tensorflow: Auto-Configuration Error: Cannot find cudnn.h at /usr/lib/x86_64-linux-gnu/include/cudnn.h

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

#4397, but it’s for 16.04

Environment info

Operating System: Ubuntu 14.04 Installed version of CUDA and cuDNN: 8.0, 5.1 (please attach the output of ls -l /path/to/cuda/lib/libcud*):

ls -l /path/to/cuda/lib/libcud*
-rw-r--r-- 1 root 543K Jan 26 15:48 /usr/local/cuda/lib64/libcudadevrt.a
lrwxrwxrwx 1 root   16 Jan 26 15:51 /usr/local/cuda/lib64/libcudart.so -> libcudart.so.8.0
lrwxrwxrwx 1 root   19 Jan 26 15:51 /usr/local/cuda/lib64/libcudart.so.8.0 -> libcudart.so.8.0.61
-rw-r--r-- 1 root 406K Jan 26 15:48 /usr/local/cuda/lib64/libcudart.so.8.0.61
-rw-r--r-- 1 root 757K Jan 26 15:48 /usr/local/cuda/lib64/libcudart_static.a

ll /usr/lib/x86_64-linux-gnu/libcudnn*                                                                                                                                                                
lrwxrwxrwx 1 root  29 Mar  1 11:58 /usr/lib/x86_64-linux-gnu/libcudnn.so -> /etc/alternatives/libcudnn_so
lrwxrwxrwx 1 root  18 Nov  6 23:19 /usr/lib/x86_64-linux-gnu/libcudnn.so.5 -> libcudnn.so.5.1.10
-rw-r--r-- 1 root 81M Nov  6 23:19 /usr/lib/x86_64-linux-gnu/libcudnn.so.5.1.10
lrwxrwxrwx 1 root  32 Mar  1 11:58 /usr/lib/x86_64-linux-gnu/libcudnn_static.a -> /etc/alternatives/libcudnn_stlib
-rw-r--r-- 1 root 68M Nov  6 23:19 /usr/lib/x86_64-linux-gnu/libcudnn_static_v5.a

If installed from source, provide

1. The commit hash (git rev-parse HEAD): e895d5ca395c2362df4f5c8f08b68501b41f8a98
2. The output of bazel version:

INFO: Reading 'startup' options from /root/.bazelrc: --batch
Build label: 0.4.2
Build target: bazel-out/local-fastbuild/bin/src/main/java/com/google/devtools/build/lib/bazel/BazelServer_deploy.jar
Build time: Wed Dec 7 18:47:11 2016 (1481136431)
Build timestamp: 1481136431
Build timestamp as int: 1481136431

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

I ran

./configure

It failed with

INFO: Reading 'startup' options from /root/.bazelrc: --batch
INFO: Starting clean (this may take a while). Consider using --expunge_async if the clean takes more than several minutes.
INFO: Reading 'startup' options from /root/.bazelrc: --batch
ERROR: package contains errors: tensorflow/compiler/tests.
ERROR: error loading package 'tensorflow/compiler/tests': Encountered error while reading extension file 'cuda/build_defs.bzl': no such package '@local_config_cuda//cuda': Traceback (most recent call last):
        File "/tensorflow/third_party/gpus/cuda_configure.bzl", line 816
                _create_cuda_repository(repository_ctx)
        File "/tensorflow/third_party/gpus/cuda_configure.bzl", line 729, in _create_cuda_repository
                _get_cuda_config(repository_ctx)
        File "/tensorflow/third_party/gpus/cuda_configure.bzl", line 585, in _get_cuda_config
                _cudnn_version(repository_ctx, cudnn_install_base..., ...)
        File "/tensorflow/third_party/gpus/cuda_configure.bzl", line 296, in _cudnn_version
                _find_cuda_define(repository_ctx, cudnn_install_base..., ...)
        File "/tensorflow/third_party/gpus/cuda_configure.bzl", line 271, in _find_cuda_define
                auto_configure_fail("Cannot find cudnn.h at %s" % st...))
        File "/tensorflow/third_party/gpus/cuda_configure.bzl", line 93, in auto_configure_fail
                fail("
%sAuto-Configuration Error:%s ...))

Auto-Configuration Error: Cannot find cudnn.h at /usr/lib/x86_64-linux-gnu/include/cudnn.h

I noticed the cudnn.h is in /usr/include/cudnn.h somehow, while its libs are in /usr/lib/x86_64-linux-gnu/.

I read cuda_configure.bzl, but didn’t find an easy way to let it find both cudnn headers and libs.