AND	$7, X10, X7
	AND	$7, X12, X8
	BNE	X7, X8, check8_unaligned
	BEQZ	X7, compare32

	// Check one byte at a time until we reach 8 byte alignment.
	SUB	X7, X0, X7
	ADD	$8, X7, X7
	SUB	X7, X5, X5
align:
	SUB	$1, X7
	MOVBU	0(X10), X8
	MOVBU	0(X12), X9
	BNE	X8, X9, cmp
	ADD	$1, X10
	ADD	$1, X12
	BNEZ	X7, align

check32:
	// X6 contains $32

	VRCP28SS X26, X2, K3, X8                           // 62126d0bcbc2 or 62126d2bcbc2 or 62126d4bcbc2
	VRCP28SS X19, X2, K3, X8                           // 62326d0bcbc3 or 62326d2bcbc3 or 62326d4bcbc3
	VRCP28SS X0, X2, K3, X8                            // 62726d0bcbc0 or 62726d2bcbc0 or 62726d4bcbc0

	MOVBU	((index*8)+1)(X29), X6; \
	MOVBU	((index*8)+2)(X29), X7; \
	MOVBU	((index*8)+3)(X29), X8; \
	SLL	$56, X5; \
	SLL	$48, X6; \
	OR	X5, X6, X5; \
	SLL	$40, X7; \
	OR	X5, X7, X5; \
	SLL	$32, X8; \
	OR	X5, X8, X5; \
	MOVBU	((index*8)+4)(X29), X9; \
	MOVBU	((index*8)+5)(X29), X6; \
	MOVBU	((index*8)+6)(X29), X7; \
	MOVBU	((index*8)+7)(X29), X8; \
	SLL	$24, X9; \
	OR	X5, X9, X5; \
	SLL	$16, X6; \
	OR	X5, X6, X5; \

		x1 := s[4] << 11
		x2 := s[6]
		x3 := s[2]
		x4 := s[1]
		x5 := s[7]
		x6 := s[5]
		x7 := s[3]

		// Stage 1.
		x8 := w7 * (x4 + x5)
		x4 = x8 + w1mw7*x4
		x5 = x8 - w1pw7*x5
		x8 = w3 * (x6 + x7)
		x6 = x8 - w3mw5*x6
		x7 = x8 - w3pw5*x7

		// Stage 2.
		x8 = x0 + x1
		x0 -= x1
		x1 = w6 * (x3 + x2)
		x2 = x1 - w2pw6*x2
		x3 = x1 + w2mw6*x3
		x1 = x4 + x6
		x4 -= x6

	VPSHUFBITQMB X24, X8, K6, K0                       // 62923d0e8fc0
	VPSHUFBITQMB X7, X8, K6, K0                        // 62f23d0e8fc7
	VPSHUFBITQMB X0, X8, K6, K0                        // 62f23d0e8fc0
	VPSHUFBITQMB (R8), X8, K6, K0                      // 62d23d0e8f00
	VPSHUFBITQMB 15(DX)(BX*2), X8, K6, K0              // 62f23d0e8f845a0f000000

	c <- a(2)
	c <- b(3)
	c <- a(4)
	c <- b(5)
}

func h() {
	var x8 byte = 100
	var x64 int64 = 200

	c <- int(x8)
	c <- int(x64)
	f := func(z int) {
		g := func() {
			c <- int(x8)
			c <- int(x64)
			c <- z
		}
		g()
		c <- int(x8)
		c <- int(x64)
		c <- int(z)
	}
	x8 = 101
	x64 = 201
	f(500)
}

	}

	type X2 struct {
		y byte
		z A2
	}
	var x2 X2
	cmpA2(&x2.z, &A2{})
	type X4 struct {
		y byte
		z A4
	}
	var x4 X4
	cmpA4(&x4.z, &A4{})
	type X8 struct {
		y byte
		z A8
	}
	var x8 X8
	cmpA8(&x8.z, &A8{})

	VPCONFLICTQ X11, K4, X8                            // 6252fd0cc4c3
	VPCONFLICTQ X16, K4, X8                            // 6232fd0cc4c0
	VPCONFLICTQ X6, K4, X8                             // 6272fd0cc4c6
	VPCONFLICTQ 15(R8)(R14*8), K4, X8                  // 6212fd0cc484f00f000000
	VPCONFLICTQ -15(R14)(R15*2), K4, X8                // 6212fd0cc4847ef1ffffff

// Test that inlining of math/bits.RotateLeft* treats those calls as intrinsics.

package p

import "math/bits"

var (
	x8  uint8
	x16 uint16
	x32 uint32
	x64 uint64
	x   uint
)

func f() { // ERROR "can inline f"
	x8 = bits.RotateLeft8(x8, 1)
	x16 = bits.RotateLeft16(x16, 1)
	x32 = bits.RotateLeft32(x32, 1)
	x64 = bits.RotateLeft64(x64, 1)
	x = bits.RotateLeft(x, 1)

	VMOVHPD 7(DX), X8, X8            // c539164207 or 6271bd28168207000000 or 6271bd48168207000000
	VMOVHPD -15(R11)(CX*1), X8, X8   // c4413916440bf1 or 6251bd2816840bf1ffffff or 6251bd4816840bf1ffffff
	VMOVHPD (SP)(AX*2), X8, X8       // c539160444 or 6271bd28160444 or 6271bd48160444
	VMOVHPD (AX), X20, X8            // 6271dd001600 or 6271dd201600 or 6271dd401600