// first k fields of t1 equals the first k fields of t2.
// SameSizeTuplePrefixComparator(k1, k2) would be a compiler error if
// k1 != k2.
template <int kSize1, int kSize2>
struct SameSizeTuplePrefixComparator;

template <>
struct SameSizeTuplePrefixComparator<0, 0> {
  template <class Tuple1, class Tuple2>
  static bool Eq(const Tuple1& /* t1 */, const Tuple2& /* t2 */) {

// first k fields of t1 equals the first k fields of t2.
// SameSizeTuplePrefixComparator(k1, k2) would be a compiler error if
// k1 != k2.
template <int kSize1, int kSize2>
struct SameSizeTuplePrefixComparator;

template <>
struct SameSizeTuplePrefixComparator<0, 0> {
  template <class Tuple1, class Tuple2>
  static bool Eq(const Tuple1& /* t1 */, const Tuple2& /* t2 */) {

      size++;
    }
    assertEquals(expectedSize, size);

    int size2 = 0;
    for (Entry<K, Collection<V>> entry2 : multimap.asMap().entrySet()) {
      size2 += entry2.getValue().size();
    }
    assertEquals(expectedSize, size2);
  }

  @CollectionSize.Require(ZERO)
  public void testIsEmptyYes() {
    assertTrue(multimap().isEmpty());
  }

      size++;
    }
    assertEquals(expectedSize, size);

    int size2 = 0;
    for (Entry<K, Collection<V>> entry2 : multimap.asMap().entrySet()) {
      size2 += entry2.getValue().size();
    }
    assertEquals(expectedSize, size2);
  }

  @CollectionSize.Require(ZERO)
  public void testIsEmptyYes() {
    assertTrue(multimap().isEmpty());
  }

    outputs=['o: T'])
def _composite_max_pool(input_, stride_w, stride_h, filter_width, filter_height,
                        padding):
  ksize = [1, filter_width, filter_height, 1]
  strides = [1, stride_w, stride_h, 1]
  return tf.raw_ops.MaxPool(
      input=input_, ksize=ksize, strides=strides, padding=padding)


@tf.RegisterGradient('NewMaxPool')
def _max_pool_grad(op: ops.Operation, grad):

  // CHECK: %[[R1:.*]] = "tf.MaxPool"(%[[R0]]) <{data_format = "NHWC", explicit_paddings = [], ksize = [1, 3, 3, 1], padding = "SAME", strides = [1, 2, 2, 1]}>
  // CHECK: %[[CST_0:.*]] = "tf.Const"() <{value = dense<[0, 3, 1, 2]> : tensor<4xi64>}>
  // CHECK: "tf.Transpose"(%[[R1]], %[[CST_0]])
  %0 = "tf.MaxPool"(%arg0)
       {
         data_format = "NCHW",
         ksize = [1, 1, 3, 3],
         padding = "SAME",
         strides = [1, 1, 2, 2]

    %0:2 = tf_executor.graph {
      %outputs:2, %control = tf_executor.island wraps "tf.MaxPoolWithArgmax"(%arg0) {T = f32, Targmax = i32, include_batch_in_index = false, ksize = [1, 2, 2, 1], padding = "SAME", strides = [1, 2, 2, 1]} : (tensor<2x4x2x2xf32>) -> (tensor<2x2x1x2xf32>, tensor<2x2x1x2xi32>)
      tf_executor.fetch %outputs#1, %outputs#0 : tensor<2x2x1x2xi32>, tensor<2x2x1x2xf32>
    }

  %8 = "tf.MaxPool"(%7)
       {
         data_format = "NCHW",
         ksize = [1, 1, 3, 3],
         padding = "SAME",
         strides = [1, 1, 2, 2]
       } : (tensor<?x64x112x112xf32>) -> tensor<?x64x56x56xf32>

  // CHECK: %[[MAX_POOL:[0-9]*]] = "tf.MaxPool"
  // CHECK-SAME: data_format = "NHWC"
  // CHECK-SAME: ksize = [1, 3, 3, 1]
  // CHECK-SAME: strides = [1, 2, 2, 1]

    %1 = "quantfork.dcast"(%0) : (tensor<*x!quant.uniform<i8:f32, 0.05:-10>>) -> tensor<*xf32>
    %2 = "tf.MaxPool"(%1) {
      data_format = "NHWC", device = "", explicit_paddings = [],
      ksize = [1, 2, 2, 1], padding = "VALID", strides = [1, 2, 2, 1]
    } : (tensor<*xf32>) -> tensor<*xf32>
    %3 = "tf.Reshape"(%2, %cst) {device = ""} : (tensor<*xf32>, tensor<2xi32>) -> tensor<*xf32>

  %1 = "quantfork.dcast"(%0) : (tensor<*x!quant.uniform<i8:f32, 5.000000e-02:-10>>) -> tensor<*xf32>
  %2 = "tf.MaxPool"(%1) {data_format = "NHWC", device = "", explicit_paddings = [], ksize = [1, 2, 2, 1], padding = "VALID", strides = [1, 2, 2, 1]} : (tensor<*xf32>) -> tensor<*xf32>
  %3 = "quantfork.qcast"(%2) {volatile} : (tensor<*xf32>) -> tensor<*x!quant.uniform<i8:f32, 5.000000e-02:-10>>