MOVD	R1, (g_sched+gobuf_sp)(g)
	MOVD	LR, R8
	MOVD	R8, (g_sched+gobuf_pc)(g)
	MOVD	R5, (g_sched+gobuf_lr)(g)
	MOVD	R11, (g_sched+gobuf_ctxt)(g)

	// Cannot grow scheduler stack (m->g0).
	MOVD	g_m(g), R7
	MOVD	m_g0(R7), R8
	CMP	g, R8
	BNE	3(PC)
	BL	runtime·badmorestackg0(SB)
	BL	runtime·abort(SB)

	// Cannot grow signal stack (m->gsignal).
	MOVD	m_gsignal(R7), R8
	CMP	g, R8
	BNE	3(PC)

	// R4: size of stack arguments (n-8)*8
	// R5: &args[8]
	// R6: loop counter, from 0 to (n-8)*8
	// R7: scratch
	// R8: copy of RSP - (R2)(RSP) assembles as (R2)(ZR)
	SUB	$8, R0, R4
	LSL	$3, R4
	ADD	$(8*8), R12, R5
	MOVD	$0, R6
	MOVD	RSP, R8
stackargs:
	MOVD	(R6)(R5), R7
	MOVD	R7, (R6)(R8)
	ADD	$8, R6
	CMP	R6, R4
	BNE	stackargs

_8args:
	MOVD	(7*8)(R12), R7
_7args:
	MOVD	(6*8)(R12), R6
_6args:

	MOVD	0(R14), R3		// arg 1 - addr
	MOVD	8(R14), R4		// arg 2 - len
	MOVW	16(R14), R5		// arg 3 - prot
	MOVW	20(R14), R6		// arg 4 - flags
	MOVW	24(R14), R7		// arg 5 - fid
	MOVW	28(R14), R8		// arg 6 - offset
	CALL	libc_mmap(SB)
	MOVD	$0, R4
	CMP	R3, $-1
	BNE	noerr
	CALL	libc_errno(SB)
	MOVW	(R3), R4		// errno
	MOVD	$0, R3
noerr:
	MOVD	R3, 32(R14)
	MOVD	R4, 40(R14)
	RET

	MOVL CX, 0(SP)
	INCL CX
	MOVL CX, 4(SP)
	INCL CX
	MOVL CX, 8(SP)
	INCL CX
	MOVL CX, 12(SP)
	MOVOU 0(SP), X12

	// Load seed words into next two rows and into DI, SI, R8..R13
	SEED(0, DI, X4)
	SEED(1, SI, X5)
	SEED(2, R8, X6)
	SEED(3, R9, X7)
	SEED(4, R10, X8)
	SEED(5, R11, X9)
	SEED(6, R12, X10)
	SEED(7, R13, X11)

	// Zeros for remaining two matrix entries.

.globl __umodti3
__umodti3:
	# specialized to u128 % u64, so verify that
	test %rcx,%rcx
	jne 1f

	# save divisor
	movq %rdx, %r8

	# reduce top 64 bits mod divisor
	movq %rsi, %rax
	xorl %edx, %edx
	divq %r8

	# reduce full 128-bit mod divisor
	# quotient fits in 64 bits because top 64 bits have been reduced < divisor.

TEXT ·Syscall9(SB),NOSPLIT,$28-52
	NO_LOCAL_POINTERS
	JAL	runtime·entersyscall(SB)
	MOVW	a1+4(FP), R4
	MOVW	a2+8(FP), R5
	MOVW	a3+12(FP), R6
	MOVW	a4+16(FP), R7
	MOVW	a5+20(FP), R8
	MOVW	a6+24(FP), R9
	MOVW	a7+28(FP), R10
	MOVW	a8+32(FP), R11
	MOVW	R8, 16(R29)
	MOVW	R9, 20(R29)
	MOVW	R10, 24(R29)
	MOVW	R11, 28(R29)
	MOVW	trap+0(FP), R2	// syscall entry
	MOVW	R0, R3	// reset R3 to zero as 1-ret SYSCALL keeps it
	SYSCALL

	// The C arguments arrive in R0, R1, R2, and R3. We want to
	// pass R0, R1, and R3 to Go, so we push those on the stack.
	// Also, save C callee-save registers R4-R12.
	MOVM.WP	[R0, R1, R3, R4, R5, R6, R7, R8, R9, g, R11, R12], (R13)
	// Finally, save the link register R14. This also puts the
	// arguments we pushed for cgocallback where they need to be,
	// starting at 4(R13).
	MOVW.W	R14, -4(R13)

	VLD1.P	(R8)(R9.SXTX<<2), [V2.B16]                       // ERROR "invalid extended register"
	VLD1.P	(R8)(R9<<2), [V2.B16]                            // ERROR "invalid extended register"
	VST1.P	[V1.B16], (R8)(R9.UXTW)                          // ERROR "invalid extended register"
	VST1.P	[V1.B16], (R8)(R9<<1)                            // ERROR "invalid extended register"

type PtraceRegsAmd64 struct {
	R15      uint64
	R14      uint64
	R13      uint64
	R12      uint64
	Rbp      uint64
	Rbx      uint64
	R11      uint64
	R10      uint64
	R9       uint64
	R8       uint64
	Rax      uint64
	Rcx      uint64
	Rdx      uint64
	Rsi      uint64
	Rdi      uint64
	Orig_rax uint64
	Rip      uint64
	Cs       uint64
	Eflags   uint64
	Rsp      uint64
	Ss       uint64

	// amd64:`MOVQ\t\$-2401018187971961171, DX`, `MOVQ\t\$-2401018187971961171, SI`, `MOVQ\t\$-2401018187971961171, DI`
	// amd64:`MOVQ\t\$-2401018187971961171, R8`, `MOVQ\t\$-2401018187971961171, R9`, `MOVQ\t\$-2401018187971961171, R10`
	// amd64:`MOVQ\t\$-2401018187971961171, R11`, `MOVQ\t\$-2401018187971961171, R12`, `MOVQ\t\$-2401018187971961171, R13`