SmallVector<Attribute, 8> shuffled(values.size());

  for (size_t i = 0; i < permutation.size(); ++i) {
    for (size_t j = 0; j < inner_size; ++j) {
      shuffled[i * inner_size + j] = values[permutation[i] * inner_size + j];
    }
  }

  return ArrayAttr::get(attr.getContext(), shuffled);
}

// Shuffle ranked tensor dimensions according to the permutation.

	return shuffledPartsMetadata
}

// shuffleDisks - shuffle input disks slice depending on the
// erasure distribution. Return shuffled slice of disks with
// their expected distribution.
func shuffleDisks(disks []StorageAPI, distribution []int) (shuffledDisks []StorageAPI) {
	if distribution == nil {
		return disks
	}
	shuffledDisks = make([]StorageAPI, len(disks))
	// Shuffle disks for expected distribution.

	}

	// already sorted
	data.initB()
	SortStableFunc(data, intPairCmp)
	if !IsSortedFunc(data, intPairCmp) {
		t.Errorf("Stable shuffled sorted %d ints (order)", n)
	}
	if !data.inOrder(false) {
		t.Errorf("Stable shuffled sorted %d ints (stability)", n)
	}

	// sorted reversed
	for i := 0; i < len(data); i++ {
		data[i].a = len(data) - i
	}
	data.initB()

	}
	fakePodControl.Clear()
}

// shuffle returns a new shuffled list of container controllers.
func shuffle(controllers []*apps.ReplicaSet) []*apps.ReplicaSet {
	numControllers := len(controllers)
	randIndexes := rand.Perm(numControllers)
	shuffled := make([]*apps.ReplicaSet, numControllers)
	for i := 0; i < numControllers; i++ {
		shuffled[i] = controllers[randIndexes[i]]
	}
	return shuffled
}

	verifyUpdates(t, m, 0)
}

// shuffle returns a new shuffled list of container statuses.
func shuffle(statuses []v1.ContainerStatus) []v1.ContainerStatus {
	numStatuses := len(statuses)
	randIndexes := rand.Perm(numStatuses)
	shuffled := make([]v1.ContainerStatus, numStatuses)
	for i := 0; i < numStatuses; i++ {
		shuffled[i] = statuses[randIndexes[i]]
	}
	return shuffled
}

// first pass over the data. It's an argument here because we merge the first
// pass with the Huffman decoding.
//
// This also implements the “single array” method from the bzip2 source code
// which leaves the output, still shuffled, in the bottom 8 bits of tt with the
// index of the next byte in the top 24-bits. The index of the first byte is
// returned.
func inverseBWT(tt []uint32, origPtr uint, c []uint) uint32 {
	sum := uint(0)

		// symbols reliably. Sort by name but use a stable sort, so that
		// any original entries with the same name (all DWARFVAR symbols
		// have empty names but different relocation sets) are not shuffled.
		// TODO: Find a better place and optimize to only sort TOC symbols.
		sort.SliceStable(ctxt.Data, func(i, j int) bool {
			return ctxt.Data[i].Name < ctxt.Data[j].Name
		})
	}

      return no_op();

    // For vectors, shuffle values by sorting instead of the obvious
    // Fisher-Yates algorithm. Fisher-Yates is simple to implement and correct,
    // but not easily parallelizable. For a sufficiently parallel architecture,
    // it is faster to sort many times, than Fisher-Yates shuffle once.
    if (input_rank == 1) {
      // Shuffle values by assigning each value a random key and sorting the

		t.Fatal(err)
	}
}

func TestDash(t *testing.T) {
	m := MapFS{
		"a-b/a": {Data: []byte("a-b/a")},
	}
	if err := TestFS(m, "a-b/a"); err != nil {
		t.Error(err)
	}
}

type shuffledFS MapFS

func (fsys shuffledFS) Open(name string) (fs.File, error) {
	f, err := MapFS(fsys).Open(name)
	if err != nil {
		return nil, err
	}
	return &shuffledFile{File: f}, nil
}

type shuffledFile struct{ fs.File }

	for i := range p {
		p[i] = i
	}
	r.Shuffle(len(p), func(i, j int) { p[i], p[j] = p[j], p[i] })
	return p
}

// Shuffle pseudo-randomizes the order of elements.
// n is the number of elements. Shuffle panics if n < 0.
// swap swaps the elements with indexes i and j.
func (r *Rand) Shuffle(n int, swap func(i, j int)) {
	if n < 0 {
		panic("invalid argument to Shuffle")
	}