legs=56 rungs=14 reps=[1, 2] slice=1048576 m=[1000, 2200, 3200, 4000, 4700, 5300, 5900] rung n m k slab W thr wall med cpu med syn med bsub med peak MB path window sources m1000k1 4 1000 1 1 512 8 30.06 207.02 12.50 4.81 10459 ntt (9200,) m1000k2 4 1000 2 1 512 8 29.97 207.02 12.68 4.80 7066 ntt (5704, 3496) m2200k1 4 2200 1 1 512 8 36.69 252.52 11.82 11.22 10392 ntt (8000,) m2200k2 4 2200 2 1 512 8 39.69 276.40 13.84 12.72 7462 ntt (4944, 3056) m3200k1 4 3200 1 1 512 8 39.29 270.56 12.22 13.47 10446 ntt (7000,) m3200k2 4 3200 2 1 512 8 39.20 272.52 14.40 11.36 7886 ntt (4312, 2688) m4000k1 4 4000 1 1 512 8 39.97 260.67 11.76 12.78 10490 ntt (6200,) m4000k2 4 4000 2 1 512 8 42.89 285.86 14.50 12.79 8227 ntt (3808, 2392) m4700k1 4 4700 1 1 512 8 48.27 319.83 10.14 21.24 10521 ntt (5500,) m4700k2 4 4700 2 1 512 8 52.96 349.38 15.21 20.48 9433 ntt (3376, 2124) m5300k1 4 5300 1 1 512 8 49.44 323.07 9.75 22.65 10634 ntt (4900,) m5300k2 4 5300 2 1 512 8 55.69 368.14 13.53 23.53 10634 ntt (3024, 1876) m5900k1 4 5900 1 1 512 8 50.40 327.20 9.40 22.49 11834 ntt (4300,) m5900k2 4 5900 2 1 512 8 52.94 348.31 13.44 21.29 11834 ntt (2664, 1636) A/A floor (|a - a_aa| / min), per rep rung rep wall % cpu % syn % m1000k1 1 5.13 3.70 11.99 m1000k1 2 0.15 2.65 13.39 m1000k2 1 4.25 3.88 4.91 m1000k2 2 9.43 3.34 8.16 m2200k1 1 5.36 6.78 11.97 m2200k1 2 4.68 4.16 3.37 m2200k2 1 15.34 13.16 14.26 m2200k2 2 6.52 4.16 4.87 m3200k1 1 7.15 5.31 4.65 m3200k1 2 0.88 2.31 5.00 m3200k2 1 10.78 12.42 15.10 m3200k2 2 2.38 3.04 4.54 m4000k1 1 2.35 5.50 11.30 m4000k1 2 0.97 0.78 4.34 m4000k2 1 2.68 2.25 2.80 m4000k2 2 2.40 2.05 5.43 m4700k1 1 1.29 0.77 4.36 m4700k1 2 0.91 1.96 0.49 m4700k2 1 0.34 6.18 10.85 m4700k2 2 2.60 2.30 11.77 m5300k1 1 0.28 3.66 4.23 m5300k1 2 4.83 3.24 0.76 m5300k2 1 9.58 11.12 2.05 m5300k2 2 20.99 23.60 0.68 m5900k1 1 8.97 5.67 2.19 m5900k1 2 5.94 7.95 2.10 m5900k2 1 3.50 0.02 1.11 m5900k2 2 0.55 0.44 2.21 MEDIAN 3.88 3.68 4.60 <- median over 28 pair(s) WORST 20.99 23.60 15.10 <- worst PER-WINDOW TREE CHARGE, by difference T(m) = syn(k=2) - syn(k=1) m S_total syn k=1 syn k=2 T(m) s T/m ms c_l ms/src nonxf cpu 1000 9200 12.50 12.68 0.180 0.1800 1.339 105.3 2200 8000 11.82 13.84 2.015 0.9159 1.226 162.7 3200 7000 12.22 14.40 2.180 0.6813 1.434 172.0 4000 6200 11.76 14.50 2.745 0.6862 1.454 167.5 4700 5500 10.14 15.21 5.070 1.0787 0.923 235.9 5300 4900 9.75 13.53 3.785 0.7142 1.216 246.1 5900 4300 9.40 13.44 4.040 0.6847 1.247 241.8 FIT BELOW 4369 T = 0.000805 * m -0.313 (c_w = 0.8050 ms/row, r2=0.88109, 4 rung(s)) FIT ABOVE 4369 T = -0.000858 * m +8.847 (c_w = -0.8583 ms/row, r2=0.57302, 3 rung(s)) SLOPE RATIO ABOVE 4369 / BELOW 4369 = -1.0663 (the model says 0 - a flat plateau at 4369) KNEE BELOW 4369 x ABOVE 4369 cross at m = 5507.4, against the tile's 4369 (+26.06%) m T meas line departure (against the BELOW-plateau line, extrapolated) 1000 0.180 0.492 -0.312 (-63.4%) 2200 2.015 1.458 +0.557 (+38.2%) 3200 2.180 2.263 -0.083 (-3.7%) 4000 2.745 2.907 -0.162 (-5.6%) 4700 5.070 3.470 +1.600 (+46.1%) 5300 3.785 3.953 -0.168 (-4.3%) 5900 4.040 4.436 -0.396 (-8.9%) LEAF TERM c_l = -0.000034 * m +1.3921 ms/source (r2=0.11327) - flat if the model holds c_l spans 0.9227 to 1.4540 ms/source, 42.6% of the median WITHIN-LEG CROSS-CHECK (k=2 legs only; solves t_i = c_l*S_i + T inside one leg) m legs S1 S2 c_l ms T s 1000 4 5704 3496 1.196 0.835 2200 4 4944 3056 1.231 1.992 3200 4 4312 2688 1.007 3.519 4000 4 3808 2392 1.321 3.076 4700 4 3376 2124 1.354 3.499 5300 4 3024 1876 1.093 3.994 5900 4 2664 1636 1.017 4.597