|
thehustler posted:So any resolution will require a major reworking unless I can find space (either a register or another chip or something) to hold the current state. But then it would need to run every tick and not every chip does. Gah! code:Either way, your output-controlling MC also does a lot of unnecessary work, like a "tcp" whose flags result is never significant, and consider if you can't use two registers with "dst" instead of testing the value of one register and then jumping around. If you save enough instructions in your output-controlling MC, you could perhaps add a special command meaning "shut everything off".
|
#
?
Dec 4, 2016 17:36
|
|
|
|
| # ? Apr 26, 2024 16:23 |
|
|
I see that I can save space there. Thanks. I have real trouble unravelling the logic into something clever and always start with a methodical branching system which just eats lines up. Then I forget to fix it  That's some power saved but doesn't solve the problem of where to hold that current state though. I'd have to store it from chip two which has a few lines left...
|
|
#
?
Dec 4, 2016 18:03
|
|
|
zedprime posted:It seems reading on the xbus side clears the IO expander's previous xbus input which means you can stick a not gate on the end and make a volatile flipflop that reinitializes after reading. I bet that's useful... somewhere? VR Buzzer maybe? I've been on a quest to solve it in 2 or maybe even 1 line, I really want to believe it's possible. I've been trying to leverage the PGA and so far it's been a journey to nowhere, but that tip there gives me an idea.
|
|
#
?
Dec 4, 2016 18:40
|
|
|
thehustler posted:I see that I can save space there. Thanks. I have real trouble unravelling the logic into something clever and always start with a methodical branching system which just eats lines up. You're way overthinking it. You need one bitfield worth of state (the current outputs) and one register of temporary storage (for the first value arriving from input reader MC). Here's a very useful sequence for that second MC: mov x0 dat; dst x0 dat; mov acc x3 - you can keep the current outputs state in acc all the time, you never need to use it as temporary storage for anything. And you don't need any conditionals unless you add an extra "command" to that MC, to reset all. nielsm fucked around with this message at 19:07 on Dec 4, 2016 |
|
#
?
Dec 4, 2016 19:05
|
|
|
 That's insane. I get so hung up on bringing in a variable and doing conditionals that I never see an opportunity to just skip that step.
|
|
#
?
Dec 4, 2016 19:34
|
|
|
nielsm posted:Either way, your output-controlling MC also does a lot of unnecessary work, like a "tcp" whose flags result is never significant, and consider if you can't use two registers with "dst" instead of testing the value of one register and then jumping around. If you save enough instructions in your output-controlling MC, you could perhaps add a special command meaning "shut everything off". Yeah, understood. I managed to rework it and it wasn't too big a hint. With these things you're usually half way there already.
|
|
#
?
Dec 4, 2016 20:25
|
|
|
Wow. Considering how hard some of the other puzzles are around it Carbine Target Illuminator was piss easy. It's odd.
|
|
#
?
Dec 5, 2016 22:10
|
|
|
thehustler posted:Wow. Considering how hard some of the other puzzles are around it Carbine Target Illuminator was piss easy. It's odd. I keep optimizing puzzles as goons talk about them and got remote kill switch down to 188 for best power, 9 lines for line count. (Someone got 7  )
|
|
#
?
Dec 5, 2016 22:16
|
|
|
I think it's different for everyone, some of the "hard" puzzles like original sandwich maker i found easy, but i only just realised how to do infrared sensor after the updates
|
|
#
?
Dec 6, 2016 03:22
|
|
|
Ugh bitcoin ATM is kicking my arse. I have an MC6000 to count the money and store it in acc (runs every tick though) and an MC6000 to read the card number and store it in RAM. I need to get this all synced up so that the -1 can get all the info and send it on to the 3rd chip which can send the data, but I'm not sure how to trigger that. If the I/O expander was blocking and only read inputs I think it'd work fine but nope.
|
|
#
?
Dec 7, 2016 00:40
|
|
|
By the time you use RAM, you might as well put everything that needs communicated into it and then a simple memory read program can be triggered by a slx (or embedded in a controller you're already using, reading memory of arbitrary length is like 3 or 4 lines of code?) and do the shuffle out of RAM and into the output.
|
|
#
?
Dec 7, 2016 01:14
|
|
|
zedprime posted:By the time you use RAM, you might as well put everything that needs communicated into it and then a simple memory read program can be triggered by a slx (or embedded in a controller you're already using, reading memory of arbitrary length is like 3 or 4 lines of code?) and do the shuffle out of RAM and into the output. Yeah that was my plan but I ran out of room. My counting chip could send an update of the count each tick to the mag stripe reader and then after a -1 is sent it could write it to ram and minuses itself from itself? But that would need an extra register and I use acc already for the card reader loop. Or it could read an updated count every tick from another chip and send it directly to its slot in RAM... hmm. An extra MC just to get another register is a bitter pill to swallow though.
|
|
#
?
Dec 7, 2016 03:31
|
|
|
No wait I can just read the count each tick in the mag stripe chip I think I have room. The counter is doing the addition already. Ok that's my plan for tomorrow
|
|
#
?
Dec 7, 2016 03:32
|
|
|
Without giving too much away, the most useful feature of RAM is reading the pointer address because it increments on every read or write. In other words its an incremental register if you have the spare ports to connect to the address port.
|
|
#
?
Dec 7, 2016 03:38
|
|
|
Man I wanna know how the hell people have under 150 on remote kill switch. I did find the 7 instruction solution but drat, must be a trick I'm missing.
|
|
#
?
Dec 7, 2016 05:35
|
|
|
zedprime posted:Without giving too much away, the most useful feature of RAM is reading the pointer address because it increments on every read or write. In other words its an incremental register if you have the spare ports to connect to the address port. Can I connect two things to the same side of a RAM chip? Already got my packet sender connected on the other side. I guess I'll have issues with my SLXs if I do. Gah this is the same with every puzzle, always so close to doing it but frustratingly far away. thehustler fucked around with this message at 12:17 on Dec 7, 2016 |
|
#
?
Dec 7, 2016 09:22
|
|
|
You can totally connect two things to the same side of a memory chip, you just have to be careful that they don't step on each other's toes (since they'll be sharing the address pointer on their side of the memory chip) And as you've noted, you can't really "double up" and use that same connection for communicating between the two chips.
|
|
#
?
Dec 7, 2016 09:55
|
|
|
I was so frustrated when I realized the Printer puzzle needs to handle the print head going backwards.
|
|
#
?
Dec 9, 2016 01:28
|
|
|
Guy Axlerod posted:I was so frustrated when I realized the Printer puzzle needs to handle the print head going backwards. I solved that one by looking at the save file of someone who did beat it just to see what they put in a ROM. Knowing just that was enough to figure it out for myself. And like everything it seemed obvious in retrospect.
|
|
#
?
Dec 9, 2016 07:09
|
|
|
I did all but the last puzzle right after launch and took a break. Today re-ran my designs and pretty much everything had scores in the middle or better and I didn't even have to redo the card reader one because I'm not a dirty cheater. But then I found this spaghetti masterpiece: I tried making a better one and welp. Either I'm rusty because of the break or this puzzle is incompatible with my brain because I'm having real trouble with this. I know it's easy and I kinda know what I should do but somehow it's really loving hard. I love this game.
|
|
#
?
Dec 9, 2016 12:54
|
|
|
Gah, still cant do the bitcoin ATM. Nothing syncs up right. I think I need to possibly rethink the loops/timing logic for it all.It's so frustrating since each individual chip seems to do its own task just fine. I did go back and change the loops/logic for a bunch of earlier levels though and improved a ton 
|
|
#
?
Dec 9, 2016 15:36
|
|
|
Elos posted:I did all but the last puzzle right after launch and took a break. Today re-ran my designs and pretty much everything had scores in the middle or better and I didn't even have to redo the card reader one because I'm not a dirty cheater. But then I found this spaghetti masterpiece: I've actually been working through the bonus campaign and finding a number of the puzzles very easy, but for some reason I was never able to make a working solution to that drat scale. I sometimes will open it and stare at it for a while, stick a few parts on... and get absolutely nowhere. It is kind of hilarious since a lot of the harder ones I looked at the specs and was able to write all the code in my head.
|
|
#
?
Dec 10, 2016 18:14
|
|
|
I feel that the fact that 400/50 = 8 is useful in the pollution sensing window puzzle, but I don't know why. I know I will have to use RAM to hold the last 8 values and then have an MC to loop through and do some kind of sum, but I'm not sure exactly WHAT sum I need to do to determine the value. I fear I may have hit my limit. Meat Printer was so very easy as well 
|
|
#
?
Dec 11, 2016 19:56
|
|
|
thehustler posted:I feel that the fact that 400/50 = 8 is useful in the pollution sensing window puzzle, but I don't know why. I know I will have to use RAM to hold the last 8 values and then have an MC to loop through and do some kind of sum, but I'm not sure exactly WHAT sum I need to do to determine the value.
|
|
#
?
Dec 11, 2016 20:12
|
|
|
homeless snail posted:What if I told you, that you didn't need RAM at all. Ok so we're looking at a running count in a register of a chip then
|
|
#
?
Dec 11, 2016 20:15
|
|
|
homeless snail posted:What if I told you, that you didn't need RAM at all. You'd still need a delay line at least?
|
|
#
?
Dec 11, 2016 20:28
|
|
|
If I remember the first number in the last 8 with a pointer on one side of RAM I can subtract it later and add the latest value to the end of the chain. Is that the sliding window thing mentioned previously? The other side of RAM can be offset before reading for the size of the range to check?
|
|
#
?
Dec 11, 2016 21:02
|
|
|
(And then I see some simple transposition shows that I can check for a sum > 399 each tick. Neat.)
|
|
#
?
Dec 11, 2016 21:03
|
|
|
homeless snail posted:What if I told you, that you didn't need RAM at all. I assume llamadeus is someone here, they actually crushed everyone on my list on this one recently. I was able to improve mine to get ahead of the pack but I'm nowhere near him. Jeffrey of YOSPOS fucked around with this message at 21:20 on Dec 11, 2016 |
|
#
?
Dec 11, 2016 21:15
|
|
|
e: quote is not edit
|
|
#
?
Dec 11, 2016 21:19
|
|
|
Jeffrey of YOSPOS posted:uhh....uhh....really? Are you using a bunch of MCs instead? I'm intrigued... Each round is only 2 bits of information, but I think you need at least 14 bits of buffer which is a lot for just MCs.
|
|
#
?
Dec 11, 2016 23:28
|
|
|
http://imgur.com/a/25G8t So close I can almost smell it. Problem: I need to trigger the last chip somehow so that it can read and output the data, but it has to do it in sequence after the keypad receives the -1 but also after the current money total has been sent to RAM. Therefore I think that I need to trigger the sending from the counting chip. Except there's no pins left. Or space, really. What can I do here? I feel I'm so close so I want to make this solution work and then go back to improve it.
|
|
#
?
Dec 12, 2016 23:02
|
|
|
thehustler posted:http://imgur.com/a/25G8t As far as space goes: you're checking dat against all three inputs (001, 010, and 100) in that middle chip. Any way you can do it all at once with one check? tcp dat 10 add 20 -add 30 +add 80
|
|
#
?
Dec 12, 2016 23:14
|
|
|
poo poo, yes! Because before I was sending a command into that pin as well so I needed to explicitly check for them all. But now I don't need to! Still short on board space but that gives me code space to work with. Does that handle the zero values between money inputs as well? Anyway, no, it doesn't handle zero values and adding a check for that would take up lines again.
thehustler fucked around with this message at 23:54 on Dec 12, 2016 |
|
#
?
Dec 12, 2016 23:26
|
|
|
I moved some things around and now it works great! http://imgur.com/a/Zu5dw Managed to do a sneaky send/receive in the card reader chip to send the value, saving me an xbus chip to initiate the sending process. The ocean sensor grid looks very doable. I already have some values going to RAM and I think the sliding window trick I learnt in this thread will help to read those values. Only caveat is that it wants the current one as well but the values get written so early in the step that I don't think that's an issue. Edit: indeed it isn't. My pointer is at the right place and ready to go in each RAM chip. Now to see if I can implement a read/send function in a single MC6000 Edit 2: ahahah massively ran out of lines. OK. this is pretty tough... thehustler fucked around with this message at 01:10 on Dec 13, 2016 |
|
#
?
Dec 13, 2016 00:42
|
|
|
My Aquaponics solution: Had many solutions for this but all ended up one line short because of resetting to 50 on p1. Feels good to find a solution in these pieces. No idea how someone did it for less than this.
|
|
#
?
Dec 13, 2016 19:15
|
|
|
i started a chatroom on SynIRC for talk about cheap virtual electronics chat, join #goontronics and talk about how slx is the worst instruction ever.
|
|
#
?
Dec 14, 2016 03:17
|
|
|
Finished the last puzzle. I generally got frustrated and gave up on the last puzzle in previous games.
|
|
#
?
Dec 14, 2016 23:12
|
|
|
It seems that on some puzzles where you use RAM you can replace using ACC in a loop as a counter with a simple conditional on the address line. That way you need no add 1 and you save a bit of power. Also you gain ACC back. Annoyingly I can't seem to get this working on every puzzle I try it with. Seems to be depending on whether I also need to do something with that number. Should I be able to use similar tricks with every puzzle involving RAM?
|
|
#
?
Dec 17, 2016 21:44
|
|
|
|
| # ? Apr 26, 2024 16:23 |
|
|
In fact is there a repository of standard and useful code fragments that you can incorporate into your designs?
|
|
#
?
Dec 17, 2016 21:46
|
|