[SPOILER] My Lowest Cycle Solutions

[skyhrg]

Hello

If anyone is interested, I have posted below my solutions for very fast cycle solutions. The key here is parallel computation and computation distribution to avoid bottlenecks and do simultaneous computations. Calculate multiple values at the same time.

Here are the results and the solutions. As you can see, the results in cycles are very optimal compared to the multiplayer solution distribution:

Differential Converter (133 cycles)
http://steamcommunity.com/sharedfiles/filedetails/?id=518436393
http://steamcommunity.com/sharedfiles/filedetails/?id=518436423
Somehow a guy named Entity managed to shave this down to 133 cycles. Crazy, I know.
Again, I used 3 cores.
EDIT: I got it down from 146 cycles to 133 using the "sub" trick from the original neg method.

Signal Multiplexer (176 cycles)
http://steamcommunity.com/sharedfiles/filedetails/?id=510196828
http://steamcommunity.com/sharedfiles/filedetails/?id=510196738

Sequence Generator (86 cycles)
http://steamcommunity.com/sharedfiles/filedetails/?id=510196814
http://steamcommunity.com/sharedfiles/filedetails/?id=510196780

Signal Edge Detector (164 cycles; 3 Core)
http://steamcommunity.com/sharedfiles/filedetails/?id=510800370
http://steamcommunity.com/sharedfiles/filedetails/?id=510800386
The 3 core method is actually better than having 4 cores because of the "Communication Failure" node on the bottom left. This causes a huge bottleneck in the input and output distribution. (If you read the comment on the 3 core picture, switch the appropriate input lines to shave cycles to 164.)

Signal Edge Detector (4 Core)
http://steamcommunity.com/sharedfiles/filedetails/?id=510801210
http://steamcommunity.com/sharedfiles/filedetails/?id=510801240

[Cynthaer]

This is pretty great. After I replicated your solution in Signal Multiplexer, I was able to apply it myself in Sequence Generator and come up with essentially the same solution you did (though I didn't think of moving the SWP commands around to shave that last dozen cycles off).

[skyhrg]

Originally posted by Cognition Attrition:

This is pretty great. After I replicated your solution in Signal Multiplexer, I was able to apply it myself in Sequence Generator and come up with essentially the same solution you did (though I didn't think of moving the SWP commands around to shave that last dozen cycles off).

Thanks!

I'm working on more and it's really fun! I'll be sure to post more good results if I come up with any :)

EDIT: I added one more. Check it out if you want~ :)

[GuiltyBystander]

Those are pretty good. Better than some of mine I think. If you want to compare to some other fast times, check out the unofficial leaderboard on the subreddit. You're just behind the ones there.

https://www.reddit.com/r/tis100/comments/391heb/table_of_lowest_cyclesnodesinstructions/

[skyhrg]

Originally posted by GuiltyBystander:

Those are pretty good. Better than some of mine I think. If you want to compare to some other fast times, check out the unofficial leaderboard on the subreddit. You're just behind the ones there.

https://www.reddit.com/r/tis100/comments/391heb/table_of_lowest_cyclesnodesinstructions/

Yeah, some of those results pretty crazy. I have no idea how they shave off like 40~50 cycles

[Brian]

Originally posted by skyhrg:

Somehow a guy named Entity managed to shave this down to 133 cycles. Crazy, I know.

You are very close to matching the 133 cycle record here; there is just one more simple trick that you need to find. If you make this change to all 3 cores, each one should be able to run with no idle cycles while they are processing inputs. Let me know if you want a bigger hint!

[skyhrg]

Originally posted by Brian:

Originally posted by skyhrg:

Somehow a guy named Entity managed to shave this down to 133 cycles. Crazy, I know.

You are very close to matching the 133 cycle record here; there is just one more simple trick that you need to find. If you make this change to all 3 cores, each one should be able to run with no idle cycles while they are processing inputs. Let me know if you want a bigger hint!

Yeah, I JUST got it. Instead of using "neg" if you just subtract the other value, I get 133 cycles. I've updated the main post :)