00:22 <+bridge> [ddnet] @deen hey, could you please allow cross-origin for https://ddnet.tw/testmaps/ and https://ddnet.tw/mappreview/ for my site https://mapview.patiga.eu ? :) 00:37 <+bridge> [ddnet] should work now @Patiga 00:37 <+bridge> [ddnet] thanks, will test in a sec! :heartw: 00:38 <+bridge> [ddnet] Seems to work: https://mapview.patiga.eu/?map=Tutorial.map 00:38 <+bridge> [ddnet] I'd suggest to focus on the start tile initially 00:39 <+bridge> [ddnet] those are local, `?ddnet=Tutorial` should work once I upload 00:39 <+bridge> [ddnet] thought the same, it will also center on the middle of all spawns 00:40 <+bridge> [ddnet] Doesn‘t work on Safari mobile for me, Firefox desktop works fine 00:40 <+bridge> [ddnet] https://cdn.discordapp.com/attachments/293493549758939136/995097151405309973/image.png 00:41 <+bridge> [ddnet] huh, interesting 00:41 <+bridge> [ddnet] Also, Rust to wasm, cool 🙂 00:43 <+bridge> [ddnet] aye, the most pain I had is figuring out mobile html ^^ 00:47 <+bridge> [ddnet] yay, works now https://mapview.patiga.eu/?ddnet=Tutorial 🥳 00:50 <+bridge> [ddnet] https://mapview.patiga.eu/?ddnet-testing=Umbrella testing works too, noice 00:58 <+bridge> [ddnet] and for sites that allow cross-origin, this works https://mapview.patiga.eu/?url=https://heinrich5991.de/teeworlds/maps/maps/TEST5_7d7362b95d75318445ae42fd4722814e6aaa1fc81048ce0207fed91ac92cf02f.map 01:05 <+bridge> [ddnet] @Not Keks now that it works nicely on mobile, wanna have a look at my fully undocumented twgpu rust crate? ^^ 01:06 <+bridge> [ddnet] still only map renderer so far 01:07 <+bridge> [ddnet] will be away for the next few weeks, so I don't really have much time for implementing stuff, but feature requests are still very welcome :) 01:20 <+bridge> [ddnet] is the site down rn? 01:20 <+bridge> [ddnet] shows a white page for me 01:21 <+bridge> [ddnet] should be up 01:21 <+bridge> [ddnet] which link are you on, are you on mobile and on which browser? 01:21 <+bridge> [ddnet] pixel 4a https://mapview.patiga.eu/?map=Tutorial.map and on chrome 01:22 <+bridge> [ddnet] ah that is a dead link now, it happens so fast 😅 01:22 <+bridge> [ddnet] try https://mapview.patiga.eu/?ddnet=Tutorial 01:23 <+bridge> [ddnet] still white 01:24 <+bridge> [ddnet] ouch, I will need to have a look at it then 01:25 <+bridge> [ddnet] ah, there was also someone else for who the tilemaps didn't render 01:26 <+bridge> [ddnet] I'll have to learn what kind of devices have which problems, I think @Not Keks has experience in that direction ^^ 01:27 <+bridge> [ddnet] i lied it works now 01:27 <+bridge> [ddnet] yay :happy: 01:52 <+bridge> [ddnet] I don't see any tiles, firefox windows 02:00 <+bridge> [ddnet] don't see them in chrome either 02:00 <+bridge> [ddnet] did you wait? try scrolling maybe 02:01 <+bridge> [ddnet] i wanted to say that but i didnt know if it was supposed to be like that for now. It doesnt show the tiles for me either (chrome) 02:01 <+bridge> [ddnet] Doesn't load for me at all now with ?ddnet @Patiga `Uncaught LinkError: import object field '__wbg_screen_04bce210c29df7c4' is not a Function ` 02:01 <+bridge> [ddnet] I did wait quite a while 02:03 <+bridge> [ddnet] works for me in firefox for linux 02:04 <+bridge> [ddnet] works for me in chrome for linux 02:05 <+bridge> [ddnet] map_inspect_web.js:1497 Uncaught (in promise) Error: Using exceptions for control flow, don't mind me. This isn't actually an error! 02:05 <+bridge> [ddnet] at imports.wbg.__wbindgen_throw (map_inspect_web.js:1497:15) 02:05 <+bridge> [ddnet] at map_inspect_web_bg.wasm:0x1fe31a 02:05 <+bridge> [ddnet] at map_inspect_web_bg.wasm:0x1eb387 02:05 <+bridge> [ddnet] at map_inspect_web_bg.wasm:0xbc8f4 02:05 <+bridge> [ddnet] at map_inspect_web_bg.wasm:0x1d44bf 02:05 <+bridge> [ddnet] at map_inspect_web_bg.wasm:0x1a220d 02:05 <+bridge> [ddnet] at map_inspect_web_bg.wasm:0x1fcaa2 02:05 <+bridge> [ddnet] at __wbg_adapter_50 (map_inspect_web.js:222:10) 02:05 <+bridge> [ddnet] at real (map_inspect_web.js:199:20) 02:05 <+bridge> [ddnet] 02:05 <+bridge> [ddnet] Well, i do mind. :feelsbadman: 02:05 <+bridge> [ddnet] works for me in firefox for linux 02:05 <+bridge> [ddnet] eh 😄 wanted to say windows, but it's actually linux like I wrote 02:06 <+bridge> [ddnet] maybe it's a gpu thing 02:06 <+bridge> [ddnet] I'm on nvidia, gtx 1080 02:11 <+bridge> [ddnet] I offer GTX 760 02:12 <+bridge> [ddnet] Windows - Chrome - GTX 1050ti 02:17 <+bridge> [ddnet] Can someone explain why there are curly brackets after the method call? (void) 02:17 <+bridge> [ddnet] https://cdn.discordapp.com/attachments/293493549758939136/995121450308608131/unknown.png 02:17 <+bridge> [ddnet] menus_ingame.cpp 02:32 <+bridge> [ddnet] I don't see curly brackets there 02:33 <+bridge> [ddnet] perhaps you could link the code on github instead of screenshotting btw 02:34 <+bridge> [ddnet] https://github.com/ddnet/ddnet/blob/master/src/game/client/components/menus_ingame.cpp 02:34 <+bridge> [ddnet] Line 403 02:35 <+bridge> [ddnet] How? 02:35 <+bridge> [ddnet] you can click on the line number 02:35 <+bridge> [ddnet] https://github.com/ddnet/ddnet/blob/5345e55b94c30b4d4d3fe9e5ec3b2c908a812f12/src/game/client/components/menus_ingame.cpp#L403= 02:35 <+bridge> [ddnet] ok thanks! 02:36 <+bridge> [ddnet] it's just a block 02:36 <+bridge> [ddnet] so you can just use curly brackets mid code without a context? 02:36 <+bridge> [ddnet] Is that just for encapsulation? 02:36 <+bridge> [ddnet] yes 02:37 <+bridge> [ddnet] it's just so that the variables inside don't live longer 02:37 <+bridge> [ddnet] Oh no wait i dont know if thats the right word. I mean like using in c#? 02:37 <+bridge> [ddnet] Oh no wait i dont know if thats the right word. I mean like the "using" statement in c#? 02:38 <+bridge> [ddnet] in c# you can also do blocks without context 02:38 <+bridge> [ddnet] Habe mit encapsulation gemeint dass der code abgeschottet vom übergeordneten code ist 02:38 <+bridge> [ddnet] didnt know that lol 02:38 <+bridge> [ddnet] aber hab gerade gegoogled und gemerkt dass encapsulation was anderes is xD 02:42 <+bridge> [ddnet] Anyways, why is only that particular part in a special block? What if you remove the braces 02:42 <+bridge> [ddnet] Anyway, why is only that particular part in a special block? What if you remove the braces 02:42 <+bridge> [ddnet] Anyway, why is only that particular part in a special block? What if you remove the brackets 02:43 <+bridge> [ddnet] like the "using" statement in c#? 02:44 <+bridge> [ddnet] it makes the `CUIRect Button` local to that block 02:47 <+bridge> [ddnet] oh so its just a way for reusing the name Button 02:55 <+bridge> [ddnet] is it possible to have a branch which represents the current deployed build of ddnet? Idk how to keep track of which commits dean cherry picks into the actual build 02:55 <+bridge> [ddnet] I imagine it would also be useful for debugging 03:05 <+bridge> [ddnet] hmm I found a solution to my issue but I still think this would be nice 03:05 <+bridge> [ddnet] I think(tm) we always run master on the servers unless something is broken in master 03:05 <+bridge> [ddnet] but client is different? 03:05 <+bridge> [ddnet] Yes, client can be behind 03:06 <+bridge> [ddnet] I guess it's not so simple then 03:06 <+bridge> [ddnet] Actually we probably don't update the servers unless there is a new commit we want in there, but if we ever do update it's always to master 03:06 <+bridge> [ddnet] we don't run a special branch with cherry picked stuff from master 03:06 <+bridge> [ddnet] Whats the solution? Can you clone a Tag like a branch? 03:07 <+bridge> [ddnet] I guess we can try to create a branch, but it'd be a manual thing we have to update, might forget, hmm 03:07 <+bridge> [ddnet] ``git fetch --tags upstream`` updates tags from upstream to my fork then I can rebase on 16.2.1 tag 03:07 <+bridge> [ddnet] but it would be nice if 16.2.1 was not a tag 03:07 <+bridge> [ddnet] just some branch that gets updated with client 03:08 <+bridge> [ddnet] you can check the deployed version by using `/info` on the server btw 03:08 <+bridge> [ddnet] it'll tell you the git commit we built 03:10 <+bridge> [ddnet] I don't really care about server build so much since it's always the most recent. but if for example I want to develop on the most recent stable release the only way to do that right now is to manually figure out which tag is the most recent and then rebase on it 03:11 <+bridge> [ddnet] I mean it could probably be automated somehow but still not ideal I think 07:33 <+bridge> [ddnet] yeah apple doesn't support webgl2 as the only player in the market as always xD 07:33 <+bridge> [ddnet] DDNet to wasm also doesn't work on safari's webkit, while it works on every other device i tested^^ 07:35 <+bridge> [ddnet] except apple devices (that use safari's webkit... so all mobile and safari desktop) 07:35 <+bridge> [ddnet] it should indeed work pretty much everywhere. The web "drivers" have some workaround for knows bugs afaik 07:35 <+bridge> [ddnet] @Patiga how can i toggle FPS 😄 07:42 <+bridge> [ddnet] I assume it does not have mipmap support yet? 😄 07:57 <+bridge> [ddnet] can't comile tho :/ 07:57 <+bridge> [ddnet] ``` 07:57 <+bridge> [ddnet] error[E0432]: unresolved import `web_sys::HtmlCanvasElement` 07:57 <+bridge> [ddnet] --> map-inspect-web/src/lib.rs:12:15 07:57 <+bridge> [ddnet] | 07:57 <+bridge> [ddnet] 12 | use web_sys::{HtmlCanvasElement, Response}; 07:57 <+bridge> [ddnet] | ^^^^^^^^^^^^^^^^^ no `HtmlCanvasElement` in the root 07:57 <+bridge> [ddnet] ``` 07:57 <+bridge> [ddnet] can't comile tho :/ 07:57 <+bridge> [ddnet] ``` 07:57 <+bridge> [ddnet] error[E0432]: unresolved import `web_sys::HtmlCanvasElement` 07:57 <+bridge> [ddnet] --> map-inspect-web/src/lib.rs:12:15 07:57 <+bridge> [ddnet] | 07:57 <+bridge> [ddnet] 12 | use web_sys::{HtmlCanvasElement, Response}; 07:57 <+bridge> [ddnet] | ^^^^^^^^^^^^^^^^^ no `HtmlCanvasElement` in the root 07:57 <+bridge> [ddnet] ``` 07:57 <+bridge> [ddnet] 07:57 <+bridge> [ddnet] I am on rust nightly 08:32 <+bridge> [ddnet] ``` 08:32 <+bridge> [ddnet] [2022-07-09T06:32:09Z ERROR wgpu_hal::vulkan::instance] VALIDATION [VUID-vkCmdCopyImage-srcImageLayout-00128 (0xef9e2bd6)] 08:32 <+bridge> [ddnet] Validation Error: [ VUID-vkCmdCopyImage-srcImageLayout-00128 ] Object 0: handle = 0x562d4c486300, name = Array Texture Creator, type = VK_OBJECT_TYPE_COMMAND_BUFFER; | MessageID = 0xef9e2bd6 | vkCmdCopyImage: Cannot use VkImage 0x1510000000151[Array Image generic_deathtiles] (layer=0 mip=0) with specific layout VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL that doesn't match the previous known layout VK_IMAGE_LAYOUT_TRANSFER_DST_OPTIMAL. The 08:32 <+bridge> [ddnet] [2022-07-09T06:32:09Z ERROR wgpu_hal::vulkan::instance] objects: (type: COMMAND_BUFFER, hndl: 0x562d4c486300, name: Array Texture Creator) 08:32 <+bridge> [ddnet] [2022-07-09T06:32:09Z ERROR wgpu_hal::vulkan::instance] VALIDATION [VUID-vkCmdCopyImage-srcImageLayout-00129 (0x85e93015)] 08:32 <+bridge> [ddnet] Validation Error: [ VUID-vkCmdCopyImage-srcImageLayout-00129 ] Object 0: handle = 0x562d4c486300, name = Array Texture Creator, type = VK_OBJECT_TYPE_COMMAND_BUFFER; | MessageID = 0x85e93015 | vkCmdCopyImage: Layout for VkImage 0x1510000000151[Array Image generic_deathtiles] is VK_IMAGE_LAYOUT_TRANSFER_SRC_OPTIMAL but can only be VK_IMAGE_LAYOUT_GENERAL or VK_IMAGE_LAYOUT_GENERAL. The Vulkan spec states: srcImageLayout must be VK_IM 08:32 <+bridge> [ddnet] [2022-07-09T06:32:09Z ERROR wgpu_hal::vulkan::instance] objects: (type: COMMAND_BUFFER, hndl: 0x562d4c486300, name: Array Texture Creator) 08:32 <+bridge> [ddnet] ``` 08:33 <+bridge> [ddnet] vulkan not happy yet 08:38 <+bridge> [ddnet] your fps calculation seems slightly off xD 08:38 <+bridge> [ddnet] https://cdn.discordapp.com/attachments/293493549758939136/995217373256753202/unknown.png 08:38 <+bridge> [ddnet] or is that CPU only? 08:39 <+bridge> [ddnet] zoom performance is really awesome, how did you do that? build the map borders as buffers on fly? 08:47 <+bridge> [ddnet] thats caching I think 08:49 <+bridge> [ddnet] weird, must've missed to add a feature to a dependency there. odd that it worked for me 08:50 <+bridge> [ddnet] yea ^^ 08:51 <+bridge> [ddnet] true, I blame wgpu 😛 (although I could probably fix it, but it works riiiight? 08:53 <+bridge> [ddnet] hm yes, for 5 seconds I bump a counter for each render process I start, I suppose that is not the right way o.o 08:54 <+bridge> [ddnet] thats my version of tilemaps. those also make it less performant on smaller zoom steps (I think) 08:56 <+bridge> [ddnet] the tilemaps are each just a single quad that cover the screen, and in the fragment shader I figure out which tile it is + orientation to figureout the pixels color 08:57 <+bridge> [ddnet] @Patiga I don't see any tiles, only the quads and background? same on firefox/chrome 08:57 <+bridge> [ddnet] https://cdn.discordapp.com/attachments/293493549758939136/995222077571342357/unknown.png 08:58 <+bridge> [ddnet] 3070/AMD 5600x if it helps 08:59 <+bridge> [ddnet] thanks! don't yet know what causes that 09:00 <+bridge> [ddnet] I only see one error in console 09:00 <+bridge> [ddnet] https://cdn.discordapp.com/attachments/293493549758939136/995222931930095686/unknown.png 09:03 <+bridge> [ddnet] https://mapview.patiga.eu/?ddnet=Tutorial empty white page for me on iOS safari 09:03 <+bridge> [ddnet] yea those are normal for now 😅 09:04 <+bridge> [ddnet] chillerdragon: safari apparently doesnt support webgl as well as the other browsers 09:06 <+bridge> [ddnet] webmetal when 09:11 <+bridge> [ddnet] but one "tilemap" is a collection of multiple tiles? 09:15 <+bridge> [ddnet] https://cdn.discordapp.com/attachments/293493549758939136/995226654546862101/screenshot_2022-07-09_08-34-48.png 09:15 <+bridge> [ddnet] https://cdn.discordapp.com/attachments/293493549758939136/995226654752378920/unknown.png 09:16 <+bridge> [ddnet] 3.5k fps is max for me on your renderer, vs 10-11k on ddnet 09:19 <+bridge> [ddnet] but urs also does the surface conversion, that probs costs quite a lot of frames, probably atleast 100% 09:19 <+bridge> [ddnet] so i assume yours can do at least 6-7k 09:21 <+bridge> [ddnet] would it be possible to do the tile rendering in ddnet by just putting all the tiles into some sort of texture atlas then running a compute/fragment shader to lookup the texture per pixel? 09:21 <+bridge> [ddnet] Wait, that's YOU?! 09:21 <+bridge> [ddnet] Wait, that's YOU?! 09:21 <+bridge> [ddnet] Wait, this is totally off topic, my bad. 09:21 <+bridge> [ddnet] lol 09:22 <+bridge> [ddnet] possible yes, if its a good solution is another question tho 09:22 <+bridge> [ddnet] it would get good zoom performance I guess 09:22 <+bridge> [ddnet] yes 09:22 <+bridge> [ddnet] but would defs kill non zoom performance 09:22 <+bridge> [ddnet] yeah 09:23 <+bridge> [ddnet] big zoom is not used that often anyway ingame 09:23 <+bridge> [ddnet] i'd defs like to improve it sooner or later tho 09:23 <+bridge> [ddnet] i see other problems with texture usage tho, e.g. it must look relativly good 09:24 <+bridge> [ddnet] but on which zoom level do you create the textures? 09:24 <+bridge> [ddnet] rerender at every zoom step change? 09:24 <+bridge> [ddnet] mipmaps will suck hard if you dont do that 09:24 <+bridge> [ddnet] but rerendering will create a lag everytime you change the zoom step 09:24 <+bridge> [ddnet] hmm 09:25 <+bridge> [ddnet] @Patiga 's solution might work 09:26 <+bridge> [ddnet] which textures? I only have an additional texture for the srgb redirection 09:26 <+bridge> [ddnet] I'm not very familiar with how the current system works tbh, I looked at it a big then gave up. 09:26 <+bridge> [ddnet] the tilemaps render directly to the target 09:26 <+bridge> [ddnet] tilemap = textures? 09:26 <+bridge> [ddnet] yea I rerender every frame 09:26 <+bridge> [ddnet] ah yes 09:26 <+bridge> [ddnet] those ones 09:26 <+bridge> [ddnet] they are just static, its basically the same tilemap as on the cpu 09:26 <+bridge> [ddnet] with some value changes 09:27 <+bridge> [ddnet] but how did you solve the border? 09:27 <+bridge> [ddnet] why is it so efficient 09:27 <+bridge> [ddnet] when I index into the tilemap, I clamp the index to the min and max index 09:27 <+bridge> [ddnet] do you use a geometry shader? 09:27 <+bridge> [ddnet] https://gitlab.com/Patiga/twgpu/-/blob/master/twgpu/src/map/tilemap_shader.wgsl#L66 09:28 <+bridge> [ddnet] i dont understand why instanced rendering is so slow vs your approach 09:28 <+bridge> [ddnet] u basically do everything on the GPU right? 09:28 <+bridge> [ddnet] yep 09:28 <+bridge> [ddnet] so it creates the borders itself 09:29 <+bridge> [ddnet] yes, it renders the same way as the normal tiles 09:29 <+bridge> [ddnet] this feels like it should be a solved problem, aren't there lots of games that use tile rendering? 09:30 <+bridge> [ddnet] i am just suprised that it works out so well compared to instanced rendering 09:30 <+bridge> [ddnet] to me it sounds relativly similar 09:30 <+bridge> [ddnet] if it creates the vertices on the GPU, or if it uses one vertex and moves it around 09:33 <+bridge> [ddnet] but you enter the fragment shader twice? 09:33 <+bridge> [ddnet] at least sounded like that last time we talked about it 09:33 <+bridge> [ddnet] once for each tilemap 09:34 <+bridge> [ddnet] the vertex buffer (that contains the quad that covers the screen) is actually shared between all tilemaps, the parallax and so on all happens in the vertex shader 09:35 <+bridge> [ddnet] ok understandable 09:36 <+bridge> [ddnet] but still i am mostly interested in the borders 09:36 <+bridge> [ddnet] are they part of the vertex buffer? 09:36 <+bridge> [ddnet] fully extended? 09:37 <+bridge> [ddnet] wait you create a texture for the whole map? 09:38 <+bridge> [ddnet] pls explain before i learn rust syntax xddd 09:47 <+bridge> [ddnet] I'm not very familiar with how the current system works tbh, I looked at it a bit then gave up. 09:58 <+bridge> [ddnet] yes, the entire cpu tilemap is also on the gpu (with some adjusted values) 09:58 <+bridge> [ddnet] yes 09:58 <+bridge> [ddnet] as texture right? 09:58 <+bridge> [ddnet] one pixel per thing 09:58 <+bridge> [ddnet] per tile 09:58 <+bridge> [ddnet] the vertex buffer is simply a quad that always covers the entire screen, no further than that 09:58 <+bridge> [ddnet] yup 09:58 <+bridge> [ddnet] ok 09:58 <+bridge> [ddnet] i understand that part 09:58 <+bridge> [ddnet] just not the borders 09:58 <+bridge> [ddnet] 😄 09:59 <+bridge> [ddnet] so for any given pixel on the screen I need to figure out which tile it is on 09:59 <+bridge> [ddnet] for that I floor the position and have that as the index for the tile 09:59 <+bridge> [ddnet] hey that's what I suggested 09:59 <+bridge> [ddnet] xd 10:00 <+bridge> [ddnet] I use that to access the tilemap 10:00 <+bridge> [ddnet] to support the border, I clamp the index between 0 and the tilemap bounds 10:00 <+bridge> [ddnet] ok i understand 10:01 <+bridge> [ddnet] nice 10:01 <+bridge> [ddnet] maybe wrote a bit hectic, gotta pack my rucksack rn ^^ 10:03 <+bridge> [ddnet] what is meant by "tilemap" ? 10:03 <+bridge> [ddnet] it's a texture? 10:04 <+bridge> [ddnet] yes 10:04 <+bridge> [ddnet] 1 pixel per tile 10:04 <+bridge> [ddnet] so in back in time 3 you'll probably get a 1500x4000 texture 10:04 <+bridge> [ddnet] it's like a lookup table for the texture? 10:05 <+bridge> [ddnet] each pixel in the tilemap contains an index to a texture? or something else 10:06 <+bridge> [ddnet] u basically see the world as pixels too 10:06 <+bridge> [ddnet] and then overlay the tilemap texture and clamp on the edges 10:06 <+bridge> [ddnet] or see the world as tiles which refer to 1 pixel on the tilemap 10:08 <+bridge> [ddnet] ok I get that but you go from screen pixel -> tile coords -> tilemap -> ?? 10:08 <+bridge> [ddnet] where do the actual tile textures come in 10:09 <+bridge> [ddnet] u go from screen -> tile that would map the screen -> some offset etc calculation to know where the tile is on the pixelmap (clamp if outside) 10:09 <+bridge> [ddnet] atleast that's how i imagine it 10:09 <+bridge> [ddnet] screen pixel -> tile coords -> tilemap -> tile id -> tilemap array texture 10:10 <+bridge> [ddnet] ok I understand 10:12 <+bridge> [ddnet] so you only render a single fullscreen quad per layer? 10:12 <+bridge> [ddnet] exactly :) 10:12 <+bridge> [ddnet] I think we could do that in ddnet client 10:12 <+bridge> [ddnet] but the computations in the fragment shader are expensive 10:13 <+bridge> [ddnet] ah 10:13 <+bridge> [ddnet] yeah it leaves everything to the GPU 10:13 <+bridge> [ddnet] 10:13 <+bridge> [ddnet] but at a certain zoom level it might still be worth it to use a similar approach for the borders only 10:13 <+bridge> [ddnet] the computations are expensive or the texture lookup is expensive? 10:13 <+bridge> [ddnet] I thought you just do simple math 10:13 <+bridge> [ddnet] it's clearly faster than instanced rendering 10:14 <+bridge> [ddnet] @Not Keks you have a quad for each tile, couldn't you also add just 8 quads for each border segment? those could just all be activated when any border is seen, that shouldn't be expensive, no? 10:15 <+bridge> [ddnet] i tried that yes 10:15 <+bridge> [ddnet] it gives bit more fps than the native driver, but still sucks too hard 10:15 <+bridge> [ddnet] i assume the vertex shader overhead is just too huge 10:15 <+bridge> [ddnet] it has to rasterize a few thousand quads 10:15 <+bridge> [ddnet] more like then thousands 10:16 <+bridge> [ddnet] that's generally the pro side of your approach 10:16 <+bridge> [ddnet] ah, okay 10:16 <+bridge> [ddnet] it's cheap on vertex stage 10:16 <+bridge> [ddnet] but expensive on fragment 10:16 <+bridge> [ddnet] so a toggle at a certain zoom level would make sense 10:17 <+bridge> [ddnet] yes its pretty simple math with a few io thingies. you want the fragment shader as short as possible though, it has to be computed for each pixel it is triggered for, so each line in the fragment shader adds up (as far as I understand it) 10:18 <+bridge> [ddnet] in my experience doing a single texture lookup is way more expensive than doing a moderately large amount of math 10:18 <+bridge> [ddnet] actually lemme just look 10:18 <+bridge> [ddnet] is that so? 10:19 <+bridge> [ddnet] i mean reading from VRAM at some level yes 10:19 <+bridge> [ddnet] but the gpu cores are also rather slow 10:19 <+bridge> [ddnet] it depends on many things 10:19 <+bridge> [ddnet] yeah probably 😄 10:20 <+bridge> [ddnet] if this is all the math you're doing then it's basically free 10:20 <+bridge> [ddnet] https://cdn.discordapp.com/attachments/293493549758939136/995243120268681247/unknown.png 10:20 <+bridge> [ddnet] idk if I'm missing something 10:21 <+bridge> [ddnet] that could be a big slowdown, not just having to sample the mapres but also indexing into the tilemap 10:21 <+bridge> [ddnet] yes 10:21 <+bridge> [ddnet] thats all the math 10:21 <+bridge> [ddnet] https://www.youtube.com/watch?v=7YPPFIqUN84 10:24 <+bridge> [ddnet] i wouldnt call it free tho 😄 10:24 <+bridge> [ddnet] not all that video is relevant 10:24 <+bridge> [ddnet] it still probs around half as fast as ddnet, which uses even less math 😄 10:24 <+bridge> [ddnet] hmm 10:25 <+bridge> [ddnet] it goes into a GPU bottleneck 10:25 <+bridge> [ddnet] i only get a CPU bottleneck with ddnet 10:25 <+bridge> [ddnet] I think it probably bottle necks on memory tho 10:26 <+bridge> [ddnet] radeontop says the pipeline is at 100% 10:26 <+bridge> [ddnet] i can certainly imagine its already on the GPU cores 10:26 <+bridge> [ddnet] well idk it doesn't even run on my pc xd 10:26 <+bridge> [ddnet] xD 10:29 <+bridge> [ddnet] @Tater if you want you could try to compile it if you haven't already 10:29 <+bridge> [ddnet] seems like jupstar already figured it out ^^ 10:29 <+bridge> [ddnet] I guess the performance also scales linearly with number of layers regardless of layer content? does ddnet renderer do that? 10:29 <+bridge> [ddnet] ok I'll try 10:29 <+bridge> [ddnet] yes, thats another big factor I think 10:30 <+bridge> [ddnet] https://cdn.discordapp.com/attachments/293493549758939136/995245523978158171/Screenshot_20220709_102921.png 10:30 <+bridge> [ddnet] https://cdn.discordapp.com/attachments/293493549758939136/995245524334690314/Screenshot_20220709_102948.png 10:31 <+bridge> [ddnet] Grafik 10:31 <+bridge> [ddnet] left you see the usage of the GPU 10:32 <+bridge> [ddnet] that's a cool tool 10:32 <+bridge> [ddnet] I guess amd only? 10:32 <+bridge> [ddnet] yes radeontop 10:32 <+bridge> [ddnet] so the GPU cores and the VRAM fetching both go up 10:32 <+bridge> [ddnet] @Not Keks have you compiled in release mode btw? 10:32 <+bridge> [ddnet] yes 10:32 <+bridge> [ddnet] cargo b --release 10:33 <+bridge> [ddnet] nice 👍 10:38 <+bridge> [ddnet] I think it's building :o 10:40 <+bridge> [ddnet] well anyway really amazing work, its cool to see it working with vulkan and webgl without major problems. I cannot judge how well it compares to native vulkan only, but it seems like the shaders are on a level at least that of higher OpenGL version like, GLES 3 maybe? 10:40 <+bridge> [ddnet] ddnet's vulkan outperforms its opengl backends just bcs of the CPU bottleneck 10:41 <+bridge> [ddnet] in fact RADV is slower than radeonsi on GPU bounds for me 10:41 <+bridge> [ddnet] aah :( 10:41 <+bridge> [ddnet] https://cdn.discordapp.com/attachments/293493549758939136/995248230877110432/unknown.png 10:41 <+bridge> [ddnet] just ignore the web inspector 10:41 <+bridge> [ddnet] it also doesnt build for me 10:41 <+bridge> [ddnet] oh 10:41 <+bridge> [ddnet] as long as the other tool runs 😄 10:44 <+bridge> [ddnet] which is the other tool? 10:44 <+bridge> [ddnet] twgpu-map-inspect 10:46 <+bridge> [ddnet] what is MAPRES_DIRECTORY 10:46 <+bridge> [ddnet] the editor mapres 10:46 <+bridge> [ddnet] in /data in ur ddnet dir 10:52 <+bridge> [ddnet] omg I works 10:52 <+bridge> [ddnet] omg it works 10:52 <+bridge> [ddnet] how do I move the camera 10:53 <+bridge> [ddnet] oh nvm the mouse moves forever 10:53 <+bridge> [ddnet] yeah the mouse bit buggy xD 10:53 <+bridge> [ddnet] bcs its not clamped inside the window 😄 10:53 <+bridge> [ddnet] perf seems good but I don't have the fancy tool 10:54 <+bridge> [ddnet] u on intel? 10:54 <+bridge> [ddnet] nvidia 10:54 <+bridge> [ddnet] which? 10:54 <+bridge> [ddnet] 3070 10:54 <+bridge> [ddnet] ok so also high end 10:54 <+bridge> [ddnet] would like to know on a "bad" GPU 😄 10:54 <+bridge> [ddnet] ah 10:57 <+bridge> [ddnet] you can double-click for fullscreen 10:57 <+bridge> [ddnet] maybe that helps with the mouse issue 10:57 <+bridge> [ddnet] I'm using desktop version 10:57 <+bridge> [ddnet] web was broke for me 10:59 <+bridge> [ddnet] ohh a rust renderer 10:59 <+bridge> [ddnet] there are some funny artifacts on the end of the map xD 10:59 <+bridge> [ddnet] https://cdn.discordapp.com/attachments/293493549758939136/995252966011437146/unknown.png 11:03 <+bridge> [ddnet] btw, I managed to bring down the rust code size when building statically 11:04 <+bridge> [ddnet] so now it's okay at a couple of hundred kilobytes 11:04 <+bridge> [ddnet] :PepeA: 11:04 <+bridge> [ddnet] the trick was enabling LTO for the final artifact… 11:04 <+bridge> [ddnet] i.e. in CMakeLists.txt 11:04 <+bridge> [ddnet] oh yeah 11:04 <+bridge> [ddnet] also mentioned on min-sized-rust doc 11:05 <+bridge> [ddnet] not really 11:05 <+bridge> [ddnet] I enabled it in Cargo.toml, but it's no good there because we're producing a static library 11:05 <+bridge> [ddnet] (min-sized-rust only talks about Cargo.toml wrt. LTO) 11:05 <+bridge> [ddnet] ah yeah 11:05 <+bridge> [ddnet] I needed it in the final artifact 11:05 <+bridge> [ddnet] i guess lto makes more sense in the final binary 11:06 <+bridge> [ddnet] ye 11:06 <+bridge> [ddnet] so, do you have an idea where I could implement something small in rust? 11:07 <+bridge> [ddnet] hmm 11:07 <+bridge> [ddnet] a chat command? 11:08 <+bridge> [ddnet] I suppose, sounds good 11:09 <+bridge> [ddnet] @heinrich5991 if u want to use the gameserver in the chat command 11:09 <+bridge> [ddnet] do u have to define the whole gameserver interface in rust? 11:09 <+bridge> [ddnet] xD 11:09 <+bridge> [ddnet] sounds like some work 11:09 <+bridge> [ddnet] ah, maybe a chat command isn't the best start 😄 11:09 <+bridge> [ddnet] hmm 11:10 <+bridge> [ddnet] u can just pass the chat text 11:10 <+bridge> [ddnet] and pass smth back 11:11 <+bridge> [ddnet] the http downloader maybe 11:11 <+bridge> [ddnet] if thats not too big already 😄 11:13 <+bridge> [ddnet] word filter for client 11:13 <+bridge> [ddnet] ez 11:23 <+bridge> [ddnet] @Patiga I think you could return 0 on air tile to skip texture, might get some performance 11:24 <+bridge> [ddnet] GPUs heavily rely on predication 11:24 <+bridge> [ddnet] they don't use jmp opcodes like CPUs do so heavily 11:25 <+bridge> [ddnet] makes sense bcs there is only 1 instruction register for multiple shader cores 11:25 <+bridge> [ddnet] oh interesting, TIL 11:25 <+bridge> [ddnet] predication? 11:25 <+bridge> [ddnet] prediction 11:25 <+bridge> [ddnet] or prediction 11:25 <+bridge> [ddnet] ah 11:25 <+bridge> [ddnet] > Definition of predication 11:25 <+bridge> [ddnet] > 1 archaic. a : an act of proclaiming or preaching. b : sermon. 2 : an act or instance of predicating: such as. a : the expression of action, state, or quality by a grammatical predicate. 11:25 <+bridge> [ddnet] xD 11:26 <+bridge> [ddnet] i think its called predication 11:26 <+bridge> [ddnet] https://en.wikipedia.org/wiki/Predication_(computer_architecture) 11:26 <+bridge> [ddnet] interesting 11:26 <+bridge> [ddnet] from predicate 11:26 <+bridge> [ddnet] then I even misunderstood you because I "autocorrected" predication to prediction. thanks for the question @Ryozuki 11:28 <+bridge> [ddnet]