VSEL X1H, X2H, SEL1, X1H
	VSEL Y1L, Y2L, SEL1, Y1L
	VSEL Y1H, Y2H, SEL1, Y1H

	VADDUBM SEL2, ONE, SEL2    // Increment SEL2 bytes by 1
	ADD     $64, P1ptr         // Next chunk
	BDNZ	loop_select

	STXVD2X X1H, (P3ptr+R0)
	STXVD2X X1L, (P3ptr+R16)
	STXVD2X Y1H, (P3ptr+R17)
	STXVD2X Y1L, (P3ptr+R18)
	RET

#undef P3ptr
#undef P1ptr
#undef COUNT
#undef X1L

// countAlloc returns the number of objects allocated in span s by
// scanning the mark bitmap.
func (s *mspan) countAlloc() int {
	count := 0
	bytes := divRoundUp(uintptr(s.nelems), 8)
	// Iterate over each 8-byte chunk and count allocations
	// with an intrinsic. Note that newMarkBits guarantees that
	// gcmarkBits will be 8-byte aligned, so we don't have to
	// worry about edge cases, irrelevant bits will simply be zero.

//
// For hashing larger byte slices, or byte streams such as those read from
// a file or socket, use Sendto with MSG_MORE to instruct the kernel to update
// the hash digest instead of creating a new one for a given chunk and finalizing it.
//
//	// Assume hashfd and addr are already configured using the setup process.
//	hash := os.NewFile(hashfd, "sha1")
//	// Hash the contents of a file.
//	f, _ := os.Open("/tmp/linux-4.10-rc7.tar.xz")

		return append(dst, prog...)
	}

	// Element is small with pointer mask; use as literal bits.
	ptrs := typ.PtrBytes / goarch.PtrSize
	mask := typ.GcSlice(0, (ptrs+7)/8)

	// Emit 120-bit chunks of full bytes (max is 127 but we avoid using partial bytes).
	for ; ptrs > 120; ptrs -= 120 {
		dst = append(dst, 120)
		dst = append(dst, mask[:15]...)
		mask = mask[15:]
	}