Reversing GoL for Art Project

[furrypony]

I intend to do a reverse Conway's Game of Life for an art project. Specifically, I want to illustrate the idea of Boltzman's Brain by drawing a comic in GoL, then finding a predecessor (about 5 steps forward would be good, I assume) of that comic, then running it forward would create the effect of a comic emerging from chaos, then rapidly back to chaos.

I have found two existing programs, https://nbickford.wordpress.com/2012/04 ... nd-profit/ is hopeful and I have tried it. It uses a few different kinds of algorithms. Duparc's Method (reversing line-by-line, merging as you go), Wood's Method (reversing cell-by-cell, merging as you go), and QuadWoods (recursively a square dividing into 4, until reaching the base case of 1 cell, then reverse the cell, then merge the results recursively upwards).
The problem is that the program is not fast enough to do it. That's expected, since it takes about a few minutes to even reverse a 10x10 pattern.
I don't want to enumerate all possible predecessors, just one, but from my understanding of QuadWoods algorithm, it seems that it would require the same amount of memory no matter how many results I want, and my computer might have run out of memory (the program crashed when I tried 10x10 patterns).

The other one is similar https://github.com/PeterBorah/atabot, though it has less algorithms and is just as slow.

Oh, the GoL field I want to atavise is on the order of magnitude of 50x50, so it's a lot bigger than all the test cases being done on the program.

My question is that are there any program that can do this? Again, I just want to find one 5-generation predecessor of a field that's about 50x50 size. Thank you.

[dvgrn]

... are there any program that can do this? Again, I just want to find one 5-generation predecessor of a field that's about 50x50 size. Thank you.

It's probably doable somehow, if you're happy with any solution and the size is allowed to increase to 60x60 or so.

This subject has been coming up quite a bit recently. The most recent discussion happened in this thread, I think.

UnLife is mentioned a few messages down from the above link. Looks like it's handling T=-1 for somewhere around a 50x50 grid without too much difficulty. With any luck you can just wait a little more than five times as long, and get a T=-5 solution.

You might possibly be able to get what you want with enough wrestling with JavaLifeSearch, which you wouldn't have to compile. I bet you'd have to fiddle with the settings quite a bit, though.

Failing that, just post the actual 50x50 pattern that you want a great-great-great-grandparent for, right here on this thread. Quite possibly somebody will not be able to resist the challenge... I would think a C/C++ rewrite of Duparc's algorithm would have a good chance of coming up with a solution fairly quickly.