Score int64
}

// BuildBrokenLinearFunction creates a function which is built using linear segments. Segments are defined via shape array.
// Shape[i].Utilization slice represents points on "Utilization" axis where different segments meet.
// Shape[i].Score represents function values at meeting points.
//
// function f(p) is defined as:
//
//	shape[0].Score for p < shape[0].Utilization

                         %weight : tensor<*xi8>, %input : tensor<*xi32>, %axis : tensor<i32>,
                         %weight_scale : tensor<*xf32>, %weight_zp : tensor<*xi32>) -> tensor<*xf32>
      attributes {tf_quant.quantized_ops = ${quantized_ops}}
  {
    %accum_out = "tf.GatherV2"(%weight, %input, %axis) {

			j++
			y++
		}
		j &= 7                               // octant modulo 2Pi radians (360 degrees)
		z = ((x - y*PI4A) - y*PI4B) - y*PI4C // Extended precision modular arithmetic
	}
	if j > 3 { // reflect in x axis
		j -= 4
		sinSign, cosSign = !sinSign, !cosSign
	}
	if j > 1 {
		cosSign = !cosSign
	}

	zz := z * z
	cos = 1.0 - 0.5*zz + zz*zz*((((((_cos[0]*zz)+_cos[1])*zz+_cos[2])*zz+_cos[3])*zz+_cos[4])*zz+_cos[5])

  input_value_shape = tf.shape(op.inputs[2])
  _, reduction_axes = tf.raw_ops.BroadcastGradientArgs(
      s0=broadcast_shape, s1=input_value_shape)
  updates_grad_reshaped = tf.reduce_sum(
      grad, axis=reduction_axes, keepdims=True)
  bias_grad = tf.reshape(updates_grad_reshaped, input_value_shape)

  dilations = [1, op.get_attr('dilation_w'), op.get_attr('dilation_h'), 1]

# CHECK:           %[[VAL_12:.*]] = "tfl.pack"(%[[VAL_10]], %[[VAL_11]]) <{axis = 0 : i32, values_count = 2 : i32}> : (tensor<5x7xf32>, tensor<5x7xf32>) -> tensor<2x5x7xf32>
# CHECK:           return %[[VAL_12]] : tensor<2x5x7xf32>

// Return a 1-D INT32 tensor containing a single value 1.
TFE_TensorHandle* TestAxisTensorHandle(TFE_Context* ctx);

// Return an op taking minimum of `input` long `axis` dimension.
TFE_Op* MinOp(TFE_Context* ctx, TFE_TensorHandle* input,
              TFE_TensorHandle* axis);

// If there is a device of type `device_type`, returns true
// and sets 'device_name' accordingly.
// `device_type` must be either "GPU" or "TPU".

    x0, x1 = array_ops.split(x, 2, 0)
    y0, y1 = array_ops.split(y, 2, 0)
    x0 += 1
    y0 += 1
    z = math_ops.matmul(x, y, name='x_y_prod')
    a = array_ops.concat([x0, y1], axis=0, name='concat_x0_y1')
    b = array_ops.concat([y0, x1], axis=0, name='concat_y0_x1')
    x = math_ops.matmul(a, b, name='a_b')
    y = math_ops.add(x, z)
  array_ops.identity(y, name='result')


def tftop_k(_):

        <script>
        var chart = c3.generate({
            data: {
                url: 'data.csv',
                type: 'bar',
                x: 'Scenario'
            },
            axis: {
                x: {
                    type: 'category'
                },
                y: {
                    label: 'ops/s',
                    tick: {

}

// UtilizationShapePoint represents a single point of a priority function shape.
type UtilizationShapePoint struct {
	// Utilization (x axis). Valid values are 0 to 100. Fully utilized node maps to 100.
	Utilization int32
	// Score assigned to a given utilization (y axis). Valid values are 0 to 10.
	Score int32
}

// ResourceSpec represents single resource.
type ResourceSpec struct {
	// Name of the resource.

//
// DESCRIPTION:
//
// If
//     z = x + iy,
//
// then
//
//           sin 2x  +  i sinh 2y
//     w  =  --------------------.
//            cos 2x  +  cosh 2y
//
// On the real axis the denominator is zero at odd multiples
// of PI/2. The denominator is evaluated by its Taylor
// series near these points.
//
// ctan(z) = -i ctanh(iz).
//
// ACCURACY:
//
//                      Relative error: