Hi Toro's Sieve test

Here's a quick analysis of the JimB timing data:

 iter  ms   ms/iter
===================
   5   477    95
  10   942    94
  15  2869    191
  20  3793    190
  25  4833    193
  30  5690    190

So the first two tests run at full speed and the rest at half speed.

This suggests a small primary cache, sufficient only for the first two. Celeron?