Performance microtweaks: Pass by reference rather than by value; pre-reserve capacity when total vectoroid size is known.

PiperOrigin-RevId: 201172723

1 files changed, 2 insertions, 0 deletions

diff --git a/tensorflow/compiler/xla/service/shape_inference.cc b/tensorflow/compiler/xla/service/shape_inference.cc
index e25f5e67c7..4606d8f202 100644
--- a/tensorflow/compiler/xla/service/shape_inference.cc
+++ b/tensorflow/compiler/xla/service/shape_inference.cc
@@ -939,6 +939,7 @@ ShapeInference::InferDegenerateDimensionBroadcastShape(HloOpcode operation,
     HloOpcode opcode,
     tensorflow::gtl::ArraySlice<const HloInstruction*> operands) {
   std::vector<const Shape*> operand_shapes;
+  operand_shapes.reserve(operands.size());
   for (const HloInstruction* operand : operands) {
     operand_shapes.push_back(&operand->shape());
   }
@@ -954,6 +955,7 @@ ShapeInference::InferDegenerateDimensionBroadcastShape(HloOpcode operation,
   switch (opcode) {
     case HloOpcode::kTuple: {
       Shape result = ShapeUtil::MakeTupleShape({});
+      result.mutable_tuple_shapes()->Reserve(operand_shapes.size());
       for (const Shape* shape : operand_shapes) {
         ShapeUtil::AppendShapeToTuple(*shape, &result);
       }