RORIW	$32, X5, X6			// ERROR "immediate out of range 0 to 31"
	SLLIW	$32, X5, X6			// ERROR "immediate out of range 0 to 31"
	SRLIW	$32, X5, X6			// ERROR "immediate out of range 0 to 31"
	SRAIW	$32, X5, X6			// ERROR "immediate out of range 0 to 31"
	RORIW	$-1, X5, X6			// ERROR "immediate out of range 0 to 31"
	SLLIW	$-1, X5, X6			// ERROR "immediate out of range 0 to 31"

	/* whether x is NaN */; \
	FEQD	F0, F0, X6;	\
	BNEZ	X6, 3(PC);	\
	/* return NaN if x is NaN */; \
	MOVD	F0, ret+8(FP); 	\
	RET;			\
	MOV	$PosInf, X6;	\
	FMVDX	X6, F1;		\
	FABSD	F0, F2;		\
	/* if abs(x) > +Inf, return Inf instead of round(x) */; \
	FLTD	F1, F2, X6;	\
	/* Inf should keep same signed with x then return */;	\
	BEQZ	X6, 3(PC); \
	FCVTLD.MODE	F0, X6;	\
	FCVTDL	X6, F1;		\

	MOV	(((index-16)&0xf)*8)(X19), X21; \
	ROR	$19, X5, X7; \
	ROR	$61, X5, X8; \
	SRL	$6, X5; \
	XOR	X7, X5; \
	XOR	X8, X5; \
	ADD	X9, X5; \
	ROR	$1, X6, X7; \
	ROR	$8, X6, X8; \
	SRL	$7, X6; \
	XOR	X7, X6; \
	XOR	X8, X6; \
	ADD	X6, X5; \
	ADD	X21, X5; \
	MOV	X5, ((index&0xf)*8)(X19)

// Calculate T1 in X5.
// h is also used as an accumulator. Wt is passed in X5.
//   T1 = h + BIGSIGMA1(e) + Ch(e, f, g) + Kt + Wt

		x1 := byteorder.LeUint32(q[4*0x1:])
		x2 := byteorder.LeUint32(q[4*0x2:])
		x3 := byteorder.LeUint32(q[4*0x3:])
		x4 := byteorder.LeUint32(q[4*0x4:])
		x5 := byteorder.LeUint32(q[4*0x5:])
		x6 := byteorder.LeUint32(q[4*0x6:])
		x7 := byteorder.LeUint32(q[4*0x7:])
		x8 := byteorder.LeUint32(q[4*0x8:])
		x9 := byteorder.LeUint32(q[4*0x9:])
		xa := byteorder.LeUint32(q[4*0xa:])
		xb := byteorder.LeUint32(q[4*0xb:])

	SHUFPS $0x10, X0, X4
	PXOR X4, X0
	SHUFPS $0x8c, X0, X4
	PXOR X4, X0
	PXOR X1, X0

	MOVAPS X2, X5
	MOVAPS X2, X6
	PSLLDQ $0x4, X5
	PSHUFD $0xff, X0, X3
	PXOR X3, X2
	PXOR X5, X2

	MOVAPS X0, X1
	SHUFPS $0x44, X0, X6
	MOVUPS X6, (BX)
	SHUFPS $0x4e, X2, X1
	MOVUPS X1, 16(BX)
	ADDQ $32, BX
	RET

TEXT _expand_key_192b<>(SB),NOSPLIT,$0
	PSHUFD $0x55, X1, X1

	p256OrdSqr(t, x, 1)          // _10100
	p256OrdMul(_10101, t, _1)    // _10101
	p256OrdSqr(x, _10101, 1)     // _101010
	p256OrdMul(_101111, _101, x) // _101111
	p256OrdMul(x, _10101, x)     // _111111 = x6
	p256OrdSqr(t, x, 2)          // _11111100
	p256OrdMul(t, t, _11)        // _11111111 = x8
	p256OrdSqr(x, t, 8)          // _ff00
	p256OrdMul(x, x, t)          // _ffff = x16
	p256OrdSqr(t, x, 16)         // _ffff0000

	// seq 1 206 | sed 's/.*/	var x& [10<<20]byte/'
	// seq 1 206 | sed 's/.*/	z = x&/'
	var x1 [10<<20]byte
	var x2 [10<<20]byte
	var x3 [10<<20]byte
	var x4 [10<<20]byte
	var x5 [10<<20]byte
	var x6 [10<<20]byte
	var x7 [10<<20]byte
	var x8 [10<<20]byte
	var x9 [10<<20]byte
	var x10 [10<<20]byte
	var x11 [10<<20]byte
	var x12 [10<<20]byte
	var x13 [10<<20]byte
	var x14 [10<<20]byte
	var x15 [10<<20]byte