// CHECK-SAME: tf_saved_model.initializer_type = "restore_op"
  // CHECK: %[[VAR:.*]] = "tf.VarHandleOp"
  // CHECK-SAME: "var1"
  // CHECK: %[[CST:.*]] = arith.constant dense<> : tensor<0xf32>
  // CHECK: "tf.AssignVariableOp"(%[[VAR]], %[[CST]])
}

// -----

// Test with an existing session init function.
module attributes {tf_saved_model.semantics, tf_saved_model.under_construction} {

		}
	}

	if cst.ts.Config.ErrorLog == nil {
		cst.ts.Config.ErrorLog = log.New(testLogWriter{t}, "", 0)
	}

	switch mode {
	case http1Mode:
		cst.ts.Start()
	case https1Mode:
		cst.ts.StartTLS()
	case http2Mode:
		ExportHttp2ConfigureServer(cst.ts.Config, nil)
		cst.ts.TLS = cst.ts.Config.TLSConfig
		cst.ts.StartTLS()
	default:
		t.Fatalf("unknown test mode %v", mode)
	}
	cst.c = cst.ts.Client()

func.func @main() -> tensor<2x2xf32> {
  // CHECK: %[[cst:.*]] = "tfl.pseudo_const"() <{value = dense<2> : tensor<2xi32>}> : () -> tensor<2xi32>
  // CHECK: %{{.*}} = "tfl.reshape"(%{{.*}}, %[[cst]]) : (tensor<4xf32>, tensor<2xi32>) -> tensor<2x2xf32>
  %cst = arith.constant dense<[2, 2]> : tensor<2xi32>
  %0 = "tfl.pseudo_const" () {value = dense<1.0> : tensor<4xf32>} : () -> tensor<4xf32> loc("Const")

// CHECK-LABEL: import_stats_skip
func.func @import_stats_skip(%arg0: tensor<4xf32>, %cst: tensor<i32>) -> (tensor<2xf32>,tensor<2xf32>) {
  %0:2 = "tfl.split"(%cst, %arg0) {num_splits = 2 : i32} : (tensor<i32>, tensor<4xf32>) -> (tensor<2xf32>, tensor<2xf32>)
    loc(fused["skip1", "skip2.cc":10:8, callsite("op" at "skip3.cc":10:8)])

// CHECK-DAG: %[[CST:.*]] = "tf.Const"() {{.*}} : () -> tensor<2x3x3x3x2xf32>

// CHECK: %[[PARTITIONEDCALL_0:.*]] = "tf.PartitionedCall"(%arg0, %[[CST]])
// CHECK-SAME: f = @composite_conv3d_with_relu_fn_1}>
// CHECK-SAME: {_tfl_quant_trait = "fully_quantizable"

// CHECK: %[[PARTITIONEDCALL_1:.*]] = "tf.PartitionedCall"(%arg0, %[[CST]])
// CHECK-SAME: f = @composite_conv3d_with_relu6_fn_1

   (IsIntSplatValueEqual<"int32_t", "-128"> $min_value),
   (IsIntSplatValueEqual<"int32_t", "127"> $max_value)]>;

// This pattern optimizes:
//   (x + cst1) + cst2 -> x + cst
//   (x - cst1) - cst2 -> x - cst
// Where: cst = cst1 + cst2
foreach BinaryOp = [TF_AddV2Op, TF_SubOp] in {
  def OptimizeConsecutive#BinaryOp : Pat<
    (BinaryOp
      (BinaryOp $x, (TF_ConstOp:$cst1 $cst1_value)),

func.func @preparePrelu(%arg0: tensor<1x10x10x3xf32>) -> tensor<1x10x10x3xf32> {
  %cst = "arith.constant"() {value = dense<[[[1.66394591, 3.61694336, 2.0382936]]]> : tensor<1x1x3xf32>} : () -> tensor<1x1x3xf32>
  %prelu = "tfl.prelu"(%arg0, %cst) : (tensor<1x10x10x3xf32>, tensor<1x1x3xf32>) -> tensor<1x10x10x3xf32>
  func.return %prelu : tensor<1x10x10x3xf32>

  EXPECT_TRUE(const_op);

  return const_op;
}

TEST_F(GetSizeInBytesTest, Int32ScalarConstOpSizeInBytes) {
  constexpr absl::string_view kConstOpExpr =
      R"mlir(%cst = "tf.Const"() {value = dense<1> : tensor<i32>} : () -> tensor<i32>)mlir";

  Block block{};
  TF::ConstOp int_tensor_const_op = ParseConstOp(kConstOpExpr, block, ctx_);

  // CHECK:         [[cst:%.*]] = "tf.Const{{.*}} dense<{{\[}}2, 4]> : tensor<2xi32>
  // CHECK-NEXT:    return [[cst]] : tensor<2xi32>
  func.return %0: tensor<2xi32>
}

func.func @RemoveTrivialAdd(%arg0: tensor<2x2xf32>) -> tensor<2x2xf32> {
  %cst = arith.constant dense<0.0> : tensor<2x2xf32>
  %0 = "tf.Add"(%arg0, %cst) : (tensor<2x2xf32>, tensor<2x2xf32>) -> tensor<2x2xf32>

	"Korea Standard Time":             {"KST", "KST"},     // Asia/Seoul
	"China Standard Time":             {"CST", "CST"},     // Asia/Shanghai
	"Singapore Standard Time":         {"+08", "+08"},     // Asia/Singapore
	"Russia Time Zone 10":             {"+11", "+11"},     // Asia/Srednekolymsk
	"Taipei Standard Time":            {"CST", "CST"},     // Asia/Taipei
	"West Asia Standard Time":         {"+05", "+05"},     // Asia/Tashkent