Add support to TPU automatic profiler.

PiperOrigin-RevId: 275086152
Change-Id: Iedbb55bbf3fd2da4f9678e61dc79d7919c33ae64

tensorflow/core/profiler/BUILD

@@ -43,13 +43,22 @@ tf_proto_library(
     cc_api_version = 2,
 )
 
+tf_proto_library(
+    name = "profiler_service_monitor_result_proto",
+    srcs = ["profiler_service_monitor_result.proto"],
+    cc_api_version = 2,
+)
+
 tf_proto_library(
     name = "profiler_service_proto",
     srcs = ["profiler_service.proto"],
     has_services = 1,
     cc_api_version = 2,
     cc_grpc_version = 1,
-    protodeps = [":op_profile_proto"] + tf_additional_all_protos(),
+    protodeps = [
+        ":op_profile_proto",
+        ":profiler_service_monitor_result_proto",
+    ] + tf_additional_all_protos(),
 )
 
 tf_proto_library(
@@ -69,6 +78,7 @@ tf_proto_library(
             "profiler_service.proto",
             "profiler_analysis.proto",
             "op_profile.proto",
+            "profiler_service_monitor_result.proto",
         ],
     ),
     cc_api_version = 2,

tensorflow/core/profiler/profiler_service.proto

@@ -5,6 +5,7 @@ package tensorflow;
 import "tensorflow/core/framework/graph.proto";
 import "tensorflow/core/profiler/op_profile.proto";
 import "tensorflow/core/protobuf/config.proto";
+import "tensorflow/core/profiler/profiler_service_monitor_result.proto";
 
 // The ProfilerService service retrieves performance information about
 // the programs running on connected devices over a period of time.
@@ -128,23 +129,9 @@ message MonitorRequest {
 message MonitorResponse {
   // Properly formatted string data that can be directly returned back to user.
   string data = 1;
-  // The following are the raw components used to construct field data.
-  // Percentage of time when device is idle.
-  double device_idle_time_percent = 2;
-  // TPU matrix unit utilization percentage.
-  double matrix_unit_utilization_percent = 3;
-  // Average step time in millisecond.
-  double step_time_ms_avg = 4;
-  // Minimum step time in millisecond.
-  double step_time_ms_min = 5;
-  // Maximum step time in millisecond.
-  double step_time_ms_max = 6;
-  // Average infeed percentage.
-  double infeed_percent_avg = 7;
-  // Minimum infeed percentage.
-  double infeed_percent_min = 8;
-  // Maximum infeed percentage.
-  double infeed_percent_max = 9;
 
-  // next-field: 10
+  // A collection of monitoring results for each field show in data.
+  ProfilerServiceMonitorResult monitor_result = 2;
+
+  // next-field: 3
 }

tensorflow/core/profiler/profiler_service_monitor_result.proto

@@ -0,0 +1,39 @@
+syntax = "proto3";
+
+package tensorflow;
+
+message ProfilerServiceMonitorResult {
+  // Represents the different types of responses from the profiling service.
+  enum ResponseType {
+    // No result is returned from the profiling service.
+    EMPTY_RESULT = 0;
+    // Only device utilization is available.
+    UTIL_ONLY = 1;
+    // Both device utilization and device idle time are available.
+    UTIL_IDLE = 2;
+    // Device utilization, device idle time, step time, and infeed percentage
+    // are all available.
+    UTIL_IDLE_STEP = 3;
+  }
+
+  // Type of profiling responses.
+  ResponseType response_type = 1;
+  // Percentage of time when device is idle.
+  double device_idle_time_percent = 2;
+  // TPU matrix unit utilization percentage.
+  double matrix_unit_utilization_percent = 3;
+  // Average step time in millisecond.
+  double step_time_ms_avg = 4;
+  // Minimum step time in millisecond.
+  double step_time_ms_min = 5;
+  // Maximum step time in millisecond.
+  double step_time_ms_max = 6;
+  // Average infeed percentage.
+  double infeed_percent_avg = 7;
+  // Minimum infeed percentage.
+  double infeed_percent_min = 8;
+  // Maximum infeed percentage.
+  double infeed_percent_max = 9;
+
+  // next-field: 10
+}