tensorflow: Bazel & cmake compilations for Windows are completely broken

Can’t compile tensorflow revision 1.3 using Bazel (revision of the last successful nightly build using cmake). Error is the following:

ERROR: C:/tensorflow-1.3rc1/tensorflow/stream_executor/BUILD:39:1: C++ compilation of rule '//tensorflow/stream_executor:cuda_platform' failed (Exit 2): cl.exe failed: error executing command
  cd C:/msys64/tmp/_bazel_dmitry/ehgyfc-k/execroot/org_tensorflow
  SET CUDA_COMPUTE_CAPABILITIE=None
    SET CUDA_PATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0
    SET CUDA_TOOLKIT_PATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0
    SET CUDNN_INSTALL_PATH=C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v8.0
    SET INCLUDE=C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\INCLUDE;C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\ATLMFC\INCLUDE;C:\Program Files (x86)\Windows Kits\10\include\10.0.10240.0\ucrt;C:\Program Files (x86)\Windows Kits\NETFXSDK\4.6.1\include\um;C:\Program Files (x86)\Windows Kits\8.1\include\\shared;C:\Program Files (x86)\Windows Kits\8.1\include\\um;C:\Program Files (x86)\Windows Kits\8.1\include\\winrt;
    SET LIB=C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\LIB\amd64;C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\ATLMFC\LIB\amd64;C:\Program Files (x86)\Windows Kits\10\lib\10.0.10240.0\ucrt\x64;C:\Program Files (x86)\Windows Kits\NETFXSDK\4.6.1\lib\um\x64;C:\Program Files (x86)\Windows Kits\8.1\lib\winv6.3\um\x64;
    SET NO_WHOLE_ARCHIVE_OPTION=1
    SET PATH=C:\Program Files (x86)\Microsoft Visual Studio 14.0\Common7\IDE\CommonExtensions\Microsoft\TestWindow;C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\BIN\amd64;C:\WINDOWS\Microsoft.NET\Framework64\v4.0.30319;C:\Program Files (x86)\Microsoft Visual Studio 14.0\VC\VCPackages;C:\Program Files (x86)\Microsoft Visual Studio 14.0\Common7\IDE;C:\Program Files (x86)\Microsoft Visual Studio 14.0\Common7\Tools;C:\Program Files (x86)\Microsoft Visual Studio 14.0\Team Tools\Performance Tools\x64;C:\Program Files (x86)\Microsoft Visual Studio 14.0\Team Tools\Performance Tools;C:\Program Files (x86)\Windows Kits\8.1\bin\x64;C:\Program Files (x86)\Windows Kits\8.1\bin\x86;C:\Program Files (x86)\Microsoft SDKs\Windows\v10.0A\bin\NETFX 4.6.1 Tools\x64\;C:\msys64\mingw64\bin;C:\msys64\usr\local\bin;C:\msys64\usr\bin;C:\msys64\usr\bin;C:\Windows\System32;C:\Windows;C:\Windows\System32\Wbem;C:\Windows\System32\WindowsPowerShell\v1.0\;C:\msys64\usr\bin\site_perl;C:\msys64\usr\bin\vendor_perl;C:\msys64\usr\bin\core_perl;C:\WINDOWS\system32
    SET PWD=/proc/self/cwd
    SET PYTHON_BIN_PATH=C:/Users/Dmitry/AppData/Local/Programs/Python/Python36/python.exe
    SET PYTHON_LIB_PATH=C:/Users/Dmitry/AppData/Local/Programs/Python/Python36/lib/site-packages
    SET TEMP=C:\Users\Dmitry\AppData\Local\Temp
    SET TF_CUDA_CLANG=0
    SET TF_CUDA_COMPUTE_CAPABILITIES=3.5,5.2
    SET TF_CUDA_VERSION=8.0
    SET TF_CUDNN_VERSION=6
    SET TF_NEED_CUDA=1
    SET TF_NEED_OPENCL=0
    SET TMP=C:\Users\Dmitry\AppData\Local\Temp
  C:/Program Files (x86)/Microsoft Visual Studio 14.0/VC/bin/amd64/cl.exe /DCOMPILER_MSVC /DNOMINMAX /D_WIN32_WINNT=0x0600 /D_CRT_SECURE_NO_DEPRECATE /D_CRT_SECURE_NO_WARNINGS /D_SILENCE_STDEXT_HASH_DEPRECATION_WARNINGS /bigobj /Zm500 /J /Gy /GF /EHsc /wd4351 /wd4291 /wd4250 /wd4996 -march=native -march=native /nologo /I. /Ibazel-out/msvc_x64-py3-opt/genfiles /Iexternal/bazel_tools /Ibazel-out/msvc_x64-py3-opt/genfiles/external/bazel_tools /Iexternal/protobuf /Ibazel-out/msvc_x64-py3-opt/genfiles/external/protobuf /Iexternal/eigen_archive /Ibazel-out/msvc_x64-py3-opt/genfiles/external/eigen_archive /Iexternal/local_config_sycl /Ibazel-out/msvc_x64-py3-opt/genfiles/external/local_config_sycl /Iexternal/gif_archive /Ibazel-out/msvc_x64-py3-opt/genfiles/external/gif_archive /Iexternal/jpeg /Ibazel-out/msvc_x64-py3-opt/genfiles/external/jpeg /Iexternal/com_googlesource_code_re2 /Ibazel-out/msvc_x64-py3-opt/genfiles/external/com_googlesource_code_re2 /Iexternal/farmhash_archive /Ibazel-out/msvc_x64-py3-opt/genfiles/external/farmhash_archive /Iexternal/fft2d /Ibazel-out/msvc_x64-py3-opt/genfiles/external/fft2d /Iexternal/highwayhash /Ibazel-out/msvc_x64-py3-opt/genfiles/external/highwayhash /Iexternal/png_archive /Ibazel-out/msvc_x64-py3-opt/genfiles/external/png_archive /Iexternal/zlib_archive /Ibazel-out/msvc_x64-py3-opt/genfiles/external/zlib_archive /Iexternal/snappy /Ibazel-out/msvc_x64-py3-opt/genfiles/external/snappy /Iexternal/local_config_cuda /Ibazel-out/msvc_x64-py3-opt/genfiles/external/local_config_cuda /Iexternal/bazel_tools/tools/cpp/gcc3 /Iexternal/protobuf/src /Ibazel-out/msvc_x64-py3-opt/genfiles/external/protobuf/src /Iexternal/eigen_archive /Ibazel-out/msvc_x64-py3-opt/genfiles/external/eigen_archive /Iexternal/gif_archive/lib /Ibazel-out/msvc_x64-py3-opt/genfiles/external/gif_archive/lib /Iexternal/gif_archive/windows /Ibazel-out/msvc_x64-py3-opt/genfiles/external/gif_archive/windows /Iexternal/farmhash_archive/src /Ibazel-out/msvc_x64-py3-opt/genfiles/external/farmhash_archive/src /Iexternal/png_archive /Ibazel-out/msvc_x64-py3-opt/genfiles/external/png_archive /Iexternal/zlib_archive /Ibazel-out/msvc_x64-py3-opt/genfiles/external/zlib_archive /Iexternal/local_config_cuda/cuda /Ibazel-out/msvc_x64-py3-opt/genfiles/external/local_config_cuda/cuda /Iexternal/local_config_cuda/cuda/cuda/include /Ibazel-out/msvc_x64-py3-opt/genfiles/external/local_config_cuda/cuda/cuda/include /showIncludes /DEIGEN_MPL2_ONLY /DSNAPPY /MT /O2 /c tensorflow/stream_executor/cuda/cuda_gpu_executor.cc /Fobazel-out/msvc_x64-py3-opt/bin/tensorflow/stream_executor/_objs/cuda_platform/tensorflow/stream_executor/cuda/cuda_gpu_executor.o.
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(274): error C2589: 'constant': illegal token on right side of '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(274): error C2059: syntax error: '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(289): error C2589: 'constant': illegal token on right side of '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(289): error C2059: syntax error: '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(372): error C2589: 'constant': illegal token on right side of '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(372): error C2059: syntax error: '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(620): error C2589: 'constant': illegal token on right side of '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(620): error C2059: syntax error: '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(639): error C2589: 'constant': illegal token on right side of '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(639): error C2059: syntax error: '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(666): error C2589: 'constant': illegal token on right side of '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(666): error C2059: syntax error: '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(680): error C2589: 'constant': illegal token on right side of '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(680): error C2059: syntax error: '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(694): error C2589: 'constant': illegal token on right side of '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(694): error C2059: syntax error: '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(708): error C2589: 'constant': illegal token on right side of '::'
tensorflow/stream_executor/cuda/cuda_gpu_executor.cc(708): error C2059: syntax error: '::'
cl : Command line warning D9002 : ignoring unknown option '-march=native'
cl : Command line warning D9002 : ignoring unknown option '-march=native'
____Building complete.
Target //tensorflow/tools/pip_package:build_pip_package failed to build
____Elapsed time: 216.281s, Critical Path: 141.66s

If you try to compile master revision then compilation even won’t start with the error:

ERROR: C:/tensorflow/tensorflow/core/kernels/BUILD:1932:1: C++ compilation of rule '//tensorflow/core/kernels:resize_bilinear_op_gpu' failed (Exit 2).
.\tensorflow/core/util/cuda_kernel_helper.h(359): error C3861: 'atomicAdd': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(360): error C3861: 'atomicAdd': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(361): error C3861: 'atomicAdd': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(362): error C3861: 'atomicAdd': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(365): error C3861: 'atomicMax': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(366): error C3861: 'atomicMax': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(378): error C3861: 'max': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(378): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(394): error C3861: '__longlong_as_double': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(394): error C3861: '__double_as_longlong': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(393): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(399): error C3861: '__longlong_as_double': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(445): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(462): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(498): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(507): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(516): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(527): error C3861: '__int_as_float': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(527): error C3861: '__float_as_int': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(526): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(529): error C3861: '__int_as_float': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(538): error C3861: '__longlong_as_double': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(538): error C3861: '__double_as_longlong': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(537): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(540): error C3861: '__longlong_as_double': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(548): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(557): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(566): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(577): error C3861: '__int_as_float': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(577): error C3861: '__float_as_int': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(576): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(579): error C3861: '__int_as_float': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(588): error C3861: '__longlong_as_double': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(588): error C3861: '__double_as_longlong': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(587): error C3861: 'atomicCAS': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(590): error C3861: '__longlong_as_double': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(617): error C2065: 'warpSize': undeclared identifier
.\tensorflow/core/util/cuda_kernel_helper.h(619): error C2059: syntax error: 'volatile'
.\tensorflow/core/util/cuda_kernel_helper.h(620): error C3861: '__shfl': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(621): error C3861: '__shfl': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(622): error C2059: syntax error: 'volatile'
.\tensorflow/core/util/cuda_kernel_helper.h(637): error C2065: 'warpSize': undeclared identifier
.\tensorflow/core/util/cuda_kernel_helper.h(639): error C2059: syntax error: 'volatile'
.\tensorflow/core/util/cuda_kernel_helper.h(640): error C3861: '__shfl_up': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(641): error C3861: '__shfl_up': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(642): error C2059: syntax error: 'volatile'
.\tensorflow/core/util/cuda_kernel_helper.h(657): error C2065: 'warpSize': undeclared identifier
.\tensorflow/core/util/cuda_kernel_helper.h(659): error C2059: syntax error: 'volatile'
.\tensorflow/core/util/cuda_kernel_helper.h(660): error C3861: '__shfl_down': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(661): error C3861: '__shfl_down': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(662): error C2059: syntax error: 'volatile'
.\tensorflow/core/util/cuda_kernel_helper.h(677): error C2065: 'warpSize': undeclared identifier
.\tensorflow/core/util/cuda_kernel_helper.h(679): error C2059: syntax error: 'volatile'
.\tensorflow/core/util/cuda_kernel_helper.h(680): error C3861: '__shfl_xor': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(681): error C3861: '__shfl_xor': identifier not found
.\tensorflow/core/util/cuda_kernel_helper.h(682): error C2059: syntax error: 'volatile'
cl : Command line warning D9002 : ignoring unknown option '-march=native'
cl : Command line warning D9002 : ignoring unknown option '-march=native'
cl : Command line warning D9002 : ignoring unknown option '-x'
cl : Command line warning D9002 : ignoring unknown option '-nvcc_options=relaxed-constexpr'
cl : Command line warning D9002 : ignoring unknown option '-nvcc_options=ftz=true'
cl : Command line warning D9024 : unrecognized source file type 'cuda', object file assumed
cl : Command line warning D9027 : source file 'cuda' ignored
Target //tensorflow/tools/pip_package:build_pip_package failed to build
Use --verbose_failures to see the command lines of failed build steps.
____Elapsed time: 3817.327s, Critical Path: 143.61s

I’ll be glad to help to fix the issues. But I absolutely has no idea what’s going on here. Can anybody help to fix Windows compilation please? Currently here is no build for Windows with AVX, AVX2 support in the whole world!

About this issue

Original URL
State: closed
Created 7 years ago
Comments: 44 (17 by maintainers)

Links to this issue

performance - How faster is tensorflow-gpu with AVX and AVX2 compared with it without AVX and AVX2? - Stack Overflow

Most upvoted comments

Hi @zyavrik ,

Please set USE_MSVC_WRAPPER=1

e.g.

bazel --output_base C:\t build --action_env=USE_MSVC_WRAPPER=1 //tensorflow/tools/pip_package:build_pip_package

snnn on Sep 12, 2017

Just FYI, I’ve successfully compile v1.3 with CMake + Visual Studio 2015 + CUDA 8 + CUDNN 6.1 on Windows 10. Had to include various hacks to fix several compilation error related to Eigen and CUDA though 😒 But at least it works now 😄 Feel free to ask me about issues you run into if you are compiling with similar set up.

I follow this post to get started, but it doesn’t show how to fix the errors I got. https://joe-antognini.github.io/machine-learning/build-windows-tf This post is also very helpful for the next step if you want to build a standalone application: https://joe-antognini.github.io/machine-learning/windows-tf-project

phg1024 on Sep 15, 2017

For anyone stumbling on this while googling ignoring unknown option '-march=native', the issue is because march=native is a gcc flag. I’ve got my TF+GPU 1.14 to build in Windows by using /arch:AVX2. See MSVC docs

stevenwong on Aug 12, 2019

@snnn I think you would be interested in https://github.com/meteorcloudy/tensorflow/commits/fix_contrib_with_cc_import

I can now use Bazel@HEAD to build tensorflow on Windows with //tensorflow/contrib/... enabled.

The fix requires a new rule, cc_import, that will come out with Bazel 0.10.0, so I’ll send to PR to fix tensorflow/contrib/… after 0.10,0 is released.

meteorcloudy on Dec 14, 2017

@gunan, could you look at what could be done to streamline/fix this to work out of the box?

aselle on Nov 30, 2017

The only one reason to compile it is to enable AVX, AVX2 support…

zyavrik on Sep 19, 2017

It’s one of the external dependencies. Try to find nsync under cmake\build\external and also the cmake\build\nsync. You should be able to find compiled .lib files and header files. Link them to the library you were trying to compile.

phg1024 on Sep 18, 2017

I fixed this error in https://github.com/phg1024/tensorflow/commit/291f29533a510797f64e1f646b38aeea62238389

You need to explicitly instantiate it

template struct MaxPoolForwardWithOptionalArgmax<Eigen::QInt8>;

Then you can just compile this very file in Visual studio and see if you get other errors. If not, go ahead and do the linking stage only.

phg1024 on Sep 18, 2017

I incorrectly merged your changes to the repository. The same files were changed recently. Thanks for your help!

zyavrik on Sep 18, 2017

Can anybody create pull request for those fixes?

zyavrik on Sep 17, 2017

@zyavrik These are the changes I can find in the code: https://github.com/phg1024/tensorflow/commit/291f29533a510797f64e1f646b38aeea62238389

There are also some configuration changes about the VS project settings to resolve undefined symbols in the linking stage. You can follow take a look at https://joe-antognini.github.io/machine-learning/windows-tf-project to get some sense about how to resolve those.

phg1024 on Sep 15, 2017