xref: /aosp_15_r20/external/tensorflow/tensorflow/core/kernels/queue_base.h (revision b6fb3261f9314811a0f4371741dbb8839866f948)
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 #ifndef TENSORFLOW_CORE_KERNELS_QUEUE_BASE_H_
17 #define TENSORFLOW_CORE_KERNELS_QUEUE_BASE_H_
18 
19 #include <deque>
20 #include <vector>
21 
22 #include "absl/base/macros.h"
23 #include "tensorflow/core/framework/op_kernel.h"
24 #include "tensorflow/core/framework/queue_interface.h"
25 #include "tensorflow/core/framework/tensor.h"
26 #include "tensorflow/core/framework/tensor_shape.h"
27 #include "tensorflow/core/framework/types.h"
28 #include "tensorflow/core/lib/gtl/array_slice.h"
29 #include "tensorflow/core/platform/macros.h"
30 #include "tensorflow/core/platform/mutex.h"
31 #include "tensorflow/core/platform/types.h"
32 
33 namespace tensorflow {
34 
35 // Functionality common to asynchronous QueueInterface implementations.
36 class QueueBase : public QueueInterface {
37  public:
38   // As a possible value of 'capacity'.
39   static constexpr int32_t kUnbounded = INT_MAX;
40 
41   // Args:
42   //   component_dtypes: The types of each component in a queue-element tuple.
43   //   component_shapes: The shapes of each component in a queue-element tuple,
44   //     which must either be empty (if the shapes are not specified) or
45   //     or have the same size as component_dtypes.
46   //   name: A name to use for the queue.
47   QueueBase(int32_t capacity, const DataTypeVector& component_dtypes,
48             const std::vector<TensorShape>& component_shapes,
49             const string& name);
50 
51   // Implementations of QueueInterface methods --------------------------------
component_dtypes()52   const DataTypeVector& component_dtypes() const override {
53     return component_dtypes_;
54   }
55 
56   Status ValidateTuple(const Tuple& tuple) override;
57   Status ValidateManyTuple(const Tuple& tuple) override;
58 
59   void Close(OpKernelContext* ctx, bool cancel_pending_enqueues,
60              DoneCallback callback) override;
61 
62   // Other public methods -----------------------------------------------------
component_shapes()63   const std::vector<TensorShape>& component_shapes() const {
64     return component_shapes_;
65   }
66 
capacity()67   int32 capacity() const { return capacity_; }
68 
is_closed()69   bool is_closed() const override {
70     mutex_lock lock(mu_);
71     return closed_;
72   }
73 
74   // Copies the index^th slice (in the first dimension) of parent into element.
75   static Status CopySliceToElement(const Tensor& parent, Tensor* element,
76                                    int64_t index);
77 
78   // Copies element into the index^th slice (in the first dimension) of parent.
79   // NOTE(mrry): This method is deprecated. Use
80   // `tensorflow::batch_util::CopySliceToElement()` defined in
81   // "./batch_util.h" instead.
82   ABSL_DEPRECATED(
83       "Use `tensorflow::batch_util::CopySliceToElement()` defined in "
84       "\"./batch_util.h\" instead.")
85   static Status CopyElementToSlice(const Tensor& element, Tensor* parent,
86                                    int64_t index);
87 
88  protected:
89   enum Action { kEnqueue, kDequeue };
90   enum RunResult { kNoProgress, kProgress, kComplete };
91 
92   // Tries to enqueue/dequeue (or close) based on whatever is at the
93   // front of enqueue_attempts_/dequeue_attempts_.  Appends to
94   // *finished the callback for any finished attempt (so it may be
95   // called once mu_ is released).  Returns true if any progress was
96   // made.
97   struct CleanUp {
CleanUpCleanUp98     CleanUp(DoneCallback&& f, CancellationToken ct, CancellationManager* cm)
99         : finished(f), to_deregister(ct), cm(cm) {}
100     DoneCallback finished;
101     CancellationToken to_deregister;
102     CancellationManager* cm;
103   };
104 
105   // Returns the number of components in a queue-element tuple.
num_components()106   int32 num_components() const { return component_dtypes_.size(); }
107 
108   // True if shapes were specified.  If so, inputs will be validated
109   // against them, etc.
specified_shapes()110   bool specified_shapes() const { return component_shapes_.size() > 0; }
111 
112   // Code common to Validate*Tuple().
113   Status ValidateTupleCommon(const Tuple& tuple) const;
114 
ManyOutShape(int i,int64_t batch_size)115   TensorShape ManyOutShape(int i, int64_t batch_size) {
116     TensorShape shape({batch_size});
117     shape.AppendShape(component_shapes_[i]);
118     return shape;
119   }
120 
121   void Cancel(Action action, CancellationManager* cancellation_manager,
122               CancellationToken token);
123 
124   // Helper for cancelling all pending Enqueue(Many) operations when
125   // Close is called with cancel_pending_enqueues.
126   void CloseAndCancel();
127 
128   bool TryAttemptLocked(Action action, std::vector<CleanUp>* clean_up)
129       TF_EXCLUSIVE_LOCKS_REQUIRED(mu_);
130 
131   // Tries to make progress on the enqueues or dequeues at the front
132   // of the *_attempts_ queues.
133   void FlushUnlocked();
134 
135   ~QueueBase() override;
136 
137   // Helpers for implementing MatchesNodeDef().
138   static string ShapeListString(const gtl::ArraySlice<TensorShape>& shapes);
139   Status MatchesNodeDefOp(const NodeDef& node_def, const string& op) const;
140   Status MatchesNodeDefCapacity(const NodeDef& node_def,
141                                 int32_t capacity) const;
142   Status MatchesNodeDefTypes(const NodeDef& node_def) const;
143   Status MatchesNodeDefShapes(const NodeDef& node_def) const;
144 
145  protected:
146   const int32 capacity_;
147   const DataTypeVector component_dtypes_;
148   const std::vector<TensorShape> component_shapes_;
149   const string name_;
150   mutable mutex mu_;
151   bool closed_ TF_GUARDED_BY(mu_);
152 
153   struct Attempt;
154   typedef std::function<RunResult(Attempt*)> RunCallback;
155   struct Attempt {
156     int32 elements_requested;
157     DoneCallback done_callback;  // must be run outside mu_
158     OpKernelContext* context;
159     CancellationManager* cancellation_manager;  // not owned
160     CancellationToken cancellation_token;
161     RunCallback run_callback;  // must be run while holding mu_
162     bool is_cancelled;
163     Tuple tuple;
164     // tuples is used by some implementations allowing dynamic shapes.
165     std::vector<Tuple> tuples;
166 
AttemptAttempt167     Attempt(int32_t elements_requested, DoneCallback done_callback,
168             OpKernelContext* context, CancellationManager* cancellation_manager,
169             CancellationToken cancellation_token, RunCallback run_callback)
170         : elements_requested(elements_requested),
171           done_callback(done_callback),
172           context(context),
173           cancellation_manager(cancellation_manager),
174           cancellation_token(cancellation_token),
175           run_callback(run_callback),
176           is_cancelled(false) {}
177   };
178   std::deque<Attempt> enqueue_attempts_ TF_GUARDED_BY(mu_);
179   std::deque<Attempt> dequeue_attempts_ TF_GUARDED_BY(mu_);
180 
181   TF_DISALLOW_COPY_AND_ASSIGN(QueueBase);
182 };
183 
184 }  // namespace tensorflow
185 
186 #endif  // TENSORFLOW_CORE_KERNELS_QUEUE_BASE_H_
187