VPGATHERDQ Y0, 664(X4*8), Y6            // c4e2fd9034e598020000
	VPGATHERDQ Y0, 8(X14*1), Y6             // c4a2fd90343508000000
	VPGATHERDQ Y0, -8(X14*1), Y6            // c4a2fd903435f8ffffff
	VPGATHERDQ Y0, 0(X14*1), Y6             // c4a2fd90343500000000
	VPGATHERDQ Y0, 664(X14*1), Y6           // c4a2fd90343598020000
	VPGATHERDQ Y0, 8(X14*8), Y6             // c4a2fd9034f508000000
	VPGATHERDQ Y0, -8(X14*8), Y6            // c4a2fd9034f5f8ffffff

		y3 = (y1 + w2mw6*y3) >> 3
		y1 = y4 + y6
		y4 -= y6
		y6 = y5 + y7
		y5 -= y7

		// Stage 3.
		y7 = y8 + y3
		y8 -= y3
		y3 = y0 + y2
		y0 -= y2
		y2 = (r2*(y4+y5) + 128) >> 8
		y4 = (r2*(y4-y5) + 128) >> 8

		// Stage 4.
		s[8*0] = (y7 + y1) >> 14
		s[8*1] = (y3 + y2) >> 14
		s[8*2] = (y0 + y4) >> 14
		s[8*3] = (y8 + y6) >> 14
		s[8*4] = (y8 - y6) >> 14
		s[8*5] = (y0 - y4) >> 14

	VPSHLDQ $121, Y6, Y6, K5, Y6                       // 62f3cd2d71f679
	VPSHLDQ $121, (BX), Y6, K5, Y6                     // 62f3cd2d713379
	VPSHLDQ $121, -17(BP)(SI*1), Y6, K5, Y6            // 62f3cd2d71b435efffffff79
	VPSHLDQ $121, Y19, Y11, K5, Y6                     // 62b3a52d71f379
	VPSHLDQ $121, Y7, Y11, K5, Y6                      // 62f3a52d71f779

	VPDPBUSDS Y23, Y3, K2, Y6                          // 62b2652a51f7
	VPDPBUSDS (BX), Y3, K2, Y6                         // 62f2652a5133
	VPDPBUSDS -17(BP)(SI*1), Y3, K2, Y6                // 62f2652a51b435efffffff
	VPDPBUSDS Y28, Y8, K2, Y6                          // 62923d2a51f4
	VPDPBUSDS Y1, Y8, K2, Y6                           // 62f23d2a51f1

var y5 = x5

type T5 struct{}

func (T5) m() int {
	_ = y5
	return 0
}

// issue 4847
// simplified test case

var x6 = f6
var y6 /* ERROR "initialization cycle" */ = f6
func f6() { _ = y6 }

// full test case

type (
      E int
      S int
)

type matcher func(s *S) E

func matchList(s *S) E { return matcher(matchAnyFn)(s) }

	ORQ  R12, DI
	ADDQ R14, R9
	ADDQ R15, BX
	ADDQ R15, R9

	ADDQ DI, R9

	VPSRLQ $19, Y6, Y3
	VPSLLQ $(64-19), Y6, Y1
	VPOR   Y1, Y3, Y3
	VPXOR  Y3, Y8, Y8
	VPSRLQ $61, Y6, Y3
	VPSLLQ $(64-61), Y6, Y1
	VPOR   Y1, Y3, Y3
	VPXOR  Y3, Y8, Y8

	VPADDQ Y8, Y0, Y2

	VPBLENDD $0xF0, Y2, Y6, Y6

	MOVQ  R9, DI
	RORXQ $41, BX, R13
	RORXQ $18, BX, R14
	ADDQ  3*8+frame_YFER(SP), DX

	VORPS Y12, Y6, K7, Y6                              // 62d14c2f56f4
	VORPS -7(CX)(DX*1), Y6, K7, Y6                     // 62f14c2f56b411f9ffffff
	VORPS -15(R14)(R15*4), Y6, K7, Y6                  // 62914c2f56b4bef1ffffff
	VORPS Y11, Y26, K7, Y6                             // 62d12c2756f3
	VORPS Y26, Y26, K7, Y6                             // 62912c2756f2

L6:
	MOVW.P	4(R2), R6
	ORR	R6<<R4, R7
	MOVW.P	R7, 4(R1)
	MOVW	R6>>R3, R7
E6:
	TEQ	R1, R5
	BNE	L6

	MOVW	R7, 0(R1)
	RET

Y6:	// copy loop, because shift 0 == shift 32
	MOVW.P	4(R2), R6
	MOVW.P	R6, 4(R1)
	TEQ R1, R5
	BNE Y6

X6:
	MOVW	$0, R1
	MOVW	R1, c+28(FP)
	RET


// func mulAddVWW(z, x []Word, y, r Word) (c Word)
TEXT ·mulAddVWW(SB),NOSPLIT,$0
	MOVW	$0, R0

	VPGATHERDQ Y2, (BP)(AX*2), Y1   // ERROR "invalid instruction"
	VPGATHERDQ Y5, 664(DX*8), Y6    // ERROR "invalid instruction"
	VPGATHERDQ Y5, (DX), Y0         // ERROR "invalid instruction"
	// VM/X rejects Y index register.
	VPGATHERDQ Y5, 664(Y14*8), Y6   // ERROR "invalid instruction"
	VPGATHERQQ X2, (BP)(Y7*2), X1   // ERROR "invalid instruction"
	// VM/Y rejects X index register.

	PCALIGN $16
hugeloop_avx2:
	CMPQ	BX, $64
	JB	bigloop_avx2
	VMOVDQU	(SI), Y0
	VMOVDQU	(DI), Y1
	VMOVDQU	32(SI), Y2
	VMOVDQU	32(DI), Y3
	VPCMPEQB	Y1, Y0, Y4
	VPCMPEQB	Y2, Y3, Y5
	VPAND	Y4, Y5, Y6
	VPMOVMSKB Y6, DX
	ADDQ	$64, SI
	ADDQ	$64, DI
	SUBQ	$64, BX
	CMPL	DX, $0xffffffff
	JEQ	hugeloop_avx2
	VZEROUPPER
	XORQ	AX, AX	// return 0
	RET

bigloop_avx2:
	VZEROUPPER