Sped up the computation of the gradients of the convolution layer wrt the

inputs by 30% on GPU.
Change: 118221540

tensorflow/core/kernels/eigen_backward_spatial_convolutions.h

@@ -38,13 +38,44 @@ namespace Eigen {
   */
 
 template <typename OutputBackward, typename Kernel>
-EIGEN_ALWAYS_INLINE
-static const typename internal::conditional<
-  internal::traits<OutputBackward>::Layout == ColMajor,
-  TensorReshapingOp<const DSizes<typename internal::traits<OutputBackward>::Index, internal::traits<OutputBackward>::NumDimensions>, const TensorContractionOp<const array<IndexPair<typename internal::traits<OutputBackward>::Index>, 2>, const TensorReshapingOp<const DSizes<typename internal::traits<OutputBackward>::Index, 3>, const TensorReverseOp<const array<bool, 4>, const Kernel> >, const TensorReshapingOp<const DSizes<typename internal::traits<OutputBackward>::Index, 3>, const TensorImagePatchOp<Dynamic, Dynamic, const OutputBackward> > > >,
-  TensorReshapingOp<const DSizes<typename internal::traits<OutputBackward>::Index, internal::traits<OutputBackward>::NumDimensions>, const TensorContractionOp<const array<IndexPair<typename internal::traits<OutputBackward>::Index>, 2>, const TensorReshapingOp<const DSizes<typename internal::traits<OutputBackward>::Index, 3>, const TensorImagePatchOp<Dynamic, Dynamic, const OutputBackward> >, const TensorReshapingOp<const DSizes<typename internal::traits<OutputBackward>::Index, 3>, const TensorReverseOp<const array<bool, 4>, const Kernel> > > > >::type
-SpatialConvolutionBackwardInput(const Kernel& kernel, const OutputBackward& output_backward, typename internal::traits<OutputBackward>::Index inputRows, typename internal::traits<OutputBackward>::Index inputCols, const DenseIndex stride = 1, const DenseIndex in_stride = 1) {
-
+EIGEN_ALWAYS_INLINE static const typename internal::conditional<
+    internal::traits<OutputBackward>::Layout == ColMajor,
+    TensorReshapingOp<
+        const DSizes<typename internal::traits<OutputBackward>::Index,
+                     internal::traits<OutputBackward>::NumDimensions>,
+        const TensorContractionOp<
+            const array<
+                IndexPair<typename internal::traits<OutputBackward>::Index>, 2>,
+            const Eigen::TensorForcedEvalOp<const TensorReshapingOp<
+                const DSizes<typename internal::traits<OutputBackward>::Index,
+                             3>,
+                const TensorReverseOp<const array<bool, 4>, const Kernel> > >,
+            const TensorReshapingOp<
+                const DSizes<typename internal::traits<OutputBackward>::Index,
+                             3>,
+                const TensorImagePatchOp<Dynamic, Dynamic,
+                                         const OutputBackward> > > >,
+    TensorReshapingOp<
+        const DSizes<typename internal::traits<OutputBackward>::Index,
+                     internal::traits<OutputBackward>::NumDimensions>,
+        const TensorContractionOp<
+            const array<
+                IndexPair<typename internal::traits<OutputBackward>::Index>, 2>,
+            const TensorReshapingOp<
+                const DSizes<typename internal::traits<OutputBackward>::Index,
+                             3>,
+                const TensorImagePatchOp<Dynamic, Dynamic,
+                                         const OutputBackward> >,
+            const Eigen::TensorForcedEvalOp<const TensorReshapingOp<
+                const DSizes<typename internal::traits<OutputBackward>::Index,
+                             3>,
+                const TensorReverseOp<const array<bool, 4>,
+                                      const Kernel> > > > > >::type
+SpatialConvolutionBackwardInput(
+    const Kernel& kernel, const OutputBackward& output_backward,
+    typename internal::traits<OutputBackward>::Index inputRows,
+    typename internal::traits<OutputBackward>::Index inputCols,
+    const DenseIndex stride = 1, const DenseIndex in_stride = 1) {
   typedef typename internal::traits<OutputBackward>::Index TensorIndex;
   typedef typename internal::traits<OutputBackward>::Scalar OutScalar;
   TensorRef<Tensor<typename internal::traits<Kernel>::Scalar, internal::traits<Kernel>::NumDimensions, internal::traits<Kernel>::Layout, TensorIndex> > kern(kernel);
@@ -167,9 +198,28 @@ SpatialConvolutionBackwardInput(const Kernel& kernel, const OutputBackward& outp
     }
   }
 
-  return choose(Cond<internal::traits<OutputBackward>::Layout == ColMajor>(),
-                kernel.reverse(kernel_reverse).reshape(kernel_dims).contract(output_backward.extract_image_patches(kernelRows, kernelCols, 1, 1, in_stride, in_stride, stride, stride, padding_top, padding_bottom, padding_left, padding_right, OutScalar(0)).reshape(pre_contract_dims), contract_dims).reshape(post_contract_dims),
-                output_backward.extract_image_patches(kernelRows, kernelCols, 1, 1, in_stride, in_stride, stride, stride, padding_top, padding_bottom, padding_left, padding_right, OutScalar(0)).reshape(pre_contract_dims).contract(kernel.reverse(kernel_reverse).reshape(kernel_dims), contract_dims).reshape(post_contract_dims));
+  return choose(
+      Cond<internal::traits<OutputBackward>::Layout == ColMajor>(),
+      kernel.reverse(kernel_reverse)
+          .reshape(kernel_dims)
+          .eval()
+          .contract(output_backward
+                        .extract_image_patches(
+                            kernelRows, kernelCols, 1, 1, in_stride, in_stride,
+                            stride, stride, padding_top, padding_bottom,
+                            padding_left, padding_right, OutScalar(0))
+                        .reshape(pre_contract_dims),
+                    contract_dims)
+          .reshape(post_contract_dims),
+      output_backward
+          .extract_image_patches(kernelRows, kernelCols, 1, 1, in_stride,
+                                 in_stride, stride, stride, padding_top,
+                                 padding_bottom, padding_left, padding_right,
+                                 OutScalar(0))
+          .reshape(pre_contract_dims)
+          .contract(kernel.reverse(kernel_reverse).reshape(kernel_dims).eval(),
+                    contract_dims)
+          .reshape(post_contract_dims));
 }