final:
	// Capture CA
	SUBE  R4, R4
	NEG   R4, R4

done:
	MOVD  R4, c+56(FP)
	RET

//func shlVU(z, x []Word, s uint) (c Word)
TEXT ·shlVU(SB), NOSPLIT, $0
	MOVD    z+0(FP), R3
	MOVD    x+24(FP), R6
	MOVD    s+48(FP), R9
	MOVD    z_len+8(FP), R4
	MOVD    x_len+32(FP), R7
	CMP     R9, $0          // s==0 copy(z,x)
	BEQ     zeroshift
	CMP     R4, $0          // len(z)==0 return
	BEQ     done

	MOVD $ret+0(FP), R1 // address of 16-byte return value
	KMCTR R2, R4, R4    // cipher message with counter (KMCTR)
	RET

// func kmaQuery() queryResult
TEXT ·kmaQuery(SB), NOSPLIT|NOFRAME, $0-16
	MOVD $0, R0         // set function code to 0 (KMA-Query)
	MOVD $ret+0(FP), R1 // address of 16-byte return value
	KMA  R2, R6, R4     // cipher message with authentication (KMA)
	RET

// func kimdQuery() queryResult

	RET

zero_by_16:
	// n greater than 16 bytes, check if the start address is aligned
	NEG	R0, R4
	ANDS	$15, R4, R4
	// Try zeroing using zva if the start address is aligned with 16
	BEQ	try_zva

	// Non-aligned store
	STP	(ZR, ZR), (R0)
	// Make the destination aligned
	SUB	R4, R1, R1
	ADD	R4, R0, R0
	B	try_zva

tail_maybe_long:
	CMP	$64, R1
	BHS	no_zva

tail63:
	ANDS	$48, R1, R3

TEXT ·rawVforkSyscall(SB),NOSPLIT,$0-48
	MOVV	a1+8(FP), R4
	MOVV	a2+16(FP), R5
	MOVV	a3+24(FP), R6
	MOVV	$0, R7
	MOVV	$0, R8
	MOVV	$0, R9
	MOVV	trap+0(FP), R11	// syscall entry
	SYSCALL
	MOVW	$-4096, R12
	BGEU	R12, R4, ok
	MOVV	$-1, R12
	MOVV	R12, r1+32(FP)	// r1
	SUBVU	R4, R0, R4
	MOVV	R4, err+40(FP)	// errno
	RET
ok:
	MOVV	R4, r1+32(FP)	// r1
	MOVV	R0, err+40(FP)	// errno
	RET

	JAL	runtime·save_g(SB)

	MOVV	gobuf_sp(R4), R3
	MOVV	gobuf_lr(R4), R1
	MOVV	gobuf_ret(R4), R19
	MOVV	gobuf_ctxt(R4), REGCTXT
	MOVV	R0, gobuf_sp(R4)
	MOVV	R0, gobuf_ret(R4)
	MOVV	R0, gobuf_lr(R4)
	MOVV	R0, gobuf_ctxt(R4)
	MOVV	gobuf_pc(R4), R6
	JMP	(R6)

// void mcall(fn func(*g))
// Switch to m->g0's stack, call fn(g).
// Fn must never return. It should gogo(&g->sched)
// to keep running g.

Examples:

	MOVD (R3), R4		<=>	ld r4,0(r3)
	MOVW (R3), R4		<=>	lwa r4,0(r3)
	MOVWZU 4(R3), R4		<=>	lwzu r4,4(r3)
	MOVWZ (R3+R5), R4		<=>	lwzx r4,r3,r5
	MOVHZ  (R3), R4		<=>	lhz r4,0(r3)
	MOVHU 2(R3), R4		<=>	lhau r4,2(r3)
	MOVBZ (R3), R4		<=>	lbz r4,0(r3)

	MOVD R4,(R3)		<=>	std r4,0(r3)
	MOVW R4,(R3)		<=>	stw r4,0(r3)
	MOVW R4,(R3+R5)		<=>	stwx r4,r3,r5
	MOVWU R4,4(R3)		<=>	stwu r4,4(r3)

	// Synchronous initialization.
	MOVD	$runtime·libpreinit(SB), R4
	BL	(R4)

	// Create a new thread to do the runtime initialization and return.
	MOVD	_cgo_sys_thread_create(SB), R4
	CBZ	R4, nocgo
	MOVD	$_rt0_arm64_netbsd_lib_go(SB), R0
	MOVD	$0, R1
	SUB	$16, RSP		// reserve 16 bytes for sp-8 where fp may be saved.
	BL	(R4)
	ADD	$16, RSP
	B	restore

nocgo:

	r3 = addMul64(r3, a1, b2)
	r3 = addMul64(r3, a2, b1)
	r3 = addMul64(r3, a3, b0)
	r3 = addMul64(r3, a4_19, b4)

	// r4 = a0×b4 + a1×b3 + a2×b2 + a3×b1 + a4×b0
	r4 := mul64(a0, b4)
	r4 = addMul64(r4, a1, b3)
	r4 = addMul64(r4, a2, b2)
	r4 = addMul64(r4, a3, b1)
	r4 = addMul64(r4, a4, b0)

	// After the multiplication, we need to reduce (carry) the five coefficients

	// Case for >= 64 bytes
	// Process 64 bytes per iteration
	// Load 4 vectors of a and b
	// XOR the corresponding vectors
	// from a and b and store the result
loop64:
	LXVD2X	(R4)(R8), VS32
	LXVD2X	(R4)(R10), VS34
	LXVD2X	(R4)(R14), VS36
	LXVD2X	(R4)(R15), VS38
	LXVD2X	(R5)(R8), VS33
	LXVD2X	(R5)(R10), VS35
	LXVD2X	(R5)(R14), VS37
	LXVD2X	(R5)(R15), VS39
	XXLXOR	VS32, VS33, VS32
	XXLXOR	VS34, VS35, VS34

	// Synchronous initialization.
	MOVD	$runtime·libpreinit(SB), R4
	BL	(R4)

	// Create a new thread to do the runtime initialization and return.
	MOVD	_cgo_sys_thread_create(SB), R4
	CBZ	R4, nocgo
	MOVD	$_rt0_arm64_openbsd_lib_go(SB), R0
	MOVD	$0, R1
	SUB	$16, RSP		// reserve 16 bytes for sp-8 where fp may be saved.
	BL	(R4)
	ADD	$16, RSP
	B	restore

nocgo: