Benchmark: timing 1000000 iterations of i_noop, i_peers, i_sep...
    i_noop: 13.0096 wallclock secs (10.13 usr +  2.89 sys = 13.02 CPU) @ 76804.92/s (n=1000000)
   i_peers: 13.0293 wallclock secs (10.17 usr +  2.85 sys = 13.02 CPU) @ 76804.92/s (n=1000000)
     i_sep: 12.6381 wallclock secs ( 9.72 usr +  2.92 sys = 12.64 CPU) @ 79113.92/s (n=1000000)
           Rate i_peers  i_noop   i_sep
i_peers 76805/s      --     -0%     -3%
i_noop  76805/s      0%      --     -3%
i_sep   79114/s      3%      3%      --

Benchmark: timing 1000000 iterations of i_noop, i_peers, i_sep...
    i_noop: 14.0463 wallclock secs (11.13 usr +  2.91 sys = 14.04 CPU) @ 71225.07/s (n=1000000)
   i_peers: 14.6951 wallclock secs (11.36 usr +  3.34 sys = 14.70 CPU) @ 68027.21/s (n=1000000)
     i_sep: 15.484 wallclock secs (12.20 usr +  3.29 sys = 15.49 CPU) @ 64557.78/s (n=1000000)
           Rate   i_sep i_peers  i_noop
i_sep   64558/s      --     -5%     -9%
i_peers 68027/s      5%      --     -4%
i_noop  71225/s     10%      5%      --

This one is from the Windows box:
Benchmark: timing 1000000 iterations of i_noop, i_peers, i_sep...
    i_noop: 26 wallclock secs (15.97 usr +  7.00 sys = 22.97 CPU) @ 43536.94/s (n=1000000)
   i_peers: 26 wallclock secs (16.84 usr +  6.30 sys = 23.14 CPU) @ 43215.21/s (n=1000000)
     i_sep: 23 wallclock secs (16.16 usr +  6.61 sys = 22.76 CPU) @ 43927.08/s (n=1000000)
           Rate i_peers  i_noop   i_sep
i_peers 43215/s      --     -1%     -2%
i_noop  43537/s      1%      --     -1%
i_sep   43927/s      2%      1%      --