Benchmark: timing 1500 iterations of bw, tied...
        bw:  1 wallclock secs ( 0.36 usr +  0.30 sys =  0.66 CPU) @ 2283.11/s (n=1500)
      tied:  5 wallclock secs ( 2.05 usr +  3.31 sys =  5.36 CPU) @ 279.90/s (n=1500)