var sink *uint64

func main() {
	var a1, a2 A
	var b1, b2, b3, b4 B
	var x1, x2, x3, x4 uint64 // ERROR "moved to heap: x1" "moved to heap: x3"
	b1.b[0] = &x1
	b2.b[0] = &x2
	b3.b[0] = &x3
	b4.b[0] = &x4
	f(a1)
	g(&a2)
	sink = h(&b1)
	h(&b2)
	sink = h2(&b1)
	h2(&b4)
	x1 = 17
	println("*sink=", *sink) // Verify that sink addresses x1
	x3 = 42
	sink = k(b3)
	println("*sink=", *sink) // Verify that sink addresses x3

move_17through32:
	MOVOU	(SI), X0
	MOVOU	-16(SI)(BX*1), X1
	MOVOU	X0, (DI)
	MOVOU	X1, -16(DI)(BX*1)
	RET
move_33through64:
	MOVOU	(SI), X0
	MOVOU	16(SI), X1
	MOVOU	-32(SI)(BX*1), X2
	MOVOU	-16(SI)(BX*1), X3
	MOVOU	X0, (DI)
	MOVOU	X1, 16(DI)
	MOVOU	X2, -32(DI)(BX*1)
	MOVOU	X3, -16(DI)(BX*1)
	RET
move_65through128:
	MOVOU	(SI), X0
	MOVOU	16(SI), X1
	MOVOU	32(SI), X2
	MOVOU	48(SI), X3

	CMPQ    AX, BX
	JLT     isMaxNaN
	MOVQ    R9, CX
	ANDQ    DX, CX // y = |y|
	CMPQ    AX, CX
	JLT     isMaxNaN
	// ±0 special cases
	ORQ     CX, BX
	JEQ     isMaxZero

	MOVQ    R8, X0
	MOVQ    R9, X1
	MAXSD   X1, X0
	MOVSD   X0, ret+16(FP)
	RET
isMaxNaN: // return NaN
	MOVQ	$NaN, AX
isPosInf: // return +Inf
	MOVQ    AX, ret+16(FP)
	RET
isMaxZero:
	MOVQ    $(1<<63), AX // -0.0
	CMPQ    AX, R8

	VPMADD52HUQ X7, X11, K1, X1                        // 62f2a509b5cf
	VPMADD52HUQ X0, X11, K1, X1                        // 62f2a509b5c8
	VPMADD52HUQ 17(SP)(BP*2), X11, K1, X1              // 62f2a509b58c6c11000000
	VPMADD52HUQ -7(DI)(R8*4), X11, K1, X1              // 62b2a509b58c87f9ffffff
	VPMADD52HUQ X7, X31, K1, X1                        // 62f28501b5cf
	VPMADD52HUQ X0, X31, K1, X1                        // 62f28501b5c8

	JMP	main(SB)

// When building with -buildmode=c-shared, this symbol is called when the shared
// library is loaded.
TEXT _rt0_riscv64_freebsd_lib(SB),NOSPLIT,$224
	// Preserve callee-save registers, along with X1 (LR).
	MOV	X1, (8*3)(X2)
	MOV	X8, (8*4)(X2)
	MOV	X9, (8*5)(X2)
	MOV	X18, (8*6)(X2)
	MOV	X19, (8*7)(X2)
	MOV	X20, (8*8)(X2)
	MOV	X21, (8*9)(X2)
	MOV	X22, (8*10)(X2)
	MOV	X23, (8*11)(X2)
	MOV	X24, (8*12)(X2)

	MOVD $(NUM_ROUNDS/2), R1

loop:
	ROUND4(X0, X4, X12,  X8, X1, X5, X13,  X9, X2, X6, X14, X10, X3, X7, X15, X11)
	ROUND4(X0, X5, X15, X10, X1, X6, X12, X11, X2, X7, X13, X8,  X3, X4, X14, X9)

	ADD $-1, R1
	BNE loop

	// decrement length
	ADD $-256, R4

	// rearrange vectors
	SHUFFLE(X0, X1, X2, X3, M0, M1, M2, M3)
	ADDV(J0, X0, X1, X2, X3)
	SHUFFLE(X4, X5, X6, X7, M0, M1, M2, M3)

	V4FNMADDPS 99(R15)(R15*1), [Z21-Z24], K3, Z12      // 62125743aaa43f63000000
	V4FNMADDPS (DX), [Z21-Z24], K3, Z12                // 62725743aa22
	V4FNMADDSS -17(BP)(SI*8), [X1-X4], K4, X11         // 6272770cab9cf5efffffff or 6272772cab9cf5efffffff or 6272774cab9cf5efffffff
	V4FNMADDSS (R15), [X1-X4], K4, X11                 // 6252770cab1f or 6252772cab1f or 6252774cab1f

func Mul64(x, y uint64) (hi, lo uint64) {
	const mask32 = 1<<32 - 1
	x0 := x & mask32
	x1 := x >> 32
	y0 := y & mask32
	y1 := y >> 32
	w0 := x0 * y0
	t := x1*y0 + w0>>32
	w1 := t & mask32
	w2 := t >> 32
	w1 += x0 * y1
	hi = x1*y1 + w2 + w1>>32
	lo = x * y
	return
}

// Add64 returns the sum with carry of x, y and carry: sum = x + y + carry.

// Copyright 2019 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.

package main

import "./embed0"

type X1 struct{}

func (X1) Foo() {}

type X2 struct{}

func (X2) foo() {}

type X3 struct{}

func (X3) foo(int) {}

type X4 struct{ p.M1 }

type X5 struct{ p.M1 }

func (X5) foo(int) {}

		for x := x0; x != x1; x, yi, ci = x+4, yi+1, ci+1 {
	`,
	"422": `
		ciBase := (sy-src.Rect.Min.Y)*src.CStride - src.Rect.Min.X/2
		for x, sx := x0, sp.X; x != x1; x, sx, yi = x+4, sx+1, yi+1 {
			ci := ciBase + sx/2
	`,
	"420": `
		ciBase := (sy/2-src.Rect.Min.Y/2)*src.CStride - src.Rect.Min.X/2
		for x, sx := x0, sp.X; x != x1; x, sx, yi = x+4, sx+1, yi+1 {
			ci := ciBase + sx/2
	`,