TEXT ·testBGE(SB),NOSPLIT,$0-17
	MOV	a+0(FP), X5
	MOV	b+8(FP), X6
	MOV	$1, X7
	BGE	X5, X6, b
	MOV	$0, X7
b:
	MOV	X7, r+16(FP)
	RET

// func testBGEU(a, b int64) (r bool)
TEXT ·testBGEU(SB),NOSPLIT,$0-17
	MOV	a+0(FP), X5
	MOV	b+8(FP), X6
	MOV	$1, X7
	BGEU	X5, X6, b
	MOV	$0, X7
b:
	MOV	X7, r+16(FP)
	RET

// func testBGEZ(a int64) (r bool)
TEXT ·testBGEZ(SB),NOSPLIT,$0-9

	MULHU	X5, X6, X7				// b3335302
	MULHSU	X5, X6, X7				// b3235302
	MULW	X5, X6, X7				// bb035302
	DIV	X5, X6, X7				// b3435302
	DIVU	X5, X6, X7				// b3535302
	REM	X5, X6, X7				// b3635302
	REMU	X5, X6, X7				// b3735302
	DIVW	X5, X6, X7				// bb435302
	DIVUW	X5, X6, X7				// bb535302
	REMW	X5, X6, X7				// bb635302
	REMUW	X5, X6, X7				// bb735302

	// 8.2: Load-Reserved/Store-Conditional

//     Maj(x, y, z) = (x AND y) XOR (x AND z) XOR (y AND z)
#define SHA512T2(a, b, c) \
	ROR	$28, a, X6; \
	ROR	$34, a, X7; \
	XOR	X7, X6; \
	ROR	$39, a, X8; \
	XOR	X8, X6; \
	AND	a, b, X7; \
	AND	a, c, X8; \
	XOR	X8, X7; \
	AND	b, c, X9; \
	XOR	X9, X7; \
	ADD	X7, X6

// Calculate T1 and T2, then e = d + T1 and a = T1 + T2.
// The values for e and a are stored in d and h, ready for rotation.

	AND	$7, X10, X7
	AND	$7, X12, X8
	BNE	X7, X8, check8_unaligned
	BEQZ	X7, compare32

	// Check one byte at a time until we reach 8 byte alignment.
	SUB	X7, X0, X7
	ADD	$8, X7, X7
	SUB	X7, X5, X5
align:
	SUB	$1, X7
	MOVBU	0(X10), X8
	MOVBU	0(X12), X9
	BNE	X8, X9, cmp
	ADD	$1, X10
	ADD	$1, X12
	BNEZ	X7, align

check32:
	// X6 contains $32

TEXT addMulVVWx(SB),NOFRAME|NOSPLIT,$0
	MOV	z+0(FP), X5
	MOV	x+8(FP), X7
	MOV	y+16(FP), X6
	MOV	$0, X29

	BEQZ	X30, done
loop:
	MOV	0*8(X5), X10	// z[0]
	MOV	1*8(X5), X13	// z[1]
	MOV	2*8(X5), X16	// z[2]
	MOV	3*8(X5), X19	// z[3]

	MOV	0*8(X7), X8	// x[0]
	MOV	1*8(X7), X11	// x[1]
	MOV	2*8(X7), X14	// x[2]
	MOV	3*8(X7), X17	// x[3]

	MULHU	X8, X6, X9	// z_hi[0] = x[0] * y

	VREPF $1, KEY0, X5
	VREPF $2, KEY0, X6
	VREPF $3, KEY0, X7
	VREPF $0, KEY1, X8
	VREPF $1, KEY1, X9
	VREPF $2, KEY1, X10
	VREPF $3, KEY1, X11
	VLR   CTR, X12
	VREPF $1, NONCE, X13
	VREPF $2, NONCE, X14
	VREPF $3, NONCE, X15

	MOVD $(NUM_ROUNDS/2), R1

loop:
	ROUND4(X0, X4, X12,  X8, X1, X5, X13,  X9, X2, X6, X14, X10, X3, X7, X15, X11)

	x6, x5 = bits.Mul64(x4, arg2[3])
	var x7 uint64
	var x8 uint64
	x8, x7 = bits.Mul64(x4, arg2[2])
	var x9 uint64
	var x10 uint64
	x10, x9 = bits.Mul64(x4, arg2[1])
	var x11 uint64
	var x12 uint64
	x12, x11 = bits.Mul64(x4, arg2[0])
	var x13 uint64
	var x14 uint64
	x13, x14 = bits.Add64(x12, x9, uint64(0x0))
	var x15 uint64
	var x16 uint64
	x15, x16 = bits.Add64(x10, x7, uint64(fiatScalarUint1(x14)))

	_ = x1
	var x2 C.jclass = 0
	_ = x2
	var x3 C.jthrowable = 0
	_ = x3
	var x4 C.jstring = 0
	_ = x4
	var x5 C.jarray = 0
	_ = x5
	var x6 C.jbooleanArray = 0
	_ = x6
	var x7 C.jbyteArray = 0
	_ = x7
	var x8 C.jcharArray = 0
	_ = x8
	var x9 C.jshortArray = 0
	_ = x9
	var x10 C.jintArray = 0
	_ = x10
	var x11 C.jlongArray = 0
	_ = x11
	var x12 C.jfloatArray = 0
	_ = x12

	MOVOU	48(SI), X3
	MOVOU	-64(SI)(BX*1), X4
	MOVOU	-48(SI)(BX*1), X5
	MOVOU	-32(SI)(BX*1), X6
	MOVOU	-16(SI)(BX*1), X7
	MOVOU	X0, (DI)
	MOVOU	X1, 16(DI)
	MOVOU	X2, 32(DI)
	MOVOU	X3, 48(DI)
	MOVOU	X4, -64(DI)(BX*1)
	MOVOU	X5, -48(DI)(BX*1)
	MOVOU	X6, -32(DI)(BX*1)
	MOVOU	X7, -16(DI)(BX*1)
	RET
move_129through256:
	MOVOU	(SI), X0
	MOVOU	16(SI), X1
	MOVOU	32(SI), X2
	MOVOU	48(SI), X3

	// come in on the m->g0 stack already. Or we might already
	// be on the m->gsignal stack.
	MOV	g_m(g), X6
	MOV	m_gsignal(X6), X7
	BEQ	X7, g, g0
	MOV	m_g0(X6), X7
	BEQ	X7, g, g0

	CALL	gosave_systemstack_switch<>(SB)
	MOV	X7, g
	CALL	runtime·save_g(SB)
	MOV	(g_sched+gobuf_sp)(g), X2

	// Now on a scheduling stack (a pthread-created stack).
g0: