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
	XXLXOR	VS36, VS37, VS36
	XXLXOR	VS38, VS39, VS38
	STXVD2X	VS32, (R3)(R8)
	STXVD2X	VS34, (R3)(R10)
	STXVD2X	VS36, (R3)(R14)
	STXVD2X	VS38, (R3)(R15)
	ADD	$64, R8
	ADD	$64, R10
	ADD	$64, R14
	ADD	$64, R15

	// Clear the keys
	XXLXOR	VS0, VS0, VS0
	XXLXOR	VS1, VS1, VS1
	XXLXOR	VS2, VS2, VS2
	XXLXOR	VS3, VS3, VS3
	XXLXOR	VS4, VS4, VS4
	XXLXOR	VS5, VS5, VS5
	XXLXOR	VS6, VS6, VS6
	XXLXOR	VS7, VS7, VS7
	XXLXOR	VS8, VS8, VS8
	XXLXOR	VS9, VS9, VS9
	XXLXOR	VS10, VS10, VS10
	XXLXOR	VS11, VS11, VS11
	XXLXOR	VS12, VS12, VS12
	XXLXOR	VS13, VS13, VS13
	XXLXOR	VS14, VS14, VS14

	MOVW  R0, 0(R3)     // zero 4 bytes
	ADD   $4, R3        // bump ptr by 4
	ADD   $-4, R4
	BR    zero512xsetup // ptr should now be 8 byte aligned

under512:
	SRDCC $3, R6, R7  // 64 byte chunks?
	XXLXOR VS32, VS32, VS32 // clear VS32 (V0)
	BEQ   lt64gt8

	// Prepare to clear 64 bytes at a time.

zero64setup:
	DCBTST (R3)             // prepare data cache
	MOVD   R7, CTR          // number of 64 byte chunks

#endif

	CMPU	R4, $32		// Check if it's a small string (<32 bytes)
	BLT	tail		// Jump to the small string case
	SRD	$5, R4, R20
	MOVD	R20, CTR
	MOVD	$16, R21
	XXLXOR	V4, V4, V4
	XXLXOR	V5, V5, V5

	PCALIGN	$16
cmploop:
	LXVD2X	(R0)(R3), V0	// Count 32B per loop with two vector accumulators.
	LXVD2X	(R21)(R3), V2
	VCMPEQUB V2, V1, V2
	VCMPEQUB V0, V1, V0

	"MTVSRD",
	"MTFPRD",
	"MTVRD",
	"MTVSRWA",
	"MTVSRWZ",
	"MTVSRDD",
	"MTVSRWS",
	"XXLAND",
	"XXLANDC",
	"XXLEQV",
	"XXLNAND",
	"XXLOR",
	"XXLORC",
	"XXLNOR",
	"XXLORQ",
	"XXLXOR",
	"XXSEL",
	"XXMRGHW",
	"XXMRGLW",
	"XXSPLTW",
	"XXSPLTIB",
	"XXPERM",
	"XXPERMDI",
	"XXSLDWI",
	"XXBRQ",
	"XXBRD",
	"XXBRW",
	"XXBRH",
	"XSCVDPSP",
	"XSCVSPDP",

		// power9. The Short variation is used
		// if no loop is generated.

		// sizes >= 64 generate a loop as follows:

		// Set up loop counter in CTR, used by BC
		// XXLXOR clears VS32
		//       XXLXOR VS32,VS32,VS32
		//       MOVD len/64,REG_TMP
		//       MOVD REG_TMP,CTR
		//       loop:
		//       STXV VS32,0(R20)
		//       STXV VS32,16(R20)
		//       STXV VS32,32(R20)

		// arg0 = address of memory to zero (in R3, changed as side effect)
		// returns mem
		//
		// a loop is generated when there is more than one iteration
		// needed to clear 4 doublewords
		//
		//	XXLXOR	VS32,VS32,VS32
		// 	MOVD	$len/32,R31
		//	MOVD	R31,CTR
		//	MOVD	$16,R31
		//	loop:
		//	STXVD2X VS32,(R0)(R3)
		//	STXVD2X	VS32,(R31)(R3)
		//	ADD	R3,32
		//	BC	loop

	XXLNOR VS0, VS1, VS32           // f0000d11
	XXLOR VS1, VS2, VS3             // f0611490
	XXLORC VS1, VS2, VS3            // f0611550
	XXLORQ VS1, VS2, VS3            // f0611490
	XXLXOR VS1, VS2, VS3            // f06114d0
	XXSEL VS1, VS2, VS3, VS4        // f08110f0
	XXSEL VS33, VS34, VS35, VS36    // f08110ff
	XXSEL V1, V2, V3, V4            // f08110ff
	XXMRGHW VS1, VS2, VS3           // f0611090

	XVTSQRTDP:      "xvtsqrtdp",
	XVTSQRTSP:      "xvtsqrtsp",
	XXLAND:         "xxland",
	XXLANDC:        "xxlandc",
	XXLNOR:         "xxlnor",
	XXLOR:          "xxlor",
	XXLXOR:         "xxlxor",
	XXMRGHW:        "xxmrghw",
	XXMRGLW:        "xxmrglw",
	XXPERMDI:       "xxpermdi",
	XXSEL:          "xxsel",
	XXSLDWI:        "xxsldwi",
	XXSPLTW:        "xxspltw",
	CMPB:           "cmpb",

			o2 = loadl16(REGTMP, d)
			o3 = LOP_RRR(c.oprrr(p.As), uint32(p.To.Reg), REGTMP, uint32(r))
		}
		if p.From.Sym != nil {
			c.ctxt.Diag("%v is not supported", p)
		}

	case 24: /* lfd fA,float64(0) -> xxlxor xsA,xsaA,xsaA + fneg for -0 */
		o1 = AOP_XX3I(c.oprrr(AXXLXOR), uint32(p.To.Reg), uint32(p.To.Reg), uint32(p.To.Reg), uint32(0))
		// This is needed for -0.
		if o.size == 8 {