// CHECK: [[ACCUM_NEW:%.*]] = "tf.AddV2"([[ACCUM_MOMENTUM]], [[GRAD]])
    // CHECK: "tf.AssignVariableOp"([[ACCUM_HANDLE]], [[ACCUM_NEW]])
    // CHECK: [[GRAD_LR:%.*]] = "tf.Mul"([[GRAD]], [[LR]])
    // CHECK: [[MOMENTUM_LR:%.*]] = "tf.Mul"([[MOMENTUM]], [[LR]])
    // CHECK: [[ACCUM_NEW_MOMENTUM_LR:%.*]] = "tf.Mul"([[ACCUM_NEW]], [[MOMENTUM_LR]])
    // CHECK: [[DELTA:%.*]] = "tf.AddV2"([[GRAD_LR]], [[ACCUM_NEW_MOMENTUM_LR]])

// CHECK-NEXT: [[loop_counter:%.*]] = "tf.AddV2"(%arg2, [[cst_1]])
// CHECK-NEXT: [[weight:%.*]] = "tf.ReadVariableOp"(%arg5)
// CHECK-NEXT: [[mpy:%.*]] = "tf.MatMul"(%arg6, [[weight]])
// CHECK-NEXT: [[element_index:%.*]] = "tf.AddV2"(%arg3, [[cst_1]])
// CHECK-NEXT: [[bias:%.*]] = "tf.GatherV2"(%arg7, %arg3, [[cst_0]])
// CHECK-NEXT: [[res:%.*]] = "tf.AddV2"([[mpy]], [[bias]])

//
// is lowered to:
//
//   %sum0 = "tf.AddV2"(%0, %1)
//   %result = "tf.AddV2"(%sum0, %2)
//
// While
//
//   %result = "tf.AddN"(%0, %1, %2, %3, %4)
//
// is lowered to:
//
//   %sum0 = "tf.AddV2"(%0, %1)
//   %sum1 = "tf.AddV2"(%2, %3)
//   %sum2 = "tf.AddV2"(%sum0, %sum1)
//   %result = "tf.AddV2"(%sum2, %4)
//
class LowerAddNOp : public RewritePattern {
 public:

  %0 = "tf.AddV2"(%arg0, %arg1) : (tensor<ui32>, tensor<ui32>) -> tensor<ui32>
  func.return %0 : tensor<ui32>

  // CHECK-LABEL: add_v2_uint32
  // CHECK:  %[[CAST:.*]] = "tf.Cast"(%arg0) <{Truncate = false}> : (tensor<ui32>) -> tensor<i32>
  // CHECK:  %[[CAST1:.*]] = "tf.Cast"(%arg1) <{Truncate = false}> : (tensor<ui32>) -> tensor<i32>

		ssa.OpPPC64LoweredAtomicAdd64:
		// LWSYNC
		// LDAR/LWAR    (Rarg0), Rout
		// ADD		Rarg1, Rout
		// STDCCC/STWCCC Rout, (Rarg0)
		// BNE         -3(PC)
		// MOVW		Rout,Rout (if Add32)
		ld := ppc64.ALDAR
		st := ppc64.ASTDCCC
		if v.Op == ssa.OpPPC64LoweredAtomicAdd32 {
			ld = ppc64.ALWAR
			st = ppc64.ASTWCCC
		}
		r0 := v.Args[0].Reg()
		r1 := v.Args[1].Reg()
		out := v.Reg0()

// GOPPC64 values indicate power8, power9, etc.
// That means the code is compiled for that target,
// and will not run on earlier targets.
//
(Add(Ptr|64|32|16|8) ...) => (ADD ...)
(Add64F ...) => (FADD ...)
(Add32F ...) => (FADDS ...)

(Sub(Ptr|64|32|16|8) ...) => (SUB ...)
(Sub32F ...) => (FSUBS ...)
(Sub64F ...) => (FSUB ...)

(Min(32|64)F x y) && buildcfg.GOPPC64 >= 9 => (XSMINJDP x y)