// loong64: "ADDV", "SGTU"
	// ppc64x: "ADDC", "ADDE", "ADDZE"
	// s390x:"ADDE","ADDC\t[$]-1,"
	// mips64:"ADDV","SGTU"
	// riscv64: "ADD","SLTU"
	return bits.Add(x, 7, ci)
}

func AddZ(x, y uint) (r, co uint) {
	// arm64:"ADDS","ADC",-"ADCS",-"ADD\t",-"CMP"
	// amd64:"ADDQ","SBBQ","NEGQ",-"NEGL",-"ADCQ"
	// loong64: "ADDV", "SGTU"
	// ppc64x: "ADDC", -"ADDE", "ADDZE"
	// s390x:"ADDC",-"ADDC\t[$]-1,"

// CHECK:     %[[ADD1:.*]], %[[ADD1_control:.*]] = tf_executor.island wraps "tf.Add"(%arg0, %arg1)
// CHECK:     %[[ADD2:.*]], %[[ADD2_control:.*]] = tf_executor.island wraps "tf.Add"(%[[ADD1]], %arg1)
// CHECK:     %[[PRINT:.*]], %[[PRINT_control:.*]] = tf_executor.island wraps "tf.Print"(%[[ADD2]]) <{message = "add2 result"}>
// CHECK:     tf_executor.fetch %[[ADD1]], %[[ADD2]], %[[PRINT_control]] :
// CHECK:   }

  %graph:2 = tf_executor.graph {
    %island1:3 = tf_executor.island {
      %add1 = "tf.Add"(%arg0, %arg1) : (tensor<*xi32>, tensor<i32>) -> tensor<*xi32>
      %add2 = "tf.Add"(%add1, %arg1) : (tensor<*xi32>, tensor<i32>) -> tensor<*xi32>
      %res = "tf.Print"(%add2) { message = "add result" } : (tensor<*xi32>) -> (tensor<*xi32>)
      tf_executor.yield %add1, %add2 : tensor<*xi32>, tensor<*xi32>
    }

	MULHDU R9, R6
	MOVD   (R2)(R1*1), R10
	ADDC   R10, R11        // add to low order bits
	ADDE   R0, R6
	ADDC   R4, R11
	ADDE   R0, R6
	MOVD   R6, R4
	MOVD   R11, (R2)(R1*1)

	MOVD   (8)(R8)(R1*1), R6
	MULHDU R9, R6
	MOVD   (8)(R2)(R1*1), R10
	ADDC   R10, R11           // add to low order bits
	ADDE   R0, R6
	ADDC   R4, R11
	ADDE   R0, R6
	MOVD   R6, R4
	MOVD   R11, (8)(R2)(R1*1)

  // CHECK-NOT: "tf.AssignVariableOp"
  // CHECK: %[[CONST:.*]] = "tf.Const"()
  // CHECK: %[[ADD1:[0-9]*]] = "tf.AddV2"(%arg1, %[[CONST]])
  // CHECK: %[[ADD2:[0-9]*]] = "tf.AddV2"(%[[ADD1]], %[[ADD1]])
  // CHECK: %[[PACK:[0-9]*]] = "tf.Pack"(%arg1, %[[ADD2]])
  // CHECK: return %[[PACK]], %[[ADD1]]

  %0 = "tf.Const"() {value = dense<4.200000e+01> : tensor<f32>} : () -> tensor<f32>

	MULHDU	R5, R14, R11	// high x[i]*y
	ADDC	R15, R10
	ADDZE	R11
	ADDC	R9, R10
	ADDZE	R11, R9
	MULLD	R5, R16, R14	// low x[i+1]*y
	MULHDU	R5, R16, R15	// high x[i+1]*y
	ADDC	R17, R14
	ADDZE	R15
	ADDC	R9, R14
	ADDZE	R15, R9
	MULLD	R5, R18, R16    // low x[i+2]*y
	MULHDU	R5, R18, R17    // high x[i+2]*y
	ADDC	R19, R16
	ADDZE	R17
	ADDC	R9, R16
	ADDZE	R17, R9

package foo

import (
	"errors"
	"runtime"
	"unsafe"
)

func add2(p *byte, n uintptr) *byte { // ERROR "can inline add2" "leaking param: p to result"
	return (*byte)(add1(unsafe.Pointer(p), n)) // ERROR "inlining call to add1"
}

func add1(p unsafe.Pointer, x uintptr) unsafe.Pointer { // ERROR "can inline add1" "leaking param: p to result"
	return unsafe.Pointer(uintptr(p) + x)
}

package foo

import (
	"errors"
	"runtime"
	"unsafe"
)

func add2(p *byte, n uintptr) *byte { // ERROR "can inline add2" "leaking param: p to result"
	return (*byte)(add1(unsafe.Pointer(p), n)) // ERROR "inlining call to add1"
}

func add1(p unsafe.Pointer, x uintptr) unsafe.Pointer { // ERROR "can inline add1" "leaking param: p to result"
	return unsafe.Pointer(uintptr(p) + x)
}

  TF_Operation* add1 = Add(params_->body_inputs[0], params_->body_inputs[1],
                           params_->body_graph, s_, "add1");
  ASSERT_EQ(TF_OK, TF_GetCode(s_)) << TF_Message(s_);
  TF_Operation* one = ScalarConst(1, params_->body_graph, s_);
  ASSERT_EQ(TF_OK, TF_GetCode(s_)) << TF_Message(s_);
  TF_Operation* add2 = Add(add1, one, params_->body_graph, s_, "add2");

	MOVBZ	R1, 4096(R2)(R3)       // e31320000172

	ADD	R1, R2                // b9e81022
	ADD	R1, R2, R3            // b9e81032
	ADD	$8192, R1             // a71b2000
	ADD	$8192, R1, R2         // ec21200000d9
	ADD	$32768, R1            // c21800008000
	ADD	$32768, R1, R2        // b9040021c22800008000
	ADDC	R1, R2                // b9ea1022
	ADDC	$1, R1, R2            // ec21000100db
	ADDC	$-1, R1, R2           // ec21ffff00db