From b883c28ffedcc648d476cd80be2854605c60c336 Mon Sep 17 00:00:00 2001
From: "A. Unique TensorFlower" <gardener@tensorflow.org>
Date: Wed, 4 Sep 2019 10:51:02 -0700
Subject: [PATCH] Don't push down constant tensors if they increase the amount
 of broadcasting, unless they are guaranteed to be constant folded.

PiperOrigin-RevId: 267185933
---
 .../grappler/optimizers/constant_folding.cc   | 41 +++++++++++++++++--
 .../grappler/optimizers/constant_folding.h    |  3 +-
 2 files changed, 40 insertions(+), 4 deletions(-)

diff --git a/tensorflow/core/grappler/optimizers/constant_folding.cc b/tensorflow/core/grappler/optimizers/constant_folding.cc
index f3a3e4c188a..7987089f8ea 100644
--- a/tensorflow/core/grappler/optimizers/constant_folding.cc
+++ b/tensorflow/core/grappler/optimizers/constant_folding.cc
@@ -1987,7 +1987,8 @@ Status ConstantFolding::SimplifyNode(bool use_shape_info, NodeDef* node,
   RETURN_IF_ERROR_OR_MODIFIED(SimplifyArithmeticOperations(
       *properties, use_shape_info, optimized_graph, node));
   SET_AND_RETURN_IF_MODIFIED(ReduceDivToReciprocalMul(optimized_graph, node));
-  SET_AND_RETURN_IF_MODIFIED(ConstantPushDown(optimized_graph, node));
+  SET_AND_RETURN_IF_MODIFIED(
+      ConstantPushDown(*properties, optimized_graph, node));
   SET_AND_RETURN_IF_MODIFIED(
       MulConvPushDown(optimized_graph, node, *properties));
   SET_AND_RETURN_IF_MODIFIED(PartialConstPropThroughIdentityN(node));
@@ -2840,7 +2841,8 @@ bool ConstantFolding::ReduceDivToReciprocalMul(GraphDef* optimized_graph,
   return false;
 }
 
-bool ConstantFolding::ConstantPushDown(GraphDef* optimized_graph,
+bool ConstantFolding::ConstantPushDown(const GraphProperties& properties,
+                                       GraphDef* optimized_graph,
                                        NodeDef* node) {
   // Consider the transformation
   //
@@ -2853,7 +2855,7 @@ bool ConstantFolding::ConstantPushDown(GraphDef* optimized_graph,
   // where C is constant, X is non-constant, Y may be constant or non-constant,
   // and '+' denotes an associative and commutative operator like addition or
   // multiplication. This optimization pushes constants down in the tree to
-  // canonicalize it. Moreoever, in cases where the child node has a second
+  // canonicalize it. Moreover, in cases where the child node has a second
   // constant input Y we will create a leaf node that can be folded, e.g.
   //
   //    Add(C1, Add(C2, X)) -> Add(X, Add(C1, C2)) -> Add(X, C1 + C2)
@@ -2944,6 +2946,7 @@ bool ConstantFolding::ConstantPushDown(GraphDef* optimized_graph,
       node->device() != right_leaf->device()) {
     return false;
   }
+
   // Get the node names corresponding to X, Y, and C.
   const string input_x =
       left_leaf_is_constant ? op_child->input(1) : op_child->input(0);
@@ -2957,6 +2960,38 @@ bool ConstantFolding::ConstantPushDown(GraphDef* optimized_graph,
   VLOG(1) << "\n++++++++ Reordering node " << node->name() << ": " << node->op()
           << "(" << left_child->op() << ", " << right_child->op() << ")\n";
 
+  const NodeDef* y_node = left_leaf_is_constant ? left_leaf : right_leaf;
+  if (!IsReallyConstant(*y_node)) {
+    // Now make sure that we do not push a tensor that is larger than the tensor
+    // it replaces down, since that would create more broadcasting and increase
+    // work.
+    const std::vector<OpInfo::TensorProperties>& root_props =
+        properties.GetInputProperties(node->name());
+    const std::vector<OpInfo::TensorProperties>& op_props =
+        properties.GetInputProperties(op_child->name());
+    if (!root_props.empty() && !op_props.empty()) {
+      DCHECK_EQ(2, root_props.size()) << node->DebugString();
+      DCHECK_EQ(2, op_props.size()) << op_child->DebugString();
+      const PartialTensorShape c_shape(
+          root_props[left_child_is_constant ? 0 : 1].shape());
+      const PartialTensorShape x_shape(
+          op_props[left_leaf_is_constant ? 0 : 1].shape());
+      if (c_shape.IsFullyDefined() && x_shape.IsFullyDefined() &&
+          c_shape.num_elements() > x_shape.num_elements()) {
+        return false;
+      } else if (!c_shape.unknown_rank() && !x_shape.unknown_rank() &&
+                 c_shape.dims() > 0) {
+        for (int idx = 0; idx < std::min(x_shape.dims(), c_shape.dims());
+             ++idx) {
+          if (x_shape.dim_size(idx) >= 0 &&
+              c_shape.dim_size(idx) > x_shape.dim_size(idx)) {
+            return false;
+          }
+        }
+      }
+    }
+  }
+
   // Now we have identified the nodes to swap (non_const_leaf_input and
   // const_child).
   node_map_->UpdateInput(node->name(), input_c, input_x);
diff --git a/tensorflow/core/grappler/optimizers/constant_folding.h b/tensorflow/core/grappler/optimizers/constant_folding.h
index 37c5674a011..21ad5144c24 100644
--- a/tensorflow/core/grappler/optimizers/constant_folding.h
+++ b/tensorflow/core/grappler/optimizers/constant_folding.h
@@ -145,7 +145,8 @@ class ConstantFolding : public GraphOptimizer {
 
   // Pushes down constants on '+' and '*' operators if applicable. Returns true
   // the transformation applied successfully.
-  bool ConstantPushDown(GraphDef* optimized_graph, NodeDef* node);
+  bool ConstantPushDown(const GraphProperties& properties,
+                        GraphDef* optimized_graph, NodeDef* node);
 
   // Aggregate constants present around a conv operator. Returns true if the
   // transformation was applied successfully.