// RX (12-bit displacement) and RXY (20-bit displacement) instruction encoding (e.g: ST vs STY)
	MOVW	R1, 4095(R2)(R3)       // 50132fff
	MOVW	R1, 4096(R2)(R3)       // e31320000150
	MOVWZ	R1, 4095(R2)(R3)       // 50132fff
	MOVWZ	R1, 4096(R2)(R3)       // e31320000150
	MOVH	R1, 4095(R2)(R3)       // 40132fff
	MOVHZ   R1, 4095(R2)(R3)       // 40132fff
	MOVH	R1, 4096(R2)(R3)       // e31320000170

	// ppc64x: "RLWNM\t[$]25, R[0-9]+, [$]27, [$]29, R[0-9]+"
	z[2] = uint32(uint8(a>>7)) & 0x1c
	// ppc64x: -"MOVWZ"
	y[0] = uint64((a >> 6) & 0x1c)
	// ppc64x: -"MOVWZ"
	y[1] = uint64(uint32(b)<<6) + 1
	// ppc64x: -"MOVHZ", -"MOVWZ"
	y[2] = uint64((uint16(a) >> 9) & 0x1F)
	// ppc64x: -"MOVHZ", -"MOVWZ", -"ANDCC"
	y[3] = uint64(((uint16(a) & 0xFF0) >> 9) & 0x1F)
}

// 128 bit shifts

	// ppc64le:`MOVWZ\s`,-`MOV[BH]Z\s`
	// ppc64:`MOVWBR\s`,-`MOV[BH]Z\s`
	return binary.LittleEndian.Uint32(b)
}

func load_le32_idx(b []byte, idx int) uint32 {
	// amd64:`MOVL\s\(.*\)\(.*\*1\),`,-`MOV[BW]`,-`OR`
	// 386:`MOVL\s\(.*\)\(.*\*1\),`,-`MOV[BW]`,-`OR`
	// s390x:`MOVWBR\s\(.*\)\(.*\*1\),`
	// arm64:`MOVWU\s\(R[0-9]+\)\(R[0-9]+\),`,-`MOV[BH]`
	// ppc64le:`MOVWZ\s`,-`MOV[BH]Z\s`

	MOVW (R3)(R4), R5               // 7ca41aaa
	MOVW (R3)(R0), R5               // 7ca01aaa
	MOVW (R3), R5                   // e8a30002
	MOVWZ 4(R3), R4                 // 80830004
	MOVWZ (R3)(R4), R5              // 7ca4182e
	MOVWZ (R3)(R0), R5              // 7ca0182e
	MOVWZ (R3), R5                  // 80a30000
	MOVH 4(R3), R4                  // a8830004
	MOVH (R3)(R4), R5               // 7ca41aae

	NOPH
	MOVWZ 0(R3), R3
	MOVD  R3, err+48(FP)
	MOVD  zosLibVec<>(SB), R8
	ADD   $(__err2ad), R8
	LMG   0(R8), R5, R6
	LE_CALL                   // balr R7, R6 __err2ad (return #2)
	NOPH
	MOVW  (R3), R2            // retrieve errno2
	MOVD  R2, errno2+40(FP)   // store in return area
	XOR   R2, R2
	MOVWZ R2, (R3)            // clear errno2

done:

	// set 1 to indicate end of msg.
	BNE  carry
	MOVD $1, R17
	BR   carry

less8:
	MOVD  $0, R16   // h0
	MOVD  $0, R22   // shift count
	CMP   R5, $4
	BLT   less4
	MOVWZ (R4), R16
	ADD   $4, R4
	ADD   $-4, R5
	MOVD  $32, R22

less4:
	CMP   R5, $2
	BLT   less2
	MOVHZ (R4), R21
	SLD   R22, R21, R21
	OR    R16, R21, R16
	ADD   $16, R22
	ADD   $-2, R5

	LHARX:     "LHAR",
	LBARX:     "LBAR",
	LWAX:      "MOVW",
	LHAX:      "MOVH",
	LWAUX:     "MOVWU",
	LHAU:      "MOVHU",
	LHAUX:     "MOVHU",
	LDX:       "MOVD",
	LDUX:      "MOVDU",
	LWZX:      "MOVWZ",
	LWZUX:     "MOVWZU",
	LHZX:      "MOVHZ",
	LHZUX:     "MOVHZU",
	LBZX:      "MOVBZ",
	LBZUX:     "MOVBZU",
	LDBRX:     "MOVDBR",
	LWBRX:     "MOVWBR",
	LHBRX:     "MOVHBR",
	MCRF:      "MOVFL",

	BLT	tail_4
	MOVD	(R3), R12
	CMPB	R12, R5, R12
	POPCNTD	R12, R12
	ADD	R12, R18, R18
	ADD	$8, R3, R3
	ADD	$-8, R4, R4

tail_4:	// Count the remaining 0 - 7 bytes.
	CMP	R4, $4
	BLT	tail_2
	MOVWZ	(R3), R12
	CMPB	R12, R5, R12
	SLD	$32, R12, R12	// Remove non-participating matches.
	POPCNTD	R12, R12
	ADD	R12, R18, R18
	ADD	$4, R3, R3
	ADD	$-4, R4, R4

tail_2:	// Count the remaining 0 - 3 bytes.

	"EXTSW",
	"EXTSWCC",
	"FCFID",
	"FCFIDCC",
	"FCFIDU",
	"FCFIDUCC",
	"FCFIDS",
	"FCFIDSCC",
	"FCTID",
	"FCTIDCC",
	"FCTIDZ",
	"FCTIDZCC",
	"LDAR",
	"MOVD",
	"MOVDU",
	"MOVWZ",
	"MOVWZU",
	"MULHD",
	"MULHDCC",
	"MULHDU",
	"MULHDUCC",
	"MULLD",
	"MULLDCC",
	"MULLDVCC",
	"MULLDV",
	"RFID",
	"RLDMI",
	"RLDMICC",
	"RLDIMI",
	"RLDIMICC",
	"RLDC",

	"MULLD",
	"MULHD",
	"MULHDU",
	"MLGR",
	"SUB",
	"SUBC",
	"SUBV",
	"SUBE",
	"SUBW",
	"NEG",
	"NEGW",
	"MOVWBR",
	"MOVB",
	"MOVBZ",
	"MOVH",
	"MOVHBR",
	"MOVHZ",
	"MOVW",
	"MOVWZ",
	"MOVD",
	"MOVDBR",
	"MOVDEQ",
	"MOVDGE",
	"MOVDGT",
	"MOVDLE",
	"MOVDLT",
	"MOVDNE",
	"LOCR",
	"LOCGR",
	"FLOGR",
	"POPCNT",
	"AND",
	"ANDW",
	"OR",
	"ORW",
	"XOR",