off  atomic.Uintptr
}

const traceRegionAllocBlockData = 64<<10 - unsafe.Sizeof(traceRegionAllocBlockHeader{})

// alloc allocates n-byte block. The block is always aligned to 8 bytes, regardless of platform.
func (a *traceRegionAlloc) alloc(n uintptr) *notInHeap {
	n = alignUp(n, 8)
	if n > traceRegionAllocBlockData {
		throw("traceRegion: alloc too large")
	}
	if a.dropping.Load() {
		throw("traceRegion: alloc with concurrent drop")
	}

	DMB	MB_ISHST

	STREXD	R2, (R1), R0	// stores R2 and R3
	CMP	$0, R0
	BNE	store64loop

	DMB	MB_ISH
	RET

// The following functions all panic if their address argument isn't
// 8-byte aligned. Since we're calling back into Go code to do this,
// we have to cooperate with stack unwinding. In the normal case, the
// functions tail-call into the appropriate implementation, which

    // The cache is effectively disabled, so we pass the read through to the
    // fetcher without breaking it up into blocks.
    return block_fetcher_(filename, offset, n, buffer, status);
  }
  // Calculate the block-aligned start and end of the read.
  size_t start = block_size_ * (offset / block_size_);
  size_t finish = block_size_ * ((offset + n) / block_size_);
  if (finish < offset + n) {
    finish += block_size_;
  }

	MOVV	$1, R1
	MOVB	R1, ret+24(FP)
	RET
byte_test:
	MOVBU	(R1), R6
	ADDV	$1, R1
	MOVBU	(R2), R7
	ADDV	$1, R2
	BEQ	R6, R7, byte_loop
	JMP	not_eq

chunk_entry:
	// make sure both a and b are aligned
	OR	R1, R2, R9
	AND	$0x7, R9
	BNE	R0, R9, byte_loop
	JMP	chunk_loop_1

chunk_loop:
	// chunk size is 16
	SGTU	$16, R3, R8
	BNE	R0, R8, chunk_tail_8
chunk_loop_1:
	MOVV	(R1), R6
	MOVV	(R2), R7

_back:
	ADD	N, FROM		/* from end pointer */
	CMP	$4, N		/* need at least 4 bytes to copy */
	BLT	_b1tail

_b4align:				/* align destination on 4 */
	AND.S	$3, TE, TMP
	BEQ	_b4aligned

	MOVBU.W	-1(FROM), TMP	/* pre-indexed */
	MOVBU.W	TMP, -1(TE)	/* pre-indexed */
	B	_b4align

_b4aligned:				/* is source now aligned? */
	AND.S	$3, FROM, TMP
	BNE	_bunaligned

	ADD	$31, TS, TMP	/* do 32-byte chunks if possible */

	RET

// void	Or8(byte volatile*, byte);
TEXT ·Or8(SB), NOSPLIT, $0-9
	MOVV	ptr+0(FP), R4
	MOVBU	val+8(FP), R5
	// Align ptr down to 4 bytes so we can use 32-bit load/store.
	MOVV	$~3, R6
	AND	R4, R6
	// R7 = ((ptr & 3) * 8)
	AND	$3, R4, R7
	SLLV	$3, R7
	// Shift val for aligned ptr. R5 = val << R4
	SLLV	R7, R5

	DBAR
	LL	(R6), R7
	OR	R5, R7
	SC	R7, (R6)
	BEQ	R7, -4(PC)
	DBAR
	RET

// check is negligible since it is only required for large copies.
//
// Large copies use a software pipelined loop processing 64 bytes per iteration.
// The destination pointer is 16-byte aligned to minimize unaligned accesses.
// The loop tail is handled by always copying 64 bytes from the end.

// func memmove(to, from unsafe.Pointer, n uintptr)
TEXT runtime·memmove<ABIInternal>(SB), NOSPLIT|NOFRAME, $0-24
	CBZ	R2, copy0

}

type ucontext struct {
	uc_flags    uint64
	uc_link     *ucontext
	uc_stack    stackt
	uc_sigmask  uint64
	_pad        [(1024 - 64) / 8]byte
	_pad2       [8]byte // sigcontext must be aligned to 16-byte
	uc_mcontext sigcontext

// using the initial stack allocated by the OS.
// It calls back into standard C using the BL below.
// To do that, the stack pointer must be 8-byte-aligned.
TEXT runtime·_initcgo(SB),NOSPLIT|NOFRAME,$0
	MOVM.DB.W [R4, R14], (R13)	// push {r4, lr}

	// Ensure stack is 8-byte aligned before calling C code
	MOVW	R13, R4
	BIC	$0x7, R13

	// Allocate a TLS slot to hold g across calls to external code

	minRoutingSockaddrLen = rsaAlignOf(0)
)

// Round the length of a raw sockaddr up to align it properly.
func rsaAlignOf(salen int) int {
	salign := sizeofPtr
	if darwin64Bit {
		// Darwin kernels require 32-bit aligned access to
		// routing facilities.
		salign = 4
	} else if netbsd32Bit {
		// NetBSD 6 and beyond kernels require 64-bit aligned
		// access to routing facilities.
		salign = 8
	} else if runtime.GOOS == "freebsd" {