Leverage index list to further speedup the computation of the convolution

gradients by 1 to 10% depending on the size of the convolution kernel.
Change: 118505660

tensorflow/core/kernels/eigen_backward_spatial_convolutions.h

@@ -44,6 +44,16 @@ namespace Eigen {
  * that the same order is used in the input, the kernel, and the output.
   *
   */
+#ifdef EIGEN_HAS_INDEX_LIST
+typedef IndexList<type2index<0>, type2index<0>, type2index<1>, type2index<1> >
+    ReverseColMajor;
+typedef IndexList<type2index<1>, type2index<1>, type2index<0>, type2index<0> >
+    ReverseRowMajor;
+#else
+typedef array<bool, 4> ReverseColMajor;
+typedef array<bool, 4> ReverseRowMajor;
+#endif
+
 template <typename OutputBackward, typename Kernel>
 EIGEN_ALWAYS_INLINE static const typename internal::conditional<
     internal::traits<OutputBackward>::Layout == ColMajor,
@@ -59,7 +69,7 @@ EIGEN_ALWAYS_INLINE static const typename internal::conditional<
                 const TensorShufflingOp<
                     const array<
                         typename internal::traits<OutputBackward>::Index, 4>,
-                    const TensorReverseOp<const array<bool, 4>,
+                    const TensorReverseOp<const ReverseColMajor,
                                           const Kernel> > > >,
             const TensorReshapingOp<
                 const DSizes<typename internal::traits<OutputBackward>::Index,
@@ -83,7 +93,7 @@ EIGEN_ALWAYS_INLINE static const typename internal::conditional<
                 const TensorShufflingOp<
                     const array<
                         typename internal::traits<OutputBackward>::Index, 4>,
-                    const TensorReverseOp<const array<bool, 4>,
+                    const TensorReverseOp<const ReverseRowMajor,
                                           const Kernel> > > > > > >::type
 SpatialConvolutionBackwardInput(
     const Kernel& kernel, const OutputBackward& output_backward,
@@ -161,7 +171,11 @@ SpatialConvolutionBackwardInput(
   // TODO(yangke): we can make things slightly faster by collapsing the
   // dimensions
   // where we don't reverse. Try that once we have a faster compiler.
-  array<bool, 4> kernel_reverse;
+  typedef typename internal::conditional<isColMajor, ReverseColMajor,
+                                         ReverseRowMajor>::type Reverse;
+  Reverse kernel_reverse;
+
+#ifndef EIGEN_HAS_INDEX_LIST
   if (isColMajor) {
     kernel_reverse[0] = false;
     kernel_reverse[1] = false;
@@ -173,6 +187,7 @@ SpatialConvolutionBackwardInput(
     kernel_reverse[2] = false;
     kernel_reverse[3] = false;
   }
+#endif
 
   // Reorder the dimensions to filters X patch_rows X patch_cols X channels
   array<TensorIndex, 4> kernel_shuffle;