Compacting and Optimisating - Shapez 2 MAM Build EP08
ฝัง
- เผยแพร่เมื่อ 18 ก.ย. 2024
- In this episode of our Shapez 2 MAM build, after realising our MAM's performance is extremely slow, we assess how we can redesign some sections to be more space efficient. In this episode we re-order out operations to minimise fluid input and optimise shape selection
FULL PLAYLIST: • Shapez 2 - MaM Build
BLUEPRINTS: drive.google.c...
![วันเกิดเธอ (TO YOU) - PHUWIN [ OFFICIAL MV ]](http://i.ytimg.com/vi/_ROx1Bj3T0Y/mqdefault.jpg)
![[FULL EP.16] เซียนพาลุย "เผือก-ฟรอยด์" กรี๊ดลั่น จนปูหนี | เฮ็ดอย่างเซียนหรั่ง | One Playground](http://i.ytimg.com/vi/FsqUvwS3b-Y/mqdefault.jpg)
'someone said I talk to much' I like your talking, it shows your thinking of the whole process. And when you say things like 'I lost my brain' it make me chuckle - never change who you are it's what makes your videos personal to you 🙂
I think he zooms and moves the camera too much, gives the impression he's impatient and just wants to move on. I get he's probably trying to reduce 'dead air' but he's made a lot of mistakes in the past that way, just zipping around the camera instead of looking at what he's just built for 15-30 seconds.
The shape codes are:
R = Rectangle
C = Circle
S = Star
W = Wedge
R = Rectangle
C = Circle
S = Star
W = Wumbo
Just wanted to say I'm very new to this genre and only about 25 hours into this game and I'm loving this series. It's fascinating hearing you talk through the logic and I love seeing you redesign and rebuild. I can't wait to see it all come together!
Thanks! Hope you're enjoying your playthrough!
Allan, Thanks for the shout out. There's a huge flaw in your design though: You have a 10k l/m space belt that you think is feeding twelve 7.2k pipes... BUT each pipe is going through a single 1800 l/m fluid launcher at the door. That is your bottleneck. That is your choke point. On your shape asteroids, in the extractor, there are 4 pumps that feed the shape cable (aka belt). On your platforms, you need 4 fluid pumps (aka launchers) to feed the fluid cable (aka pipe). Each fluid launcher hop in the room just reiterates this choke point. Band the 4 launchers into one pipe and then route them up a floor just like what you do when not duplicating logic circuitry and tracing wires. 3 pipes will pass through 21k l/m, and you can "peel" off any paint you need inside the room.
THIS IS NOT A SYMETRIC BUILD. THIS IS NOT A SINGLE FLOOR TESSELATION. *Don't use hoppers*. Route Blue pipes on the third floor. Green pipes on the 2nd floor, and red pipes on the first floor. Move up and down accordingly. On the third-floor, route the green and red pipes along the south wall and give yourself as much room to work with blue as possible. On the First-floor, green and blue go along the north wall.
Mixers have 2 inputs: Blue always goes top. Red always goes bottom. Green connects to both inputs but is gated for bottom if the color is cyan. (yellow top, white top)
The final edits of deleting pipes on later platforms is bad. It should always be 12 lanes banded into 3 pipes in, and 12 lanes/3 pipes out.
Let me type this out to get it right in my head...
A space pipe can carry 10800L x 12 right which is 6 x 12 x 1.8L launchers worth.
Whenever you bring a space pipe into a platform you only bring in 1 x 12 x 1.8L of fluid. Hence. One space pipe can support 6 platforms worth of operations...
In my mixer design those 4 x 3/3 platforms are basically 1 platform... so it would be using a total of 12x1.8L worth of throughput. Where's my math wrong?
@@DataEngineerPlays i hope this is what he was getting at, but your math is correct, but with what you have as a setup right now, it would run slow because of all the hoppers. I pretty much rebuilt this setup, and it only ran at like 90-95%% because of the time lost from the liquid launcher buffers.
@@DataEngineerPlays Each fluid launcher is only 1 pump in the extractor. It's not the full lane passing items at the same speed like belts. Each fluid launcher connection handles 1 dot of paint on the space pipe, not the full column of paint. You're trying to cram an ocean of paint through a squirt gun. You're draining a swimming pool through a pinhole. Draining a swimming pool (space pipe) through the pin hole (launcher) into the storm drain (pipe) doesn't make it drain any faster than the pin hole will allow.
It's not a capacity problem. It's irrelevant that the swimming pool can water 6 lawns. It's a throughput problem. Need more holes to drain through.
@@DataEngineerPlays "Where's my math wrong?" Wrong metric. You're calculating Volume, capacity, Storage. I'm talking about Flow, rate of change, Bandwidth. You're figuring out how big the bear is. I'm saying how fast your slowest friend is.
@@WAF74 I'm sure you're right as always. I need some time for my brain to digest what you're telling me. Stay tuned in a couple of episodes when I speed test my MAM and recall this explanation and go oh right WAF did tell me about liquid throughput issues xD
Another great video!
At (21:11 - 21:13) wire was disconnected from second layer because they behave like conveyors when crossing layers they can't be split to the side only vertically
Yes I am silly sometimes
I would be lying if I didn't shout a few times during this video. Nevertheless it was pretty entertaining :)
The very intensive job you had to do on the pipes/wires to be working on mixing box 2, 3 and 4, could easily have been ignored, if you just shifted the pipes either up and down (color dependent) at the far left end on box 1, so it would match the "untouched" input to the next box. No need to be re-piping or re-cabling every next box.
My next "brainfart" was..ok, you are making 3 pipes of a specific-on-demand color pr box. Why not just bring in 3 pipes of each color pr box and do the logic pipe switching that way. Premix all in advance (saves time) and probably 80% of the space you just used. Hope this helps.
You have many valid points!
You are processing 3 belts of full shape. you have 3 layers available. 3 = 3. Filter, Paint/bypass and Cut a single belt per layer.
Painting a single shape would only take one pair of mixers per secondary color and plus one pair for white. With good piping you can just feed all mixers and only gate the mixers output into a single pipe for the painters. (per layer)
Remember, mixers only consumes half of it's inputs, so a single jumper can feed one side of a pair of mixers.
btw, Good show. You are the most entertaining factory fumbler around. CW vs CCW, losing your brain at least once per episode, can't hold shift key, it's so bad it's great. Hah
Glad you're enjoying! Cool tip by the way. I've said in my videos many times that my commentors are all more clever than me and I love that
Looks good. Much neater than before. It's going to be interesting to see everything connected.
I don't quite understand why mixing the colours on demand was the better choice here, rather than just picking off a colour for each corner from your already existing space belts of every colour in the same reformatted way you did the shape corner picker
I think he wants to minimise the input
Yes that was the point, minimising the input. But also, I thought it was a cool concept so just wanted to test it out. It's good to have options. Who knows maybe later we will go back to premixing if dynamic mixing becomes a bottleneck
Just thought id let you know there’s settings on some for the logic gates so instead of doing an equal gate and then a not gate you can change the setting of the equal gate to just be a not equal gate. Hope that helps :D
Nice job!
I haven't noticed earlier also why the order of operations is important.
Now I think the best option is to build each quarter isolated and separately literally all 4 layers then use the switcher (idk if they break the crystal) to merge them.
I guess if you're building in quarters you can also optimize and build 2 quarters at the same time in oposite corners.
Yes and you can use a 3/4 mould for crystal making that way you can make multilayer multicolor crystal shapes!
@@DataEngineerPlays that's what I was thinking; using any unused shape as a filler to then cut out xD
Always feed your RGB to all the mixers and the the throughput line.
Shut down everything after the mixer and the main throughput.
If you ARE the color I need. Allow to pass.
For anything shape, paint or process to - Always come through but if your not what I need. Stop at this point.
It would be easier.
Think about it like this. Your not letting things through.
Your stopping what you don't want and you can get away with 1 not gate.
at 1:27:09 about switching the pipes on the second platform. You know that one pipe has 4x throughput than the jumpers and the gates? You could just have all jumpers connect to one pipe going across the platform and the gate at this single pipe.
20:00 - look right in front of the jumper; you can put trash and the divider facing left; the circuit reaches every layer and the middle road becomes straight.
Interesting
for the shape codes the r is probably rectangle and the w is probably a different shape from the first game called a windmill which they just changed to an emerald edit: nvm the w is for wedge
for the quadrant selectors, you coulve removed the jumpers to the bypass, this way you could still route belt 2 to slot 1, 3>2 4>3, and keep the modularity.
the "first platform" couldve been split up in two (one logic, one selector. this way the blueprint couldve been as simple as possible and keep it's modularity too, where only the logic foundation needed to change
Do the fluid jumpers have the same throughput as the mixers? If they are lower throughput, then it may be worth having multiple jumpers per pipe. For example (using made up numbers), if the mixing array needs to take in 300 units of paint, and the jumpers only jump 100 units at a time, then splitting the pipe into 3 jumpers, then re-merging on the other side ensures that the mixers are being supplied a steady 300 units without interruption.
So this is the episode
I thought pipe throughput allowed for 4 inputs to be merged into 1 pipe, so instead of using 3 levels you can take your 3 inputs, merge all to the bottom level, then split them back out to 3 outputs at the end.
In your final color picker, you have missed the red unmixed color in the output for the top left square.
Thanks I'll check it out
Feel like you forgot you could just have combined the pipes together? Could has just ran a single pipe across then split out to 4 launchers at the end of each platform.
Probably, it's hard to remember which pipes you can mix. eg. I could of mixed the RGB because only 1 of those 3 would ever be selected
Am I crazy or did you do the color picking logic for each color and then do it again to mix? Couldn’t you just use the pipes from the first logic check, and then create the if logic for white?
yes, yes I did
nice game
Ya so small. 🌟🌟🌟🌟🌟💖
You're doing way too much with your initial logic. Blue either goes forward, or it goes to the alternate path. The logic later will decide if it's still being used or not.
And blue jumps twice to get over the green pipes. Jump it once and then do your logic.
Now I think about it. You're right, the first logic section doesn't need to check for whether blue needs to be a mixed color or not because the later logic gates do that anyway!
Why R is Square? i think it's referring to Rectangle rather than Square 😅
Haha I know its super confusing
Why dont you input all 7 colors with 3 mixed space pipes per layer to your MAM? That color mixing inside MaM is cool idea and fun to play but it gonna slow down your MaM. I am bringing 6 pipes for Main colors + Yellow and 4 pipes for other 3 colors with 3 space pipes to my MaM. Beacuse of space pipe capacity decreased to 10.8k, yes I am short on other colors in case all 4 quarters of layer has same color but i did not input 4th mixed space pipe for the sake of symmetry of my MAM. It is rare possibility anyway:)
You're right it was an exercise just to see if it was possible and yes it does slow down your MAM rather than pre mixing. But it's not the bottleneck at the moment. If it does become a bottleneck one day, I'll switch back to preloading 7 colors, its good to understand different approaches to the problem so you can decide on the best approach!
@@DataEngineerPlays you r right, you need to try and observe if its working or not. Most of other youtubers do not share experimentations and their fails. Like your aproach.
28:24 Well, I haven’t watched a MAM video but I cheated anyway 😂... I build one on Shapez 1, a big part is the similar
Mam is easy but your not dum
Username checks out... DataEngineerPlays instead of DataEngineerEngineeringData. Truth table, man, if you want to build the logic!
A color unmixer (like the unstacker) would be nice - input a color and it outputs the rgb components. So we need to build one as part of the 'brain'.
The brain has one input (the color of the current layer+quadrant) and three outputs (control signals for the r, g, b pipes). Each output can have one of three values (null, 0, or 1). Null = the base color is not used - both gates are blocked (no painting, no mixing). 0 = paint using this base color (paint gate open; mix gate closed). 1 = paint using a mixed color that needs the base color (paint gate closed; mix gate open).
in color -> RGB components-> Result -> outR, outG, outB
u -> 000 -> the shape bypasses the painting station -> null, null, null
b -> 001 -> use blue paint (no mixing) -> null, null, 0
g -> 010 -> use green paint (no mixing) -> null, 0, null
c -> 011 -> mix G+B, use resulting paint = cyan -> null, 1, 1
r -> 100 -> use red paint (no mixing) -> 0, null, null
m -> 101 -> mix R+B, use resulting paint = magenta -> 1, null, 1
y -> 110 -> mix R+G, use resulting paint = yellow -> 1, 1, null
w -> 111 -> mix R+G+B, use resulting paint = white -> 1, 1, 1
no. of RGB components used - > Action
0 -> shape belts go around the painting station
1 -> paint using a base color; no mixing station used, send the paint directly to painter
2 -> mix two base color; use one mixing station; send the result to painter
3 -> mix all base colors; use two mixing stations; send the result to painter
Good luck!
That's amazing! I was never good at binary maths. Unfortunately I'm at data engineer with very rudimentary software engineering experience, I've never leveraged the power of bit operations.