// CHECK:   return %[[MUL_0]] : tensor<1xf32>
// CHECK: }

  func.func @mul2(%arg0: tensor<1xf32> {tf_saved_model.index_path = ["y"]}, %arg1: tensor<1xf32> {tf_saved_model.index_path = ["x"]}) -> (tensor<1xf32> {tf_saved_model.index_path = ["output_0"]}) attributes {tf.entry_function = {inputs = "mul2_y:0,mul2_x:0", outputs = "PartitionedCall_1:0"}, tf_saved_model.exported_names = ["mul2"]} {

        // CHECK-NOT:"tf.Const"
        // CHECK: %[[MUL1:.*]] = "tf.Mul"(%arg0, %[[CST2]])
        // CHECK-NEXT: %[[MUL2:.*]] = "tf.Mul"(%[[MUL1]], %[[CST2]])
        // CHECK-NEXT: %[[MUL3:.*]] = "tf.Mul"(%[[MUL2]], %[[CST3]])
        // CHECK-NEXT: = "tf.Mul"(%[[MUL3]], %[[CST4]])
        %3 = "tf.Mul"(%arg0, %0) : (tensor<16xf32>, tensor<f32>) -> tensor<16xf32>
        %4 = "tf.Mul"(%3, %0) : (tensor<16xf32>, tensor<f32>) -> tensor<16xf32>

# CHECK-DAG: %[[MUL1:.*]] = "tf.MatMul"(%[[ARG0]], %[[ARG1]])
# CHECK-DAG: %[[MUL2:.*]] = "tf.MatMul"(%[[ARG1]], %[[ARG0]])
# CHECK:  %[[IDENTITY:.*]]:2 = "tf.IdentityN"(%[[MUL1]], %[[MUL2]])
# CHECK: return %[[IDENTITY]]#1, %[[IDENTITY]]#0


def Test():

  x = tf.constant(1.0, shape=(5, 3))
  y = tf.constant(1.0, shape=(3, 5))

  s = tf.matmul(x, y)
  t = tf.matmul(y, x)

// simplifications and optimizations on float types.
// For codegen tests on integer types, see arithmetic.go.

// --------------------- //
//    Strength-reduce    //
// --------------------- //

func Mul2(f float64) float64 {
	// 386/sse2:"ADDSD",-"MULSD"
	// amd64:"ADDSD",-"MULSD"
	// arm/7:"ADDD",-"MULD"
	// arm64:"FADDD",-"FMULD"
	// ppc64x:"FADD",-"FMUL"
	// riscv64:"FADDD",-"FMULD"
	return f * 2.0
}

// CHECK:  %[[SUB:.*]] = "tf.Sub"(%[[ARG2:.*]], %[[MUL3]])
//              x * scale * rsqrt(variance + epsilon) +
//              offset - mean * scale * rsqrt(variance + epsilon)
// CHECK:  %[[ADD2:.*]] = "tf.Add"(%[[MUL2]], %[[SUB]])
// CHECK:  %[[BATCHNORM1_a:[^,]+]], {{.*}} = "tf.FusedBatchNormV3"(%[[ADD2]], %[[CONSTANT1]], %[[ARG2]], %[[ARG3]], %[[ARG4]])

    // CHECK: %[[scr2:.*]] = mhlo.convert %[[red1]] : tensor<8xf32>

    // CHECK: %[[mul2:.*]] = mhlo.multiply %arg2, %[[scr1]] : tensor<8xf32>
    // CHECK: %[[bcast_mul2:.+]] = "mhlo.dynamic_broadcast_in_dim"(%[[mul2]], {{.*}}) <{broadcast_dimensions = dense<3> : tensor<1xi64>}> : (tensor<8xf32>, tensor<4xindex>) -> tensor<8x8x8x8xf32>

  // CHECK-NEXT: %[[scr2:.*]] = mhlo.convert %[[red1]] : tensor<8xf32>

  // CHECK-NEXT: %[[mul2:.*]] = mhlo.multiply %arg2, %[[scr1]] : tensor<8xf32>
  // CHECK:      %[[bcast_mul2:.+]] = "mhlo.dynamic_broadcast_in_dim"(%[[mul2]], {{.*}}) <{broadcast_dimensions = dense<3> : tensor<1xi64>}> : (tensor<8xf32>, tensor<4xindex>) -> tensor<8x8x8x8xf32>

  // CHECK: %[[MUL2:.*]] = "tf.Mul"(%[[FLOOR]], %[[TWO]]) : (tensor<2xf32>, tensor<f32>) -> tensor<2xf32>
  // CHECK: %[[NEAREST_EVEN_INT:.*]] = "tf.Sub"(%[[ROUND_VAL]], %[[MUL2]]) : (tensor<2xf32>, tensor<2xf32>) -> tensor<2xf32>
  // CHECK: %[[IS_ODD:.*]] = "tf.Equal"(%[[NEAREST_EVEN_INT]], %[[ONE]]) <{incompatible_shape_error = true}> : (tensor<2xf32>, tensor<f32>) -> tensor<2xi1>

  // CHECK: %[[CONST:.*]] = "tf.Const"()
  // CHECK-DAG: %[[MUL1:.*]] = "tf.Mul"(%arg0, %arg2)
  // CHECK-DAG: %[[MUL2:.*]] = "tf.Mul"(%arg1, %arg3)
  // CHECK: "tf.ConcatV2"(%[[MUL1]], %[[MUL2]], %[[CONST]])
  %0 = "tf.Const"() { value = dense<1> : tensor<i32> } : () -> tensor<i32>
  %1 = "tf.Mul"(%arg0, %arg2) : (tensor<?x2xf32>, tensor<f32>) -> tensor<?x2xf32>

							if($.isArray(js.children) && js.children.length) {
								tmp = this._parse_json(js.children, obj, true);
								if(tmp.length) {
									ul2 = $("<ul />");
									ul2.append(tmp);
									d.append(ul2);
								}
							}
						}
					}
				}
				if(!is_callback) {
					ul1 = $("<ul />");
					ul1.append(d);
					d = ul1;
				}
				return d;