bin/tensorflow/python/framework/fast_tensor_util.so-2.params).

Tags (1)
0 Kudos
0 Replies