Chat:World/2021-08-01
srinivasTheDeveloper: https://www.codingame.com/clashofcode/clash/18937070d4951513b610bf8cb5127c9d9f0cc28
srinivasTheDeveloper: https://www.codingame.com/clashofcode/clash/18937070d4951513b610bf8cb5127c9d9f0cc28
Bentel_Inside: reeeeeeeeeeeeeeeeeeeeee
Bentel_Inside: :heart_eyes:
aetrnm: Moderators should reach an agreement on whether duplicate versions of a puzzle can be accepted or not.
aetrnm: Is there a way to call moderators on a CoC?
aetrnm: Just check it: https://www.codingame.com/contribute/view/7170e2aa0081c811d945e5e90a5d08c92a2c and this https://www.codingame.com/contribute/view/71972d794a714711601e2052f8325ceb5f51
eulerscheZahl: duplicates in clashes are allowed
eulerscheZahl: CoC close duplicates are not forbidden. https://www.codingame.com/playgrounds/40701/help-center/contribution-guidelines
aetrnm: It's up to moderators, as I said before
aetrnm: ok then
eulerscheZahl: it's not up to moderators. it's an official statement from CodinGame that duplicate clashes are explicitly allowed
aetrnm: Moderators should reach an agreement on whether duplicate versions of a puzzle can be accepted or not.
**eulerscheZahl stopped moderating because I personally disagree on this rule
aetrnm: Haha, ok
dbdr: this statement says it's up to moderators, no?
aetrnm: yeah it does
eulerscheZahl: wait, did this get changed? i could swear this rule was stronger in the past
eulerscheZahl: However, moderators should reach an agreement on whether a trivial CoC contribution is worth being added to the pool of CoCs.
that's new too
eulerscheZahl: maybe after Boulet created that "add nothing" clash
eulerscheZahl: just to prove a point
dbdr: :D
eulerscheZahl: sounds very much like him, doesn't it? :D
dbdr: imo a few clashes with the same task expressed in very different words is not useless, readhing comprehension is important
dbdr: yep :)
eulerscheZahl: also spelling abilities
juice0: lol
juice0: gotem
dbdr: for spelling abilities, seed FB/sogeti ;)
dbdr: *see
eulerscheZahl: why? did CG sneak in a few typos there?
**dbdr hints there is a joke in there
eulerscheZahl: because I didn't register and can't see it?
eulerscheZahl: aaaah
eulerscheZahl: spell
dbdr: :+1:
EdwinJagger: Is there a way to turn off the autocomplete suggestion box in the editor?
eulerscheZahl: no
masongtxr: hi
martinpapa69: any1, who is good at running sse/avx code on codingame here?
dbdr: what's the question martinpapa69?
martinpapa69: how do i get rid of this kind of errors, when i run avx code directly on cg :
martinpapa69: error: inlining failed to call always_inline
martinpapa69: error: inlining failed in call to ‘always_inline’ ‘float _cvtsh_ss(short unsigned int)’: target specific option mismatch
dbdr: seems C/C++ specific maybe pragmas are needed
darkhorse64: I already got this error with gcc when trying to setup a constant simd value using an intrinsic function eg static __m256i s= _mm256_set_epi64x (a, b, c, d);*
darkhorse64: This is OK with MSVC
martinpapa69: yep. works with clang too
darkhorse64: I guess the compiler is unable to fill the SIMD register using the intrinsic during compilation and complains
martinpapa69: gcc is lame
martinpapa69: ok i needed the #pragma GCC target("f16c")
darkhorse64: half precision weights ? :slight_smile:
martinpapa69: ye. I wanted to make an alpha_zero, that i can run on CG without precompilating, and compressing it on my pc
jacek: is it faster?
jacek: like quantization?
martinpapa69: nah, i think its a little slower actually. but you cant submit precompiled code on contests, because of obfuscation rules
martinpapa69: I planned to use it on the current contest, but it took way too much to build the framework, no time for the contest itself
jacek: the spaghetti one?
martinpapa69: ye i wanted to give it a try
struct: darkhorse64 I think I found how to do it during compilation
struct: constexpr __m256i a = (__m256){1,2,3,4};
struct: for msvc the (__m256i) is not needed
struct: I mean it doesnt work for mscvc
struct: msvc*
struct: http://chat.codingame.com/pastebin/35d45563-56aa-4a77-827a-75105622bead
darkhorse64: nice
MSmits: sigh.... so I was trying this new nimsolver method that makes cutoffs like abpruning. From a paper
MSmits: I test it, for some reason it's 5x slower
MSmits: so i thought it was a bug
MSmits: then I looked closely in the paper and it turns out the pruning method becomes worse after around 200 TT-size
MSmits: from their test
MSmits: who has a TT smaller than 200 ??
MSmits: paper from 2018 too
struct: lol
MSmits: i guess the 200 is for small solvable games, but for larger games they have a TT limit too. Since I use a TT of 400k or so, it's never gonna be worth it i think
MSmits: I guess it's similar to the effect of ab pruning weakening when you add TT. You get "lower" and "upper" limits instead of exact values when you cutoff. Except with nimsolver you don't even have these
MSmits: any cutoff prevents an entry to TT
MSmits: maybe I could adapt their method to make it better
struct: cache misses will always be a problem
MSmits: dont think it's that
struct: how big is each entry?
MSmits: tiny
MSmits: the smallest possible TT
MSmits: TT entry i mean
MSmits: it has nothing to do with cache really
struct: Im not really familiar with TT
MSmits: it has to do with storing stuff in TT or not
struct: ah ok
MSmits: when you do a cutoff halfway through the function, you save having to do the function
MSmits: but you dont store the function result in the TT
MSmits: so next time you're in the same state
MSmits: you again have to do half the function to get the same cutoff
MSmits: if you had finished it the first time
MSmits: you would have a TT entry and immediately leave the function on the next time
MSmits: so basically, i have to somehow store the half-finished work in the TT too
MSmits: so i can resume the next time i hit this state
MSmits: in ab pruning + TT this is akin to storing upper and lower limits
struct: you store, but you dont have to access it?
MSmits: a transposition table basically gives you the immediate value of a state
MSmits: if you look it up
MSmits: but the state has to actually be in there, or it won't help you
MSmits: so if it's not in there, you have the analyze the entire state
MSmits: if, halfwaythrough the analysis, you realize this state is pointless to finish analyzing. You save time
struct: yeah, but checking 500k entries might be slow no?
MSmits: you dont check 500k entries
MSmits: it's a hashtable
MSmits: like unordered_set or map
MSmits: or set() in python
MSmits: or dictionary in c#
MSmits: key-value pair
MSmits: key = hashed state
struct: I see
MSmits: so anyways, like ab-pruning causes early cutoff, this method does too. Which means the state doesn't get fully analyzed and doesn't enter the TT
struct: are you using unordered_map?
MSmits: no a much faster way
struct: custom?
MSmits: yeah, but very simple
struct: buckets are fixed?
struct: You dont move stuff from buckets right?
MSmits: no
MSmits: i am not sure there are buckets even, if you could call them that
MSmits: it's just an array
MSmits: i make a hashkey
MSmits: then do % HASH_CAPACITY
MSmits: then if that spot is already taken i look within 5 entries to find an empty spot
MSmits: if i can't find it, i randomly put it in around those 5
MSmits: overwriting previous entry
MSmits: when looking stuff up, i check if an entry is already filled by a different state and then also check 5 items to see if it got put somewhere else
MSmits: it's just an easy way to solve most collisions
struct: I see
struct: I once tried to implement a structure like map
struct: Its not easy
MSmits: not if you're not allowing any loss
MSmits: if it can be lossy, it's easy to do
MSmits: for a negamax with ab pruning + TT, lossy is good
MSmits: lossy makes it faster
MSmits: sometimes you have to recalc something due to a collision, but they are rare enough that the performance increase compensates for it
jacek: :drooling_face:
darkhorse64: still very far from the 4M rollouts in C4 but this version looks a bit faster
struct: I would not stress about it
struct: it performs the same
struct: very rps
darkhorse64: yep same code, two submits end with 2 ELO points diff
jacek: ELO?
jacek: thats not much
darkhorse64: but every optimization bit is worth trying. Maybe I should say TrueSkill points
jacek: phew, thats better
struct: also dont forget that my rollout might be a little less smart than yours
struct: ill submit my bot with 1/10 of the time again, just to see where it places
geppoz: in a CoC statement you find: "rounded down to the nearest integer"
geppoz: do you know any more confused definition ?
geppoz: "rounded" "down" "nearest"
geppoz: loooool
eulerscheZahl: i don't find this confusing
eulerscheZahl: "rounded" => how the rounding shall work
eulerscheZahl: "down" => ok
eulerscheZahl: "nearest": don't go in steps of 0.1 or 10
eulerscheZahl: of course one could just say floor
geppoz: in a "fastest" CoC, of course as you read "rounded", you are already reading the next line ;)
eulerscheZahl: of, you are cegpoz
geppoz: so you spend 2 min wondering how you fail tests, then you see it :D
Astrobytes: 'cegpoz' lol
KiwiTae: :joy:
jacek: oO
dbdr: > The Danish Meteorological Institute reported temperatures of more than 20 degrees Celsius, more than twice the normal average summer temperature, in northern Greenland.
dbdr: twice the temperature? :thinking:
jacek: 0C * 2 is 0C, phew
dbdr: 0C * 2 is 273.15C :)
eulerscheZahl: doesn't even make sense when we assume that the initial sentence was in fahrenheit
dbdr: it's probably 10C -> 20C, just very incorrect wording
eulerscheZahl: yes, makes little sense to multiply values with an arbitrary definition of 0
eulerscheZahl: in German the word "Quantensprung" (quantum jump) is popular to describe some great change or progress made. while the actual meaning is the exact opposite
dbdr: it might make sense in K. twice the kelvins means two times more heat energy? a physicist can confirm?
eulerscheZahl: not when there are aggregate changes involved
jacek: if we only had a physicist in the chat...
eulerscheZahl: "state of matter" shouldn't try to translate 1:1
dbdr: "quantum leap" is the en expression
eulerscheZahl: i looked that one up :/ https://en.wikipedia.org/wiki/Quantum_jump
dbdr: well, quantum physics was a big leap in knowledge ;)
dbdr: oh ok, but that's the scientific expression, right?
eulerscheZahl: yes
dbdr: I hear quantum leap in the vernacular meaning of a big jump
eulerscheZahl: in german we (at least the less knowledgeable of us :P ) use the scientific word to describe the exact opposite
jacek: schmetterling!
eulerscheZahl: i don't find anything odd about that word
jacek: you hear german all you life, its not odd to you
jacek: https://i.imgur.com/CBYMf0x.png
jsonwebdev: Is it possible to turn off getting any of the shortest clashes? I personally don't find them useful
Xascoria: only if private clash i think
KiwiTae: jsonwebdev other clashes are useful?
FalINTOblivion0112: i like shortest mode
jacek: oh my
jsonwebdev: yeah I don't like shortest mode because it's not per programing languae... some programming languages have features that others don't
Saint_Rose: Agreed. Plus i think its great, cause we can learn on how to make our code more efficient.
FalINTOblivion0112: i always use python or ruby when doing shortest
martinpapa69: what could be the worst lang for that? c# i guess ?
AllYourTrees: i enjoy shortest mode
AllYourTrees: helps me learn things about python i never knew about!
eulerscheZahl: with C#9 this will change martinpapa69
eulerscheZahl: i would say Java
FalINTOblivion0112: yes
martinpapa69: Java is also bad for your mental health
eulerscheZahl: as long as it doesn't harm my dental health
PatrickMcGinnisII: eulerscheZahl good catch on the conjecture... i'll remove it...wanted to make sure I gave you props first
eulerscheZahl: sorry for the bad news though
eulerscheZahl: hope it didn't take you too long to create it
PatrickMcGinnisII: it's all good. watched a veritasium video... thought it would make a puzzle. I musta forgot I solved the other puzzle b4. Nah, good practice for some better ideas tho. ;)
FalINTOblivion0112: the collatz conjecture one?
eulerscheZahl: yes
FalINTOblivion0112: I also made a contribution on that
AllYourTrees: just watched that too
PatrickMcGinnisII: yeah it has like 5 differrent names
AllYourTrees: great video
FalINTOblivion0112: if someone can upvote any of my contributions that would make my day
FalINTOblivion0112: because i need to level up a skill tree
AllYourTrees: FalINTOblivion0112 how the heck did you get code size 30 in this clash
PatrickMcGinnisII: I'll drop it now
eulerscheZahl: https://eulerschezahl.herokuapp.com/codingame/puzzles/?q=collatz&category=ANY&title=on&statement=on&tests=on&comments=on&tags=on&author=on that's how I found it
AllYourTrees: im sitting at 51 :scream:
eulerscheZahl: what was the task?
eulerscheZahl: "if someone can upvote any of my contributions" these statements always motivate me to downvote
FalINTOblivion0112: get 5 upvotes on your contributions
FalINTOblivion0112: its for collaboration gold level
jsonwebdev: Meh I just think shortest overall is a bad idea. It rewards unreadable code, which is very bad in real environments
eulerscheZahl: make a contribution worthy of upvotes and you'll get the achievement with ease
FalINTOblivion0112: well i made a quadratic formula one
AllYourTrees: not so sure about that, for instance i just learned about the built in python function str.title(), id argue that is way more readable than the solution i came up with lol
FalINTOblivion0112: yes
FalINTOblivion0112: .title()
FalINTOblivion0112: python is so broken at shortest mode
eulerscheZahl: what does that do? something like capitalize()?
FalINTOblivion0112: the python devs should nerf it
FalINTOblivion0112: it capitalizes every first letter of every word
AllYourTrees: credit to ^ for teaching me
FalINTOblivion0112: no problem
FalINTOblivion0112: there is legit a entire python library for algebra
FalINTOblivion0112: remember how i solved the equation clash under 30 sec
FalINTOblivion0112: thats how i used the library
Anduril: https://www.codingame.com/clashofcode/clash/1894487439fba99c189dfdc18ba15d49663b5c3
jacek: :upside_down:
Kittim: hey folks, how do you get better at python 3
Neehier: practice makes perfect brother
EdwinJagger: Is there a way to turn off the auto complete suggestion prompt in the editor?
oxydowe: I also would like to turn it off!
jacek: no, there isnt ~
Uljahn: yes, there is, but it's kinda hacky
Uljahn: you should block lcp-community.codingame.com host using umatrix addon or hosts file
reCurse: [2 months later] "Why is this feature not working?" :P
oxydowe: messing with hosts files? LGTM
Astrobytes: Use an external IDE with one of the sync plugins
BlaiseEbuth: Use notepad. No autocompletion, no bullshits.
reCurse: Why not just use a hex editor and put in the assembly directly, no bullshit
jacek: oO
BlaiseEbuth: Why not, hey. But perhaps a little tricky for beginners...
Husoski: Machine code would be entered in hex. Assembly needs an assembler.
jstl3ARNINGKNWLE: hello im new to this but i having trouble understanding why the easy and optimal angle test are not working can i get some help
jstl3ARNINGKNWLE: and this is for the Power of Thor
martinpapa69: ye assambly is a bloatware
Husoski: Expecially the way I code it... !!!
reCurse: Sure, opcodes
Husoski: I use those curious 20th century things called "comments"...lots of them.
reCurse: Where I'm from it's called "bloat"
Husoski: Yep. 21st century, alrightie
Husoski: PS: In olden times, "bloat" referred to the size of the object program, not the source...
theycallmedavid: I remember that once a puzzle is solved it could be shared. But now I don't see the option. Did that change?
theycallmedavid: I mean share the solution
jacek: huh? dbdr is from france?
theycallmedavid: I found the button
theycallmedavid: No, that's just social media share
jacek: you mean puzzle or clash
theycallmedavid: Puzzle
jacek: in profile settings there is auto publish solutions
theycallmedavid: I see. That is turned on. Thanks
Husoski: you can publish without auto-publish
Husoski: @jacek Can auto-publish be turned off once it's turned on?
jacek: its on by default
jacek: so you cant turn it off ~
jacek: can*
jacek: argh
Husoski: It's off for me and I don't recall turning it off. Oh well.
Husoski: Thanks, btw. Appreciate the info.
Husoski: Okay, now I see. In "profile settings" just like you said! (I only remembered seeing that setting in the Results tab after submitting a solution.)
jacek: :tada:
FalINTOblivion0112: recurse has not played a single game of clash of code?
jacek: oh my!
reCurse: If I waste my time I usually try to get something out of it :P
Illedan: Never gamed something repetitive?
Illedan: :tada:
martinpapa69: good old metin2 days
Illedan: Soon Legend :) :dance:
KiwiTae: oh i played that tiii martinpapa69
martinpapa69: metin2 is basically the definition of a repetitive game
KiwiTae: i even got married in that game :joy:
martinpapa69: so lucky
BlaiseEbuth: Woa! Kiwi python furious ninja :o https://chadok.info/codingame/leaderboard_languages.html
KiwiTae: ><
KiwiTae: its on puzzle solving?
BlaiseEbuth: Yup
KiwiTae: I been on a row solving all the easy ones since friday lol
KiwiTae: im almost at 80%
BlaiseEbuth: Well 203 puzzles/week is a pretty good yield.
KiwiTae: its this week?
KiwiTae: lol
KiwiTae: no way
BlaiseEbuth: Don't know exactly, but this leaderboard is only a few hours old, and you're the only one with this progression ^^
KiwiTae: hehe
Illedan: Promotion to Legend League in : 05H 55MN 23SC -.-
KiwiTae: gg Illedan which one?
Illedan: FB
KiwiTae: hoho
BlaiseEbuth: :santa:
KiwiTae: with NN?
Illedan: 3 points above gold boss :P
Illedan: nah
Illedan: Just moving and flipp spell.
AntiSquid: let's clash ! yah ! Automaton2000
Automaton2000: or does it depend on the game
martinpapa69: is AutomatonNN still alive ?
martinpapa69: or is he called AutomatoNN?
struct: NN is not available
AntiSquid: it dieded long ago, why no fix
BlaiseEbuth: The toad's lobby use its power to keep him dead.
AntiSquid: shit . now i have clash on my landing page's activity board :/
AntiSquid: how do i get rid of it ...
BlaiseEbuth: Do 4 other things to shift clashs out :p
AntiSquid: it doesn't shift it
AntiSquid: ok it did now ... that gave me a heart attack
BlaiseEbuth: :disappointed_relieved:
FalINTOblivion0112: you can swear in chat
FalINTOblivion0112: interesting
FalINTOblivion0112: new moderators in codingame?
Astrobytes: you can swear in chat contextually
Astrobytes: preferably when there are no minors present
tommaso.castello: can anyone recommend me a fun challenge for bot programming?
Astrobytes: Illedan: gratz dude!
FalINTOblivion0112: fuck yes astrobytes
FalINTOblivion0112: i can swear now *when no minors present*
Astrobytes: tommaso.castello: it's kinda subjective. You like board games? Physics games? Soemthing inbetween?
tommaso.castello: I loved coders strike back and was looking for a similar one
Astrobytes: FalINTOblivion0112: that was without context, I will kick you next time
FalINTOblivion0112: oh?
FalINTOblivion0112: sry
Astrobytes: tommaso.castello: there's Fantastic Bits, Mean Max, Poker Chip Race, Bit Runner 2048 or Search Race (that one's in Optim games)
tommaso.castello: thanks I'll try them
Astrobytes: enjoy ;)
FalINTOblivion0112: i really like mean max
FalINTOblivion0112: coders strike back is really simple
FalINTOblivion0112: and ghost in the cell is also very fun
FalINTOblivion0112: doing clash of code can also optimize your coding speed
ericlovesmath: Why can't I hover over people to see their info anymore?
FalINTOblivion0112: i dont know, same happened to me
BlaiseEbuth: A CG's technique to avoid you seeing my new bio...
FalINTOblivion0112: uh-
FalINTOblivion0112: but i can still search you up
FalINTOblivion0112: t(ಠ_ಠt)
BlaiseEbuth: Of course, but then YOU searched me. But you can't see it hovering my pic by mistake. :smirk:
AntiSquid: FalINTOblivion0112 i suggest real world work experience, have a good day
FalINTOblivion0112: oh okay
FalINTOblivion0112: thanks
BlaiseEbuth: There's 3 things I don't understand in this sentence: "real world", "work" and "day" Oo
FalINTOblivion0112: ???
AntiSquid: his referencing his elite programmer status
AntiSquid: he's *
FalINTOblivion0112: are you assuming my gender
AntiSquid: wasn't even part of the discussion
FalINTOblivion0112: huh?
AntiSquid: is it relevant ?
FalINTOblivion0112: i am very confused
AntiSquid: about what?
FalINTOblivion0112: whos elite programmer status
BlaiseEbuth: You can't be non-binary on a computer...
FalINTOblivion0112: true
AntiSquid: was answering your ambiguous triple question marks, that was an indirect question asking to clarify blaise's comment @_@
BlaiseEbuth: Questionception!
AntiSquid: my fidget's still spinning
FalINTOblivion0112: ohhhh
FalINTOblivion0112: now i get it
BlaiseEbuth: https://imgur.com/gallery/s44i2t8
BlaiseEbuth: Sorry ^^'
FalINTOblivion0112: hold up-
AntiSquid: https://youtu.be/Hhavsmsi_5M?t=52 this kind of fidget
grateful_tomato: is it just me, or is this puzzle broken? https://www.codingame.com/contribute/view/5665ce9c4cad0e677b11263274dcce354cf2 when I click "test in IDE" I get an error message "couldn't find the question related to this contribution".
BlaiseEbuth: Try it now grateful_tomato
grateful_tomato: thanks, now it works
BlaiseEbuth: If you have the contribution moderation rights, just click 'edit' and then 'submit' without changing anything...
FalINTOblivion0112: this is huge im top 3000 in clash of code
FalINTOblivion0112: i need to get to top 1000
AllYourTrees: gl
FalINTOblivion0112: thank you!
BlaiseEbuth: AllYourLuck
aloui: how i can have the solution ?? :)
BlaiseEbuth: 42
BlaiseEbuth: That's the Solution.
aloui: 42 ?? i search the solution for clash code
BlaiseEbuth: There's no such thing there, I'm afraid...
BlaiseEbuth: My pleasure.
FalINTOblivion0112: 4 2 i s a v e r y s p e c i a l n u m b e r
FalINTOblivion0112: plus you guys google up solutions?
FalINTOblivion0112: huh, doesn't that break the codingame spirit?
BlaiseEbuth: :ghost::hammer:
FalINTOblivion0112: :money_mouth:
Wontonimo: I've released my Connect 4 bot
Wontonimo: it's doing okay so far
PatrickMcGinnisII: oh no
Wontonimo: ?
Wontonimo: ah, I'm ahead of you ;)
PatrickMcGinnisII: i hope so
Wontonimo: why do you say that?
Phoenix.: https://escape.codingame.com/game-session/sjC-LLw-ML5-r7Z