Currently, hdf5-udf writes the UDF in the HDF5 file by:

1. Registering the UDF filter
2. Declaring a dataset with resolution N x M x sizeof(datatype)
3. Trying to fit the payload (bytecode and metadata) in the memory allocated for that dataset
4. Calling H5Dwrite with that chunk of memory

When the dataset is big enough it's likely that the payload will fit just fine. Take a small v