RET

// func rawSyscallNoError(trap, a1, a2, a3 uintptr) (r1, r2 uintptr)
TEXT ·rawSyscallNoError(SB),NOSPLIT,$0-48
	MOVQ	a1+8(FP), DI
	MOVQ	a2+16(FP), SI
	MOVQ	a3+24(FP), DX
	MOVQ	trap+0(FP), AX	// syscall entry
	SYSCALL
	MOVQ	AX, r1+32(FP)
	MOVQ	DX, r2+40(FP)
	RET

// func gettimeofday(tv *Timeval) (err uintptr)
TEXT ·gettimeofday(SB),NOSPLIT,$0-16
	MOVQ	tv+0(FP), DI
	MOVQ	$0, SI

#define INVOKE_SYSCALL	INT	$0x80

// func Syscall6(num, a1, a2, a3, a4, a5, a6 uintptr) (r1, r2, errno uintptr)
//
// Syscall # in AX, args in BX CX DX SI DI BP, return in AX
TEXT ·Syscall6(SB),NOSPLIT,$0-40
	MOVL	num+0(FP), AX	// syscall entry
	MOVL	a1+4(FP), BX
	MOVL	a2+8(FP), CX
	MOVL	a3+12(FP), DX
	MOVL	a4+16(FP), SI
	MOVL	a5+20(FP), DI
	MOVL	a6+24(FP), BP
	INVOKE_SYSCALL
	CMPL	AX, $0xfffff001
	JLS	ok
	MOVL	$-1, r1+28(FP)

      // Construct backward graph.
      auto dz = Const(scope, {{1.0, 1.0}, {1.0, 1.0}});
      auto dx = MatMul(scope, dz, y, MatMul::TransposeB(true));
      auto dy = MatMul(scope, x, dz, MatMul::TransposeA(true));

      auto du = MatMul(scope, dx, v, MatMul::TransposeB(true));
      auto dv = MatMul(scope, u, dx, MatMul::TransposeA(true));
    } else {
      // Call AddSymbolicGradients.

	PUSHAL				// save registers
	PUSHL	BP
	CALL	lwp_setprivate<>(SB)
	POPL	AX
	POPAL

	// Now segment is established. Initialize m, g.
	get_tls(AX)
	MOVL	DX, g(AX)
	MOVL	BX, g_m(DX)

	CALL	runtime·stackcheck(SB)	// smashes AX, CX
	MOVL	0(DX), DX		// paranoia; check they are not nil
	MOVL	0(BX), BX

	// more paranoia; check that stack splitting code works
	PUSHAL
	CALL	runtime·emptyfunc(SB)
	POPAL

		lines(
			"#define LOAD(off, reg) \\",
			"\tMOVBLZX	(off*4)(R12),	reg \\",
			"\tADDB	reg,		DX",
			"",
			"LOAD(8, AX)",
		),
		"\n.\n.MOVBLZX.(.8.*.4.).(.R12.).,.AX.\n.ADDB.AX.,.DX.\n",
	},
	{
		"nested multiline macro",
		lines(
			"#define KEYROUND(xmm, load, off, r1, r2, index) \\",
			"\tMOVBLZX	(BP)(DX*4),	R8 \\",
			"\tload((off+1), r2) \\",
			"\tMOVB	R8,		(off*4)(R12) \\",

// of the args of the ANDQ needs to be saved so it can be used as the arg to TESTQ.
func andWithUse(x, y int) int {
	z := x & y
	// amd64:`TESTQ\s(AX, AX|BX, BX|CX, CX|DX, DX|SI, SI|DI, DI|R8, R8|R9, R9|R10, R10|R11, R11|R12, R12|R13, R13|R15, R15)`
	if z == 0 {
		return 77
	}
	// use z by returning it
	return z
}

// Verify (OR x (NOT y)) rewrites to (ORN x y) where supported

	MOVL eaxArg+0(FP), AX
	MOVL ecxArg+4(FP), CX
	CPUID
	MOVL AX, eax+8(FP)
	MOVL BX, ebx+12(FP)
	MOVL CX, ecx+16(FP)
	MOVL DX, edx+20(FP)
	RET

// func xgetbv() (eax, edx uint32)
TEXT ·xgetbv(SB),NOSPLIT,$0-8
	MOVL $0, CX
	XGETBV
	MOVL AX, eax+0(FP)
	MOVL DX, edx+4(FP)

TEXT	·Syscall9(SB),NOSPLIT,$32-104
	NO_LOCAL_POINTERS
	CALL	runtime·entersyscall<ABIInternal>(SB)
	MOVQ	num+0(FP), AX	// syscall entry
	MOVQ	a1+8(FP), DI
	MOVQ	a2+16(FP), SI
	MOVQ	a3+24(FP), DX
	MOVQ	a4+32(FP), R10
	MOVQ	a5+40(FP), R8
	MOVQ	a6+48(FP), R9
	MOVQ	a7+56(FP), R11
	MOVQ	a8+64(FP), R12
	MOVQ	a9+72(FP), R13

	// only the first 6 arguments can be passed in registers,

	{"X%s", "X 27 ", &stringVal, "27"},

	{"X%dX", "X27X", &intVal, 27},
	{"X%dX", "X27 X", &intVal, nil}, // input does not match format
	{"X%dX", "X 27X", &intVal, 27},
	{"X%dX", "X 27 X", &intVal, nil}, // input does not match format

	{"X%dX", "X27X", &intVal, 27},
	{"X%dX", "X27X ", &intVal, 27},
	{"X%dX", " X27X", &intVal, nil},  // input does not match format
	{"X%dX", " X27X ", &intVal, nil}, // input does not match format

		x = x[1:] // skip - or .
		y = y[1:] // skip - or .
		var dx, dy string
		dx, x = nextIdent(x)
		dy, y = nextIdent(y)
		if dx != dy {
			ix := isNum(dx)
			iy := isNum(dy)
			if ix != iy {
				if ix {
					return -1
				} else {
					return +1
				}
			}
			if ix {
				if len(dx) < len(dy) {
					return -1
				}
				if len(dx) > len(dy) {
					return +1
				}
			}