VMOV V18.B[10], V27                                         // 5b06155e
	VDUP V18.B[10], V27                                         // 5b06155e
	VMOV V12.B[2], V28.B[12]                                    // 9c15196e
	VMOV R30, V4.B[13]                                          // c41f1b4e
	VMOV V2.B16, V4.B16                                         // 441ca24e

	VCMEQ	V1.D2, V5.D2, V9.D2
	VCMEQ	V2.D2, V6.D2, V10.D2
	VCMEQ	V3.D2, V7.D2, V11.D2
	VAND	V8.B16, V9.B16, V8.B16
	VAND	V8.B16, V10.B16, V8.B16
	VAND	V8.B16, V11.B16, V8.B16
	CMP	R0, R6
	VMOV	V8.D[0], R4
	VMOV	V8.D[1], R5
	CBZ	R4, not_equal
	CBZ	R5, not_equal
	BNE	chunk64_loop
	AND	$0x3f, R2, R2
	CBZ	R2, equal
chunk16:
	// work with 16-byte chunks
	BIC	$0xf, R2, R3
	CBZ	R3, tail

	"VDUP",
	"VEOR",
	"VEOR3",
	"VEXT",
	"VFMLA",
	"VFMLS",
	"VLD1",
	"VLD1R",
	"VLD2",
	"VLD2R",
	"VLD3",
	"VLD3R",
	"VLD4",
	"VLD4R",
	"VMOV",
	"VMOVD",
	"VMOVI",
	"VMOVQ",
	"VMOVS",
	"VORR",
	"VPMULL",
	"VPMULL2",
	"VRAX1",
	"VRBIT",
	"VREV16",
	"VREV32",
	"VREV64",
	"VSHL",
	"VSLI",
	"VSRI",
	"VST1",
	"VST2",
	"VST3",

	{VCMP_E_EQ_F64, []int{1, 0}, "VCMP.E", "CMPD"},
	{VLDR_EQ, []int{1}, "VLDR", "MOV"},
	{VSTR_EQ, []int{1}, "VSTR", "MOV"},
	{VMOV_EQ_F32, []int{1, 0}, "VMOV", "MOVF"},
	{VMOV_EQ_F64, []int{1, 0}, "VMOV", "MOVD"},
	{VMOV_EQ_32, []int{1, 0}, "VMOV", "MOVW"},
	{VMOV_EQ, []int{1, 0}, "VMOV", "MOVW"},
	{VCVT_EQ_F64_F32, []int{1, 0}, "VCVT", "MOVFD"},
	{VCVT_EQ_F32_F64, []int{1, 0}, "VCVT", "MOVDF"},

	  RET

7. Move large constants to vector registers.

Go asm uses VMOVQ/VMOVD/VMOVS to move 128-bit, 64-bit and 32-bit constants into vector registers, respectively.
And for a 128-bit integer, it take two 64-bit operands, for the low and high parts separately.

Examples:

	VMOVS $0x11223344, V0
	VMOVD $0x1122334455667788, V1
	VMOVQ $0x1122334455667788, $0x99aabbccddeeff00, V2   // V2=0x99aabbccddeeff001122334455667788

	rounds12to59	(m0,4,m3,m1,sha256msg1,vmov)    // 16-19
	rounds12to59	(m1,5,m0,m2,sha256msg1,vmov)    // 20-23
	rounds12to59	(m2,6,m1,m3,sha256msg1,vmov)    // 24-27
	rounds12to59	(m3,7,m2,m0,sha256msg1,vmov)    // 28-31
	rounds12to59	(m0,8,m3,m1,sha256msg1,vmov)    // 32-35
	rounds12to59	(m1,9,m0,m2,sha256msg1,vmov)    // 36-39
	rounds12to59	(m2,10,m1,m3,sha256msg1,vmov)   // 40-43

	FMOVD	F4, (R2)(R6<<3)                 // 447826fc

// vmov
	VMOV	V8.S[1], R1           // 013d0c0e
	VMOV	V0.D[0], R11          // 0b3c084e
	VMOV	V0.D[1], R11          // 0b3c184e
	VMOV	R20, V1.S[0]          // 811e044e
	VMOV	R20, V1.S[1]          // 811e0c4e
	VMOV	R1, V9.H4             // 290c020e
	VDUP	R1, V9.H4             // 290c020e
	VMOV	R22, V11.D2           // cb0e084e
	VDUP	R22, V11.D2           // cb0e084e

	CBZ	R10, noaes
	MOVD	$runtime·aeskeysched+0(SB), R3

	VEOR	V0.B16, V0.B16, V0.B16
	VLD1	(R3), [V2.B16]
	VLD1	(R0), V0.S[1]
	VMOV	R1, V0.S[0]

	AESE	V2.B16, V0.B16
	AESMC	V0.B16, V0.B16
	AESE	V2.B16, V0.B16
	AESMC	V0.B16, V0.B16
	AESE	V2.B16, V0.B16

	VMOV	V0.D[0], R0
	RET
noaes:
	B	runtime·memhash32Fallback<ABIInternal>(SB)

// func memhash64(p unsafe.Pointer, h uintptr) uintptr

		o1 |= (uint32(p.To.Reg) & 15) << 12
		o1 |= ((uint32(p.Scond) & C_SCOND) ^ C_SCOND_XOR) << 28

	case 80: /* fmov zfcon,freg */
		if p.As == AMOVD {
			o1 = 0xeeb00b00 // VMOV imm 64
			o2 = c.oprrr(p, ASUBD, int(p.Scond))
		} else {
			o1 = 0x0eb00a00 // VMOV imm 32
			o2 = c.oprrr(p, ASUBF, int(p.Scond))
		}

		v := int32(0x70) // 1.0
		r := (int(p.To.Reg) & 15) << 0

	{AVMOV, C_ELEM, C_NONE, C_NONE, C_ZREG, C_NONE, 73, 4, 0, 0, 0},
	{AVMOV, C_ELEM, C_NONE, C_NONE, C_ELEM, C_NONE, 92, 4, 0, 0, 0},
	{AVMOV, C_ELEM, C_NONE, C_NONE, C_VREG, C_NONE, 80, 4, 0, 0, 0},
	{AVMOV, C_ZREG, C_NONE, C_NONE, C_ARNG, C_NONE, 82, 4, 0, 0, 0},
	{AVMOV, C_ZREG, C_NONE, C_NONE, C_ELEM, C_NONE, 78, 4, 0, 0, 0},
	{AVMOV, C_ARNG, C_NONE, C_NONE, C_ARNG, C_NONE, 83, 4, 0, 0, 0},