1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
|
/* Copyright 2017 The TensorFlow Authors. All Rights Reserved.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
==============================================================================*/
#include <memory>
#include <utility>
#include "absl/memory/memory.h"
#include "tensorflow/compiler/xla/client/xla_builder.h"
#include "tensorflow/compiler/xla/literal_util.h"
#include "tensorflow/compiler/xla/service/cpu/custom_call_target_registry.h"
#include "tensorflow/compiler/xla/service/hlo_computation.h"
#include "tensorflow/compiler/xla/service/hlo_instruction.h"
#include "tensorflow/compiler/xla/service/hlo_module.h"
#include "tensorflow/compiler/xla/service/hlo_opcode.h"
#include "tensorflow/compiler/xla/shape_util.h"
#include "tensorflow/compiler/xla/tests/client_library_test_base.h"
#include "tensorflow/compiler/xla/tests/hlo_test_base.h"
#include "tensorflow/compiler/xla/tests/literal_test_util.h"
#include "tensorflow/compiler/xla/tests/test_macros.h"
#include "tensorflow/compiler/xla/xla_data.pb.h"
#include "tensorflow/core/platform/dynamic_annotations.h"
#include "tensorflow/core/platform/macros.h"
#include "tensorflow/core/platform/test.h"
namespace {
void R0F32Add2(float* out, float** in) {
TF_ANNOTATE_MEMORY_IS_INITIALIZED(in, sizeof(float*));
*out = **in + 2.0f;
}
void R2F32ReduceSum(float* out, float** in) {
TF_ANNOTATE_MEMORY_IS_INITIALIZED(in, sizeof(float) * 4);
float* array = in[0];
*out = array[0] + array[1] + array[2] + array[3];
}
void Add1ToValues(float* out, float** in) {
TF_ANNOTATE_MEMORY_IS_INITIALIZED(in, sizeof(float) * 4);
float* array = in[0];
out[0] = array[0] + 1;
out[1] = array[1] + 1;
out[2] = array[2] + 1;
out[3] = array[3] + 1;
}
} // namespace
REGISTER_CUSTOM_CALL_TARGET(R0F32Add2);
REGISTER_CUSTOM_CALL_TARGET(R2F32ReduceSum);
REGISTER_CUSTOM_CALL_TARGET(Add1ToValues);
namespace xla {
namespace {
class CustomCallTest : public HloTestBase {
protected:
Shape r0f32_ = ShapeUtil::MakeShape(F32, {});
Shape r2f32_ = ShapeUtil::MakeShape(F32, {2, 2});
};
XLA_TEST_F(CustomCallTest, DISABLED_ON_GPU(CustomCallR0F32Add2)) {
auto module = CreateNewModule();
auto builder = HloComputation::Builder(TestName());
auto constant = builder.AddInstruction(
HloInstruction::CreateConstant(LiteralUtil::CreateR0<float>(42.0f)));
builder.AddInstruction(
HloInstruction::CreateCustomCall(r0f32_, {constant}, "R0F32Add2"));
module->AddEntryComputation(builder.Build());
Literal result = ExecuteAndTransfer(std::move(module), {});
LiteralTestUtil::ExpectR0Near<float>(44.0f, result, error_spec_);
}
XLA_TEST_F(CustomCallTest, DISABLED_ON_GPU(CustomCallR2F32Reduce)) {
auto module = CreateNewModule();
auto builder = HloComputation::Builder(TestName());
Array2D<float> array(2, 2);
array(0, 0) = 1.0f;
array(0, 1) = 2.0f;
array(1, 0) = 3.0f;
array(1, 1) = 4.0f;
auto constant = builder.AddInstruction(
HloInstruction::CreateConstant(LiteralUtil::CreateR2FromArray2D(array)));
builder.AddInstruction(
HloInstruction::CreateCustomCall(r0f32_, {constant}, "R2F32ReduceSum"));
module->AddEntryComputation(builder.Build());
Literal result = ExecuteAndTransfer(std::move(module), {});
LiteralTestUtil::ExpectR0Near<float>(10.0f, result, error_spec_);
}
XLA_TEST_F(CustomCallTest,
DISABLED_ON_GPU(CustomCall_UsedInOtherComputations)) {
auto module = CreateNewModule();
auto b = HloComputation::Builder(TestName());
auto input = b.AddInstruction(
HloInstruction::CreateConstant(LiteralUtil::CreateR2FromArray2D(
Array2D<float>{{1.0f, 2.0f}, {3.0f, 4.0f}})));
auto incremented = b.AddInstruction(HloInstruction::CreateCustomCall(
ShapeUtil::MakeShape(F32, {1, 2, 2}), {input}, "Add1ToValues"));
auto incremented_again = b.AddInstruction(HloInstruction::CreateCustomCall(
ShapeUtil::MakeShape(F32, {1, 2, 2}), {incremented}, "Add1ToValues"));
// Concatenate the values along first dim.
b.AddInstruction(
HloInstruction::CreateConcatenate(ShapeUtil::MakeShape(F32, {2, 2, 2}),
{incremented, incremented_again}, 0));
module->AddEntryComputation(b.Build());
Literal result = ExecuteAndTransfer(std::move(module), {});
LiteralTestUtil::ExpectR3EqualArray3D<float>(
Array3D<float>{{{2, 3}, {4, 5}}, {{3, 4}, {5, 6}}}, result);
}
class CustomCallClientAPITest : public ClientLibraryTestBase {};
// When using the client API, CustomCall targets can't begin with '$' -- these
// are reserved for internal use.
XLA_TEST_F(CustomCallClientAPITest, IllegalCustomCallTarget) {
XlaBuilder builder(TestName());
CustomCall(&builder, "$illegal", /*operands=*/{},
ShapeUtil::MakeShape(F32, {1}));
StatusOr<std::unique_ptr<GlobalData>> result =
Execute(&builder, /*arguments=*/{});
EXPECT_FALSE(result.ok());
}
} // namespace
} // namespace xla
|
