JMP	f_words_ua

f_tail_ua:
	BEQ	R1, R5, ret
	MOVWHI	-4(R4), R8
	MOVWLO	-1(R4), R8
	MOVWLO	R8, -1(R5)
	JMP	ret

f_small_copy:
	BEQ	R1, R5, ret
	ADDU	$1, R1
	MOVB	0(R2), R6
	ADDU	$1, R2
	MOVB	R6, -1(R1)
	JMP	f_small_copy

backward:
	SGTU	$4, R3, R6
	BNE	R6, b_small_copy

	AND	$3, R5, R6
	BEQ	R6, b_dest_aligned
	MOVWHI	-4(R4), R7
	SUBU	R6, R3
	MOVWLO	-1(R4), R7

	JMP	diff16
diff48:
	ADDQ	$32, SI
	ADDQ	$32, DI
	JMP	diff16
diff32:
	ADDQ	$16, SI
	ADDQ	$16, DI
	// AX = bit mask of differences
diff16:
	BSFQ	AX, BX	// index of first byte that differs
	XORQ	AX, AX
	MOVB	(SI)(BX*1), CX
	CMPB	CX, (DI)(BX*1)
	SETHI	AX
	LEAQ	-1(AX*2), AX	// convert 1/0 to +1/-1
	RET

	// 0 through 16 bytes left, alen>=8, blen>=8
_0through16:
	CMPQ	R8, $8
	JBE	_0through8
	MOVQ	(SI), AX

#include "textflag.h"

TEXT ·IndexByte(SB),NOSPLIT,$0-40
	MOVD	b_base+0(FP), R0
	MOVD	b_len+8(FP), R2
	MOVBU	c+24(FP), R1
	MOVD	$ret+32(FP), R8
	B	indexbytebody<>(SB)

TEXT ·IndexByteString(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	indexbytebody<>(SB)

// input:
//   R0: data
//   R1: byte to search
//   R2: data len

#include "funcdata.h"
#include "textflag.h"

// If !iscgo, this is a no-op.
//
// NOTE: mcall() assumes this clobbers only X31 (REG_TMP).
TEXT runtime·save_g(SB),NOSPLIT|NOFRAME,$0-0
#ifndef GOOS_openbsd
	MOVB	runtime·iscgo(SB), X31
	BEQZ	X31, nocgo
#endif
	MOV	g, runtime·tls_g(SB)
nocgo:
	RET

TEXT runtime·load_g(SB),NOSPLIT|NOFRAME,$0-0
	MOV	runtime·tls_g(SB), g
	RET

	// Finally, save the link register R14. This also puts the
	// arguments we pushed for cgocallback where they need to be,
	// starting at 4(R13).
	MOVW.W	R14, -4(R13)

	// Skip floating point registers if goarmsoftfp!=0.
	MOVB    runtime·goarmsoftfp(SB), R11
	CMP     $0, R11
	BNE     skipfpsave
	MOVD	F8, (13*4+8*1)(R13)
	MOVD	F9, (13*4+8*2)(R13)
	MOVD	F10, (13*4+8*3)(R13)
	MOVD	F11, (13*4+8*4)(R13)
	MOVD	F12, (13*4+8*5)(R13)

	JMP	·Store64(SB)

// void	·Or8(byte volatile*, byte);
TEXT ·Or8(SB), NOSPLIT, $0-9
	MOVQ	ptr+0(FP), AX
	MOVB	val+8(FP), BX
	LOCK
	ORB	BX, (AX)
	RET

// void	·And8(byte volatile*, byte);
TEXT ·And8(SB), NOSPLIT, $0-9
	MOVQ	ptr+0(FP), AX
	MOVB	val+8(FP), BX
	LOCK
	ANDB	BX, (AX)
	RET

// func Or(addr *uint32, v uint32)
TEXT ·Or(SB), NOSPLIT, $0-12
	MOVQ	ptr+0(FP), AX

	"LDPSW",
	"LDPW",
	"LDXP",
	"LDXPW",
	"LDXR",
	"LDXRB",
	"LDXRH",
	"LDXRW",
	"LSL",
	"LSLW",
	"LSR",
	"LSRW",
	"MADD",
	"MADDW",
	"MNEG",
	"MNEGW",
	"MOVB",
	"MOVBU",
	"MOVD",
	"MOVH",
	"MOVHU",
	"MOVK",
	"MOVKW",
	"MOVN",
	"MOVNW",
	"MOVP",
	"MOVPD",
	"MOVPQ",
	"MOVPS",
	"MOVPSW",
	"MOVPW",
	"MOVW",
	"MOVWU",
	"MOVZ",

#include "go_asm.h"
#include "go_tls.h"
#include "funcdata.h"
#include "textflag.h"

// If !iscgo, this is a no-op.
//
// NOTE: mcall() assumes this clobbers only R30 (REGTMP).
TEXT runtime·save_g(SB),NOSPLIT|NOFRAME,$0-0
	MOVB	runtime·iscgo(SB), R30
	BEQ	R30, nocgo

	MOVV	g, runtime·tls_g(SB)

nocgo:
	RET

TEXT runtime·load_g(SB),NOSPLIT|NOFRAME,$0-0
	MOVV	runtime·tls_g(SB), g
	RET

loop:
	MOVD	0(R2)(R5*1), R7
	MOVD	0(R3)(R5*1), R8
	XOR	R7, R8
	MOVD	R8, 0(R1)(R5*1)
	LAY	8(R5), R5
	SUB	$8, R4
	CMPBGE	R4, $8, loop
tail:
	CMPBEQ	R4, $0, done
	MOVB	0(R2)(R5*1), R7
	MOVB	0(R3)(R5*1), R8
	XOR	R7, R8
	MOVB	R8, 0(R1)(R5*1)
	LAY	1(R5), R5
	SUB	$1, R4
	BR	tail
done:
	RET

// func cryptBlocksGCM(fn code, key, dst, src, buf []byte, cnt *[16]byte)
TEXT ·cryptBlocksGCM(SB),NOSPLIT,$0-112

// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

// Issue 4396. Arrays of bytes are not required to be
// word aligned. 5g should use MOVB to load the address
// of s.g[0] for its nil check.
//
// This test _may_ fail on arm, but requires the host to 
// trap unaligned loads. This is generally done with
//
// echo "4" > /proc/cpu/alignment