return %0 : tensor<2x2xf32>
  }
  func.func private @composite_matmul_fn_1(%arg0: tensor<2x2xf32>, %arg1: tensor<2x2xf32>) -> tensor<2x2xf32> attributes {tf_quant.composite_function} {
    %0 = "tf.MatMul"(%arg0, %arg1) {attr_map = "0:transpose_a,1:transpose_b", device = "", transpose_a = false, transpose_b = false} : (tensor<2x2xf32>, tensor<2x2xf32>) -> tensor<2x2xf32>
    return %0 : tensor<2x2xf32>
  }

// CHECK:           return %[[VAL_3]] : tensor<3x6xf32>
// CHECK:         }
func.func @concat_v2_1d_axis(%arg0: tensor<3x3xf32>, %arg1: tensor<3x3xf32>) -> tensor<3x6xf32> {
  %2 = "mhlo.concatenate"(%arg0, %arg1) <{dimension = 1 : i64}> : (tensor<3x3xf32>, tensor<3x3xf32>) -> tensor<3x6xf32>
  func.return %2 : tensor<3x6xf32>
}

//CHECK-NEXT:}

func.func @iota() -> tensor<3x4xf32> {
 %0 = "vhlo.iota_v1" () <{iota_dimension = #vhlo.integer_v1<0 : i64>}> : () -> tensor<3x4xf32>
 return %0 : tensor<3x4xf32>
}

//CHECK:func.func private @iota() -> tensor<3x4xf32> {
//CHECK-NEXT: %0 = "vhlo.iota_v1"() <{iota_dimension = #vhlo.integer_v1<0 : i64>}> : () -> tensor<3x4xf32>
//CHECK-NEXT: return %0 : tensor<3x4xf32>
//CHECK-NEXT:}

  %1 = "tfl.quantize"(%0) {qtype = tensor<2x1x!quant.uniform<i8:f32, 0.024986599940879671:92>>} : (tensor<2x1xf32>) -> tensor<2x1x!quant.uniform<i8:f32, 0.024986599940879671:92>>
  %2 = "tfl.dequantize"(%1) : (tensor<2x1x!quant.uniform<i8:f32, 0.024986599940879671:92>>) -> tensor<2x1xf32>

  %outputs =  "tf.Const"() {device = "/job:localhost/replica:0/task:0/device:CPU:0", value = dense<0> : tensor<i32>} : () -> tensor<i32>...

  %one = "glue.constant"() { value = 1: i32 } : () -> i32
  %done = "glue.compare" (%one, %one) { predicate = #glue<"compare LTE"> } : (i32, i32) -> i1
  %2 = mhlo.constant dense<[[1.1]]> : tensor<1x1xf32>
  %3 = mhlo.multiply %2, %2 : tensor<1x1xf32>
  %cst = "tf.Const"() {value = dense<0.0> : tensor<f32>} : () -> tensor<f32>
  %0 = "tf.AddV2"(%arg0, %cst) {device = "/device:CPU:0"} : (tensor<f32>, tensor<f32>) -> tensor<f32>

    %0 = "tf.MatMul"(%arg0, %arg1) {attr_map = "0:transpose_a,1:transpose_b", device = "", transpose_a = false, transpose_b = false} : (tensor<2x2xf32>, tensor<2x2xf32>) -> tensor<2x2xf32>
    return %0 : tensor<2x2xf32>
  }

func.func @select_float(%arg0: tensor<1x3xi1>, %arg1: tensor<1x3xf32>, %arg2: tensor<1x3xf32>) -> tensor<1x3xf32> {
  %0 = "stablehlo.select"(%arg0, %arg1, %arg2) : (tensor<1x3xi1>, tensor<1x3xf32>, tensor<1x3xf32>) -> tensor<1x3xf32>
  return %0 : tensor<1x3xf32>
}
// CHECK-LABEL: select_float
// CHECK-NOT: tfl.select_v2
// CHECK: stablehlo.select

// -----

    %2:2 = "tf.RecvTPUEmbeddingActivations"() {_tpu_embedding_layer = "call1", config = "\0A\0B\0C\0D"} : () -> (tensor<2x2xf32>, tensor<4x4xf32>)
    "tf.SendTPUEmbeddingGradients"(%2#0, %2#1) {_tpu_embedding_layer = "call1", config = "\0A\0B\0C\0D", operandSegmentSizes = array<i32: 2, 0>} : (tensor<2x2xf32>, tensor<4x4xf32>) -> ()
    tf_device.return

  %2 = mhlo.constant dense<6.400000e+01> : tensor<64xf32>
  %3 = mhlo.constant dense<3.200000e+01> : tensor<64xf32>
  %4 = mhlo.constant dense<5.000000e-01> : tensor<64xf32>
  %5 = "mhlo.iota"() <{iota_dimension = 0 : i64}> : () -> tensor<64xf32>
  %6 = mhlo.add %5, %4 : tensor<64xf32>
  %7 = mhlo.multiply %6, %3 : tensor<64xf32>
  %8 = mhlo.divide %7, %2 : tensor<64xf32>
  %9 = mhlo.floor %8 : tensor<64xf32>