tensorflow--tensorflow/tensorflow/python/framework/test_ops.cc

/* Copyright 2015 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.
==============================================================================*/

// This file defines ops and op kernels that are only used by Python tests.

#include "tensorflow/python/framework/test_ops.h"

#include "absl/time/clock.h"
#include "absl/time/time.h"
#include "tensorflow/core/framework/common_shape_fns.h"
#include "tensorflow/core/framework/op.h"
#include "tensorflow/core/framework/op_kernel.h"
#include "tensorflow/core/framework/resource_handle.pb.h"
#include "tensorflow/core/framework/resource_mgr.h"
#include "tensorflow/core/lib/core/status.h"
#include "tensorflow/core/platform/errors.h"
#include "tensorflow/core/platform/tensor_float_32_utils.h"
#include "tensorflow/core/public/version.h"

namespace tensorflow {

REGISTER_OP("KernelLabel")
    .Output("result: string")
    .SetShapeFn(shape_inference::ScalarShape);

REGISTER_OP("KernelLabelRequired")
    .Input("input: int32")
    .Output("result: string")
    .SetShapeFn([](shape_inference::InferenceContext* c) {
      shape_inference::ShapeHandle out;
      TF_RETURN_IF_ERROR(c->WithRank(c->input(0), 1, &out));
      c->set_output(0, c->Scalar());
      return absl::OkStatus();
    });

REGISTER_OP("GraphDefVersion")
    .Output("version: int32")
    .SetIsStateful()
    .SetShapeFn(shape_inference::ScalarShape);

REGISTER_OP("RequiresOlderGraphVersion")
    .Output("version: int32")
    .SetIsStateful()
    .SetShapeFn([](shape_inference::InferenceContext* c) {
      if (c->graph_def_version() != TF_GRAPH_DEF_VERSION - 1) {
        return absl::InvalidArgumentError("Wrong graph version for shape");
      }
      return shape_inference::ScalarShape(c);
    });

REGISTER_OP("Old")
    .SetShapeFn(shape_inference::UnknownShape)
    .Deprecated(8, "For reasons");

REGISTER_OP("GetDeadline")
    .Output("deadline_from_epoch_micros: int64")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("SleepOp")
    .Input("sleep_seconds: int32")
    .SetIsStateful()
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("SleepIdentityOp")
    .Input("sleep_seconds: int32")
    .Input("input: T")
    .Output("output: T")
    .Attr("T: type")
    .SetIsStateful()
    .SetShapeFn(shape_inference::UnchangedShape);

REGISTER_RESOURCE_HANDLE_OP(StubResource);

REGISTER_OP("ResourceInitializedOp")
    .Input("resource: resource")
    .Output("initialized: bool")
    .SetShapeFn(shape_inference::ScalarShape);

REGISTER_OP("ResourceCreateOp")
    .Input("resource: resource")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("ResourceUsingOp")
    .Input("resource: resource")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("IsResourceHandleRefCounting")
    .Input("handle: resource")
    .Output("result: bool")
    .SetShapeFn(shape_inference::ScalarShape);

REGISTER_OP("MakeWeakResourceHandle")
    .Input("handle: resource")
    .Output("dup: resource")
    .SetIsStateful()
    .SetShapeFn(tensorflow::shape_inference::ScalarShape);

REGISTER_OP("TestStringOutput")
    .Input("input: float")
    .Output("output1: float")
    .Output("output2: string")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("Namespace>TestStringOutput")
    .Input("input: float")
    .Output("output1: float")
    .Output("output2: string")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("TestAttr")
    .Output("out: T")
    .Attr("T: {float, double}")
    .SetDoNotOptimize()
    .SetShapeFn(shape_inference::UnknownShape);

namespace {
enum KernelLabel { DEFAULT_LABEL, OVERLOAD_1_LABEL, OVERLOAD_2_LABEL };
}  // namespace

template <KernelLabel KL>
class KernelLabelOp : public OpKernel {
 public:
  using OpKernel::OpKernel;

  void Compute(OpKernelContext* ctx) override {
    Tensor* output;
    OP_REQUIRES_OK(ctx,
                   ctx->allocate_output("result", TensorShape({}), &output));
    switch (KL) {
      case DEFAULT_LABEL:
        output->scalar<tstring>()() = "My label is: default";
        break;
      case OVERLOAD_1_LABEL:
        output->scalar<tstring>()() = "My label is: overload_1";
        break;
      case OVERLOAD_2_LABEL:
        output->scalar<tstring>()() = "My label is: overload_2";
        break;
    }
  }
};

REGISTER_KERNEL_BUILDER(Name("KernelLabel").Device(DEVICE_CPU),
                        KernelLabelOp<DEFAULT_LABEL>);
REGISTER_KERNEL_BUILDER(
    Name("KernelLabel").Device(DEVICE_CPU).Label("overload_1"),
    KernelLabelOp<OVERLOAD_1_LABEL>);
REGISTER_KERNEL_BUILDER(
    Name("KernelLabel").Device(DEVICE_CPU).Label("overload_2"),
    KernelLabelOp<OVERLOAD_2_LABEL>);

// All "KernelLabelRequired" kernels have labels
REGISTER_KERNEL_BUILDER(
    Name("KernelLabelRequired").Device(DEVICE_CPU).Label("overload_1"),
    KernelLabelOp<OVERLOAD_1_LABEL>);
REGISTER_KERNEL_BUILDER(
    Name("KernelLabelRequired").Device(DEVICE_CPU).Label("overload_2"),
    KernelLabelOp<OVERLOAD_2_LABEL>);

class GraphDefVersionOp : public OpKernel {
 public:
  explicit GraphDefVersionOp(OpKernelConstruction* ctx)
      : OpKernel(ctx), graph_def_version_(ctx->graph_def_version()) {}

  void Compute(OpKernelContext* ctx) override {
    Tensor* output = nullptr;
    OP_REQUIRES_OK(ctx, ctx->allocate_output(0, TensorShape({}), &output));
    output->scalar<int>()() = graph_def_version_;
  }

 private:
  const int graph_def_version_;
};

REGISTER_KERNEL_BUILDER(Name("GraphDefVersion").Device(DEVICE_CPU),
                        GraphDefVersionOp);

class OldOp : public OpKernel {
 public:
  explicit OldOp(OpKernelConstruction* ctx) : OpKernel(ctx) {}

  void Compute(OpKernelContext* ctx) override {}
};

REGISTER_KERNEL_BUILDER(Name("Old").Device(DEVICE_CPU), OldOp);

class GetDeadlineOp : public OpKernel {
 public:
  explicit GetDeadlineOp(OpKernelConstruction* ctx) : OpKernel(ctx) {}

  void Compute(OpKernelContext* ctx) override {
    if (!ctx->deadline()) {
      ctx->SetStatus(absl::InvalidArgumentError("Deadline has not ben set."));
      return;
    }
    Tensor* output;
    OP_REQUIRES_OK(ctx, ctx->allocate_output(0, TensorShape({}), &output));
    output->scalar<int64_t>()() = absl::ToUnixMicros(*ctx->deadline());
  }
};

REGISTER_KERNEL_BUILDER(Name("GetDeadline").Device(DEVICE_CPU), GetDeadlineOp);

class SleepOp : public OpKernel {
 public:
  explicit SleepOp(OpKernelConstruction* ctx) : OpKernel(ctx) {}

  void Compute(OpKernelContext* ctx) override {
    OP_REQUIRES(ctx, TensorShapeUtils::IsScalar(ctx->input(0).shape()),
                absl::InvalidArgumentError(
                    absl::StrCat("Expected argument 0 to be a scalar. Received",
                                 ctx->input(0).DebugString())));
    absl::SleepFor(absl::Seconds(ctx->input(0).scalar<int>()()));
  }
};

REGISTER_KERNEL_BUILDER(Name("SleepOp").Device(DEVICE_CPU), SleepOp);

#if GOOGLE_CUDA
class SleepGpuOp : public OpKernel {
 public:
  explicit SleepGpuOp(OpKernelConstruction* ctx) : OpKernel(ctx) {}

  void Compute(OpKernelContext* ctx) override {
    GpuSleep(ctx, ctx->input(0).scalar<int>()());
  }
};

REGISTER_KERNEL_BUILDER(
    Name("SleepOp").Device(DEVICE_GPU).HostMemory("sleep_seconds"), SleepGpuOp);
#endif  // GOOGLE_CUDA

class SleepIdentityOp : public OpKernel {
 public:
  explicit SleepIdentityOp(OpKernelConstruction* ctx) : OpKernel(ctx) {}

  void Compute(OpKernelContext* ctx) override {
    OP_REQUIRES(ctx, TensorShapeUtils::IsScalar(ctx->input(0).shape()),
                absl::InvalidArgumentError(
                    absl::StrCat("Expected argument 0 to be a scalar. Received",
                                 ctx->input(0).DebugString())));
    absl::SleepFor(absl::Seconds(ctx->input(0).scalar<int>()()));
    ctx->set_output(0, ctx->input(1));
  }
};

REGISTER_KERNEL_BUILDER(Name("SleepIdentityOp").Device(DEVICE_CPU),
                        SleepIdentityOp);

// Stubbed-out resource to test resource handle ops.
class StubResource : public ResourceBase {
 public:
  std::string DebugString() const override { return ""; }
};

REGISTER_RESOURCE_HANDLE_KERNEL(StubResource);

REGISTER_KERNEL_BUILDER(Name("ResourceInitializedOp").Device(DEVICE_CPU),
                        IsResourceInitialized<StubResource>);

class ResourceCreateOp : public OpKernel {
 public:
  explicit ResourceCreateOp(OpKernelConstruction* c) : OpKernel(c) {}

  void Compute(OpKernelContext* c) override {
    OP_REQUIRES_OK(c,
                   CreateResource(c, HandleFromInput(c, 0), new StubResource));
  }
};

REGISTER_KERNEL_BUILDER(Name("ResourceCreateOp").Device(DEVICE_CPU),
                        ResourceCreateOp);

// Uses a ResourceHandle to check its validity.
class ResourceUsingOp : public OpKernel {
 public:
  explicit ResourceUsingOp(OpKernelConstruction* context) : OpKernel(context) {}

  void Compute(OpKernelContext* ctx) override {
    StubResource* unused;
    OP_REQUIRES_OK(ctx, LookupResource<StubResource>(
                            ctx, HandleFromInput(ctx, 0), &unused));
  }
};

REGISTER_KERNEL_BUILDER(Name("ResourceUsingOp").Device(DEVICE_CPU),
                        ResourceUsingOp);

class IsResourceHandleRefCountingOp : public OpKernel {
 public:
  explicit IsResourceHandleRefCountingOp(OpKernelConstruction* ctx)
      : OpKernel(ctx) {}

  void Compute(OpKernelContext* ctx) override {
    const auto& handle = HandleFromInput(ctx, 0);
    Tensor* output;
    OP_REQUIRES_OK(ctx, ctx->allocate_output(0, {}, &output));
    output->flat<bool>()(0) = handle.IsRefCounting();
  }
};

REGISTER_KERNEL_BUILDER(Name("IsResourceHandleRefCounting").Device(DEVICE_CPU),
                        IsResourceHandleRefCountingOp);

// Duplicates a ResourceHandle as a weak ResourceHandle.
class MakeWeakResourceHandleOp : public OpKernel {
 public:
  explicit MakeWeakResourceHandleOp(OpKernelConstruction* c) : OpKernel(c) {}

  void Compute(OpKernelContext* ctx) override {
    Tensor tensor;
    ResourceHandleProto proto;
    HandleFromInput(ctx, 0).AsProto(&proto);

    AllocatorAttributes attr;
    attr.set_on_host(true);
    OP_REQUIRES_OK(
        ctx, ctx->allocate_temp(DT_RESOURCE, TensorShape({}), &tensor, attr));
    tensor.scalar<ResourceHandle>()() = ResourceHandle{proto};
    ctx->set_output(0, tensor);
  }
};

REGISTER_KERNEL_BUILDER(Name("MakeWeakResourceHandle").Device(DEVICE_CPU),
                        MakeWeakResourceHandleOp);
REGISTER_KERNEL_BUILDER(Name("MakeWeakResourceHandle").Device(DEVICE_DEFAULT),
                        MakeWeakResourceHandleOp);

class TestAttrOp : public OpKernel {
 public:
  explicit TestAttrOp(OpKernelConstruction* ctx) : OpKernel(ctx) {}

  void Compute(OpKernelContext* ctx) override {
    Tensor* output;
    OP_REQUIRES_OK(ctx, ctx->allocate_output(0, TensorShape({}), &output));
    output->scalar<float>()() = 1.0;
  }
};

REGISTER_KERNEL_BUILDER(
    Name("TestAttr").Device(DEVICE_CPU).TypeConstraint<float>("T"), TestAttrOp);

// Various test ops without kernels. These are used to test graph construction.

REGISTER_OP("A")
    .Output("out: float32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("B")
    .Output("out: float32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("Foo1")
    .Input("a: float32")
    .Input("b: int32")
    .Input("c: int32")
    .Output("d: float32")
    .Output("e: int32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("Foo2")
    .Input("a: float32")
    .Input("b: string")
    .Input("c: string")
    .Output("d: float32")
    .Output("e: int32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("Foo3")
    .Input("a: float32")
    .Input("b: string")
    .Input("c: float32")
    .Output("d: float32")
    .Output("e: int32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("CopyOp").Input("a: T").Output("b: T").Attr("T: type").SetShapeFn(
    shape_inference::UnknownShape);

REGISTER_OP("None").SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("IntOutput")
    .Output("a: int32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("Int64Output")
    .Output("out: int64")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("RefOutput")
    .Output("a: Ref(int32)")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("FloatOutput")
    .Output("a: float32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("TwoFloatOutputs")
    .Output("a: float32")
    .Output("b: float32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("FiveFloatOutputs")
    .Output("a: float32")
    .Output("b: float32")
    .Output("c: float32")
    .Output("d: float32")
    .Output("e: float32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("RefOutputFloatOutput")
    .Output("a: Ref(float32)")
    .Output("b: float32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("RefInputFloatInput")
    .Input("a: Ref(float)")
    .Input("b: float")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("IntInput")
    .Input("a: int32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("IntInputIntOutput")
    .Input("a: int32")
    .Output("b: int32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("FloatInput")
    .Input("a: float32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("TwoIntOutputs")
    .Output("a: int32")
    .Output("b: int32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("IntOutputFloatOutput")
    .Output("a: int32")
    .Output("b: float32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("FloatOutputStringOutput")
    .Output("a: float32")
    .Output("b: string")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("TwoIntInputs")
    .Input("a: int32")
    .Input("b: int32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("TwoFloatInputs")
    .Input("a: float32")
    .Input("b: float32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("IntInputFloatInput")
    .Input("a: int32")
    .Input("b: float32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("RefInputIntInput")
    .Input("a: Ref(int32)")
    .Input("b: int32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("TwoFloatInputsFloatOutput")
    .Input("a: float32")
    .Input("b: float32")
    .Output("c: float32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("TwoFloatInputsIntOutput")
    .Input("a: float32")
    .Input("b: float32")
    .Output("c: int32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("RefInputFloatInputIntOutput")
    .Input("a: Ref(float32)")
    .Input("b: float32")
    .Output("c: int32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("ListInput")
    .Input("a: N * T")
    .Attr("N: int >= 1")
    .Attr("T: type")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("ListOutput")
    .Output("a: T")
    .Attr("T: list(type) >= 1")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("Unary").Input("a: T").Output("b: T").Attr("T: type").SetShapeFn(
    shape_inference::UnknownShape);

REGISTER_OP("OpWithDefaultAttr")
    .Output("a: int32")
    .Attr("default_float: float = 123.0")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("OpWithFutureDefaultAttr")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("IntAttr")
    .Output("out: int64")
    .Attr("foo: int = 1")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("StringListAttr")
    .Attr("a: list(string)")
    .Attr("b: string")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("DefaultAttrs")
    .Attr("string_val: string = 'abc'")
    .Attr("string_list_val: list(string) = ['abc', '']")
    .Attr("int_val: int = 123")
    .Attr("int_list_val: list(int) = [1, 2, 3]")
    .Attr("float_val: float = 10.0")
    .Attr("float_list_val: list(float) = [10.0]")
    .Attr("bool_val: bool = true")
    .Attr("bool_list_val: list(bool) = [true, false]")
    .Attr("type_val: type = DT_INT32")
    .Attr("type_list_val: list(type) = [DT_INT32, DT_FLOAT]")
    .Attr("shape_val: shape = { dim { size: 2 } dim { size: 1 } }")
    .Attr("shape_list_val: list(shape) = [{}, { dim { size: 1} }]")
    .Attr("tensor_val: tensor = { dtype: DT_INT32 tensor_shape: {} int_val: 1}")
    .Attr(
        "tensor_list_val: list(tensor) = "
        "[{ dtype: DT_INT32 tensor_shape: {} int_val: 1}]")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("FuncAttr")
    .Attr("f: func")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("FuncListAttr")
    .Attr("f: list(func)")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("Simple")
    .Input("a: int32")
    .Output("out: float")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("OutT").Output("a: T").Attr("T: type").SetShapeFn(
    shape_inference::UnknownShape);

REGISTER_OP("ReservedInput")
    .Input("input: int32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("Polymorphic")
    .Input("a: T")
    .Output("out: T")
    .Attr("T: type")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("PolymorphicOut")
    .Output("out: T")
    .Attr("T: type")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("PolymorphicDefaultOut")
    .Output("out: T")
    .Attr("T: type = DT_STRING")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("Binary")
    .Input("a: T")
    .Input("b: T")
    .Output("out: T")
    .Attr("T: type")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("Restrict")
    .Input("a: T")
    .Output("out: T")
    .Attr("T: {string, bool}")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("TypeList")
    .Input("a: T")
    .Attr("T: list(type) >= 0")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("TypeListTwice")
    .Input("a: T")
    .Input("b: T")
    .Attr("T: list(type) >= 0")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("OutTypeList")
    .Output("out: T")
    .Attr("T: list(type) >= 0")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("TypeListRestrict")
    .Input("a: T")
    .Attr("T: list({string, bool})")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("OutTypeListRestrict")
    .Output("out: t")
    .Attr("t: list({string, bool})")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("Attr").Attr("a: int").SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrFloat")
    .Attr("a: float")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrBool")
    .Attr("a: bool")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrBoolList")
    .Attr("a: list(bool)")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrMin")
    .Attr("a: int >= 5")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrListMin")
    .Attr("a: list(int) >= 2")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrEnum")
    .Attr("a: {'apples', 'oranges'}")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrEnumList")
    .Attr("a: list({'apples', 'oranges'})")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrShape")
    .Attr("a: shape")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrShapeList")
    .Attr("a: list(shape)")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrPartialShape")
    .Attr("a: shape")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrPartialShapeList")
    .Attr("a: list(shape)")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrDefault")
    .Attr("a: string = 'banana'")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrListDefault")
    .Attr("a: list(int) = [5, 15]")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrEmptyListDefault")
    .Attr("a: list(float) = []")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("ReservedAttr")
    .Attr("range: int")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrTypeDefault")
    .Input("a: T")
    .Attr("T: type = DT_INT32")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("AttrListTypeDefault")
    .Input("a: N * T")
    .Input("b: N * T")
    .Attr("T: type = DT_INT32")
    .Attr("N: int")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("NIntsIn")
    .Input("a: N * int32")
    .Attr("N: int >= 2")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("NPolymorphicIn")
    .Input("a: N * T")
    .Attr("T: type")
    .Attr("N: int >= 2")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("NPolymorphicRestrictIn")
    .Input("a: N * T")
    .Attr("T: {string, bool}")
    .Attr("N: int >= 2")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("NInTwice")
    .Input("a: N * int32")
    .Input("b: N * string")
    .Attr("N: int >= 0")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("NInPolymorphicTwice")
    .Input("a: N * T")
    .Input("b: N * T")
    .Attr("T: type")
    .Attr("N: int >= 0")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("NInTwoTypeVariables")
    .Input("a: N * S")
    .Input("b: N * T")
    .Attr("S: type")
    .Attr("T: type")
    .Attr("N: int >= 0")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("InPolymorphicTwice")
    .Input("a: N * T")
    .Input("b: M * T")
    .Attr("T: type = DT_INT32")
    .Attr("N: int >= 0")
    .Attr("M: int >= 0")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("NIntsOut")
    .Output("a: N * int32")
    .Attr("N: int >= 2")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("NIntsOutDefault")
    .Output("a: N * int32")
    .Attr("N: int >= 2 = 3")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("NPolymorphicOut")
    .Output("a: N * T")
    .Attr("T: type")
    .Attr("N: int >= 2")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("NPolymorphicOutDefault")
    .Output("a: N * T")
    .Attr("T: type = DT_BOOL")
    .Attr("N: int >= 2 = 2")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("NPolymorphicRestrictOut")
    .Output("a: N * T")
    .Attr("T: {string, bool}")
    .Attr("N: int >= 2")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("RefIn")
    .Input("a: Ref(T)")
    .Attr("T: type")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("TwoRefsIn")
    .Input("a: Ref(T)")
    .Input("b: Ref(T)")
    .Attr("T: type")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("RefOut")
    .Output("a: Ref(T)")
    .Attr("T: type")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("SimpleStruct")
    .Output("a: n_a * int32")
    .Attr("n_a: int >= 0")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("MixedStruct")
    .Output("a: n_a * int32")
    .Output("b: float")
    .Attr("n_a: int >= 0")
    .SetShapeFn(shape_inference::UnknownShape);

REGISTER_OP("ComplexStruct")
    .Output("a: n_a * int32")
    .Output("b: n_b * int64")
    .Output("c: t_c")
    .Attr("n_a: int >= 0")
    .Attr("n_b: int >= 0")
    .Attr("t_c: list(type) >= 0")
    .SetShapeFn(shape_inference::UnknownShape);

// An op which returns its own device placement as a string, useful for testing
// where ops get placed.
REGISTER_OP("DevicePlacementOp")
    .Output("device: string")
    .SetIsStateful()
    .SetShapeFn(shape_inference::ScalarShape);

class DevicePlacementOp : public OpKernel {
 public:
  using OpKernel::OpKernel;

  void Compute(OpKernelContext* ctx) override {
    Tensor* output;
    OP_REQUIRES_OK(ctx,
                   ctx->allocate_output("device", TensorShape({}), &output));
    output->scalar<tstring>()() = ctx->device()->name();
  }
};

REGISTER_KERNEL_BUILDER(Name("DevicePlacementOp").Device(DEVICE_CPU),
                        DevicePlacementOp);
REGISTER_KERNEL_BUILDER(Name("DevicePlacementOp").Device(DEVICE_DEFAULT),
                        DevicePlacementOp);

// An op which returns the dtype of the tensor it was passed in. It expects
// DT_UINT8.
REGISTER_OP("DtypeWithDefaultOp")
    .Input("in: T")
    .Attr("T: type = DT_UINT8")
    .Output("dtype: string")
    .SetIsStateful()
    .SetShapeFn(shape_inference::ScalarShape);

class DTypeWithDefaultOp : public OpKernel {
 public:
  using OpKernel::OpKernel;

  void Compute(OpKernelContext* ctx) override {
    const Tensor& input = ctx->input(0);
    Tensor* output;
    OP_REQUIRES_OK(ctx,
                   ctx->allocate_output("dtype", TensorShape({}), &output));
    output->scalar<tstring>()() = tensorflow::DataTypeString(input.dtype());
  }
};

REGISTER_KERNEL_BUILDER(Name("DtypeWithDefaultOp").Device(DEVICE_CPU),
                        DTypeWithDefaultOp);

// An op that returns True if TensorFloat-32 execution is enabled. Useful for
// testing that enabling/disabling TensorFloat-32 works correctly, even when
// the test does not run with a GPU that supports TensorFloat-32.
REGISTER_OP("IsTensorFloat32Enabled")
    .Output("enabled: bool")
    .SetIsStateful()
    .SetShapeFn(shape_inference::ScalarShape);

class IsTensorFloat32Enabled : public OpKernel {
 public:
  using OpKernel::OpKernel;

  void Compute(OpKernelContext* ctx) override {
    Tensor* output;
    OP_REQUIRES_OK(ctx,
                   ctx->allocate_output("enabled", TensorShape({}), &output));
    output->scalar<bool>()() = tensor_float_32_execution_enabled();
  }
};

REGISTER_KERNEL_BUILDER(
    Name("IsTensorFloat32Enabled").Device(DEVICE_CPU).HostMemory("enabled"),
    IsTensorFloat32Enabled);
REGISTER_KERNEL_BUILDER(
    Name("IsTensorFloat32Enabled").Device(DEVICE_GPU).HostMemory("enabled"),
    IsTensorFloat32Enabled);
}  // end namespace tensorflow