Translated Cyclers Decider

I made a decider for translated cyclers in python today:

As intended, it seems to decide exactly the translated cyclers (i checked around 200 TMs it didn’t decide and none of them were translated cyclers, and i checked around 30 TMs it decided and all of them were translated cyclers), and it should be pretty easy to verify, unless there are mistakes like forgetting a +1 somewhere. I was told to post it here.

Hi Yto, I have also made a decider for translated cyclers. As a stress test for your decider, have you tried to give it machine number 63687188?

My decider says it has a cycle length of 9482, and a shift of 74. Can your program confirm this? It is good to have a decider give statistics about the machines as well as the decider output, as this makes confirmation much more robust.

Yes, my program did confirm that.

Thank you for your work!

A few machines for which we have stats are here:

P is the size of the period and should be correct
S is the time step when the cycling behavior starts and should not be correct in the forum post (see the posts after for why)

I had forgotten that we have this cycler with period 91,995 :smile:
(its marked as undecided because I did not bother adding it manually to the index)

Also, note that the indices of all the machines decided so far are here: arranged by decider and parameters are in the title.

Each 4-byte (big endian) chunk of these files is a uint32 giving the index of a machine. The number of machine in a file is the size of the file divided by 4.

There are 2 runs of translated cyclers with the same parameters (10,000 time 5,000 memory), they must certainly contain machines in common because of nonlinear history of when backtracking was un-applied because of a bug. I should merge them when times permits.

I’ve created another reproduction of this decider here.

I’ve verified that it produces identical results to the translated-cyclers-run-8f5b2539279a-time-1000-space-500-minIndex-14322029-maxIndex-88664064 file, with the caveat that I had to give some machines more time on my implementation. I’m unsure why the difference exists, but my algorithm was just based on the textual description on the forums so there is probably some slight difference.

1 Like

Thank you for your work, this is great news.