00:09 < bridge> money 00:09 < bridge> !sh echo 00:09 < chillerbot12> bash moved to chat.zillyhuhn.com #off-topic 00:10 < bridge> chillerdragon: why does chillerbot have 12 in the name? 00:10 < bridge> is it an easter egg 01:17 < bridge> yo chillerdragon you have a license right? 01:17 < bridge> yo chillerdragon you have a drivers license right? 01:17 < bridge> yo chillerdragon you have a drivers license right? 01:27 < bridge> <.lepinpin> Good idea to recode ddnet in ddnet with quads x) ? 01:47 < bridge> the ddnet map format is not turing complete afaik, so probably no 01:48 < bridge> <.lepinpin> Okey so bad apple ? 01:48 < bridge> <.lepinpin> Or maybe doom 01:49 < bridge> bad apple was already made. doom isn't possible either 02:23 < bridge> with music? 02:23 < bridge> with server mods you should be able to make pretty much anything 02:23 < bridge> (yeah obviously with musc u can play random sounds) 02:25 < bridge> https://www.youtube.com/watch?v=7WPTr4meLIY 02:25 < bridge> this was like a week ago how can you not remember xD 02:25 < bridge> owo 02:25 < bridge> no map download 02:27 < bridge> @essigautomat solly is doubting you 02:46 < bridge> no i just wna play it on my compooper x-x 08:04 < bridge> I want to release it with recorder tiles as fun map in the future 08:07 < bridge> you can almost do that without client changes, but good luck syncing anything 08:42 < bridge> maybe I can send you the demo tho 🤔 08:46 < bridge> maybe I can send you the demo tho 🤔 the demo is flickering and not as performant as ingame rendering 🙁 08:47 < bridge> maybe I can send you the demo tho 🤔 the demo is flickering and not as performant as ingame rendering 🙁 and it desyncs if you stop it 09:08 < bridge> btw I am not sure that doom is impossible 😉 09:51 < bridge> i think ill restart edos, i have tech debt from not knowing common patterns from start, i think i can do a way better job now 09:52 < bridge> so next you are working on edos reloaded 09:53 < bridge> I like doing this with my projects, first attempt is more like a playground where I figure things out 10:18 < bridge> I love the new blazingly fast loading times of nightly :brownbear: 11:03 < bridge> yeah, im also starting with limine now 11:04 < bridge> https://github.com/jasondyoungberg/limine-rust-template 11:05 < bridge> also on september 10 i have time off for 10 days so im gonna use it 11:05 < bridge> looking forward to my time off 11:05 < bridge> :justatest: 11:07 < bridge> nice 11:14 < bridge> I have vacation in 3 weeks, looking forward to it, I need it desperately 11:14 < bridge> I have vacation in 3 days, looking forward to it, I need it desperately 11:23 < bridge> I should read up on this limine thing 11:48 < ws-client> **** @teero777 my driver license status may or may not be confidental. What do you need? 11:49 < ws-client> **** @kollpotato irc increments a number after the name if it is already used (in the same network i think) so i assume my bot bugged and reconnected or something like that. I was worried when it reached 100 11:49 < ws-client> **** @milkeeycat I could meet in game now 11:52 < ws-client> **** @learath2 yes i want to sit down! when do we do it? how do we do it? 11:58 < bridge> y0.1k 11:58 < bridge> @chillerdragon 11:58 < bridge> same thing happened with browsers 11:58 < bridge> `"Chromium";v="139", "Not;A=Brand";v="99"` 12:04 < ws-client> **** same as what 12:05 < bridge> chiller for some reason ur bot constantly reconnects after some time, u can see that in quakenet webchat 12:06 < bridge> i assumed it was irc stuff but when i made my bot it was stable and didnt reconnect at all 12:06 < bridge> so javascript L 12:06 < bridge> or ur raspi has unstable internet 12:14 < ws-client> **** yea i am aware 12:14 < ws-client> **** no idea how to fix it 12:15 < ws-client> **** my webchat irc client is also unstable and it runs on my vps 12:15 < bridge> maybe the irc lib forgets to do ping pongs 12:15 < bridge> but how does it reconnect by itself then 12:16 < bridge> chillerdragon: I still don't like builder functions :/ imo default value consts would be enough 12:18 < bridge> builder pattern sucks 12:18 < bridge> just define everything beforehand smh 12:18 < ws-client> **** @milkeeycat yea default value consts could also work havent thought about it. I think both work the same for me. 12:18 < ws-client> **** go send pr 12:19 < ws-client> **** and replace the existing builder with default const 12:21 < bridge> anyone have expereicne getting windows runners to run 12:21 < bridge> chillerdragon: did you think about consts for messages or specific fields?(I was talking about specific fields) 12:22 < bridge> whats the issue w the runners 12:22 < bridge> just waiting forever 12:22 < bridge> yes, delete the idea, use a linux runner with MinGW instead and test with wine 12:22 < bridge> cant get it to compile 12:22 < bridge> or well cant get rust to accept the .dll 12:23 < bridge> https://github.com/SollyBunny/libccar/actions/runs/17263478129/job/48990122397 12:23 < bridge> (teeros thing) 12:24 < ws-client> **** @milkeeycat oh i thought you mean the entire message. Imo thats the most convienient and the one that actually saves the user time, lines, bugs and typing 12:26 < bridge> ccar.lib is the only thing there with no absolute path 12:26 < bridge> idk what generates that build command since you're using msvc but double check 12:27 < ws-client> **** not ddnet related 12:27 < ws-client> **** ban 12:28 < bridge> that applies to like 50% of the things in here 12:38 < bridge> I already suggested having a developer-off-topic 12:39 < bridge> but bad-apple is on topic 😄 12:46 < bridge> second time destroying @chillerdragon in the PR comments 😄 13:04 < bridge> chillerdragon: lerato said everything developa related can be talked about here 13:15 < bridge> idk what i did 13:15 < bridge> but it works now 13:15 < ws-client> **** @Assa destroy? ^^ 13:16 < bridge> somewhat 13:16 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1410221307441713242/image.png?ex=68b03a70&is=68aee8f0&hm=2a98ddae2d1cf4210221a4cd71b55260a5009b6a7b363b99cc8976ee81f6a7e2& 13:16 < bridge> delete pChillerDragon; 13:16 < ws-client> **** pff 13:16 < bridge> xD 13:16 < ws-client> **** that might cause a use after free 13:18 < bridge> ||_as if you ever get used_|| you make it too easy to answer like that 13:18 < bridge> `free(pChillerDragon); pChillerDragon = NULL`; 13:18 < bridge> `free(pChillerDragon); pChillerDragon = NULL;` 13:18 < bridge> :) 13:18 < ws-client> **** bing bong 13:19 < bridge> sollybunny1 went up in flames 13:19 < bridge> `pChillerDragon = std::make_shared(EDevRole::MAINTAINER);` 13:21 < bridge> imagine chillerdragon is an irc ai made by admins for developers not to feel lonely 13:48 < bridge> Chillerdragon AI manifested into existence from the collective consciousness of Teeworlds players 14:02 < bridge> limine is a bootloader, it supports grub protocols (multiboot 1 and 2) but it also has its own modern protocol. the thing is there is a rust crate for the limine protocol (aka limine boot protocol) which makes it a pleasure to work with in rust, im sure C has the same, the crate is rly easy 14:02 < bridge> limine is a bootloader/menu, it supports grub protocols (multiboot 1 and 2) but it also has its own modern protocol. the thing is there is a rust crate for the limine protocol (aka limine boot protocol) which makes it a pleasure to work with in rust, im sure C has the same, the crate is rly easy 14:03 < bridge> https://en.wikipedia.org/wiki/Intel_5-level_paging 14:03 < bridge> i wonder when this is used 14:05 < bridge> so that's why he is all knowing and all powerful 14:34 < bridge> setting up windows dual boot in limine was easy as hell 14:34 < bridge> Also I could fill my screen with screams and agony when selecting the windows entry 14:34 < bridge> :owo: 14:43 < bridge> :owo: 14:45 < bridge> > limine 14:45 < bridge> noted down ✍️ 14:56 < bridge> oh that is ugly doing this retroactively, you need to install ubuntu on other disks, then run limine and find windows 14:57 < bridge> oh god, looking forward to my 6 disk dual boot setup :pepeW: 14:57 < bridge> huh 14:57 < bridge> I have windows and Linux on the same disk 14:57 < bridge> :justatest: 14:58 < bridge> yeah this is possible, you don't need physicial disks don't worry 14:58 < bridge> it's just my HDD/SSD/NVMESSD is ... spicy 15:02 < bridge> limine wont find windows for u 15:02 < bridge> limine is just a config file 15:02 < bridge> https://wiki.gentoo.org/wiki/Limine#Dual-booting_with_Windows_in_Limine_.28UEFI.29 15:03 < bridge> ``` 15:03 < bridge> /Windows 15:03 < bridge> //Windows Example 15:03 < bridge> protocol: efi 15:03 < bridge> # This tells the efi protocol to call the specified EFI file and load it. 15:03 < bridge> path: boot():/EFI/Microsoft/bootmgfw.efi 15:03 < bridge> comment: Boot Microsoft Windows! 15:03 < bridge> ``` 15:04 < bridge> I didn't even have to "find" windows 15:04 < bridge> 15:04 < bridge> I just installed the bootloader files using a live USB :justatest: 15:07 < bridge> have a cat 15:07 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1410249424575397992/PXL_20250827_130659112.jpg?ex=68b0549f&is=68af031f&hm=033eb678001945d2c2b31a3f873c9fc104ebb3042044d1a9ed307207bc40aeb2& 15:08 < bridge> sir, it's #developer channel 🤓 15:09 < bridge> silly cat 15:10 < bridge> my cat codes in rust 15:14 < bridge> https://milkv.io/jupiter 15:14 < bridge> @tsfreddie buy one for me pls 15:15 < bridge> anyone know big riscv boards with sensible ram? 15:15 < bridge> 15:15 < bridge> ohi can ship with the latest option 15:15 < bridge> €70,95 15:15 < bridge> 8gb ram riscv 15:15 < bridge> not bad 15:15 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1410251450059198484/K1_2048x2048.png?ex=68b05682&is=68af0502&hm=89de0a1f49ff742432f61761dcba699f278e8b4567f8a51b0ba31e73a57c8b78& 15:24 < bridge> Jupstars alter ego mentioned 15:25 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1410253853990981692/Screenshot_20250827-152505.png?ex=68b058c0&is=68af0740&hm=885064ab2d417559420bcba5e744273a14685b3060ac7045b49b47ff0b9fe730& 15:36 < bridge> I wonder if we would improve zooming out on maps, if we start to skip clipped regions that are smaller than a pixel or start to fade away high details 15:36 < bridge> I wonder if we would improve zooming out on maps, if we start to skip clipped regions that are smaller than a pixel ~~or start to fade away high details~~ 15:36 < bridge> probably yes, but I am already 3 steps ahead of my PRs 15:43 < bridge> Render the whole map to a texture, reduce FPS of animations, uhh, some other cursed LoD tricks 16:15 < bridge> Is every tile it's own quad or how does map rendering work at all 16:16 < bridge> Is every tile its own quad or how does map rendering work at all 16:17 < bridge> its beyond our understanding 16:18 < bridge> only the true developers know how all that works 16:18 < bridge> jupstar assa patiga 16:19 < bridge> Yeah 16:19 < bridge> There should a second dev role 16:19 < bridge> "Real Developer" 16:19 < bridge> Lol 16:24 < bridge> @0xdeen can you use in your announcement? `` 16:24 < bridge> ah my bad 16:24 < bridge> wrong timestamp 16:24 < bridge> tf 16:25 < bridge> @0xdeen can you use in your announcement? `` 16:25 < bridge> @0xdeen can you use in your announcement? `` 16:26 < bridge> @0xdeen can you use in your announcement? `` 16:27 < bridge> it also should be `Prize pool` not `Price pool` 16:28 < bridge> That's a funny mistake 16:29 < bridge> prize pool was misspelled quite a lot here 16:30 < bridge> mostly by germans 16:30 < bridge> its the same word in german :justatest: 16:30 < bridge> nah you actually have to pay the admins the amount that is given there. there is no prize 16:30 < bridge> xD 16:31 < bridge> everyone pays 200€ so ddnet servers keep running 16:31 < bridge> it's an involuntary donation if you win the tournament 16:32 < bridge> if everyone payed 200€ then the servers would keep running till 2500 16:33 < bridge> sadly dreams don't come true 16:36 < bridge> the average server costs are ~1731€. **all** active players would mean all active players, currently ~6000. so (6000*200€)/1731 = ~693 years of funding 16:37 < bridge> My playtime is 1k hours 16:37 < bridge> How much is a fair price to pay? 16:37 < bridge> but who is gonna manage all those servers besides The Deen 16:37 < bridge> lets subtract 100yrs of payment because deen has to buy the immortality potion to manage the servers. seems fine 16:38 < bridge> reasonable 16:39 < bridge> DDNet is cheap 16:40 < bridge> i recently found a free ddnet server hosting 16:40 < bridge> its not 24/7 but is good enough for testing 16:41 < bridge> trashmap? 16:41 < bridge> no another one 16:41 < bridge> they host many other game servers 16:41 < bridge> You can also host a server yourself 16:41 < bridge> On a raspberry pi 16:42 < bridge> but i dont have a raspberry pi 16:42 < bridge> Unfortunate 16:42 < bridge> It's an awesome device 16:42 < bridge> i can buy a small vps for 2.5€ a month for a ddnet server 16:43 < bridge> raspberry pi would cost about 50€ 16:43 < bridge> Raspberry pi + HDD is great for torrenting too 16:46 < bridge> ig im not a real develper anymore 16:46 < bridge> real maintainer 16:46 < bridge> afaik there is a tile shader, in vulkan 16:46 < bridge> :justatest: 16:46 < bridge> built-in? 16:47 < bridge> the shader is made by jupstar 16:48 < bridge> Okay 16:48 < bridge> A shader 16:48 < bridge> I have no idea how that works but cool 16:48 < bridge> https://github.com/ddnet/ddnet/blob/master/data/shader/tile.vert 16:48 < bridge> https://github.com/ddnet/ddnet/blob/master/data/shader/tile.frag 16:48 < bridge> https://github.com/ddnet/ddnet/blob/master/data/shader/vulkan/tile.vert 16:52 < bridge> least complex shaders i have ever seen 16:53 < bridge> It's sus 16:53 < bridge> How can it do anything when it doesn't do much? 16:55 < bridge> because ddnet just renders squares from an image 16:55 < bridge> it's simple 16:57 < bridge> And yet zooming out lags 16:57 < bridge> not for me 16:57 < bridge> Supposedly, on any half decent hardware it doesn't lag lol 16:57 < bridge> what are ur specs? 16:59 < bridge> it lags because you are rendering everything and cant ignore stuff 17:00 < bridge> also remember there are animations slowing things down 17:02 < bridge> oh well i always play with entities on 17:02 < bridge> so no lags for me 17:03 < bridge> I don't think we have LoD for zoom tbh 17:06 < bridge> But actually rendering everything to a texture could work 17:06 < bridge> Or probably multiple textures 17:07 < bridge> Conceptually it makes sense 17:07 < bridge> You just need a little bit of memory i guess 17:07 < bridge> exactly twice as much texture memory 17:07 < bridge> Twice as much? 17:08 < bridge> yes, like 1 + 0.5 + 0.25 + 0.125 ... 17:10 < bridge> oh since this is 2D i am wrong, it's smaller: 17:10 < bridge> https://upload.wikimedia.org/wikipedia/commons/5/5c/MipMap_Example_STS101.jpg 17:11 < bridge> it's exactyl 1/3rd more 😄 17:11 < bridge> it's exactyl 1/3rd more data 😄 17:12 < bridge> Okay but like 17:12 < bridge> it's exactyl 1/3rd more data (for full mimapping) 😄 17:12 < bridge> This is a mip map 17:13 < bridge> I think we could get this working, but I have no idea if this would make things better tbh 17:14 < bridge> It's impossible to tell without testing it unfortunately 17:14 < bridge> It's a cool idea tho 17:15 < bridge> I was always wondering if in 3D games it would be possible to render distant terrain to a texture 17:15 < bridge> Because it doesn't change much 17:16 < bridge> And it seems dumb to do so much work every frame when the result is almost the same as the previous frame 17:16 < bridge> the vulkan backend seems to support mipmaps, I checked the code. I don't know if this is used anywhere 17:17 < bridge> But why would mip maps be useful? 17:17 < bridge> It's not a performance thing 17:17 < bridge> It makes textures look better i think 17:18 < bridge> "They are intended to increase rendering speed and reduce aliasing artifacts." 17:18 < bridge> https://github.com/ddnet/ddnet/blob/ff74738338202ba06aaf0518c89be40b65204565/src/engine/client/backend/vulkan/backend_vulkan.cpp#L2582 17:18 < bridge> "They are intended to increase rendering speed and reduce aliasing artifacts." () 17:18 < bridge> so vulkan seems to already do this as long as you don't deactivate it 17:19 < bridge> Wait so how can it make rendeing faster? 17:19 < bridge> In practice it doesn't matter i think 17:19 < bridge> opengl as well 17:19 < bridge> idk magic 17:19 < bridge> In minecraft it changes nothing as far as i can tell, textures just look better 17:21 < bridge> Also doesn't anisotropic filtering do the same thing basically 17:21 < bridge> Just better? 17:21 < bridge> Or maybe both approaches are needed? 17:22 < bridge> Anisotropic filtering also makes textures look better with minimal performance impact 17:23 < bridge> I guess mip maps are a way of doing anisotropic filtering 17:23 < bridge> That makes sense actually 17:24 < bridge> MipMapLevels are limited to 1 here 17:35 < bridge> mipmaps does make rendering faster by offering memory locality: it uses a smaller texture which makes caching more effective 17:36 < bridge> I currently don#t see any effect 17:36 < bridge> yes, otherwise you get weird artifacts 17:37 < bridge> to make the rendering faster, it would need to be a relevant bottleneck. typically, ddnet rendering is cpu bottlenecked iirc 17:40 < bridge> @essigautomat the opengl 1 implementation doesn't use mipmaps i think. you can take a look at that xd 17:42 < bridge> I now managed to get more fps when zoomed out on a map by just setting mip map level count to 10 17:42 < bridge> with vulkan 17:42 < bridge> set this to 10: 17:42 < bridge> https://github.com/ddnet/ddnet/blob/ff74738338202ba06aaf0518c89be40b65204565/src/engine/client/backend/vulkan/backend_vulkan.cpp#L2583 17:42 < bridge> 17:42 < bridge> Done 17:43 < bridge> maybe also the 1 in ImgSize below 17:43 < bridge> i know for a fact ddnet uses more than 1 mip map level in the vulkan implementation xD 17:43 < bridge> else it would look horrible 17:44 < bridge> i dont know what that variable controls 17:44 < bridge> even setting it to 4 improves this 17:45 < bridge> so then one thing is that ImgSize has a depth, and it it used to calculate mip map level, afait it always returns 1 17:46 < bridge> ``` 17:46 < bridge> static size_t ImageMipLevelCount(size_t Width, size_t Height, size_t Depth) 17:46 < bridge> { 17:46 < bridge> return std::floor(std::log2(maximum(Width, maximum(Height, Depth)))) + 1; 17:46 < bridge> } 17:46 < bridge> ``` 17:46 < bridge> 17:46 < bridge> What is this for Depth 1? 17:46 < bridge> ah it's maximum 🤔 17:47 < bridge> is the mipmap flag always off? 🤔 17:49 < bridge> wtf why does changing this work 17:53 < bridge> I think because I have entities overlay turned on, which contains text, which doesn't use mipmaps 17:55 < bridge> omg this was exactly it ._. mipmaps are implemented and working 17:55 < bridge> they are also deeply integrated so nobody needs to think about it 18:04 < bridge> I can confirm this, otherwise the culling wouldn't work 18:04 < bridge> ~~I can confirm this, otherwise the culling wouldn't work~~ 18:04 < bridge> 18:04 < bridge> Idk what I smoked today, culling improves both gpu calls and cpu 18:22 < bridge> # discordapp.com/invite/5yfVG8BewR @everyone @everyone 18:37 < ws-client> **** @learath2 where u 19:39 < bridge> time to make blog posts with the restart of edos 19:39 < bridge> as i progress 19:40 < bridge> if it's a restart here's a logo 19:40 < bridge> https://cdn.discordapp.com/attachments/293493549758939136/1410318005597896845/Screenshot_2025-08-23-17-53-02-91_40deb401b9ffe8e1df2f1cc5ba480b12.jpg?ex=68b0947e&is=68af42fe&hm=ebdc8c4254992fb88506bc9bd80ad70e8abe8b0c9c1849688693abeb6b1d3d7e& 19:40 < bridge> i made it in some random generator xd 19:40 < bridge> looks funny 19:40 < bridge> highly jpegged OS 19:42 < ws-client> **** pegged OS 19:44 < bridge> is it FAT64 time?!??!?!?!?! 19:57 < bridge> less memory allocation, better vram usage, less memory addresses need to be filled, proper texture scaling on a pixel perfect level 19:57 < bridge> why do you even question this as a programmer? 19:58 < bridge> better vram usage? 19:58 < bridge> how 19:58 < bridge> it literally uses more memory 19:58 < bridge> you use a lower version of the image, that you can store and use in the GPU's memory as cache 19:59 < bridge> since the bitmap version (which you could also just generate at runtime) will be lower res, than the original, while also allow you to run algorithms to make it less pixelated or less distorted, then* storing it to the GPU 20:00 < bridge> xdd 20:00 < bridge> implementing a bitmap system from scratch is a nightmare tho, due to CPU / GPU communication :justatest: 20:00 < bridge> mipmaps actually use more vram iirc xd 20:01 < bridge> > A mipmap is a pre-calculated sequence of images, each one half the resolution of the previous level, used to improve rendering performance and visual quality when objects appear at different distances from the camera. 20:01 < bridge> > 20:01 < bridge> > For example, if you have a 1024x1024 texture, the mipmap chain would include: 20:01 < bridge> > 20:01 < bridge> > Level 0: 1024x1024 (original) 20:01 < bridge> > Level 1: 512x512 20:01 < bridge> > Level 2: 256x256 20:01 < bridge> > Level 3: 128x128 20:01 < bridge> > And so on down to 1x1 20:01 < bridge> > 20:01 < bridge> > Yes, using mipmaps increases VRAM usage by approximately 33% compared to storing just the original texture. This is because you're storing multiple versions of the same texture. 20:01 < bridge> > 20:01 < bridge> > However, mipmaps provide significant benefits: 20:01 < bridge> > 20:01 < bridge> > Reduced texture aliasing (shimmering/flickering at distances) 20:01 < bridge> > Better performance when sampling distant objects (GPU can use smaller textures) 20:01 < bridge> > Improved cache coherency 20:01 < bridge> > 20:01 < bridge> > The memory overhead is usually worth it for the visual quality and performance gains, which is why mipmaps are enabled by default in most graphics applications. 20:01 < bridge> only when you send the full resolution image to the GPU and let the GPU handle bitmapping 20:01 < bridge> Then yes, it costs more. 20:01 < bridge> But you can assign a cache to the lower resolution image (bitmap image) and discard the full resolution one when it's not needed 20:02 < bridge> But to implement such a dynamic system while keeping a record on what bitmap texture is being cached is... eh xdd 20:02 < bridge> Possible to do for sure, but takes a lot of trial and error to complete 20:02 < bridge> Possible to do for sure, but takes a lot of trial and error to complete (all for not getting GPU memory leaks) 20:04 < bridge> I think for chunks that include textures, should be able to have a dynamic LOD system which would benefit from those bitmap changes. Otherwise handling every instance of that texture one by one is a nightmare to do 20:06 < bridge> In Teeworlds' case, every block is unique in a way, that shouldn't be chunk-ated due to the visible repetitiveness of such LOD systems. 20:06 < bridge> Also the fact that rapid zoom in, zoom out and immediate zoom scale exists, the system has to prepare for discarding / allocating memory for all involved textures and their bitmap (cached) variants, which kinda sucks in performance 20:08 < bridge> (And unless you actually keep all those cached bitmaps from the map which is just extra memory in stack, this will be like... idk, exponentially expensive based on the amount of textures in a map used) 20:08 < bridge> Rename to Erdos 20:09 < bridge> perdoles? 20:09 < bridge> :justatest: 20:10 < bridge> So based on my knowledge and my nerd factor: 20:10 < bridge> - Bitmaps are only useful if you can effectively afford textures to be cached, which will not only yield better performance, but a stable cache line that can be tracked on what to load 20:10 < bridge> - Bitmaps only suck if you want to avoid allocating more than one version of the same texture, thus essentially increasing memory, where memory is tight 20:11 < bridge> I might miss out on crucial info about this topic, in that case feel free to correct me 😄 20:11 < bridge> u are a goddamn walking encyclopedia 20:12 < bridge> (of nonsense xd) 20:12 < bridge> No, I just experienced with this exact thing a couple of weeks ago in Godot 20:12 < bridge> Rude 20:12 < bridge> In honor of Paul Erdős 20:12 < bridge> why nonsense? 20:12 < bridge> As I said, feel free to correct me 20:12 < bridge> master 20:12 < bridge> nice reference 20:13 < bridge> :feelsbadman: 🙏 20:13 < bridge> ugh we have been at the same iteration of this event multiple times 20:13 < bridge> correcting takes longer than ignoring and going on 20:13 < bridge> u started by saying mimaps use less vram, u were already wrong, so lets just stop there 20:13 < bridge> true but... eh? 20:14 < bridge> did u get ur text from a blog post or is ur own 20:14 < bridge> So you read only that part and then ignored the rest? 20:14 < bridge> damn 20:14 < bridge> i read a bit what u said, but it just makes no sense to me xD 20:14 < bridge> ig I'm cooked then 20:15 < bridge> btw was this a chatgpt response? 20:15 < bridge> Is AI determining what you know about this topic? 20:15 < bridge> If so, I think I know who is more cooked then 20:15 < bridge> btw by definition a cache also means more vram usage, if said cache is on the gpu, otherwise i guess on the memory but then i would bet my ass a texture transfer from host to gpu is slower than using more vram 20:15 < bridge> so ur just probs overengineering nonsense 20:16 < bridge> btw I already mentioned the 33%, you probably safe more vram by disabling skins. 20:16 < bridge> that was yeah, but i already knew mimaps used more ram 20:16 < bridge> anyway mipmaps arent bad u should use them probs nearly always 20:16 < bridge> from CPU to GPU is faster, than reading back from GPU 20:16 < bridge> its still a transfer 20:17 < bridge> than having all on gpu 20:17 < bridge> I already implemented gpu memory clearing so the vram Situation should be better in nightly 20:17 < bridge> and??? 20:17 < bridge> skins are really small though 20:17 < bridge> It's like "oh my pizza takes a long time cuz it's far away from me" 20:17 < bridge>