  round 1/4 done
  round 2/4 done
  round 3/4 done
  round 4/4 done

A. WHOLE solve, SEPARATE subprocesses, alternating, MIN over 4 rounds
grid   M  layers  dense [s] factored [s]  speedup  spread off  spread on
(2,2)   6      1      1.196        1.065    1.12x       14.9%      11.8%
(2,2)   7      1      1.950        1.523    1.28x        5.2%       8.5%
(2,2)   8      1      3.344        2.513    1.33x        8.9%       8.8%
(3,3)   6      1      4.494        3.355    1.34x       12.5%      12.6%
(3,3)   7      1     11.963        8.206    1.46x        3.5%      12.7%
(3,3)   8      1     30.248       20.049    1.51x        3.7%      12.0%
(2,2)   7      3      3.573        2.336    1.53x       10.3%       6.5%
(3,3)   7      3     25.334       14.099    1.80x        4.8%       6.9%

single-layer speedup 1.12 .. 1.51x   (build doc claims 1.49 - 1.55x)
three-layer speedup  1.53 .. 1.80x   (build doc claims 1.80 - 1.87x)

B. REGION solve, in-process, min over 3 timed calls after a warm-up
grid   M    4q^2   OFF [s]    ON [s]  ratio
(2,2)   6    400     0.218     0.069   3.16
(2,2)   7    576     0.605     0.175   3.46
(2,2)   8    784     1.263     0.373   3.39
(3,3)   6    900     1.817     0.566   3.21
(3,3)   7   1296     5.130     1.339   3.83
(3,3)   8   1764    12.146     2.931   4.14

region-solve ratio   3.16 .. 4.14x   (build doc claims 3.3 - 4.2x)
