[ROCm] Update eigen_contraction_kernel subroutine to make it device compatible

[jerryyin]

This PR addresses ROCm specific compilation issues. Below is the summary of the PR changes.

---

First of all, UseCustomContractionKernels() should be declared with EIGEN_DEVICE_FUNC and EIGEN_DONT_INLINE as well to be consistent with the signature from packLhs() or packRhs(), invoke().

Secondly, std::call_once() is not available in gpu device code, and this function should not be invoked in device as well. If it does happen, take the short cut and return.

---

The change has been thoroughly tested under different compilation targets in develop-upstream branch.

@tatianashp @whchung @chsigg

[jerryyin]

Looking at the test results, Linux GPU has a large number of failures on Broken by missing target @mkl_dnn//:mkldnn_single_threaded. I do realize that this change might be related because eigen_contraction_kernel_with_mkl is the only target build with mkldnn_single_threaded. However, I'm not able to view the corresponding build log online or reproduce locally.

An alternative of fixing this for the rocm target is to remove the defined macros in this target to make the chunk of code not build for rocm, i.e. "//tensorflow:using_rocm_hipcc": [], @chsigg Let me know if you prefer this alternative fix.

[deven-amd]

posting here for reference, the error(s) we see in the --config=rocm build without this fix

[[32mINFO: ^[[0mFrom Compiling tensorflow/core/kernels/conv_2d_gpu_uint64.cu.cc [for host]:                                                                                                                                                                      
clang-9: warning: /usr/bin/gcc: 'linker' input unused [-Wunused-command-line-argument]                                                                                                                                                                            
clang-9: warning: /usr/bin/gcc: 'linker' input unused [-Wunused-command-line-argument]                                                                                                                                                                            
In file included from tensorflow/core/kernels/conv_2d_gpu_uint64.cu.cc:25:                                                                                                                                                                                        
In file included from ./tensorflow/core/kernels/conv_2d.h:22:                                                                                                                                                                                                     
In file included from ./tensorflow/core/kernels/eigen_spatial_convolutions.h:26:                                                                                                                                                                                  
./tensorflow/core/kernels/eigen_contraction_kernel.h:860:1: error:  'Eigen::internal::UseCustomContractionKernels':  no overloaded function has restriction specifiers that are compatible with the ambient context 'packLhs'                                     
REGISTER_TENSOR_CONTRACTION_KERNEL_WITH_FALLBACK(float, float, float);                                                                                                                                                                                            
^                                                                                                                                                                                                                                                                 
./tensorflow/core/kernels/eigen_contraction_kernel.h:605:38: note: expanded from macro 'REGISTER_TENSOR_CONTRACTION_KERNEL_WITH_FALLBACK'                                                                                                                         
      if (UseCustomContractionKernels()) {                                     \                                                                                                                                                                                  
                                     ^                                                                                                                                                                                                                            
./tensorflow/core/kernels/eigen_contraction_kernel.h:860:1: error:  'Eigen::internal::UseCustomContractionKernels':  no overloaded function has restriction specifiers that are compatible with the ambient context 'packLhs'                                     
./tensorflow/core/kernels/eigen_contraction_kernel.h:605:38: note: expanded from macro 'REGISTER_TENSOR_CONTRACTION_KERNEL_WITH_FALLBACK'                                                                                                                         
      if (UseCustomContractionKernels()) {                                     \                                                                                                                                                                                  
                                     ^                                                                                                                                                                                                                            
./tensorflow/core/kernels/eigen_contraction_kernel.h:860:1: error:  'Eigen::internal::UseCustomContractionKernels':  no overloaded function has restriction specifiers that are compatible with the ambient context 'packRhs'                                     
./tensorflow/core/kernels/eigen_contraction_kernel.h:624:38: note: expanded from macro 'REGISTER_TENSOR_CONTRACTION_KERNEL_WITH_FALLBACK'                                                                                                                         
      if (UseCustomContractionKernels()) {                                     \                                                                                                                                                                                  
                                     ^                                                                                                                                                                                                                            
./tensorflow/core/kernels/eigen_contraction_kernel.h:860:1: error:  'Eigen::internal::UseCustomContractionKernels':  no overloaded function has restriction specifiers that are compatible with the ambient context 'packRhs'                                     
./tensorflow/core/kernels/eigen_contraction_kernel.h:624:38: note: expanded from macro 'REGISTER_TENSOR_CONTRACTION_KERNEL_WITH_FALLBACK'                                                                                                                         
      if (UseCustomContractionKernels()) {                                     \                                                                                                                                                                                  
                                     ^                                                                                                                                                                                                                            
./tensorflow/core/kernels/eigen_contraction_kernel.h:860:1: error:  'Eigen::internal::UseCustomContractionKernels':  no overloaded function has restriction specifiers that are compatible with the ambient context 'invoke'                                      
./tensorflow/core/kernels/eigen_contraction_kernel.h:644:38: note: expanded from macro 'REGISTER_TENSOR_CONTRACTION_KERNEL_WITH_FALLBACK'                                                                                                                         
      if (UseCustomContractionKernels()) {                                     \                                                                                                                                                                                  
                                     ^                                                                                                                                                                                                                            
./tensorflow/core/kernels/eigen_contraction_kernel.h:860:1: error:  'Eigen::internal::UseCustomContractionKernels':  no overloaded function has restriction specifiers that are compatible with the ambient context 'invoke'                                      
./tensorflow/core/kernels/eigen_contraction_kernel.h:644:38: note: expanded from macro 'REGISTER_TENSOR_CONTRACTION_KERNEL_WITH_FALLBACK'                                                                                                                         
      if (UseCustomContractionKernels()) {                                     \                                                                                                                                                                                  
                                     ^                                                                                                                                                                                                                            
6 errors generated.                                                                                                                                                                                                                                               
                                                                                                                                          

One curious aspect of these errors, is that the they do not cause the build to terminate, and the build actually finishes with a non-error exit status! That was the reason we failed to spot it before.

Have you (TF developers) seen this with errors (in Eigen headers) before, where in the errors do not lead to a build failure? just curious.