Is your feature request related to a problem? Please describe.
Spark-RAPIDS has an NDS-H benchmark that uses SQL inputs and the Spark-RAPIDs plugin to complete the query work. Also the input files are stored on a HDFS data source. The libcudf benchmarks are different in that they process query work using calls to the libcudf C++ API. We should rename the libcudf benchmarks to avoid confusion.