// This signals that the block is removed, and should not be inadvertently
  // reinserted into the cache in UpdateLRU.
  entry->second->timestamp = 0;
  lru_list_.erase(entry->second->lru_iterator);
  lra_list_.erase(entry->second->lra_iterator);
  cache_size_ -= entry->second->data.capacity();
  block_map_.erase(entry);
}

  // Reading at offset file_size + 4 will read the second block (since the read
  // at file_size + 4 = 28 will be aligned to an offset of 16) but will return
  // OutOfRange because the offset is past the end of the 24-byte file.
  Status status = ReadCache(&cache, "", file_size + 4, 4, &out);
  EXPECT_EQ(status.code(), error::OUT_OF_RANGE);
  EXPECT_TRUE(second_block);
  // Reading the second full block will return 8 bytes, from a cache hit.

    TF_DeleteAbstractOp(add_op);
    // Extract the resulting tensor.
    add_output1 = TF_OutputListGet(add_outputs, 0);
    TF_DeleteOutputList(add_outputs);
  }

  // Same with a second "Add" computing `arg1 + arg1`.
  TF_AbstractTensor* add_output2;
  {
    // Build an abstract operation, inputs and output.
    auto* add_op = TF_NewAbstractOp(graph_ctx);
    TF_AbstractOpSetOpType(add_op, "Add", s);

    if (attr.first == "dtype") {
      if (attr.second.type() == tensorflow::DT_INT32) {
        found_dtype = true;
      } else {
        return false;
      }
    } else if (attr.first == "value") {
      if (attr.second.has_tensor() &&
          attr.second.tensor().int_val_size() == 1 &&
          attr.second.tensor().int_val(0) == v) {
        found_value = true;
      } else {

  // [Optional]
  // Create function for optimizer.
  void* (*create_func)();

  // Optimizer function for optimizer. The first param is an optimizer created
  // by create_func. The second param is input graph. The third param is
  // GrapplerItem. The fourth param is output graph.
  void (*optimize_func)(void*, const TF_Buffer*, const TF_GrapplerItem*,
                        TF_Buffer*, TF_Status*);

-   If the PR passes all the quality checks then we go ahead and assign a
    reviewer.
-   If the PR didn't meet the validation criteria, we request for additional
    changes to be made to PR to pass quality checks and send it back or on a
    rare occasion we may reject it.

**3. Review**

-   For a valid PR, reviewer (person familiar with the code/functionality)
    checks if the PR looks good or needs additional changes.

  if (iter == registry_.end()) {
    const string error_msg = "No gradient defined for op: " + op.op_name + ".";
    return errors::NotFound(error_msg);
  }
  gradient_function->reset(iter->second(op));
  return absl::OkStatus();
}

TapeTensor::TapeTensor(AbstractTensorHandle* handle) : handle_(handle) {
  handle_->Ref();
}
TapeTensor::TapeTensor(const TapeTensor& other) {
  handle_ = other.handle_;

  TFE_OpSetAttrString(op, "tensor_name", "dummy", 5);
  TFE_OpSetAttrString(op, "send_device", send_device.c_str(),
                      send_device.size());
  TFE_OpSetAttrString(op, "recv_device", recv_device.c_str(),
                      recv_device.size());
  TFE_OpSetAttrInt(op, "send_device_incarnation", send_device_incarnation);

  return op;
}

                    /*input_shapes*/ {make_shape({3, 2}), make_shape({2, 4})},
                    /*input_tensors*/ {},
                    /*expected_shape*/ make_shape({3, 4}));

  // Infer shape when second operand has unknown shape.
  CheckOutputShapes(matmul_op,
                    /*input_shapes*/ {make_shape({3, 2}), unknown_shape()},
                    /*input_tensors*/ {},

  TFE_Op* func_op = TFE_NewOp(ctx, func.name().data(), status);
  for (const auto& attr : func.attr()) {
    if (!status->status.ok()) return nullptr;
    SetOpAttrValueScalar(ctx, func_op, attr.second, attr.first.data(), status);
    if (!status->status.ok()) return nullptr;
  }
  return func_op;
}
}  // namespace

void TFE_ContextStartStep(TFE_Context* ctx) {