Last row of RAMA on LL(n) against the zeros of zeta. PARI/GP 2.19, Python 3 with numpy.

  python3 sieve_e.py run 100000000000 out/run_1e11     scan of e(n) and M(n), all n <= 10^11
  gp -q zeros.gp < /dev/null                           zeros below height 10000 -> out/zeros.txt
  python3 waves.py out/zeros.txt out/run_1e11/samples.npz out/run_1e11/summary.json

This ZIP is flat. The commands above write to out/ and out/run_1e11/. To run waves.py on the
files of the ZIP: python3 waves.py zeros.txt samples.npz summary.json

Files of the run:
  summary.json       totals and one record per decade (extremes, weighted sums, histograms)
  samples.npz        n and M(floor(n/j)), j = 1..11, at the sample points
  segments_top.csv   the 200 segments of length 10^7 with the largest e(n)/sqrt(n), and the 200
                     with the smallest; same columns as the full table that sieve_e.py writes
  waves_1e11.json    output of waves.py
  zeros.txt          gamma, Re a, Im a with a = 1/(rho*zeta'(rho)); rounded here (gamma to 13
                     decimals, a to 11 digits; zeros.gp writes 38 digits) to fit the 2 MB limit;
                     waves.py reads both forms and works in double precision
