cpp_padding_optimizer.cc

Go to the documentation of this file.

// Protocol Buffers - Google's data interchange format
// Copyright 2008 Google Inc.  All rights reserved.
// https://developers.google.com/protocol-buffers/
//
// Redistribution and use in source and binary forms, with or without
// modification, are permitted provided that the following conditions are
// met:
//
//     * Redistributions of source code must retain the above copyright
// notice, this list of conditions and the following disclaimer.
//     * Redistributions in binary form must reproduce the above
// copyright notice, this list of conditions and the following disclaimer
// in the documentation and/or other materials provided with the
// distribution.
//     * Neither the name of Google Inc. nor the names of its
// contributors may be used to endorse or promote products derived from
// this software without specific prior written permission.
//
// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
// OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
// SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
// LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
// DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
// THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
// (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
// OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 
#include <google/protobuf/compiler/cpp/cpp_padding_optimizer.h>
 
#include <google/protobuf/compiler/cpp/cpp_helpers.h>
 
namespace google {
namespace protobuf {
namespace compiler {
namespace cpp {
 
namespace {
 
// FieldGroup is just a helper for PaddingOptimizer below. It holds a vector of
// fields that are grouped together because they have compatible alignment, and
// a preferred location in the final field ordering.
class FieldGroup {
 public:
  FieldGroup() : preferred_location_(0) {}
 
  // A group with a single field.
  FieldGroup(float preferred_location, const FieldDescriptor* field)
      : preferred_location_(preferred_location), fields_(1, field) {}
 
  // Append the fields in 'other' to this group.
  void Append(const FieldGroup& other) {
    if (other.fields_.empty()) {
      return;
    }
    // Preferred location is the average among all the fields, so we weight by
    // the number of fields on each FieldGroup object.
    preferred_location_ = (preferred_location_ * fields_.size() +
                           (other.preferred_location_ * other.fields_.size())) /
                          (fields_.size() + other.fields_.size());
    fields_.insert(fields_.end(), other.fields_.begin(), other.fields_.end());
  }
 
  void SetPreferredLocation(float location) { preferred_location_ = location; }
  const std::vector<const FieldDescriptor*>& fields() const { return fields_; }
 
  // FieldGroup objects sort by their preferred location.
  bool operator<(const FieldGroup& other) const {
    return preferred_location_ < other.preferred_location_;
  }
 
 private:
  // "preferred_location_" is an estimate of where this group should go in the
  // final list of fields.  We compute this by taking the average index of each
  // field in this group in the original ordering of fields.  This is very
  // approximate, but should put this group close to where its member fields
  // originally went.
  float preferred_location_;
  std::vector<const FieldDescriptor*> fields_;
  // We rely on the default copy constructor and operator= so this type can be
  // used in a vector.
};
 
}  // namespace
 
// Reorder 'fields' so that if the fields are output into a c++ class in the new
// order, fields of similar family (see below) are together and within each
// family, alignment padding is minimized.
//
// We try to do this while keeping each field as close as possible to its field
// number order so that we don't reduce cache locality much for function that
// access each field in order.  Originally, OptimizePadding used declaration
// order for its decisions, but generated code minus the serializer/parsers uses
// the output of OptimizePadding as well (stored in
// MessageGenerator::optimized_order_).  Since the serializers use field number
// order, we use that as a tie-breaker.
//
// We classify each field into a particular "family" of fields, that we perform
// the same operation on in our generated functions.
//
// REPEATED is placed first, as the C++ compiler automatically initializes
// these fields in layout order.
//
// STRING is grouped next, as our Clear/SharedCtor/SharedDtor walks it and
// calls ArenaStringPtr::Destroy on each.
//
// LAZY_MESSAGE is grouped next, as it interferes with the ability to memset
// non-repeated fields otherwise.
//
// MESSAGE is grouped next, as our Clear/SharedDtor code walks it and calls
// delete on each.  We initialize these fields with a NULL pointer (see
// MessageFieldGenerator::GenerateConstructorCode), which allows them to be
// memset.
//
// ZERO_INITIALIZABLE is memset in Clear/SharedCtor
//
// OTHER these fields are initialized one-by-one.
void PaddingOptimizer::OptimizeLayout(
    std::vector<const FieldDescriptor*>* fields, const Options& options) {
  // The sorted numeric order of Family determines the declaration order in the
  // memory layout.
  enum Family {
    REPEATED = 0,
    STRING = 1,
    // Laying out LAZY_MESSAGE before MESSAGE allows a single memset to zero
    // MESSAGE and ZERO_INITIALIZABLE fields together.
    LAZY_MESSAGE = 2,
    MESSAGE = 3,
    ZERO_INITIALIZABLE = 4,
    OTHER = 5,
    kMaxFamily
  };
 
  // First divide fields into those that align to 1 byte, 4 bytes or 8 bytes.
  std::vector<FieldGroup> aligned_to_1[kMaxFamily];
  std::vector<FieldGroup> aligned_to_4[kMaxFamily];
  std::vector<FieldGroup> aligned_to_8[kMaxFamily];
  for (int i = 0; i < fields->size(); ++i) {
    const FieldDescriptor* field = (*fields)[i];
 
    Family f = OTHER;
    if (field->is_repeated()) {
      f = REPEATED;
    } else if (field->cpp_type() == FieldDescriptor::CPPTYPE_STRING) {
      f = STRING;
    } else if (field->cpp_type() == FieldDescriptor::CPPTYPE_MESSAGE) {
      f = MESSAGE;
      if (IsLazy(field, options)) {
        f = LAZY_MESSAGE;
      }
    } else if (CanInitializeByZeroing(field)) {
      f = ZERO_INITIALIZABLE;
    }
 
    const int j = field->number();
    switch (EstimateAlignmentSize(field)) {
      case 1:
        aligned_to_1[f].push_back(FieldGroup(j, field));
        break;
      case 4:
        aligned_to_4[f].push_back(FieldGroup(j, field));
        break;
      case 8:
        aligned_to_8[f].push_back(FieldGroup(j, field));
        break;
      default:
        GOOGLE_LOG(FATAL) << "Unknown alignment size " << EstimateAlignmentSize(field)
                   << "for a field " << field->full_name() << ".";
    }
  }
 
  // For each family, group fields to optimize padding.
  for (int f = 0; f < kMaxFamily; f++) {
    // Now group fields aligned to 1 byte into sets of 4, and treat those like a
    // single field aligned to 4 bytes.
    for (int i = 0; i < aligned_to_1[f].size(); i += 4) {
      FieldGroup field_group;
      for (int j = i; j < aligned_to_1[f].size() && j < i + 4; ++j) {
        field_group.Append(aligned_to_1[f][j]);
      }
      aligned_to_4[f].push_back(field_group);
    }
    // Sort by preferred location to keep fields as close to their field number
    // order as possible.  Using stable_sort ensures that the output is
    // consistent across runs.
    std::stable_sort(aligned_to_4[f].begin(), aligned_to_4[f].end());
 
    // Now group fields aligned to 4 bytes (or the 4-field groups created above)
    // into pairs, and treat those like a single field aligned to 8 bytes.
    for (int i = 0; i < aligned_to_4[f].size(); i += 2) {
      FieldGroup field_group;
      for (int j = i; j < aligned_to_4[f].size() && j < i + 2; ++j) {
        field_group.Append(aligned_to_4[f][j]);
      }
      if (i == aligned_to_4[f].size() - 1) {
        if (f == OTHER) {
          // Move incomplete 4-byte block to the beginning.  This is done to
          // pair with the (possible) leftover blocks from the
          // ZERO_INITIALIZABLE family.
          field_group.SetPreferredLocation(-1);
        } else {
          // Move incomplete 4-byte block to the end.
          field_group.SetPreferredLocation(fields->size() + 1);
        }
      }
      aligned_to_8[f].push_back(field_group);
    }
    // Sort by preferred location.
    std::stable_sort(aligned_to_8[f].begin(), aligned_to_8[f].end());
  }
 
  // Now pull out all the FieldDescriptors in order.
  fields->clear();
  for (int f = 0; f < kMaxFamily; ++f) {
    for (int i = 0; i < aligned_to_8[f].size(); ++i) {
      fields->insert(fields->end(), aligned_to_8[f][i].fields().begin(),
                     aligned_to_8[f][i].fields().end());
    }
  }
}
 
}  // namespace cpp
}  // namespace compiler
}  // namespace protobuf
}  // namespace google