published table cells bad ---------------------------------------------------------------------------- 8.17.4 census, 64 KiB (9 rungs x fill-triple + additive) 18 0 8.17.5 three timing arms (2 slopes + 2 r2, x3) 12 0 8.17.5 per-window charge, 14,856..31,206 15 0 8.17.5 the two secants quoted in prose 2 0 8.18.5 the ladder (14 rungs x 5 numeric) + 14 window splits 84 0 8.18.4 diagnostics (leaf spread, above-tile slope) 2 0 8.18.6 the joint fit + both controls 11 0 8.18.6 steal ladder: legs kept and r2 10 0 8.18.6 steal ladder: best K 5 3 8.18.7 A/A floor of the day 6 0 8.21.4 charge + 3 fits + corrected knees + percentages 30 0 8.21.4 the uncorrected pair 2 1 8.16.5 wall / cpu / peak (7 rungs x 3) 21 0 8.16.5 ntt syndromes (7 rungs) 7 3 'CPU crossover at n = 16,384' (3 blocks x 2 pools) 6 0 the 1 MiB ladder in full (4 rungs x 6 columns) 24 0 256 KiB at n = 16,384 (2 crossovers, under its own filter) 2 0 'The 256 KiB figure did not hold' (6 F/T + 2 A/A + 2 x-over) 10 0 EPYC 64 KiB (10 rungs x 6) + 4 crossovers + k at 2 pools 66 0 EPYC 1 MiB (2 crossovers) 2 0 the 16 Sep nibble k cells (183 / 241 / 249 / 395) 4 0 ---------------------------------------------------------------------------- TOTAL re-derived 339 7 332 of 339 cells reproduce exactly; 7 do not. HELD OUT of the tally - re-derivable from no banked input: 15 cells 8.17.5 additive-leaves column - the per-leg .err sidecars carrying [ntt-fill] were never committed