Home | History | Annotate | Download | only in distributed_runtime 
      1 /* Copyright 2016 The TensorFlow Authors. All Rights Reserved.
      2 
      3 Licensed under the Apache License, Version 2.0 (the "License");
      4 you may not use this file except in compliance with the License.
      5 You may obtain a copy of the License at
      6 
      7     http://www.apache.org/licenses/LICENSE-2.0
      8 
      9 Unless required by applicable law or agreed to in writing, software
     10 distributed under the License is distributed on an "AS IS" BASIS,
     11 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
     12 See the License for the specific language governing permissions and
     13 limitations under the License.
     14 ==============================================================================*/
     15 
     16 #include "tensorflow/core/distributed_runtime/remote_device.h"
     17 
     18 #include <vector>
     19 
     20 #include "tensorflow/core/common_runtime/device.h"
     21 #include "tensorflow/core/common_runtime/process_util.h"
     22 #include "tensorflow/core/distributed_runtime/worker_cache.h"
     23 #include "tensorflow/core/distributed_runtime/worker_interface.h"
     24 #include "tensorflow/core/lib/core/errors.h"
     25 #include "tensorflow/core/lib/gtl/cleanup.h"
     26 #include "tensorflow/core/platform/logging.h"
     27 #include "tensorflow/core/platform/macros.h"
     28 #include "tensorflow/core/protobuf/worker.pb.h"
     29 
     30 namespace tensorflow {
     31 
     32 // TODO(zhifengc): We need to consolidate (full/partial) device name
     33 // parsing into one place.
     34 //
     35 // Parses and returns the local device part (e.g., cpu:0, gpu:4).
     36 string GetLocalDeviceName(StringPiece fullname) {
     37   auto pos = fullname.rfind('/');
     38   CHECK_NE(pos, StringPiece::npos);
     39   fullname.remove_prefix(pos + 1);
     40   return fullname.ToString();
     41 }
     42 
     43 class RemoteDevice : public Device {
     44  public:
     45   RemoteDevice(Env* env, const DeviceAttributes& da)
     46       : Device(env, da), local_dev_name_(GetLocalDeviceName(da.name())) {}
     47 
     48   Status Sync() override { return Status::OK(); }
     49   Allocator* GetAllocator(AllocatorAttributes attr) override { return nullptr; }
     50 
     51  private:
     52   const string local_dev_name_;
     53 
     54   TF_DISALLOW_COPY_AND_ASSIGN(RemoteDevice);
     55 };
     56 
     57 void NewRemoteDevices(Env* env, WorkerCacheInterface* worker_cache,
     58                       const string& worker_name, NewRemoteDevicesDone done) {
     59   WorkerInterface* wi = worker_cache->CreateWorker(worker_name);
     60   if (wi == nullptr) {
     61     std::vector<Device*> empty;
     62     done(errors::NotFound("Device ", worker_name, " is not found."), &empty);
     63     return;
     64   }
     65   struct Call {
     66     GetStatusRequest req;
     67     GetStatusResponse resp;
     68   };
     69   Call* call = new Call;
     70   auto cb = [env, worker_cache, worker_name, done, wi,
     71              call](const Status& status) {
     72     Status s = status;
     73     std::vector<Device*> remote_devices;
     74     auto cleanup = gtl::MakeCleanup(
     75         [&worker_cache, &worker_name, &wi, &done, &remote_devices, &s, call] {
     76           worker_cache->ReleaseWorker(worker_name, wi);
     77           done(s, &remote_devices);
     78           delete call;
     79         });
     80     if (s.ok()) {
     81       DeviceNameUtils::ParsedName worker_name_parsed;
     82       if (!DeviceNameUtils::ParseFullName(worker_name, &worker_name_parsed) ||
     83           !worker_name_parsed.has_job || !worker_name_parsed.has_replica ||
     84           !worker_name_parsed.has_task) {
     85         s = errors::InvalidArgument("Could not parse worker name: ",
     86                                     worker_name);
     87         LOG(WARNING) << s;
     88         return;
     89       }
     90       remote_devices.reserve(call->resp.device_attributes_size());
     91       for (const DeviceAttributes& da : call->resp.device_attributes()) {
     92         DeviceNameUtils::ParsedName device_name_parsed;
     93         CHECK(DeviceNameUtils::ParseFullName(da.name(), &device_name_parsed))
     94             << "Device attribute name '" << da.name() << "' could not be "
     95             << "parsed. Device Attribute: " << da.DebugString();
     96         // Preserve the exact name, if possible.
     97         // TODO(b/37868888): Simplify when legacy device name formats removed.
     98         if (device_name_parsed.job == worker_name_parsed.job &&
     99             device_name_parsed.replica == worker_name_parsed.replica &&
    100             device_name_parsed.task == worker_name_parsed.task) {
    101           auto d = new RemoteDevice(env, da);
    102           remote_devices.push_back(d);
    103         } else {
    104           DeviceAttributes da_rewritten = da;
    105           da_rewritten.set_name(DeviceNameUtils::FullName(
    106               worker_name_parsed.job, worker_name_parsed.replica,
    107               worker_name_parsed.task, device_name_parsed.type,
    108               device_name_parsed.id));
    109           auto d = new RemoteDevice(env, da_rewritten);
    110           remote_devices.push_back(d);
    111         }
    112       }
    113     }
    114   };
    115   wi->GetStatusAsync(&call->req, &call->resp, cb);
    116 }
    117 
    118 }  // namespace tensorflow
    119