We changed to a sequence of benchmarks that fit better our current use
case.