/* Copyright 2017 The TensorFlow Authors. All Rights Reserved.

Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at

    http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/

syntax = "proto3";

package tensorflow;
option cc_enable_arenas = true;

import "tensorflow/core/framework/tensor.proto";
import "tensorflow/core/framework/tensor_shape.proto";
import "tensorflow/core/framework/types.proto";
import "tensorflow/core/framework/attr_value.proto";
import "tensorflow/core/protobuf/device_properties.proto";

// Description of the session when an op is run.
message SessionInfo {
  int64 intra_op_parallelism = 1;
}

// Description of an operation as well as the parameters expected to impact its
// performance.
message OpInfo {
  // The operation name.  There may be custom parameters in attrs.
  string op = 1;

  // Custom parameters impacting the behavior of the op.
  map<string, AttrValue> attr = 2;

  // Input data types, shapes and values if known.
  message TensorProperties {
    DataType dtype = 1;
    TensorShapeProto shape = 2;
    TensorProto value = 3;
  };
  repeated TensorProperties inputs = 3;

  // Optional description of the op outputs
  repeated TensorProperties outputs = 5;

  // Device on which the operation is run.
  DeviceProperties device = 4;

  // Information about the session configs.
  SessionInfo session_info = 6;
}

message NormalDistribution {
  double mu = 1;
  double sigma = 2;
}

message LogNormalDistribution {
  double mu = 1;
  double sigma = 2;
}

// Performance data for tensorflow operations
message OpPerformance {
  // The op
  OpInfo op = 1;

  // Information about the session configs.
  SessionInfo session_info = 12 [deprecated = true];

  // The node name (optional). Makes it easier to associate the performance data
  // with a specific graph node.
  string node = 5;

  // Temporary memory used by this node (in bytes).
  int64 temporary_memory_size = 2;

  // Time it takes to run the op (in nanoseconds).
  int64 compute_cost = 3;

  // Analytical compute cost (in nanoseconds).
  int64 compute_time = 6;

  // Analytical memory access cost (in nanoseconds).
  int64 memory_time = 7;

  // Percentage of theoretical compute performance.
  double compute_efficiency = 4;

  // Percentage of theoretical memory performance.
  double memory_efficiency = 8;

  // Expected execution time, modeled using one of 2 possible distributions.
  oneof execution_time {
    NormalDistribution execution_time_normal = 10;
    LogNormalDistribution execution_time_log_normal = 11;
  };

  // Memory usage data for a tensorflow operation.
  message OpMemory {
    // The output information may have memory usage and output shapes.
    repeated int64 output_memory = 1;

    // Temp and persistent memory allocated by this node.
    int64 temp_memory = 2;
    int64 persistent_memory = 4;

    int64 device_temp_memory = 3 [deprecated = true];
    int64 device_persistent_memory = 5 [deprecated = true];
  }
  OpMemory op_memory = 9;
}

// A collection of OpPerformance data points.
message OpPerformanceList {
  repeated OpPerformance op_performance = 1;
}