%dconv = "tfl.depthwise_conv_2d"(%arg0, %w, %b) {depth_multiplier = 4 : i32, dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "NONE", padding = "VALID", stride_h = 4 : i32, stride_w = 5 : i32} : (tensor<1x224x224x3xf32>, tensor<64x3x3x3xf32>, tensor<64xf32>)...

  func.return %1 : tensor<1x1x1x16xf32>

              IsIntList1XY1:$ksize,
              IsIntList1XY1:$strides,
              $padding,
              $explicit_paddings,
              IsDataFormatNHWC:$format),
          (TFL_MaxPool2DOp $value,
              /*padding=*/$padding,
              /*stride_w=*/ExtractI32At<2>:$strides,
              /*stride_h=*/ExtractI32At<1>:$strides,
              /*filter_width=*/ExtractI32At<2>:$ksize,

// CHECK:           %[[VAL_5:.*]] = "tfl.average_pool_2d"(%[[VAL_4]]) <{filter_height = 3 : i32, filter_width = 3 : i32, fused_activation_function = "NONE", padding = "VALID", stride_h = 1 : i32, stride_w = 1 : i32}> : (tensor<1x6x6x3xf32>) -> tensor<1x4x4x3xf32>
// CHECK:           %[[VAL_6:.*]] = arith.constant dense<[0, 3, 1, 2]> : tensor<4xi32>

    const int stride_i = mlir::cast<IntegerAttr>(strides[i]).getInt();
    const int dilation_i = mlir::cast<IntegerAttr>(dilations[i]).getInt();
    const int filter_i = filter_shape.getDimSize(i - 1);
    Value pad_i_low, pad_i_high;
    GetSamePaddingValues(builder, loc, input_size_i, filter_i, dilation_i,
                         stride_i, pad_i_low, pad_i_high);

        offset = [0.5] * self.out_channel_size
        mean, variance = scale, offset
        out = nn_ops.conv2d(
            input_tensor,
            self.filters,
            strides=strides,
            dilations=dilations,
            padding=padding,
            data_format='NHWC',
            name='sample/conv',
        )
        if bias_fn is not None:

      }

      @Override
      public int compare(byte[] left, byte[] right) {
        int stride = 8;
        int minLength = Math.min(left.length, right.length);
        int strideLimit = minLength & ~(stride - 1);
        int i;

        /*
         * Compare 8 bytes at a time. Benchmarking on x86 shows a stride of 8 bytes is no slower

      }

      @Override
      public int compare(byte[] left, byte[] right) {
        int stride = 8;
        int minLength = Math.min(left.length, right.length);
        int strideLimit = minLength & ~(stride - 1);
        int i;

        /*
         * Compare 8 bytes at a time. Benchmarking on x86 shows a stride of 8 bytes is no slower

              mlir::cast<IntegerAttr>(op.getRhsDilationAttr()[i]).getInt());

      const int64_t output_size = (lhs_size_dilated + stride - 1) / stride;
      const int64_t total_padding = std::max(
          (output_size - 1) * stride + rhs_size_dilated - lhs_size_dilated,
          static_cast<int64_t>(0));
      const int64_t padding_begin = total_padding / 2;

	bw := blockWriter{e: e}
	bw.setup()
	lzww := lzw.NewWriter(bw, lzw.LSB, litWidth)
	if dx := b.Dx(); dx == pm.Stride {
		_, e.err = lzww.Write(pm.Pix[:dx*b.Dy()])
		if e.err != nil {
			lzww.Close()
			return
		}
	} else {
		for i, y := 0, b.Min.Y; y < b.Max.Y; i, y = i+pm.Stride, y+1 {
			_, e.err = lzww.Write(pm.Pix[i : i+dx])
			if e.err != nil {
				lzww.Close()
				return
			}
		}
	}