• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /* Copyright 2015 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/framework/op_segment.h"
17 
18 #include "tensorflow/core/framework/function.h"
19 #include "tensorflow/core/framework/op_kernel.h"
20 #include "tensorflow/core/lib/core/errors.h"
21 #include "tensorflow/core/lib/gtl/map_util.h"
22 #include "tensorflow/core/platform/logging.h"
23 #include "tensorflow/core/platform/mutex.h"
24 #include "tensorflow/core/platform/types.h"
25 
26 namespace tensorflow {
27 
~Item()28 OpSegment::Item::~Item() {
29   for (auto kv : name_kernel) delete kv.second;
30 }
31 
OpSegment()32 OpSegment::OpSegment() {}
33 
~OpSegment()34 OpSegment::~OpSegment() {
35   for (auto kv : sessions_) delete kv.second;
36 }
37 
FindOrCreate(const string & session_handle,const string & node_name,OpKernel ** kernel,CreateKernelFn create_fn)38 Status OpSegment::FindOrCreate(const string& session_handle,
39                                const string& node_name, OpKernel** kernel,
40                                CreateKernelFn create_fn) {
41   {
42     mutex_lock l(mu_);
43     auto item = gtl::FindPtrOrNull(sessions_, session_handle);
44     if (item == nullptr) {
45       return errors::NotFound("Session ", session_handle, " is not found.");
46     }
47     *kernel = gtl::FindPtrOrNull(item->name_kernel, node_name);
48     if (*kernel != nullptr) {
49       return Status::OK();
50     }
51   }
52   Status s = create_fn(kernel);
53   if (!s.ok()) {
54     LOG(ERROR) << "Create kernel failed: " << s;
55     return s;
56   }
57   {
58     mutex_lock l(mu_);
59     auto item = gtl::FindPtrOrNull(sessions_, session_handle);
60     if (item == nullptr) {
61       return errors::NotFound("Session ", session_handle, " is not found.");
62     }
63     OpKernel** p_kernel = &(item->name_kernel[node_name]);
64     if (*p_kernel == nullptr) {
65       *p_kernel = *kernel;  // Inserts 'kernel' in the map.
66     } else {
67       delete *kernel;
68       *kernel = *p_kernel;
69     }
70   }
71   return Status::OK();
72 }
73 
AddHold(const string & session_handle)74 void OpSegment::AddHold(const string& session_handle) {
75   mutex_lock l(mu_);
76   Item** item = &sessions_[session_handle];
77   if (*item == nullptr) {
78     *item = new Item;  // num_holds == 1
79   } else {
80     ++((*item)->num_holds);
81   }
82 }
83 
RemoveHold(const string & session_handle)84 void OpSegment::RemoveHold(const string& session_handle) {
85   Item* item = nullptr;
86   {
87     mutex_lock l(mu_);
88     auto siter = sessions_.find(session_handle);
89     if (siter == sessions_.end()) {
90       VLOG(1) << "Session " << session_handle << " is not found.";
91       return;
92     }
93     item = siter->second;
94     if (--(item->num_holds) > 0) {
95       return;
96     } else {
97       sessions_.erase(siter);
98     }
99   }
100   delete item;
101 }
102 
ShouldOwnKernel(FunctionLibraryRuntime * lib,const string & node_op)103 bool OpSegment::ShouldOwnKernel(FunctionLibraryRuntime* lib,
104                                 const string& node_op) {
105   // OpSegment should not own kernel if the node is stateless, or a function.
106   return lib->IsStateful(node_op) &&
107          lib->GetFunctionLibraryDefinition()->Find(node_op) == nullptr &&
108          node_op != "PartitionedCall" && node_op != "StatefulPartitionedCall";
109 }
110 
111 }  // end namespace tensorflow
112