• What Are You Working On? September 2015
    1,261 replies, posted
[QUOTE=Karmah;48728662]Well I mostly finished shadows. I somehow managed to both [I]create[/I] and[I] fix [/I]the worst Z-Fighting I've ever seen anywhere, but now I have functioning spot and point light shadows. [vid]https://my.mixtape.moe/qnzyds.mp4[/vid] I'm still somehow messing up getting the appropriate directional vector for each silhouette when emanated from a spot light, however it is kinda close. I'm still convinced there has to be a better way of doing this with deferred shading, instead of just re-rendering everything using a stencil buffer for every light source supporting shadows.[/QUOTE] How are you finding this technique so far? Do you think it's faster than shadow mapping? I'm currently implementing shadow mapping, but I'm kinda afraid it is turning out too slow. Even though I don't have deferred shading yet I want to have a few lights in scene and not just one. I can speed up the shadow map rendering a bit with parallel computing but rendering the scene each time from each light every time there's a change in the scene is still expensive I think, I haven't finished it yet but from the looks of things it will be slow. Is shadow volume faster? I haven't looked in to the matter but I know you have to use the stencil buffer? I don't have one, is it always necessary for the algorithm? I could implement one if this technique is significantly faster than shadow mapping.
Back to working on reverse engineering the D3DBSP format, now on material files [t]http://puu.sh/kj00O/84749fa537.png[/t] Puzzling like this is so much fun. But once I'm done I'll definitely publish a blogpost or something about the format. It's p smart Here's my current progress btw, same as some time ago but just a reminder. I'll be handing out brownie points to the first person guessing the game (easy) and map (harder) [IMG]https://i.imgur.com/1SRkSgW.png[/IMG] [I]Cyberuben you're banned from this quiz![/I]
[QUOTE=Handsome Matt;48730710]if you can obtain it, [URL="http://www.sweetscape.com/010editor/"]010 editor[/URL] is probably the best hex editor out there for reverse engineering (it'll save you a lot of trouble using that online hex converter)[/QUOTE] That and the binary templates seem pretty nice. I'll give it a try, thanks
[QUOTE=Asgard;48730640] I'll be handing out brownie points to the first person guessing the game (easy) and map (harder) [t]https://i.imgur.com/1SRkSgW.png[/t] [I]Cyberuben you're banned from this quiz![/I][/QUOTE] Is this Carentan from Call Of Duty 2? It looks really similar to it...
[QUOTE={TFS} Rock Su;48730783]Is this Carentan from Call Of Duty 2? It looks really similar to it...[/QUOTE] You get the brownie points
[QUOTE=Asgard;48730793]You get the brownie points[/QUOTE] How many brownie points do you need to get to be able to redeem them for actual brownies?
[QUOTE=cartman300;48730816]How many brownie points do you need to get to be able to redeem them for actual brownies?[/QUOTE] Yes
[QUOTE=Asgard;48730840]Yes[/QUOTE] No it's 2 meters I think
[QUOTE=Handsome Matt;48730710]if you can obtain it, [URL="http://www.sweetscape.com/010editor/"]010 editor[/URL] is probably the best hex editor out there for reverse engineering (it'll save you a lot of trouble using that online hex converter)[/QUOTE] Vim can do it too: [url]http://vim.wikia.com/wiki/Hex_dump[/url]
[QUOTE=Cyberuben;48730377]guys I want to create an UI designer tool. Simply speaking, I want people with no programming knowledge to be able to design i.e. a Menu with animations, transitions, etc. However, this would require a custom editor. I have heard of Qt and I'm currently reading into it, but I must say there is quite a steep learning curve. Are there any other libraries that any of you could recommend? The idea I had was as following: [IMG]http://scrnsht.me/u/r9/raw[/IMG] I really want to code this in either C++ or C# for speed reasons[/QUOTE] If you go the Qt route, I've been using it for a couple years now and have a firm understanding of how it works. Its pretty much like how Java's UI worked, except its actually really easy to use their event system. It looks intimidating, but connecting a button to a thing (or your own custom thing to another thing) is really easy. The only tough part is layouts, as you are restricted to laying things out in cells. It would be possible though.
i really hate qt because its basically a different language with all the macros and shit and it requires a custom build pipeline what is even up with that
[QUOTE=KaPow!;48730481]How are you finding this technique so far? Do you think it's faster than shadow mapping? I'm currently implementing shadow mapping, but I'm kinda afraid it is turning out too slow. Even though I don't have deferred shading yet I want to have a few lights in scene and not just one. I can speed up the shadow map rendering a bit with parallel computing but rendering the scene each time from each light every time there's a change in the scene is still expensive I think, I haven't finished it yet but from the looks of things it will be slow. Is shadow volume faster? I haven't looked in to the matter but I know you have to use the stencil buffer? I don't have one, is it always necessary for the algorithm? I could implement one if this technique is significantly faster than shadow mapping.[/QUOTE] There are pros and cons, I don't really know which is better. Shadow Mapping requires re-rendering the scene from the point of view of each light, which means spot lights require 1 rendering pass, but point lights require 6 (in in each direction). With Shadow volumes, a single re-rendering pass is done for each light, period, since the shader [B]builds[/B] an object along the direction vector of the light. The cons to shadow volumes is that Z-Fighting is a possibility, since you technically build an object to represent the shadow, it has to overlay the object its going on to. Also I'm not sure if anything can be done to the shadowed region, like softening the shadow. I don't know which is the better one. After reading up on some of these new ways of shadow mapping (like in the deus ex article below), I'm undecided in which way I want to go, so I'm going to do further research. I highly recommend switching over to deferred shading though. Instead of using a single shader, you use one per technique you wish to perform, like one for spot lights, one for point lights etc. Further, your lighting becomes so much more optimized because you don't do it on every single object in the scene (like ones that are obviously too far away to get it) because you limit it to areas of influence by using stenciling, as well as sides that aren't seen by the camera don't get filled in either. In short, don't worry about your performance yet unless you are done learning everything you want to learn. Deferred shading will increase your performance. In my case, I had a lot if problems implementing shadow mapping before I switched to deferred shading, and I recently tried it again but I'm not quite sure how to get it to work with deferred shading, so I tried something else. [editline]21st September 2015[/editline] [QUOTE=lxmach1;48729223]I don't know if it'll help, but here's an article someone wrote about reverse-engineering Deus Ex: Human Revolution's rendering. [URL]http://www.adriancourreges.com/blog/2015/03/10/deus-ex-human-revolution-graphics-study/[/URL][/QUOTE] Thanks, I completely forgot that this existed and I never got a chance to read it yet. I will definitely keep this bookmarked and try some of their techniques as I get further. If anyone has any further knowledge on Shadowing techniques, and their pros V cons, I'd love to hear them. The only reason why I tried shadow volumes is because it was the next thing on the tutorial series I'm following. However, I'm inclined to think that since DE:HR used Parallel-split shadow maps, and it was a "Recent" game, then that technique must be good, as I don't have much knowledge about them/
Anybody know a good, free, Mac OSX alternative to ShareX? I want to be able to record an .mp4 of my project. EDIT: Nevermind, it seems that quicktime player works quite well for my purposes.
What happened to that Rant Telegram bot?
Previously posted this in the Gmod WAYWO but I guess it's more appropriate here. I'm making a web-based BSP renderer. The goal is to be able to render a Source engine map without having to load any extra textures or models, hence the goofy flat shading. Here's a [URL="http://cogg.rocks/vbsp/"]demo[/URL] and the [URL="https://github.com/birdbrainswagtrain/vbsp.js"]git repo[/URL]. It also doesn't actually bother with any of the BSP stuff, it just shoves everything (including brush models) into a vertex buffer and hopes for the best.
[QUOTE=Darkwater124;48732084]What happened to that Rant Telegram bot?[/QUOTE] I think Rant projects are jinxed just like Facepunch mobile apps. Almost everyone who has done something interesting with Rant has disappeared. [I][sp]...and they were delicious[/sp][/I]
Blender isn't that bad after you start figuring out the shortcuts. Too bad nobody told me that they're required to function at all :v: Anyway, welcome to the very first real gameplay change since porting to Unity: arrows (to swipe, instead of tap) Video coming after I get my schoolwork done. (possibly 2018) [img]http://i.imgur.com/h24RjQ3.png?1[/img]
Also I started a [URL="https://github.com/RantLang/Rant3Overview"]repository of ideas[/URL] for the next version of Rant. I needed a place to put my feature ideas because I think of so many, even though I'm not programming right now. If you think of something cool I should add, tell me about it or make an issue/comment and I'll consider it.
[QUOTE=Berkin;48733635]Also I started a [URL="https://github.com/RantLang/Rant3Overview"]repository of ideas[/URL] for the next version of Rant. I needed a place to put my feature ideas because I think of so many, even though I'm not programming right now. If you think of something cool I should add, tell me about it or make an issue/comment and I'll consider it.[/QUOTE] Why not use the issues system on your existing github project? Or the wiki?
[QUOTE=DrDevil;48733777]Why not use the issues system on your existing github project? Or the wiki?[/QUOTE] I wanted to make the list separate so it would be easier for people to find. I prefer to use the issue tracker mostly for tracking actual problems with the existing code and suggestions for improvements to the current version. It's simply my preference.
[QUOTE=Karmah;48731637][...] The cons to shadow volumes is that Z-Fighting is a possibility, since you technically build an object to represent the shadow, it has to overlay the object its going on to. [...][/QUOTE] If you use the exact same coords it can't happen (between the object and shadow at least). You could still get z-fighting between different shadows though I think.
[QUOTE=Darkwater124;48732084]What happened to that Rant Telegram bot?[/QUOTE] The refurbished dual-slot Xeon server in my closet that keeps overheating and rebooting happened, combined with my friends using my torrent box to download every goddamn movie on the internet :hairpull: Anyway, if it's ever shitting itself, feel free to set up your own instance (preferably on reliable hardware): [url]https://github.com/jmazouri/rantogram[/url]
[QUOTE=Tamschi;48733944]If you use the exact same coords it can't happen (between the object and shadow at least). You could still get z-fighting between different shadows though I think.[/QUOTE] Huh? What is depth buffer precision? Shit can happen when 2 surfaces get really close, that's why a tiny epsilon value is used in this case, so that the shadow geometry is moved slightly away from the surface. Or am I misunderstanding you? In any case, although the shadows worked, it ultimately was a shit system and fucked too many other things over, so I'm amidst attempt #3 of shadow mapping. I see way more positive things about shadow mapping then stencil shadow volumes.
[QUOTE=Karmah;48734118]Huh? What is depth buffer precision? Shit can happen when 2 surfaces get really close, that's why a tiny epsilon value is used in this case, so that the shadow geometry is moved slightly away from the surface.[/QUOTE] Right, when they get [I]really close[/I], but not if they overlap [I]exactly[/I] (as long as it's guaranteed they aren't drawn simultaneously, which may cause race conditions that look very different). If it wasn't reliable for exactly identical geometry then Deus-Ex: HR (with its depth pre-pass) wouldn't render properly at all. IEEE754 floats (and any other currently commonly used decimal number implementations for that matter) are entirely deterministic, so rounding errors don't matter as long as you push the exact same values into the exact same formula. [editline]22nd September 2015[/editline] [QUOTE=Karmah;48734118][...] Or am I misunderstanding you?[/QUOTE] Kind of. Mostly you were just wrong though :wink: [QUOTE]In any case, although the shadows worked, it ultimately was a shit system and fucked too many other things over, so I'm amidst attempt #3 of shadow mapping. I see way more positive things about shadow mapping then stencil shadow volumes.[/QUOTE] The only downside that I know of is that they don't project fine detail particularly well (which is a plus in many situations anyway).
I was trying to learn SFML, so I made a thing. It's pretty lame but I had fun making it. Next up is to clean up the code so it actually makes some sort of sense, and then polish up the game. [URL="https://github.com/Dooskington/Orbs"]Here's[/URL] the source. [vid]https://my.mixtape.moe/uvfgno.mp4[/vid] Apologies, how do I make the video embed smaller?
[QUOTE=Duskling;48734396]I was trying to learn SFML, so I made a thing. It's pretty lame but I had fun making it. Next up is to clean up the code so it actually makes some sort of sense, and then polish up the game. [URL="https://github.com/Dooskington/Orbs"]Here's[/URL] the source. :snip: Apologies, how do I make the video embed smaller?[/QUOTE] Upload it to youtube.
[QUOTE=DrDevil;48734429]Upload it to youtube.[/QUOTE] Alternatively: [quote] it. The size of that [vid] is fine though.
[QUOTE=Handsome Matt;48730710]if you can obtain it, [URL="http://www.sweetscape.com/010editor/"]010 editor[/URL] is probably the best hex editor out there for reverse engineering (it'll save you a lot of trouble using that online hex converter)[/QUOTE] A close second of mine is [url=http://mh-nexus.de/en/hxd/]HxD[/url]. It's free. It also has templates that you can create yourself.
[QUOTE=Fourier;48730994]No it's 2 meters I think[/QUOTE] Is that, like, stacked end-to-end, or stacked on top of each other? Can you cleverly cut the brownies and [I]then[/I] arrange them to achieve the 2-meter requirement? Can one simply place two brownies (or one cleverly-cut-in-half brownie) 2 meters apart? I need details!
I've been experimenting with spiking neural networks.. this bastard won't paint the Mona Lisa no matter how hard I try. [IMG]https://dl.dropboxusercontent.com/u/27714141/mona-lisa-1.png[/IMG] Each network gets a 'paint brush', an algorithm tracks its progress and rates the genome based on image similarity analysis. There's a twist to this though: each neuron has a physical manifestation (circles in a physics sandbox) and they tend to seek each other out and collide when they spike in unison, forming new connections. You can call it applied Hebbian learning or something. The screenshot is only for show though, since the simulation slowed down to a crawl when I applied Dijkstra's algorithm to the Voronoi diagram's vertices to produce those fancy-looking synapses. (red is inhibitory, blue is excitatory)
Sorry, you need to Log In to post a reply to this thread.