|
im playing unavowed. cool v:tm type setting so far
|
# ? Aug 10, 2018 17:55 |
|
|
# ? Jun 8, 2024 01:28 |
|
yeah this game owns. stayed up until 1am blasting through some of the puzzles, looking forward to some of the new mechanics
|
# ? Aug 10, 2018 18:07 |
|
Pardot posted:how is activity measured? number of times you link between nodes + number of times you execute a kill instruction kinda random, but it does do the job of pushing you in a different way from what the cycles and size metrics do.
|
# ? Aug 10, 2018 18:15 |
|
there are tooltips if you hover somewhere. activity is incremented by each link and kill operation. i think abstractly you can imagine it as network impact.
|
# ? Aug 10, 2018 18:15 |
|
i adore this game so far Unknown Network 1 is the first time i've willingly made two solutions, a naive one that works all the time (BFS for any tree) and a bespoke DFS solution that's pretty tops e: here's the bespoke if anyone thinks they can squeeze a few more cycles out of this code:
|
# ? Aug 10, 2018 18:41 |
|
dude that friends list is hell of bare it's no use going up against the histograms it's all about smacking down actual ppl here's my profile put something yospos in the request https://steamcommunity.com/profiles/76561197977504666/
|
# ? Aug 10, 2018 19:16 |
|
awww there's a 1M cycle limit. I wanted to brute force tec exa-blaster™ modem by trying all the possible numbers and going to lunch or something for a few hours
|
# ? Aug 11, 2018 20:36 |
|
Fuzzy Mammal posted:dude that friends list is hell of bare it's no use going up against the histograms it's all about smacking down actual ppl sent, also people can add me with my username to make themselves feel better about their solutions
|
# ? Aug 12, 2018 01:18 |
|
im very sad that the RAND function doesnt seem to work because it was the only graceful way of solving "unknown network 1"
|
# ? Aug 12, 2018 09:32 |
|
uc berkely has broken me *had echinopsis fucked around with this message at 10:34 on Aug 12, 2018 |
# ? Aug 12, 2018 10:20 |
|
echinopsis posted:im very sad that the RAND function doesnt seem to work because it was the only graceful way of solving "unknown network 1" I still haven't had a call to use RAND Unknown Network 1 i ended up doing 2 solutions for, one was a basic naive BFS that could handle any (host format) binary tree. Then for my optimized unrolled brute force solution I saw that in every test run the file was in a leaf node host so i switched to a blunt DFS algo and got good scores id post them but I gather you're through that one anyway
|
# ? Aug 12, 2018 12:37 |
|
i dont know code lingo so no idea what BFS is i wanted to use rand like RAND 800 801 X LINK X and just send as many out as necessary to find the file on that particular run. woud have worked too but do post if you can be bothered. am i super interested in how else these things can be done
|
# ? Aug 12, 2018 13:06 |
|
echinopsis posted:i dont know code lingo so no idea what BFS is Breadth-first and depth-first search, generic tree-searching methods I'll just post both, hang on (edit) i posted the dfs solution I have above so here's the bfs (naive) solution that sucks but is generic???? code:
|
# ? Aug 12, 2018 13:08 |
|
man how am i supposed to indicate the end of a datastream instead of infinitely pausing from copy m Reading through to see how long the file is then transmitting length first is slow. there has to be a better way!
|
# ? Aug 12, 2018 13:13 |
|
Bhodi posted:man how am i supposed to indicate the end of a datastream instead of infinitely pausing from copy m TEST MRD and make sure it never tries to copy m while there is nothing to read (until the end)? and my crappy unknown network 1 solution: code:
|
# ? Aug 12, 2018 13:20 |
|
Bhodi posted:man how am i supposed to indicate the end of a datastream instead of infinitely pausing from copy m you can TEST EOF on the reader but nothin' doing on communications receipt/messaging, i struggled with that too register M is the world's worst mutex or something
|
# ? Aug 12, 2018 13:25 |
|
i like to just send -9999 and hope that whatever real data doesn't have -9999 in it
|
# ? Aug 12, 2018 13:26 |
|
This is for the copy book thing, i wanted to find a way to copy them in parallel, so i need a way to detect when the other end has died from eof. The check mem would work as long as i put the check after the read in the loop... Yeah that might work Just need to send two per book and manually bring back rather than one home base copier like i had orig planned send / look for 9999 was my first thought but it drastically slows down the copy rate i suspect running through and sending a loop val first is faster / same speed. You can also append 9999 to the file then read the whole thing but the receiver still has to test making the read half speed Bhodi fucked around with this message at 13:34 on Aug 12, 2018 |
# ? Aug 12, 2018 13:26 |
|
Bhodi posted:man how am i supposed to indicate the end of a datastream instead of infinitely pausing from copy m couple of options: - just let it pause, and have another exa come along and kill it when it's done. works pretty well if you're using local mode, since the exa that knows when it's run out of data to send is already in the same node. bad if you're trying to optimize activity i guess? - send an eof value at the end. a bit slow, since you need to copy to a register and test if it's the eof value - test mrd, and make sure whatever exa is writing to m always has data in time. fiddly.
|
# ? Aug 12, 2018 13:31 |
|
i had to use NOOP in one solution to time everything right and it felt like using a goto in c++
|
# ? Aug 12, 2018 13:34 |
|
Ciaphas posted:i had to use NOOP in one solution to time everything right and it felt like using a goto in c++ just be glad you never had to use asm::noop in c++ to time everything right
|
# ? Aug 12, 2018 13:53 |
|
blessings: counted
|
# ? Aug 12, 2018 13:54 |
|
Jabor posted:- send an eof value at the end. a bit slow, since you need to copy to a register and test if it's the eof value
|
# ? Aug 12, 2018 17:00 |
|
Zlodo posted:use 0 as the eof value and T as the transfer register on the reading side, this way you can use a FJMP to exit the loop after reading into T without an additional test instruction that's loving genius
|
# ? Aug 12, 2018 17:12 |
|
i've used rand a bit in the battle mode. that's about it though.
|
# ? Aug 12, 2018 18:25 |
|
simulate mode needs to get faster, there's been a few solutions in a row where i have to wait like 5 minutes to simulate all 100 no my solutions are fine gently caress off this is the game's problem
|
# ? Aug 12, 2018 22:17 |
|
Zlodo posted:use 0 as the eof value and T as the transfer register on the reading side, this way you can use a FJMP to exit the loop after reading into T without an additional test instruction only works if 0 isn't a legit value, but yeah if you're copying keywords or non-zero numbers this is ace
|
# ? Aug 13, 2018 00:27 |
|
most of my solutions involve copying M to T.
|
# ? Aug 13, 2018 01:19 |
|
801 cycles on the last challenge (except I have a NOOP somewhere at the beginning executed only once and it shouldn't affect anything but if I remove it it takes like 40 more cycles and I have no idea why )
|
# ? Aug 13, 2018 03:15 |
|
Something I've noticed is only one EXA can use a LINK at a time, and that created some desync that I wasn't expecting
|
# ? Aug 13, 2018 03:19 |
|
John Wilkes Booth posted:Something I've noticed is only one EXA can use a LINK at a time, and that created some desync that I wasn't expecting yeah but the EXA that replicates through the network has already been sent by that point, that NOOP is in an EXA that stays in the starting host
|
# ? Aug 13, 2018 03:25 |
|
it's "random" which exa goes through the link, there's a prng seeded with a bunch of stuff to make it consistent each time possibly that noop is changing the seed used for the randomisation, and giving you a more efficient linking order?
|
# ? Aug 13, 2018 03:28 |
|
Jabor posted:it's "random" which exa goes through the link, there's a prng seeded with a bunch of stuff to make it consistent each time maybe, I'll have to actually compare how the two versions execute in practice
|
# ? Aug 13, 2018 03:30 |
|
huh, a zach interview from gdc 2018 popped up Here's him talking about Infiniminer and Minecraft: https://www.youtube.com/watch?v=iAfcJB1WaDU&t=228s
|
# ? Aug 13, 2018 04:59 |
|
homercles posted:huh, a zach interview from gdc 2018 popped up you just know from watching that he's loving writhing inside over notch
|
# ? Aug 13, 2018 05:04 |
|
Ciaphas posted:you just know from watching that he's loving writhing inside over notch frankly i don't think i'd have handled that snub publicly with as much class Bhodi fucked around with this message at 05:24 on Aug 13, 2018 |
# ? Aug 13, 2018 05:19 |
|
Bhodi posted:tbf i'd be bitter too if someone did a iterative pass over a program i made, sold it for 2.5 billion dollars, then didn't give even 0.002% of the proceeds for me to live comfortably for the rest of my life as thanks for the inspiration oh i'm not ragging on him, i would have committed Actual Crimes if this happened to me basically as sorry as i sometimes feel for notch's evident mental health issues he can also go gently caress himself, the miserable thief
|
# ? Aug 13, 2018 05:25 |
|
Hulllooooooo, a wild Illectro appears at 26:18 to ask a question
|
# ? Aug 13, 2018 05:46 |
|
zach should've turned infiniminer into an actual product instead of abandoning it
|
# ? Aug 13, 2018 06:34 |
|
|
# ? Jun 8, 2024 01:28 |
|
I don't feel sorry for notch at all. He bought literally the most expensive house in Beverly Hills and has over a billion dollars to his name because he didn't give any of it to his employees in Sweden. He could stock the house with drugs and booze and jet skis and sports cars, pay four guys a quarter million a year each to be his permanent entourage and bring over girls and throw parties, and live the rest of his life in a hedonistic orgy that would embarrass a Roman Emperor, dying with hundreds of millions still to go. But instead he just sits in the dark on his computer in his giant empty house and makes misogynistic posts on Twitter all day, because he's a piece of poo poo. Sagebrush fucked around with this message at 06:59 on Aug 13, 2018 |
# ? Aug 13, 2018 06:57 |