FAIL: building TensorFlow on Ubuntu 16.04

MathiasMagnus commented 6 years ago

TensorFlow fails to build on Ubuntu 16.04 when doing things according to the wiki entry on Codeplays site.

System information

Have I written custom code: just building, no custom code yet
OS Platform and Distribution: Ubuntu 16.04.3 64-bit
TensorFlow installed from: this repo, branch dev/gpu_amd
TensorFlow version: ?
Python version: 2.7.12
Bazel version: 0.9.0
GCC/Compiler version (if compiling from source): gcc 5.4
CUDA/cuDNN version: -
GPU model and memory: AMD HD 7970
Exact command to reproduce: bazel build -c opt --config=sycl //tensorflow/tools/pip_package:build_pip_package

Describe the problem

Trying to build TF with OpenCL & SYCL support using ComputeCpp-CE 0.5.1.

Source code / logs

This is what's seen on console when build fails:

WARNING: Output base '/home/mnagy/.cache/bazel/_bazel_mnagy/b6c672ee274d6e3919f72d8fdb8667a3' is on NFS. This may lead to surprising failures and undetermined behavior.
WARNING: /home/mnagy/Source/Repos/tensorflow/tensorflow/core/BUILD:1832:1: in includes attribute of cc_library rule //tensorflow/core:framework_headers_lib: '../../external/nsync/public' resolves to 'external/nsync/public' not below the relative path of its package 'tensorflow/core'. This will be an error in the future. Since this rule was created by the macro 'cc_header_only_library', the error might have been caused by the macro implementation in /home/mnagy/Source/Repos/tensorflow/tensorflow/tensorflow.bzl:1150:30
WARNING: /home/mnagy/.cache/bazel/_bazel_mnagy/b6c672ee274d6e3919f72d8fdb8667a3/external/grpc/WORKSPACE:1: Workspace name in /home/mnagy/.cache/bazel/_bazel_mnagy/b6c672ee274d6e3919f72d8fdb8667a3/external/grpc/WORKSPACE (@com_github_grpc_grpc) does not match the name given in the repository's definition (@grpc); this will cause a build error in future versions
WARNING: /home/mnagy/Source/Repos/tensorflow/tensorflow/contrib/learn/BUILD:15:1: in py_library rule //tensorflow/contrib/learn:learn: target '//tensorflow/contrib/learn:learn' depends on deprecated target '//tensorflow/contrib/session_bundle:exporter': No longer supported. Switch to SavedModel immediately.
WARNING: /home/mnagy/Source/Repos/tensorflow/tensorflow/contrib/learn/BUILD:15:1: in py_library rule //tensorflow/contrib/learn:learn: target '//tensorflow/contrib/learn:learn' depends on deprecated target '//tensorflow/contrib/session_bundle:gc': No longer supported. Switch to SavedModel immediately.
INFO: Analysed target //tensorflow/tools/pip_package:build_pip_package (159 packages loaded).
INFO: Found 1 target...
ERROR: /home/mnagy/Source/Repos/tensorflow/tensorflow/core/BUILD:1529:1: undeclared inclusion(s) in rule '//tensorflow/core:lib_hash_crc32c_accelerate_internal':
this rule is missing dependency declarations for the following files included by 'tensorflow/core/lib/hash/crc32c_accelerate.cc':
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/stddef.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/__stddef_max_align_t.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/stdint.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/nmmintrin.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/smmintrin.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/tmmintrin.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/pmmintrin.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/emmintrin.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/xmmintrin.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/mmintrin.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/mm_malloc.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/f16cintrin.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/popcntintrin.h'
  '/opt/Codeplay/ComputeCpp/0.5.1/lib/clang/3.9.0/include/stdarg.h'
Target //tensorflow/tools/pip_package:build_pip_package failed to build
Use --verbose_failures to see the command lines of failed build steps.
INFO: Elapsed time: 86.109s, Critical Path: 13.77s
FAILED: Build did NOT complete successfully

lukeiwanski commented 6 years ago

hi @MathiasMagnus Thanks for the report.

I presume that you pointed ./config to /opt/Codeplay/ComputeCpp/ when it asked for ComputeCpp installation path.

Otherwise this https://github.com/lukeiwanski/tensorflow/issues/193#issuecomment-354729010 might be helpful.

The workaround on your system would be to add: cxx_builtin_include_directory: "'/opt/Codeplay/ComputeCpp" to third_party/sycl/crosstool/CROSSTOOL.tpl around line 52

MathiasMagnus commented 6 years ago

Thanks! Worked like a charm. Now comes usage. Should any issues arise, those will come as separate tickets.

Thanks again and keep up the good work! The more SYCL the merrier.

lukeiwanski / tensorflow