// total samples collected.
func (p *Profile) Merge(pb *Profile, r float64) error {
	if err := p.Compatible(pb); err != nil {
		return err
	}

	pb = pb.Copy()

	// Keep the largest of the two periods.
	if pb.Period > p.Period {
		p.Period = pb.Period
	}

	p.DurationNanos += pb.DurationNanos

	p.Mapping = append(p.Mapping, pb.Mapping...)
	for i, m := range p.Mapping {
		m.ID = uint64(i + 1)

}

TEST(CreateExportedModelTest, CreateExportedModelWithAddedAssetFileDefs) {
  AssetFileDef asset1;
  ASSERT_TRUE(
      TextFormat::ParseFromString(R"pb(filename: "fname1")pb", &asset1));

  AssetFileDef asset2;
  ASSERT_TRUE(
      TextFormat::ParseFromString(R"pb(filename: "fname2")pb", &asset2));

  const ExportedModel exported_model = CreateExportedModelFromGraphDef(
      GraphDef(), /*init_node_name=*/"", /*checkpoint_dir=*/"",

}

func BenchmarkPinnerPinUnpinParallel(b *testing.B) {
	b.RunParallel(func(pb *testing.PB) {
		p := new(obj)
		for pb.Next() {
			var pinner runtime.Pinner
			pinner.Pin(p)
			pinner.Unpin()
		}
	})
}

func BenchmarkPinnerPinUnpinParallelTiny(b *testing.B) {
	b.RunParallel(func(pb *testing.PB) {
		p := new(bool)
		for pb.Next() {
			var pinner runtime.Pinner
			pinner.Pin(p)
			pinner.Unpin()

// WholeModel-DAG: "tf.DumpTensor"(%[[matmul0_q]]) <{enabled = true, file_name = "unquantized_tensor_data.pb", func_name = "composite_dot_general_with_bias_and_relu6_dynamic_fn_2", log_dir_path = "/tmp/dumps/composite_dot_general_with_bias_and_relu6_dynamic_fn_2", node_name = "_empty_node"}> : (tensor<?x2xf32>) -> ()

	myc := make(chan int)
	b.RunParallel(func(pb *testing.PB) {
		for pb.Next() {
			select {
			case <-myc:
			default:
			}
		}
	})
}

func BenchmarkSelectUncontended(b *testing.B) {
	b.RunParallel(func(pb *testing.PB) {
		myc1 := make(chan int, 1)
		myc2 := make(chan int, 1)
		myc1 <- 0
		for pb.Next() {
			select {
			case <-myc1:
				myc2 <- 0

				b.ResetTimer()
				t := time.Now()
				var ops int64
				var lat int64
				b.SetParallelism(par)
				b.RunParallel(func(pb *testing.PB) {
					rng := rand.New(rand.NewSource(time.Now().UnixNano()))
					n := 0
					var latency int64
					managers := grid.Managers
					hosts := grid.Hosts
					for pb.Next() {
						// Pick a random manager.
						src, dst := rng.Intn(len(managers)), rng.Intn(len(managers))
						if src == dst {

	authorizationMeta = "authorization"
)

type mockCAServer struct {
	pb.UnimplementedIstioCertificateServiceServer
	Certs         []string
	Authenticator *security.FakeAuthenticator
	Err           error
}

func (ca *mockCAServer) CreateCertificate(ctx context.Context, in *pb.IstioCertificateRequest) (*pb.IstioCertificateResponse, error) {
	if ca.Authenticator != nil {

		<-c
	}
}

func BenchmarkMutexUncontended(b *testing.B) {
	b.ResetTimer()
	b.ReportAllocs()

	type PaddedMutex struct {
		*DRWMutex
	}
	b.RunParallel(func(pb *testing.PB) {
		mu := PaddedMutex{NewDRWMutex(ds, "")}
		for pb.Next() {
			mu.Lock(id, source)
			mu.Unlock(context.Background())
		}
	})
}

func benchmarkMutex(b *testing.B, slack, work bool) {
	b.ResetTimer()
	b.ReportAllocs()

}

// Next reports whether there are more iterations to execute.
func (pb *PB) Next() bool {
	if pb.cache == 0 {
		n := pb.globalN.Add(pb.grain)
		if n <= pb.bN {
			pb.cache = pb.grain
		} else if n < pb.bN+pb.grain {
			pb.cache = pb.bN + pb.grain - n
		} else {
			return false
		}
	}
	pb.cache--
	return true
}

// RunParallel runs a benchmark in parallel.

#include "tensorflow/core/framework/attr_value.pb.h"
#include "tensorflow/core/framework/function.pb.h"
#include "tensorflow/core/framework/graph.pb.h"
#include "tensorflow/core/framework/node_def_util.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/framework/tensor.pb.h"
#include "tensorflow/core/framework/tensor_shape.h"
#include "tensorflow/core/framework/types.pb.h"
#include "tensorflow/core/lib/gtl/flatmap.h"