Benchmark: timing 1000 iterations of Way1A, Way2A, Way2B, Way3A, Way4A, Way4B, Way5A, Way5B, Way6A...
     Way1A: 52 wallclock secs ( 1.46 usr +  0.28 sys =  1.74 CPU) @ 574.05/s (n=1000)
     Way2A: 30 wallclock secs ( 1.17 usr +  0.17 sys =  1.34 CPU) @ 745.16/s (n=1000)
     Way2B: 30 wallclock secs ( 1.31 usr +  0.18 sys =  1.49 CPU) @ 669.79/s (n=1000)
     Way3A: 47 wallclock secs ( 1.02 usr +  0.22 sys =  1.24 CPU) @ 805.80/s (n=1000)
     Way4A: 30 wallclock secs ( 1.31 usr +  0.18 sys =  1.49 CPU) @ 670.24/s (n=1000)
     Way4B: 29 wallclock secs ( 1.14 usr +  0.20 sys =  1.34 CPU) @ 744.60/s (n=1000)
     Way5A: 29 wallclock secs ( 1.16 usr +  0.12 sys =  1.28 CPU) @ 780.03/s (n=1000)
     Way5B: 25 wallclock secs ( 0.92 usr +  0.18 sys =  1.10 CPU) @ 908.27/s (n=1000)
     Way6A: 53 wallclock secs ( 1.30 usr +  0.27 sys =  1.57 CPU) @ 636.13/s (n=1000)
       Rate Way1A Way6A Way2B Way4A Way4B Way2A Way5A Way3A Way5B
Way1A 574/s    --  -10%  -14%  -14%  -23%  -23%  -26%  -29%  -37%
Way6A 636/s   11%    --   -5%   -5%  -15%  -15%  -18%  -21%  -30%
Way2B 670/s   17%    5%    --   -0%  -10%  -10%  -14%  -17%  -26%
Way4A 670/s   17%    5%    0%    --  -10%  -10%  -14%  -17%  -26%
Way4B 745/s   30%   17%   11%   11%    --   -0%   -5%   -8%  -18%
Way2A 745/s   30%   17%   11%   11%    0%    --   -4%   -8%  -18%
Way5A 780/s   36%   23%   16%   16%    5%    5%    --   -3%  -14%
Way3A 806/s   40%   27%   20%   20%    8%    8%    3%    --  -11%
Way5B 908/s   58%   43%   36%   36%   22%   22%   16%   13%    --