// amd64:"NEGL","ADCQ","SBBQ","NEGQ"
	// loong64: "ADDV", "SGTU"
	// ppc64x: "ADDC", "ADDE", "ADDZE"
	// s390x:"ADDE","ADDC\t[$]-1,"
	// mips64:"ADDV","SGTU"
	// riscv64: "ADD","SLTU"
	return bits.Add(x, 7, ci)
}

func AddZ(x, y uint) (r, co uint) {
	// arm64:"ADDS","ADC",-"ADCS",-"ADD\t",-"CMP"
	// amd64:"ADDQ","SBBQ","NEGQ",-"NEGL",-"ADCQ"
	// loong64: "ADDV", "SGTU"
	// ppc64x: "ADDC", -"ADDE", "ADDZE"

TEXT runtime·memmove<ABIInternal>(SB), NOSPLIT|NOFRAME, $0-24
	BNE	R6, check
	RET

check:
	SGTU	R4, R5, R7
	BNE	R7, backward

	ADDV	R4, R6, R9 // end pointer

	// if the two pointers are not of same alignments, do byte copying
	SUBVU	R5, R4, R7
	AND	$7, R7
	BNE	R7, out

	// if less than 8 bytes, do byte copying
	SGTU	$8, R6, R7
	BNE	R7, out

	// do one byte at a time until 8-aligned
	AND	$7, R4, R8

byte_loop:
	BEQ	R4, R12, samebytes
	MOVBU	(R4), R8
	ADDVU	$1, R4
	MOVBU	(R6), R9
	ADDVU	$1, R6
	BEQ	R8, R9, byte_loop

byte_cmp:
	SGTU	R8, R9, R4 // R12 = 1 if (R8 > R9)
	BNE	R0, R4, ret
	MOVV	$-1, R4
	JMP	ret

samebytes:
	SGTU	R5, R7, R8
	SGTU	R7, R5, R9
	SUBV	R9, R8, R4

ret:

	ADDV	R4, R5, R6

	// if less than 8 bytes, do one byte at a time
	SGTU	$8, R5, R8
	BNE	R8, out

	// do one byte at a time until 8-aligned
	AND	$7, R4, R8
	BEQ	R8, words
	MOVB	R0, (R4)
	ADDV	$1, R4
	JMP	-4(PC)

words:
	// do 8 bytes at a time if there is room
	ADDV	$-7, R6, R5

	PCALIGN	$16
	SGTU	R5, R4, R8
	BEQ	R8, out
	MOVV	R0, (R4)
	ADDV	$8, R4
	JMP	-4(PC)

out:

	"MULD",
	"MULF",
	"MULU",
	"MULH",
	"MULHU",
	"MULW",
	"NEGD",
	"NEGF",
	"NEGW",
	"NEGV",
	"NOOP",
	"NOR",
	"OR",
	"REM",
	"REMU",
	"RFE",
	"SC",
	"SCV",
	"SGT",
	"SGTU",
	"SLL",
	"SQRTD",
	"SQRTF",
	"SRA",
	"SRL",
	"ROTR",
	"SUB",
	"SUBD",
	"SUBF",
	"SUBU",
	"SUBW",
	"DBAR",
	"SYSCALL",
	"TEQ",
	"TNE",
	"WORD",
	"XOR",
	"MASKEQZ",

// we don't have variable-sized frames, so we use a small number
// of constant-sized-frame functions to encode a few bits of size in the pc.

#define DISPATCH(NAME,MAXSIZE)	\
	MOVW	$MAXSIZE, R23;	\
	SGTU	R1, R23, R23;	\
	BNE	R23, 3(PC);	\
	MOVW	$NAME(SB), R4;	\
	JMP	(R4)

TEXT ·reflectcall(SB),NOSPLIT|NOFRAME,$0-28
	MOVW	frameSize+20(FP), R1

	DISPATCH(runtime·call16, 16)
	DISPATCH(runtime·call32, 32)

// of constant-sized-frame functions to encode a few bits of size in the pc.
// Caution: ugly multiline assembly macros in your future!

#define DISPATCH(NAME,MAXSIZE)		\
	MOVV	$MAXSIZE, R30;		\
	SGTU	R19, R30, R30;		\
	BNE	R30, 3(PC);			\
	MOVV	$NAME(SB), R4;	\
	JMP	(R4)
// Note: can't just "BR NAME(SB)" - bad inlining results.

TEXT ·reflectcall(SB), NOSPLIT|NOFRAME, $0-48
	MOVWU frameSize+32(FP), R19

OPmyßt}æ^IªêVkq ^pu²vñ^Ty[Sxi8¿l ’Œ€à@DK~Ô^zÜJ ¨™1|vž}û\–iY»ÕmP\ep—÷SlûY¡\ïTmG_s™Ùs`H{_}©GEì´Œ»SqòSa¸ïL}ÂbDR^Ì ÕO„V^e¼Bñjhñqkxé°{W_÷E@JˆVk<‚Åò y¦Õ^kh~ ô[Rg¢?qr› †ÕsHrïpŠQR¾xPe¨¹yndñ E¾Ri]Ü ª m|jqme¨àr 2Jdª}gVPªzæi¾LK^j¼VU VdC²HHg@e¼Yp8ì[Negz{Öa€uV¢Qrûcf¼csM¼SGTܤ%ûX; „^y²IRñre¸fO^ š -oCˆuMIV„åjypge†À¬õ Èû8}ûz}ô¼o~ t{_qLOrs|ôŠLurmmNmX˾@{ jsò*ñCixW{V sB ^øG²zjïûZk{2OlqhJ¦GÈ}CßTi@fa¨ª ñ~ EFL€B\G paxz¸D{¸ÜRCR„ns\Iå|i\I†Vñ¬B`w¾{ËKvK÷„jf²* 6 hj†^ÐJrh_`ÌkqTçå&»œé&Õ%fzwH„ hôDHpÔúè¬õ a!` Cþì%tvWaSeò ¾]Vã p_„sGq\ ...

			},
			outputs: []outputInfo{
				{0, 1071644664}, // R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 R23 R24 R25 R26 R27 R28 R29 R31
			},
		},
	},
	{
		name:   "SGTU",
		argLen: 2,
		asm:    loong64.ASGTU,
		reg: regInfo{
			inputs: []inputInfo{
				{0, 1073741816}, // R4 R5 R6 R7 R8 R9 R10 R11 R12 R13 R14 R15 R16 R17 R18 R19 R20 R21 g R23 R24 R25 R26 R27 R28 R29 R31