can't bazel build benchmark_model with quantization deps due to -lpthread linking error

[dwightcrow]

Trying to build //tensorflow/tools/benchmark:benchmark_model per instructions at this stackoveflow post http://stackoverflow.com/questions/37953369/tensorflow-quantized-graph-for-android. Am getting error that cannot find -lpthread despite my belief that lib is unneeded and best efforts to remove it.

external/androidndk/ndk/toolchains/arm-linux-androideabi-4.9/prebuilt/linux-x86_64/bin/../lib/gcc/arm-linux-androideabi/4.9/../../../../arm-linux-androideabi/bin/ld: error: cannot find -lpthread
collect2: error: ld returned 1 exit status

//tensorflow/tools/benchmark/BUILD cc_binary

deps = [":benchmark_model_lib",
        "//tensorflow/contrib/quantization/kernels:quantized_ops",
            ],

//tensorflow/contrib/quantization/kernels/BUILD:

deps = [
    "//tensorflow/contrib/quantization:cc_array_ops",
    "//tensorflow/contrib/quantization:cc_math_ops",
    "//tensorflow/contrib/quantization:cc_nn_ops",
    #"//tensorflow/core",
    #"//tensorflow/core:framework",
    #"//tensorflow/core:lib",
    #"//tensorflow/core/kernels:concat_lib_hdrs",
    #"//tensorflow/core/kernels:conv_ops",
    #"//tensorflow/core/kernels:eigen_helpers",
    #"//tensorflow/core/kernels:ops_util",
    #"//tensorflow/core/kernels:pooling_ops",
    "//third_party/eigen3",
    "@gemmlowp//:eight_bit_int_gemm",
],

Then run:
bazel build -c opt --cxxopt='-std=gnu++11'--crosstool_top=//external:android/crosstool --cpu=armeabi-v7a --host_crosstool_top=@bazel_tools//tools/cpp:toolchain tensorflow/tools/benchmark:benchmark_model --verbose_failures

Which (with following all other instructions in linked post) succeeds with the exception that it fails to link against pthread.

I've tried removing -lpthread in tensorflow/tensorflow.bzl, tensorflow/tools/proto_text/BUILD, tensorflow/tools/proto_text/BUILD, and tensorflow/cc/BUILD but this still results in -lpthread being linked to in final compilation.

I've also looked at the fixes such as this one #419 which either remove -lpthread from google protobuf or link to the dummy lib but these seem to only apply to earlier versions of TensorFlow.

Apologies if this is a basic config error, I'm new to Bazel and am pretty stuck. benchmark_model and quantize_graph both compile well independently but not yet for me when combined.

What related GitHub issues or StackOverflow threads have you found by searching the web for your problem?

http://stackoverflow.com/questions/37953369/tensorflow-quantized-graph-for-android
#419
protocolbuffers/protobuf#1373

Environment info

Operating System: Ubuntu 14.04

Installed version of CUDA and cuDNN:
(ml) socialh@socialh:~$ echo $LD_LIBRARY_PATH
/home/socialh/cuda/lib64:/usr/local/cuda/lib64:

(ml) socialh@socialh:~/otf$ ll /usr/local/cuda/lib/libcud*
-rw-r--r-- 1 root root 189170 Aug 31 15:22 /usr/local/cuda/lib/libcudadevrt.a
lrwxrwxrwx 1 root root 16 Aug 31 15:22 /usr/local/cuda/lib/libcudart.so -> libcudart.so.7.5*
lrwxrwxrwx 1 root root 19 Aug 31 15:22 /usr/local/cuda/lib/libcudart.so.7.5 -> libcudart.so.7.5.18*
-rwxr-xr-x 1 root root 311596 Aug 31 15:22 /usr/local/cuda/lib/libcudart.so.7.5.18*
-rw-r--r-- 1 root root 558020 Aug 31 15:22 /usr/local/cuda/lib/libcudart_static.a

(ml) socialh@socialh:~$ ll /home/socialh/cuda/lib64/
lrwxrwxrwx 1 socialh socialh 13 Jun 10 01:20 libcudnn.so -> libcudnn.so.5*
lrwxrwxrwx 1 socialh socialh 17 Jun 10 01:20 libcudnn.so.5 -> libcudnn.so.5.1.3*
-rwxr-xr-x 1 socialh socialh 60696704 Jun 10 01:18 libcudnn.so.5.1.3*
-rw-r--r-- 1 socialh socialh 59715990 Jun 10 01:18 libcudnn_static.a

If installed from source, provide
git rev-parse head
640353d
bazel version
Build target: bazel-out/local-fastbuild/bin/src/main/java/com/google/devtools/build/lib/bazel/BazelServer_deploy.jar
Build time: Thu Jan 01 00:00:00 1970 (0)
Build timestamp: Thu Jan 01 00:00:00 1970 (0)
Build timestamp as int: 0

If possible, provide a minimal reproducible example (We usually don't have time to read hundreds of lines of your code)

bazel build -c opt --cxxopt '-std=gnu++11' --crosstool_top=//external:android/crosstool --cpu=armeabi-v7a --host_crosstool_top=@bazel_tools//tools/cpp:toolchain tensorflow/tools/benchmark:benchmark_model --verbose_failures
with other steps taken in first StackOverflow example

What other attempted solutions have you tried?

Tried removing all -lpthreads references, also tried removing lpthreads in bazel-out files but did not help.

Logs or other output that would be helpful

(If logs are large, please upload as attachment or provide link).

[andrewharp]

To work around the pthread issue, use the @gemmlowp//:eight_bit_int_gemm_sources filegroup in your srcs rather than the cc_library target as a dep.

e.g.:

cc_library(
    name = "android_tensorflow_lib",
    srcs = if_android([":android_op_registrations_and_gradients",
        "//tensorflow/contrib/quantization:android_ops",
        "//tensorflow/contrib/quantization/kernels:android_ops",
        "@gemmlowp//:eight_bit_int_gemm_sources"]),
    copts = tf_copts() + ["-Iexternal/gemmlowp"],
    linkopts = ["-lz"],
    tags = [
        "manual",
        "notap",
    ],
    visibility = ["//visibility:public"],
    deps = [
        ":android_tensorflow_lib_lite",
        ":protos_cc",
        "//tensorflow/core/kernels:android_tensorflow_kernels",
        "//third_party/eigen3",
    ],
    alwayslink = 1,
)

[dwightcrow]

Worked great, thanks so much @andrewharp!!!!

Really appreciate prompt reply, transcribed answer (along with git pull --recurse-submodules and removing all earlier "fixes") to StackOverflow so hopefully will prevent other n00bs hitting you with same question.

Thanks again!