Large amounts of dynamic memory allocation.

[NeilZhy]

The officially provided sample code is as follows:

draco::DecoderBuffer buffer;
buffer.Init(data.data(), data.size());

const draco::EncodedGeometryType geom_type =
    draco::GetEncodedGeometryType(&buffer);
if (geom_type == draco::TRIANGULAR_MESH) {
  unique_ptr<draco::Mesh> mesh = draco::DecodeMeshFromBuffer(&buffer);
} else if (geom_type == draco::POINT_CLOUD) {
  unique_ptr<draco::PointCloud> pc = draco::DecodePointCloudFromBuffer(&buffer);
}

After calling unique_ptr<draco::PointCloud> pc = draco::DecodePointCloudFromBuffer(&buffer);, the code: DRACO_ASSIGN_OR_RETURN(std::unique_ptr<PointCloudDecoder> decoder, CreatePointCloudDecoder(header.encoder_method)) will be called.

Eventually the following code will be called:

bool DataBuffer::Update(const void *data, int64_t size, int64_t offset) {
  if (data == nullptr) {
    if (size + offset < 0)
      return false;
    // If no data is provided, just resize the buffer.
    data_.resize(size + offset);
  } else {
    if (size < 0)
      return false;
    if (size + offset > static_cast<int64_t>(data_.size())) {
      data_.resize(size + offset);
    }
    const uint8_t *const byte_data = static_cast<const uint8_t *>(data);
    std::copy(byte_data, byte_data + size, data_.data() + offset);
  }
  descriptor_.buffer_update_count++;
  return true;
}

Each time a PointCloudDecoder object is created, it will eventually result in calling data_.resize(size + offset);, resulting in a large amount of dynamic memory allocation. Why do we need to create a new PointCloudDecoder object every time?

[ocaspi-sc]

+1 I also suffer from this issue