• Home
  • Line#
  • Scopes#
  • Navigate#
  • Raw
  • Download
1 /* Copyright 2019 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_COMPILER_XLA_SERVICE_DYNAMIC_PADDER_H_
17 #define TENSORFLOW_COMPILER_XLA_SERVICE_DYNAMIC_PADDER_H_
18 
19 #include "tensorflow/compiler/xla/service/dynamic_dimension_inference.h"
20 #include "tensorflow/compiler/xla/service/hlo_pass_interface.h"
21 
22 namespace xla {
23 
24 // With bounded shapes, only part of the shape contains effective data and the
25 // rest contains padded data, whose value can be anything depending on the
26 // source of the data. When a bounded shape is directly consumed by an
27 // instruction that collapses dimensions (reduce for example), the padding data
28 // would affect result of the instruction.
29 //
30 // DynamicPadder uses DynamicDimensionInference to detect bounded shapes in a
31 // hlo module, it then inserts certain instructions to reset the padding into an
32 // identity value so that in doesn't affect the result of subsequent
33 // instruction. For example, it'd reset the padding to 0 before a bounded shape
34 // is consumed by a reduce-sum.
35 //
36 // Dynamic_padder removes dynamic shapes from the entry computation, and inserts
37 // custom calls (with dynamic shapes), which are lowered by specialized
38 // emitters: PadToStatic and SliceToDynamic.
39 
40 // Each instruction can have one of the three modes in supporting dynamic
41 // lowering.
42 enum OpDynamismSupport {
43   // There is no support for dynamic lowering -- dynamic padder will make sure
44   // the input to that op has static bound by rewriting the op (e.g, extra space
45   // in reduce_sum will be padded with 0).
46   kNoSupport = 0,
47   // The op can take either dynamic input or static input.
48   kOptional,
49   // The op only has a dynamic lowering, dynamic padder will make sure the input
50   // to this op is in dynamic form.
51   kRequired,
52 };
53 
54 class DynamicPadder : public HloModulePass {
55  public:
56   // Returns true if given instruction supports native dynamic lowering. If so,
57   // dynamic padder will not attempt to pad it.
58   using OpSupportsDynamismHandler =
59       std::function<OpDynamismSupport(HloInstruction*)>;
60 
61   // If `slice_dynamic_output` is true, insert 'slice_to_dynamic' ops to all
62   // outputs that are inferred to be dynamic.
63   explicit DynamicPadder(
64       bool slice_dynamic_output = true,
65       DynamicDimensionInference::CustomCallInferenceHandler
66           custom_call_handler = nullptr,
67       OpSupportsDynamismHandler op_supports_dynamism_handler = nullptr)
slice_dynamic_output_(slice_dynamic_output)68       : slice_dynamic_output_(slice_dynamic_output),
69         custom_call_handler_(custom_call_handler),
70         op_supports_dynamism_handler_(op_supports_dynamism_handler) {}
71 
name()72   absl::string_view name() const override { return "dynamic_padder"; }
73 
74   StatusOr<bool> Run(HloModule* module) override;
75 
76  private:
77   // Insert 'slice_to_dynamic' ops to all outputs that are inferred to be
78   // dynamic.
79   bool slice_dynamic_output_;
80 
81   // A handler for dynamic dimension inference of custom calls.
82   DynamicDimensionInference::CustomCallInferenceHandler custom_call_handler_;
83 
84   // A handler to indicate if a given hlo instruction support native dynamism
85   // lowering.
86   OpSupportsDynamismHandler op_supports_dynamism_handler_;
87 };
88 
89 }  // namespace xla
90 
91 #endif  // TENSORFLOW_COMPILER_XLA_SERVICE_DYNAMIC_PADDER_H_
92