Relied on streams more instead of lists, trying to get faster running
of generated benchmarks

Removed double-generation of benchmarks