init.Append(as)
	return res
}

// Returns the data word (the second word) used to represent conv.X in
// an interface.
func dataWord(conv *ir.ConvExpr, init *ir.Nodes) ir.Node {
	pos, n := conv.Pos(), conv.X
	fromType := n.Type()

	// If it's a pointer, it is its own representation.
	if types.IsDirectIface(fromType) {
		return n
	}

        def file3 = new File("artifact-3.zip")

        def conf1 = dependencyScope("conf1")
        def conf2 = dependencyScope("conf2")
        def child1 = consumable("child1", [conf1, conf2])
        consumable("child2", [conf1])

        addArtifact(conf1, artifact1, file1)
        addArtifact(conf2, artifact2, file2)
        addArtifact(child1, artifact3, file3)

        when:

    }
  }

  // Handle Conv2D input, stride and filter.
  HandleConv2DInput(conv2d, block_size);
  HandleConv2DStride(conv2d);
  HandleConv2DFilter(conv2d, block_size);

  // Book keeping new filter shape for backprop filter rewrite.
  // Filter shape is defined in HandleConv2DFilter, thus it is RankedTensorType.
  filter_shape =
      mlir::cast<RankedTensorType>(conv2d.getFilter().getType()).getShape();

  }
  func.func private @composite_conv2d_with_bias_and_relu6_fn_2(%arg0: tensor<*xf32>, %arg1: tensor<*xf32>, %arg2: tensor<2xf32>) -> tensor<*xf32> attributes {tf_quant.composite_function} {

//
//
//   SpaceToBatchND -> Expand -> Conv2D -> Squeeze -> BatchToSpaceND -> BiasAdd
//
//   SpaceToBatchND -> Expand -> Conv2D -> Squeeze -> Pad -> BatchToSpaceND ->
//   BiasAdd
//
//   SpaceToBatchND -> Expand -> Conv2D -> Squeeze -> BiasAdd -> BatchToSpaceND
//
//   SpaceToBatchND -> Conv2D -> Pad -> BatchToSpaceND -> BiasAdd
//
//   SpaceToBatchND -> Conv2D -> BatchToSpaceND -> BiasAdd
//
//

    %zp_offset: tensor<?x2x2x1xi32>, %bias: tensor<1xi32>
  ) -> tensor<?x2x2x1xi32> {
  // CHECK-DAG: %[[conv:.*]] = mhlo.convolution
  // CHECK-DAG: %[[combined:.*]] = chlo.broadcast_add %[[zp_offset:.*]], %[[bias:.*]]
  // CHECK-DAG: %[[result:.*]] = chlo.broadcast_add %[[conv]], %[[combined]]
  // CHECK: return %[[result]]
  %0 = mhlo.convolution(%lhs, %rhs)
      dim_numbers = [b, 0, 1, f]x[0, 1, i, o]->[b, 0, 1, f],

  %b = arith.constant dense<-1.23697901> : tensor<64xf32>
  %conv = "tfl.conv_2d"(%arg0, %w, %b) {dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "SAME", stride_h = 2 : i32, stride_w = 2 : i32} : (tensor<1x224x224x3xf32>, tensor<64x3x3x3xf32>, tensor<64xf32>) -> tensor<1x112x112x64xf32>
  func.return %conv : tensor<1x112x112x64xf32>

  }
  func.func private @composite_conv2d_with_bias_and_relu6_fn_1(%arg0: tensor<*xf32>, %arg1: tensor<*xf32>, %arg2: tensor<2xf32>) -> tensor<*xf32> attributes {tf_quant.composite_function} {

  %cst = "tf.Const"() {value = dense<0.000000e+00> : tensor<2xf32>} : () -> tensor<2xf32>
  %cst_1 = "tf.Const"() {value = dense<3.000000e+00> : tensor<2x3x3x2xf32>} : () -> tensor<2x3x3x2xf32>
  %0 = "tf.Conv2D"(%arg0, %cst_1) {
    data_format = "NHWC", device = "", dilations = [1, 1, 1, 1], explicit_paddings = [],
    padding = "SAME", strides = [1, 1, 2, 1], use_cudnn_on_gpu = true

		return mkcall("countrunes", n.Type(), init, typecheck.Conv(n.X.(*ir.ConvExpr).X, types.Types[types.TSTRING]))
	}
	if isByteCount(n) {
		conv := n.X.(*ir.ConvExpr)
		walkStmtList(conv.Init())
		init.Append(ir.TakeInit(conv)...)
		_, len := backingArrayPtrLen(cheapExpr(conv.X, init))
		return len
	}
	if isChanLenCap(n) {
		name := "chanlen"
		if n.Op() == ir.OCAP {