// -----

// CHECK: tac.cost = 0x4B673001
func.func @quantize_ops_CPU_QUANTIZED_INT8(%arg0: tensor<384x512x!quant.uniform<i8:f32, 0.1>>, %arg1: tensor<128x512x!quant.uniform<i8<-127:127>:f32, 0.1>>, %arg2: tensor<128x!quant.uniform<i8:f32, 0.2:-128>>, %arg3: tensor<128x!quant.uniform<i8:f32, 0.2:-4>>) -> tensor<1x384x128x!quant.uniform<i8:f32, 0.3:-3>>  attributes {tac.device = "CPU", tac.inference_type = "QUANTIZED_INT8", tac.interface_name = "quantize_ops"} {

  %requant = stablehlo.uniform_quantize %conv : (tensor<1x3x4x2x!quant.uniform<i32:f32, 5.8949912267181218E-5>>) -> tensor<1x3x4x2x!quant.uniform<i8:f32, 0.045673100153605144:-62>>

  %0 = "tfl.pseudo_qconst"() { qtype = tensor<3x!quant.uniform<i8:f32, 0.1>>, value = dense<2> : tensor<3xi8>} : () -> tensor<3x!quant.uniform<i8:f32, 0.1>>
  %1 = "tfl.mul"(%arg0, %0) {fused_activation_function = "NONE"} : (tensor<3x!quant.uniform<i8:f32, 0.1>>, tensor<3x!quant.uniform<i8:f32, 0.1>>) -> tensor<3x!quant.uniform<i8:f32, 0.1>> loc("mul")
  func.return %1 : tensor<3x!quant.uniform<i8:f32, 0.1>>

  %0 = "tfl.pseudo_qconst"() { qtype = tensor<3x!quant.uniform<i8:f32, 1.0>>, value = dense<2> : tensor<3xi8>} : () -> tensor<3x!quant.uniform<i8:f32, 1.0>>
  %1 = "tfl.mul"(%arg0, %0) {fused_activation_function = "NONE"} : (tensor<3x!quant.uniform<i8:f32, 1.0>>, tensor<3x!quant.uniform<i8:f32, 1.0>>) -> tensor<3x!quant.uniform<i8:f32, 1.0>> loc("mul")
  func.return %1 : tensor<3x!quant.uniform<i8:f32, 1.0>>

// CHECK: %[[BITCAST_CONVERT_0:.*]] = stablehlo.bitcast_convert %[[UQ]] : (tensor<1x!quant.uniform<i8:f32, 1.000000e+00:3>>) -> tensor<1xi8>
// CHECK: %[[BITCAST_CONVERT_1:.*]] = stablehlo.bitcast_convert %[[BITCAST_CONVERT_0]] : (tensor<1xi8>) -> tensor<1x!quant.uniform<i8:f32, 1.000000e+00:3>>

// CHECK: %[[q1:.*]] = "quantfork.qcast"(%[[maxpool]])
// CHECK-SAME: quant.uniform<i8:f32, 5.000000e-02:-10>
// CHECK: %[[dq1:.*]] = "quantfork.dcast"(%[[q1]])
// CHECK-SAME: quant.uniform<i8:f32, 5.000000e-02:-10>
// CHECK: %[[reshape:.*]] = "tf.Reshape"(%[[dq1]]
// CHECK: %[[q2:.*]] = "quantfork.qcast"(%[[reshape]])
// CHECK-SAME: quant.uniform<i8:f32, 5.000000e-02:-10>
// CHECK: %[[dq2:.*]] = "quantfork.dcast"(%[[q2]])

}

func main() {
	var ia [20]I0
	var b bool
	var s string
	var i8 int8
	var i16 int16
	var i32 int32
	var i64 int64
	var u8 uint8
	var u16 uint16
	var u32 uint32
	var u64 uint64

	f()

	ia[0] = "xxx"
	ia[1] = 12345
	ia[2] = true

	s = "now is"
	ia[3] = s
	b = false
	ia[4] = b

	i8 = 29
	ia[5] = i8
	i16 = 994
	ia[6] = i16
	i32 = 3434
	ia[7] = i32
	i64 = 1234567

  %1 = "quantfork.qcast"(%0) : (tensor<8xf32>) -> tensor<8x!quant.uniform<i8:f32, 1.000000e+00:-128>>
  func.return %1 : tensor<8x!quant.uniform<i8:f32, 1.000000e+00:-128>>

// CHECK:  %0 = "tf.FakeQuantWithMinMaxVars"(%arg0, %cst, %cst_0) <{narrow_range = false, num_bits = 5 : i64}>
// CHECK:  %1 = "quantfork.qcast"(%0) : (tensor<8xf32>) -> tensor<8x!quant.uniform<i8:f32, 1.000000e+00:-128>>
// CHECK:  return %1
}

    %cst = stablehlo.constant dense<3.000000e-01> : tensor<2x3xf32>
    %0 = "quantfork.qcast"(%cst) : (tensor<2x3xf32>) -> tensor<2x3x!quant.uniform<i8:f32, 6.000000e-03>>
    %1 = "quantfork.dcast"(%0) : (tensor<2x3x!quant.uniform<i8:f32, 6.000000e-03>>) -> tensor<2x3xf32>

func main() {
	chki8(int8(i8), ci8&0xff-1<<8)
	chki8(int8(i16), ci16&0xff)
	chki8(int8(i32), ci32&0xff-1<<8)
	chki8(int8(i64), ci64&0xff)
	chki8(int8(u8), cu8&0xff-1<<8)
	chki8(int8(u16), cu16&0xff)
	chki8(int8(u32), cu32&0xff)
	chki8(int8(u64), cu64&0xff)
	//	chki8(int8(f32), 0)
	//	chki8(int8(f64), 0)

	chki16(int16(i8), ci8&0xffff-1<<16)
	chki16(int16(i16), ci16&0xffff-1<<16)