Bob Showalter wrote:
>
> This will be quicker than the other methods, especially as the array size
> grows. Try your benchmark with an array of say 10, or 100 elements.
>
yes, it's faster. i didn't include the code since this algr. is well known.
just some number for those who are interested:
Benchmark: timing 99999 iterations of reg_ref_s, reg_s, split_ref_s,
split_s, sr_s, st_s...
reg_ref_s: 155 wallclock secs (68.77 usr + 0.12 sys = 68.89 CPU) @
1451.57/s (n=99999)
reg_s: 204 wallclock secs (90.51 usr + 0.11 sys = 90.62 CPU) @
1103.50/s (n=99999)
split_ref_s: 139 wallclock secs (63.22 usr + 0.07 sys = 63.29 CPU) @
1580.01/s (n=99999)
split_s: 254 wallclock secs (125.16 usr + 0.12 sys = 125.28 CPU) @
798.20/s (n=99999)
sr_s: 114 wallclock secs (51.76 usr + 0.06 sys = 51.82 CPU) @
1929.74/s (n=99999)
st_s: 74 wallclock secs (35.84 usr + 0.06 sys = 35.90 CPU) @
2785.49/s (n=99999)
st_s (Schwartzian Transform) is the winner.
david
--
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
