/external/tensorflow/tensorflow/core/distributed_runtime/ |
D | partial_run_mgr_test.cc | 27 int step_id = 1; in TEST() local 29 partial_run_mgr.FindOrCreate(step_id, &cancellation_manager); in TEST() 36 int step_id = 1; in TEST() local 38 partial_run_mgr.FindOrCreate(step_id, &cancellation_manager); in TEST() 41 partial_run_mgr.FindOrCreate(step_id, &found_cancellation_manager); in TEST() 48 int step_id = 1; in TEST() local 50 partial_run_mgr.FindOrCreate(step_id, &cancellation_manager); in TEST() 62 int step_id = 1; in TEST() local 64 partial_run_mgr.FindOrCreate(step_id, &cancellation_manager); in TEST() 68 step_id, [&called](Status status) { called++; }, Status::OK()); in TEST() [all …]
|
D | worker.cc | 108 void Worker::AbortStep(int64 step_id) { in AbortStep() argument 109 Rendezvous* rendez = env_->rendezvous_mgr->Find(step_id); in AbortStep() 110 SchedNonBlockingClosureAfter(1000000, [rendez, step_id]() { in AbortStep() 114 rendez->StartAbort(errors::Aborted("Step ", step_id, in AbortStep() 164 const int64 step_id = request->step_id(); in DoRunGraph() local 165 TRACEPRINTF("RunGraph: %lld", step_id); in DoRunGraph() 206 opts->SetCancelCallback([this, cm, step_id]() { in DoRunGraph() 209 AbortStep(step_id); in DoRunGraph() 225 request->graph_handle(), step_id, session.get(), request->exec_opts(), in DoRunGraph() 227 [this, step_id, response, session, cm, out, token, collector, in DoRunGraph() [all …]
|
D | partial_run_mgr.cc | 22 bool PartialRunMgr::FindOrCreate(int step_id, in FindOrCreate() argument 25 auto it = step_id_to_partial_run_.find(step_id); in FindOrCreate() 36 step_id_to_partial_run_[step_id] = std::move(partial_run); in FindOrCreate() 40 void PartialRunMgr::ExecutorDone(int step_id, const Status& executor_status) { in ExecutorDone() argument 45 auto run_it = step_id_to_partial_run_.find(step_id); in ExecutorDone() 63 step_id_to_partial_run_.erase(step_id); in ExecutorDone() 67 void PartialRunMgr::PartialRunDone(int step_id, StatusCallback done, in PartialRunDone() argument 72 auto run_it = step_id_to_partial_run_.find(step_id); in PartialRunDone() 88 step_id_to_partial_run_.erase(step_id); in PartialRunDone()
|
D | graph_mgr.cc | 145 [this, session](const int64 step_id, const DeviceMgr*, in InitItem() 147 auto* remote_r = this->worker_env_->rendezvous_mgr->Find(step_id); in InitItem() 152 [this](const int64 step_id) { in InitItem() argument 153 this->worker_env_->rendezvous_mgr->Cleanup(step_id); in InitItem() 352 Status GraphMgr::SendInputs(const int64 step_id, const NamedTensors& in) { in SendInputs() argument 353 Rendezvous* rendezvous = worker_env_->rendezvous_mgr->Find(step_id); in SendInputs() 371 Status GraphMgr::RecvOutputs(const int64 step_id, NamedTensors* out) { in RecvOutputs() argument 372 Rendezvous* rendezvous = worker_env_->rendezvous_mgr->Find(step_id); in RecvOutputs() 378 s = errors::Internal("Failed to fetch outputs for step ", step_id, in RecvOutputs() 389 void GraphMgr::RecvOutputsAsync(const int64 step_id, NamedTensors* out, in RecvOutputsAsync() argument [all …]
|
D | worker_cache_logger.cc | 59 bool WorkerCacheLogger::RetrieveLogs(int64 step_id, StepStats* ss) { in RetrieveLogs() argument 61 LogMap::iterator iter = log_map_.find(step_id); in RetrieveLogs() 71 void WorkerCacheLogger::Save(const string& device, int64 step_id, in Save() argument 74 StepLog* sl = &log_map_[step_id]; in Save() 85 void WorkerCacheLogger::RecordRecvTensor(int64 step_id, int64 start_usecs, in RecordRecvTensor() argument 91 RecordDataTransfer(step_id, start_usecs, end_usecs, tensor_name, src_device, in RecordRecvTensor() 95 void WorkerCacheLogger::RecordDataTransfer(int64 step_id, int64 start_usecs, in RecordDataTransfer() argument 130 Save(dst_device, step_id, ns); in RecordDataTransfer()
|
D | rpc_collective_executor_mgr.cc | 50 CollectiveExecutor* RpcCollectiveExecutorMgr::Create(int64 step_id) { in Create() argument 53 work_queue_, worker_cache_, step_id, in Create() 55 return new BaseCollectiveExecutor(this, rma, step_id, dev_mgr_, in Create() 64 int64 step_id = random::New64(); in NewRandomStepId() local 66 step_id &= kStepIdMask; in NewRandomStepId() 67 return step_id; in NewRandomStepId() 158 void RpcCollectiveExecutorMgr::RetireStepId(int64 graph_key, int64 step_id) { in RetireStepId() argument 162 if (step_id == it->second->next_step_id_) { in RetireStepId()
|
D | base_rendezvous_mgr.cc | 56 RemoteRendezvous* BaseRendezvousMgr::Find(int64 step_id) { in Find() argument 57 return FindOrCreate(step_id); in Find() 60 BaseRemoteRendezvous* BaseRendezvousMgr::FindOrCreate(int64 step_id) { in FindOrCreate() argument 62 auto iter = table_.find(step_id); in FindOrCreate() 64 auto rr = Create(step_id, worker_env_); in FindOrCreate() 65 iter = table_.insert({step_id, rr}).first; in FindOrCreate() 71 void BaseRendezvousMgr::RecvLocalAsync(int64 step_id, in RecvLocalAsync() argument 74 auto rendez = FindOrCreate(step_id); in RecvLocalAsync() 85 Status BaseRendezvousMgr::RecvLocal(int64 step_id, in RecvLocal() argument 90 RecvLocalAsync(step_id, parsed, in RecvLocal() [all …]
|
D | base_rendezvous_mgr.h | 72 RemoteRendezvous* Find(int64 step_id) override; 78 void RecvLocalAsync(int64 step_id, const Rendezvous::ParsedKey& parsed, 82 Status RecvLocal(int64 step_id, const Rendezvous::ParsedKey& parsed, 89 void Cleanup(int64 step_id) override; 92 virtual BaseRemoteRendezvous* Create(int64 step_id, 105 BaseRemoteRendezvous* FindOrCreate(int64 step_id); 118 BaseRemoteRendezvous(const WorkerEnv* env, int64 step_id);
|
D | rendezvous_mgr_interface.h | 77 virtual RemoteRendezvous* Find(int64 step_id) = 0; 83 virtual void RecvLocalAsync(int64 step_id, 88 virtual Status RecvLocal(int64 step_id, const Rendezvous::ParsedKey& parsed, 95 virtual void Cleanup(int64 step_id) = 0;
|
D | worker_cache_logger.h | 48 bool RetrieveLogs(int64 step_id, StepStats* ss); 59 void RecordRecvTensor(int64 step_id, int64 start_usecs, int64 end_usecs, 65 void RecordDataTransfer(int64 step_id, int64 start_usecs, int64 end_usecs, 84 void Save(const string& device, int64 step_id, NodeExecStats* ns);
|
D | master_session.cc | 159 void RetrieveLogs(int64 step_id, StepStats* ss) { in RetrieveLogs() argument 161 worker_cache_->RetrieveLogs(step_id, ss); in RetrieveLogs() 165 req.add_fetch_step_id(step_id); in RetrieveLogs() 174 [step_id, ss, resp, &scoped_mu, &all_done](const Status& s) { in RetrieveLogs() 179 if (step_id != lss.step_id()) { in RetrieveLogs() 204 Status RunPartitions(const MasterEnv* env, int64 step_id, 209 Status RunPartitions(const MasterEnv* env, int64 step_id, 216 void CleanupPartitionsAsync(int64 step_id, StatusCallback done); 219 void ProcessStats(int64 step_id, PerStepState* pss, ProfileHandler* ph, 323 const FetchListType& fetches, const MasterEnv* env, int64 step_id, [all …]
|
D | graph_mgr.h | 90 void ExecuteAsync(const string& handle, const int64 step_id, 97 Status SendInputs(const int64 step_id, const NamedTensors& in); 98 Status RecvOutputs(const int64 step_id, NamedTensors* out); 99 void RecvOutputsAsync(const int64 step_id, NamedTensors* out, 166 void StartParallelExecutors(const string& handle, int64 step_id, Item* item,
|
D | partial_run_mgr.h | 53 bool FindOrCreate(int step_id, CancellationManager** cancellation_manager); 58 void ExecutorDone(int step_id, const Status& executor_status); 66 void PartialRunDone(int step_id, StatusCallback done, const Status& status);
|
/external/tensorflow/tensorflow/core/distributed_runtime/rpc/ |
D | rpc_rendezvous_mgr_test.cc | 140 const int64 step_id = 123; in TEST_F() local 145 RemoteRendezvous* rendez = rmgr_.Find(step_id); in TEST_F() 154 TF_ASSERT_OK(rmgr_.RecvLocal(step_id, key, &val, &val_dead)); in TEST_F() 157 rmgr_.Cleanup(step_id); in TEST_F() 165 const int64 step_id = 123; in TEST_F() local 166 RemoteRendezvous* rendez = rmgr_.Find(step_id); in TEST_F() 179 const int64 step_id = 321; in TEST_F() local 180 RemoteRendezvous* rendez = rmgr_.Find(step_id); in TEST_F() 182 SchedClosure([this, step_id]() { in TEST_F() 184 rmgr_.Cleanup(step_id); in TEST_F() [all …]
|
D | grpc_response_cache.cc | 22 bool GrpcResponseCache::QueueRequest(int64 request_id, int64 step_id, in QueueRequest() argument 52 entry.step_id = step_id; in QueueRequest() 101 void GrpcResponseCache::CleanEntriesForStep(int64 step_id) { in CleanEntriesForStep() argument 106 if (it->second.step_id == step_id) { in CleanEntriesForStep()
|
D | grpc_worker_service.cc | 446 const int64 step_id = request->step_id(); in GrpcRecvTensorAsync() local 464 response_cache_->QueueRequest(request_id, step_id, do_response)) { in GrpcRecvTensorAsync() 491 TRACEPRINTF("RecvTensor: %lld %s", step_id, key.c_str()); in GrpcRecvTensorAsync() 510 [step_id]() { LOG(WARNING) << "RecvTensor cancelled for " << step_id; }); in GrpcRecvTensorAsync() 512 step_id, parsed, in GrpcRecvTensorAsync() 589 const int64 step_id = request->step_id(); in RecvBufAsync() local 608 response_cache_->QueueRequest(request_id, step_id, do_response)) { in RecvBufAsync() 635 env_->collective_executor_mgr->FindOrCreate(step_id), true); in RecvBufAsync() 673 "GrpcWorker::RecvBufAsync::consumer_callback", request->step_id(), in RecvBufAsync() 719 for (const auto& step_id : request->fetch_step_id()) { in LoggingAsync() local [all …]
|
/external/tensorflow/tensorflow/core/framework/ |
D | log_memory.cc | 40 void LogMemory::RecordStep(const int64 step_id, const string& handle) { in RecordStep() argument 42 step.set_step_id(step_id); in RecordStep() 48 const int64 step_id, in RecordTensorAllocation() argument 51 allocation.set_step_id(step_id); in RecordTensorAllocation() 66 const int64 step_id, const int index, in RecordTensorOutput() argument 69 output.set_step_id(step_id); in RecordTensorOutput() 77 const int64 step_id, size_t num_bytes, in RecordRawAllocation() argument 80 allocation.set_step_id(step_id); in RecordRawAllocation() 90 const int64 step_id, void* ptr, in RecordRawDeallocation() argument 93 deallocation.set_step_id(step_id); in RecordRawDeallocation()
|
D | log_memory.h | 62 static void RecordStep(int64 step_id, const std::string& handle); 71 int64 step_id, const Tensor& tensor); 83 static void RecordTensorOutput(const std::string& kernel_name, int64 step_id, 95 static void RecordRawAllocation(const std::string& operation, int64 step_id, 104 static void RecordRawDeallocation(const std::string& operation, int64 step_id,
|
D | rendezvous.h | 151 cleanup_fn_([](const int64 step_id) { return Status::OK(); }) {} in Factory() 155 Status operator()(const int64 step_id, const DeviceMgr* device_mgr, in operator() 157 return create_fn_(step_id, device_mgr, rendez); in operator() 160 Status CleanUp(const int64 step_id) const { return cleanup_fn_(step_id); } in CleanUp() argument
|
/external/tensorflow/tensorflow/core/common_runtime/ |
D | collective_executor_mgr.cc | 46 CollectiveExecutor* CollectiveExecutorMgr::FindOrCreate(int64 step_id) { in FindOrCreate() argument 50 auto it = executor_table_.find(step_id); in FindOrCreate() 54 ce = Create(step_id); in FindOrCreate() 55 executor_table_[step_id] = ce; in FindOrCreate() 62 CollectiveExecutor* CollectiveExecutorMgr::Create(int64 step_id) { in Create() argument 64 new CollectiveRemoteAccessLocal(dev_mgr_, dev_resolver_.get(), step_id); in Create() 65 return new BaseCollectiveExecutor(this, rma, step_id, dev_mgr_, in Create() 69 void CollectiveExecutorMgr::Cleanup(int64 step_id) { in Cleanup() argument 73 auto it = executor_table_.find(step_id); in Cleanup()
|
D | scoped_allocator_mgr.cc | 129 void ScopedAllocatorMgr::Cleanup(int64 step_id) { in Cleanup() argument 131 auto it = per_step_map_.find(step_id); in Cleanup() 138 ScopedAllocatorContainer* ScopedAllocatorMgr::GetContainer(int64 step_id) { in GetContainer() argument 139 VLOG(2) << "GetContainer " << step_id << " on " << device_name(); in GetContainer() 142 auto it = per_step_map_.find(step_id); in GetContainer() 144 sac = new ScopedAllocatorContainer(this, step_id); in GetContainer() 145 per_step_map_[step_id] = sac; in GetContainer() 153 const Tensor& backing_tensor, int64 step_id, int32 scope_id, in AddScopedAllocator() argument 157 ScopedAllocatorContainer* sac = GetContainer(step_id); in AddScopedAllocator()
|
D | test_collective_executor_mgr.h | 72 CollectiveExecutor* FindOrCreate(int64 step_id) override { in FindOrCreate() argument 75 auto iter = table_.find(step_id); in FindOrCreate() 80 table_[step_id] = ce; in FindOrCreate() 86 void Cleanup(int64 step_id) override { in Cleanup() argument 88 auto iter = table_.find(step_id); in Cleanup() 123 void RetireStepId(int64 graph_key, int64 step_id) override {} in RetireStepId() argument
|
D | scoped_allocator_mgr.h | 48 ScopedAllocatorContainer(const ScopedAllocatorMgr* mgr, int64 step_id) in ScopedAllocatorContainer() argument 49 : mgr_(mgr), step_id_(step_id) {} in ScopedAllocatorContainer() 80 ScopedAllocatorContainer* GetContainer(int64 step_id); 84 const Tensor& backing_tensor, int64 step_id, int32 scope_id, 89 void Cleanup(int64 step_id);
|
D | collective_executor_mgr.h | 37 CollectiveExecutor* FindOrCreate(int64 step_id) override; 39 void Cleanup(int64 step_id) override; 64 void RetireStepId(int64 graph_key, int64 step_id) override {} in RetireStepId() argument 68 virtual CollectiveExecutor* Create(int64 step_id);
|
/external/tensorflow/tensorflow/compiler/mlir/tensorflow/transforms/ |
D | test_visitor_util.cc | 46 int step_id = 0; in runOnFunction() local 48 op->emitRemark() << step_id++ << ": " << get_stage_description(stage); in runOnFunction() 53 op.emitRemark() << step_id++ << ": " << get_stage_description(stage); in runOnFunction() 63 int step_id = 0; in runOnFunction() local 82 op->emitRemark() << step_id++ << ": " << get_stage_description(stage); in runOnFunction() 90 func.emitRemark() << step_id++ << ": walk was interrupted"; in runOnFunction() 99 func.emitRemark() << step_id++ << ": walk was interrupted"; in runOnFunction()
|