// a4  | SI          | R10
// a5  | R8          | R8
// a6  | R9          | R9
//
// r1  | AX          | AX
// r2  | BX          | DX
// err | CX          | part of AX
//
// Note that this differs from "standard" ABI convention, which would pass 4th
// arg in CX, not R10.
TEXT ·Syscall6<ABIInternal>(SB),NOSPLIT,$0
	// a6 already in R9.
	// a5 already in R8.
	MOVQ	SI, R10 // a4
	MOVQ	DI, DX  // a3

	MOVL	d,		R9; \
	XORL	c,		R9; \
	XORL	b,		R9; \
	ADDL	$const,	a; \
	ADDL	R8,		a; \
	MOVL	(index*4)(SI),R8; \
	ADDL	R9,		a; \
	ROLL	$shift,		a; \
	ADDL	b,		a

#define ROUND3(a, b, c, d, index, const, shift) \
	XORL	a,		R9; \
	XORL	b,		R9; \
	ADDL	$const,	a; \
	ADDL	R8,		a; \
	MOVL	(index*4)(SI),R8; \
	ADDL	R9,		a; \
	ROLL	$shift,		a; \
	ADDL	b,		a

	JMP		addMulVVWx(SB)

TEXT addMulVVWx(SB), NOFRAME|NOSPLIT, $0
	MOVD z+0(FP), R2
	MOVD x+8(FP), R8
	MOVD y+16(FP), R9

	MOVD $0, R1 // i*8 = 0
	MOVD $0, R7 // i = 0
	MOVD $0, R0 // make sure it's zero
	MOVD $0, R4 // c = 0

	MOVD   R5, R12
	AND    $-2, R12
	CMPBGE R5, $2, A6
	BR     E6

A6:
	MOVD   (R8)(R1*1), R6
	MULHDU R9, R6
	MOVD   (R2)(R1*1), R10
	ADDC   R10, R11        // add to low order bits

	// R7 = byte to find
	AND	$0xff, R7
	MOVV	R4, R6		// store base for later
	ADDV	R4, R5		// end
	ADDV	$-1, R4

	PCALIGN	$16
loop:
	ADDV	$1, R4
	BEQ	R4, R5, notfound
	MOVBU	(R4), R8
	BNE	R7, R8, loop

	SUBV	R6, R4		// remove base
	RET

notfound:
	MOVV	$-1, R4
	RET

TEXT ·IndexByteString<ABIInternal>(SB),NOSPLIT,$0-32
	// R4 = s_base
	// R5 = s_len
	// R6 = byte to find

	RET

gcm_ghash_p8_4x:
	LVSL     (R8)(R0), T0      // 0x0001..0e0f
	MOVD     $0x70, R8
	LXVD2X   (HTBL)(R9), VH2
	MOVD     $0x80, R9
	VSPLTISB $8, T1            // 0x0808..0808
	MOVD     $0x90, R10
	LXVD2X   (HTBL)(R8), VH3L  // load H^3
	MOVD     $0xa0, R8
	LXVD2X   (HTBL)(R9), VH3
	MOVD     $0xb0, R9
	LXVD2X   (HTBL)(R10), VH3H
	MOVD     $0xc0, R10
	LXVD2X   (HTBL)(R8), VH4L  // load H^4
	MOVD     $0x10, R8

// performing [R4:R12:R11:R10:R9] = R4 + R3 * [R8:R7:R6:R5] + [R12:R11:R10:R9]
// where R4 is carried from the previous iteration, R8:R7:R6:R5 hold the next
// 4 words of x, R3 is y and R12:R11:R10:R9 are part of the result z.
loop:
	CBZ	R0, done

	LDP.P	16(R2), (R5, R6)
	LDP.P	16(R2), (R7, R8)

	LDP	(R1), (R9, R10)
	ADDS	R4, R9
	MUL	R6, R3, R14
	ADCS	R14, R10
	MUL	R7, R3, R15

	MOVD	num+0(FP), R1	// syscall entry
	MOVD	a1+8(FP), R2
	MOVD	a2+16(FP), R3
	MOVD	a3+24(FP), R4
	MOVD	a4+32(FP), R5
	MOVD	a5+40(FP), R6
	MOVD	a6+48(FP), R7
	SYSCALL
	MOVD	$0xfffffffffffff001, R8
	CMPUBLT	R2, R8, ok
	MOVD	$-1, r1+56(FP)
	MOVD	$0, r2+64(FP)
	NEG	R2, R2
	MOVD	R2, errno+72(FP)
	RET
ok:
	MOVD	R2, r1+56(FP)
	MOVD	R3, r2+64(FP)
	MOVD	$0, errno+72(FP)

end:
	MOVW	R1, ret+4(FP)
	RET

TEXT	·Store(SB),NOSPLIT,$0-8
	MOVW	addr+0(FP), R1
	MOVW	v+4(FP), R2

	MOVB	runtime·goarm(SB), R8
	CMP	$7, R8
	BGE	native_barrier
	BL	memory_barrier<>(SB)
	B	store
native_barrier:
	DMB	MB_ISH

store:
	MOVW	R2, (R1)

	CMP	$7, R8
	BGE	native_barrier2
	BL	memory_barrier<>(SB)
	RET
native_barrier2:
	DMB	MB_ISH
	RET

TEXT	·Load8(SB),NOSPLIT,$0-5

	MOVD	b_base+0(FP), R0
	MOVD	b_len+8(FP), R2
	MOVBU	c+24(FP), R1
	MOVD	$ret+32(FP), R8
	B	countbytebody<>(SB)

TEXT ·CountString(SB),NOSPLIT,$0-32
	MOVD	s_base+0(FP), R0
	MOVD	s_len+8(FP), R2
	MOVBU	c+16(FP), R1
	MOVD	$ret+24(FP), R8
	B	countbytebody<>(SB)

// input:
//   R0: data
//   R2: data len
//   R1: byte to find
//   R8: address to put result
TEXT countbytebody<>(SB),NOSPLIT,$0

	MOVW	12(R13), R0
	MOVW	$1234, R1
	CMP	R0, R1
	BEQ	2(PC)
	BL	runtime·abort(SB)

	MOVW	0(R13), R8    // m
	MOVW	4(R13), R0    // g

	CMP	$0, R8
	BEQ	nog
	CMP	$0, R0
	BEQ	nog

	MOVW	R0, g
	MOVW	R8, g_m(g)

	// paranoia; check they are not nil
	MOVW	0(R8), R0
	MOVW	0(g), R0

	BL	runtime·emptyfunc(SB)	// fault if stack check is wrong