// value from the input block.
	VCIPHERLAST8_XOR_INPUT
	// Store the results (8*16) and update BLK_OUT by 128.
	STORE_OUTPUT_BLOCK128(BLK_OUT)
	ADD	$-128, IN_LEN	// input size
	CMP     IN_LEN, $128	// check if >= blocksize
	BGE	block128_loop	// next input block
	CMP	IN_LEN, $0
	BEQ	done
block64:
	CMP	IN_LEN, $64	// Check if >= 64
	BLT	block16_loop
	// Do 4 encryptions in parallel by setting

	var erasure Erasure
	if !latestMeta.Deleted && !latestMeta.IsRemote() {
		// Initialize erasure coding
		erasure, err = NewErasure(ctx, latestMeta.Erasure.DataBlocks,
			latestMeta.Erasure.ParityBlocks, latestMeta.Erasure.BlockSize)
		if err != nil {
			return result, err
		}
	}

	result.ObjectSize, err = latestMeta.ToObjectInfo(bucket, object, true).GetActualSize()
	if err != nil {
		return result, err
	}

	writeBlocks(ctxt, out, ctxt.outSem, ctxt.loader, ctxt.Textp, addr, size, pad)
}

const blockSize = 1 << 20 // 1MB chunks written at a time.

// writeBlocks writes a specified chunk of symbols to the output buffer. It
// breaks the write up into ≥blockSize chunks to write them out, and schedules
// as many goroutines as necessary to accomplish this task. This call then

	},
	"crypto/sha1": {
		{"BlockSize", Const, 0},
		{"New", Func, 0},
		{"Size", Const, 0},
		{"Sum", Func, 2},
	},
	"crypto/sha256": {
		{"BlockSize", Const, 0},
		{"New", Func, 0},
		{"New224", Func, 0},
		{"Size", Const, 0},
		{"Size224", Const, 0},
		{"Sum224", Func, 2},
		{"Sum256", Func, 2},
	},
	"crypto/sha512": {
		{"BlockSize", Const, 0},
		{"New", Func, 0},

    }
    TF_SetStatus(status, TF_OK, "");
    return bytes_transferred;
  };
  for (size_t block_size = 2; block_size <= 4; block_size++) {
    // Make a cache of N-byte block size (1 block) and verify that reads of
    // varying offsets and lengths return correct data.
    tf_gcs_filesystem::RamFileBlockCache cache(block_size, block_size, 0,
                                               fetcher);

                 uint64_t block_size, size_t max_bytes, uint64_t max_staleness,
                 uint64_t stat_cache_max_age, size_t stat_cache_max_entries)
    : gcs_client(gcs_client),
      compose(compose),
      block_cache_lock(),
      block_size(block_size) {
  file_block_cache = std::make_unique<RamFileBlockCache>(
      block_size, max_bytes, max_staleness,

  int64_t out_channel = filter_shape[3];
  // Value/Op before reshape op.
  Value before_reshape_value = filter;
  if (height % block_size != 0 || width % block_size != 0) {
    // Calculate paddings for height and width.
    int32_t pad_h = block_size - height % block_size;
    int32_t pad_w = block_size - width % block_size;
    auto pad_op =
        GetPadOpForConv2DFilter(filter_shape, filter, &builder, pad_h, pad_w);

  float curr_sparsity = 0;
  std::vector<int> selected_block_size;
  result.needs_densify = true;
  for (const auto& block_size : supported_block_size) {
    curr_sparsity = CalculateBlockSparsity(attr, type, block_size);
    if (curr_sparsity / random_sparsity > ratio_threshold) {
      selected_block_size = block_size;
      result.can_compress = true;
      result.needs_densify = false;

  bool compose;
  absl::Mutex block_cache_lock;
  std::shared_ptr<RamFileBlockCache> file_block_cache
      ABSL_GUARDED_BY(block_cache_lock);
  uint64_t block_size;  // Reads smaller than block_size will trigger a read
                        // of block_size.
  std::unique_ptr<ExpiringLRUCache<GcsFileStat>> stat_cache;
  GCSFile(google::cloud::storage::Client&& gcs_client);
  // This constructor is used for testing purpose only.

	MRS	DCZID_EL0, R3
	TBZ	$4, R3, init
	// ZVA not available
	MOVW	$~0, R5
	MOVW	R5, block_size<>(SB)
	B	no_zva

init:
	MOVW	$4, R9
	ANDW	$15, R3, R5
	LSLW	R5, R9, R5
	MOVW	R5, block_size<>(SB)

	ANDS	$63, R5, R9
	// Block size is less than 64.
	BNE	no_zva

zero_by_line:
	CMP	R5, R1
	// Not enough memory to reach alignment