configurations {
                conf1
                conf2
            }

            dependencies {
                conf1 'org:toplevel1:1.0'
                conf2 'org:toplevel2:1.0'
            }
        """

        when:
        run "dependencies", "--configuration", "conf2"

        then:
        output.contains """
conf2
\\--- org:toplevel2:1.0
     +--- org:leaf3:1.0

      mlir::mhlo::createLegalizeDotToDotGeneralPass());
  // Unfuse mhlo BatchNorm to primitive ops.
  pm.addNestedPass<mlir::func::FuncOp>(mlir::odml::createUnfuseBatchNormPass());
  // Fuse Conv + Mul to Conv.
  pm.addNestedPass<mlir::func::FuncOp>(mlir::odml::createFuseConvolutionPass());
  // Fold broadcast_in_dim + Mul.
  pm.addNestedPass<mlir::func::FuncOp>(mlir::odml::createFoldBroadcastPass());

In this pass, every op will be targeted with the user specified targets based on
the device capabilites. For example, If the user specified the desired targets
are "GPU", "CPU", `conv2d` can run on both "GPU" and "CPU", we will annotate
the op `conv2d` with "GPU" since it's preferred; `pack` can only run on "CPU",
so we will annotate the op with "CPU" since "GPU" does not support this op.

#### Raise Target Subgraphs Pass

                           op.input(2), strides, padding, filter_attrs));

  Conv2D::Attrs conv_attrs;
  conv_attrs.use_cudnn_on_gpu_ = use_cudnn_on_gpu;
  conv_attrs.explicit_paddings_ = explicit_paddings;
  conv_attrs.data_format_ = data_format;
  conv_attrs.dilations_ = dilations;
  grad_outputs->push_back(
      Conv2D(scope, grad_inputs[0], op.input(1), strides, padding, conv_attrs));
  return scope.status();
}

  ASSERT_EQ(status.message(),
            "'optimize_func' field in TP_Optimizer must be set.");
}

TEST(TF_GrapplerItem, NodesToPreserve) {
  GrapplerItem item;
  item.fetch = std::vector<string>{"Conv", "BiasAdd"};
  std::unordered_set<string> nodes_preserved = item.NodesToPreserve();
  TF_GrapplerItem* c_item = reinterpret_cast<TF_GrapplerItem*>(&item);

  int list_total_size = 0;

    %5 = "tf.Pad"(%arg0, %3) : (tensor<2x224x224x3xf32>, tensor<4x2xi32>) -> tensor<2x230x230x3xf32>
    // CHECK: "tf.Conv2D"
    // CHECK-SAME: strides = [1, 1, 1, 1]
    // CHECK-SAME: (tensor<2x115x115x12xf32>, tensor<4x4x12x64xf32>) -> tensor<2x112x112x64xf32>

	if types.IsComplex[et] && n.Op() == ir.ODIV {
		t := n.Type()
		call := mkcall("complex128div", types.Types[types.TCOMPLEX128], init, typecheck.Conv(n.X, types.Types[types.TCOMPLEX128]), typecheck.Conv(n.Y, types.Types[types.TCOMPLEX128]))
		return typecheck.Conv(call, t)
	}

	// Nothing to do for float divisions.
	if types.IsFloat[et] {
		return n
	}

	// rewrite 64-bit div and mod on 32-bit architectures.

// issue 7205

var i9 interface{}

func f9() bool {
	g8()
	x := i9
	y := interface{}(g18()) // ERROR "live at call to convT: x.data$" "live at call to g18: x.data$" "stack object .autotmp_[0-9]+ \[2\]string$"
	i9 = y                  // make y escape so the line above has to call convT
	return x != y
}

// liveness formerly confused by UNDEF followed by RET,
// leading to "live at entry to f10: ~r1" (unnamed result).

// CHECK: %[[DEQUANTIZE:.*]] = "tfl.dequantize"(%[[QUANTIZE]])
// CHECK: %[[CONV:.*]] = "tfl.conv_2d"(%arg0, %[[DEQUANTIZE]], %[[CONSTANT]])
// CHECK: return %[[CONV]]
}

// CHECK-LABEL: perChannelFakeQuantWithConv2D
func.func @perChannelFakeQuantWithConv2D(tensor<256x32x32x3xf32>) -> (tensor<256x8x7x16xf32>) {

#include "tensorflow/core/lib/core/status_test_util.h"
#include "tensorflow/core/lib/random/random.h"

namespace tensorflow {
namespace {

using ops::AvgPool;
using ops::AvgPool3D;
using ops::BiasAdd;
using ops::Conv2D;
using ops::Conv2DBackpropInput;
using ops::DepthwiseConv2dNative;
using ops::Elu;
using ops::FractionalAvgPool;
using ops::FractionalMaxPool;
using ops::FusedBatchNormV3;
using ops::L2Loss;
using ops::LogSoftmax;