} else {
				ss = z / cc
			}
		}

		// j1(x) = 1/sqrt(pi) * (P(1,x)*cc - Q(1,x)*ss) / sqrt(x)
		// y1(x) = 1/sqrt(pi) * (P(1,x)*ss + Q(1,x)*cc) / sqrt(x)

		var z float64
		if x > Two129 {
			z = (1 / SqrtPi) * cc / Sqrt(x)
		} else {
			u := pone(x)
			v := qone(x)
			z = (1 / SqrtPi) * (u*cc - v*ss) / Sqrt(x)
		}
		if sign {
			return -z
		}
		return z
	}

	"NMULAF",
	"NMULAD",
	"MULSF",
	"MULSD",
	"NMULSF",
	"NMULSD",
	"FMULAF",
	"FMULAD",
	"FNMULAF",
	"FNMULAD",
	"FMULSF",
	"FMULSD",
	"FNMULSF",
	"FNMULSD",
	"DIVF",
	"DIVD",
	"SQRTF",
	"SQRTD",
	"ABSF",
	"ABSD",
	"NEGF",
	"NEGD",
	"SRL",
	"SRA",
	"SLL",
	"MULU",
	"DIVU",
	"MUL",
	"MMUL",
	"DIV",
	"MOD",
	"MODU",
	"DIVHW",
	"DIVUHW",
	"MOVB",
	"MOVBS",

		{name: "NEGF", argLength: 1, reg: fp11, asm: "NEGF"},   // -arg0, float32
		{name: "NEGD", argLength: 1, reg: fp11, asm: "NEGD"},   // -arg0, float64
		{name: "SQRTD", argLength: 1, reg: fp11, asm: "SQRTD"}, // sqrt(arg0), float64
		{name: "SQRTF", argLength: 1, reg: fp11, asm: "SQRTF"}, // sqrt(arg0), float32

		{name: "MASKEQZ", argLength: 2, reg: gp21, asm: "MASKEQZ"}, // returns 0 if arg1 == 0, otherwise returns arg0

		{name: "NEGD", argLength: 1, reg: fp11, asm: "NEGD"},   // -arg0, float64
		{name: "ABSD", argLength: 1, reg: fp11, asm: "ABSD"},   // abs(arg0), float64
		{name: "SQRTD", argLength: 1, reg: fp11, asm: "SQRTD"}, // sqrt(arg0), float64
		{name: "SQRTF", argLength: 1, reg: fp11, asm: "SQRTF"}, // sqrt(arg0), float32

		// shifts
		{name: "SLLV", argLength: 2, reg: gp21, asm: "SLLV"},                    // arg0 << arg1, shift amount is mod 64

}

func TestDecimalConstants(t *testing.T) {
	sqrtM1String := "19681161376707505956807079304988542015446066515923890162744021073123829784752"
	if exp := new(Element).fromDecimal(sqrtM1String); sqrtM1.Equal(exp) != 1 {
		t.Errorf("sqrtM1 is %v, expected %v", sqrtM1, exp)
	}
	// d is in the parent package, and we don't want to expose d or fromDecimal.
	// dString := "37095705934669439343138083508754565189542113879843219016388785533085940283555"

		if x >= Two302 { // x > 2**302

			// (x >> n**2)
			//          Jn(x) = cos(x-(2n+1)*pi/4)*sqrt(2/x*pi)
			//          Yn(x) = sin(x-(2n+1)*pi/4)*sqrt(2/x*pi)
			//          Let s=sin(x), c=cos(x),
			//              xn=x-(2n+1)*pi/4, sqt2 = sqrt(2),then
			//
			//                 n    sin(xn)*sqt2    cos(xn)*sqt2
			//              ----------------------------------
			//                 0     s-c             c+s

	// amd64/v3:"BLSRL",-"TESTL",-"CALL"
	if isNotPowerOfTwo32(x) {
		a(true)
	} else {
		b("false")
	}
}

func sarx64(x, y int64) int64 {
	// amd64/v3:"SARXQ"
	return x >> y
}

func sarx32(x, y int32) int32 {
	// amd64/v3:"SARXL"
	return x >> y
}

func sarx64_load(x []int64, i int) int64 {
	// amd64/v3: `SARXQ\t[A-Z]+[0-9]*, \([A-Z]+[0-9]*\)\([A-Z]+[0-9]*\*8\), [A-Z]+[0-9]*`

	DIVF.NE	F0, F2        // 002a821e
	DIVD	F3, F5        // 035b85ee
	NEGF	F0, F1        // 401ab1ee
	NEGD	F4, F5        // 445bb1ee
	ABSF	F0, F1        // c01ab0ee
	ABSD	F4, F5        // c45bb0ee
	SQRTF	F0, F1        // c01ab1ee
	SQRTD	F4, F5        // c45bb1ee
	MOVFD	F0, F1        // c01ab7ee
	MOVDF	F4, F5        // c45bb7ee

	LDREX	(R8), R9      // 9f9f98e1
	LDREXD	(R11), R12    // 9fcfbbe1

TAC_REGISTER_GPU_OP(SinOp, CreateBasicOpNoCost);
TAC_REGISTER_GPU_OP(SliceOp, CreateBasicOpNoCost);
TAC_REGISTER_GPU_OP(SoftmaxOp, CreateBasicOpNoCost);
TAC_REGISTER_GPU_OP(SpaceToDepthOp, CreateBasicOpNoCost);
TAC_REGISTER_GPU_OP(SqrtOp, CreateBasicOpNoCost);
TAC_REGISTER_GPU_OP(SquareOp, CreateBasicOpNoCost);
TAC_REGISTER_GPU_OP(SquaredDifferenceOp, CreateBasicOpNoCost);
TAC_REGISTER_GPU_OP(StridedSliceOp, CreateBasicOpNoCost);

		{name: "BSWAPL", argLength: 1, reg: gp11, asm: "BSWAPL", resultInArg0: true}, // arg0 swap bytes

		{name: "SQRTSD", argLength: 1, reg: fp11, asm: "SQRTSD"}, // sqrt(arg0)
		{name: "SQRTSS", argLength: 1, reg: fp11, asm: "SQRTSS"}, // sqrt(arg0), float32

		{name: "SBBLcarrymask", argLength: 1, reg: flagsgp, asm: "SBBL"}, // (int32)(-1) if carry is set, 0 if carry is clear.