Equation for computing remaining time during epsilon initialization is wrong

[electroflow]

Thank you for your amazing software package. It is no big issue but I am always a bit irritated when seeing the progress estimates when running meep e.g.

subpixel-averaging is 6.68749% done, 55.8381 s remaining
subpixel-averaging is 12.2589% done, 28.6371 s remaining
subpixel-averaging is 17.3983% done, 19.0087 s remaining
subpixel-averaging is 22.3127% done, 13.9325 s remaining
subpixel-averaging is 27.155% done, 10.7399 s remaining
subpixel-averaging is 31.8264% done, 8.57822 s remaining
subpixel-averaging is 36.3807% done, 7.00073 s remaining
subpixel-averaging is 40.8811% done, 5.78909 s remaining
subpixel-averaging is 45.3184% done, 4.82662 s remaining
subpixel-averaging is 49.7377% done, 4.04236 s remaining
subpixel-averaging is 54.1751% done, 3.38955 s remaining
subpixel-averaging is 58.4594% done, 2.84773 s remaining
subpixel-averaging is 62.9417% done, 2.35727 s remaining
subpixel-averaging is 67.2891% done, 1.9467 s remaining
subpixel-averaging is 71.8794% done, 1.56742 s remaining
subpixel-averaging is 76.7488% done, 1.21236 s remaining
subpixel-averaging is 81.4291% done, 0.913055 s remaining
subpixel-averaging is 86.4875% done, 0.62555 s remaining
subpixel-averaging is 91.9509% done, 0.350244 s remaining
subpixel-averaging is 98.1164% done, 0.0767999 s remaining

The reason the remaining time does not fall linearly as expected is equation

meep/src/anisotropic_averaging.cpp

Line 276 in 0d4050c

(npixels - ipixel) * (wall_time() - last_output_time) / ipixel);

I think it should be normalized not to ipixel but rather the amount of pixels since the last pixel. I think something like

    if (verbosity > 0 && (ipixel + 1) % 1000 == 0 &&
        wall_time() > last_output_time + MEEP_MIN_OUTPUT_TIME) {
      master_printf("%s is %g%% done, %g s remaining\n",
                    use_anisotropic_averaging ? "subpixel-averaging" : "grid initialization",
                    ipixel * 100.0 / npixels,
                    (npixels - ipixel) * (wall_time() - last_output_time) / (ipixel - last_ipixel));
      last_output_time = wall_time();
      last_ipixel = ipixel;
    }

might do the job.

[stevengj]

Good catch. Could you submit a PR?