next »

[SuperDisk]

I decided to take a break on my main project and work on a recreation of Pyoro (in part to see how easy my newly created engine was to use, and also as a tech-demo for a pseudo browser plugin I'm making)

The engine was just ripped straight out of my main project and stripped of all game specific stuff...

I've gotten the basics done in about 10 minutes. I'm proud of you, engine.  

[_Tommo_]

This is more like "happy designer", but:

The gameplay modes in our game were becoming quickly unmanageable, each one overlapping the others that little bit that made if/else chains based on the mode a mess.
And what's worse, changing one mode unvariably broke some other mode!

So, I've broke down each mode to a simple list of boolean flags enabling each single "atomic" feature that makes up the whole mode.

Result: modes are now loaded from file and can be defined by just adding a new flag set that makes sense, if/else chains are destroyed, it is saner and everything worked after a few tries (despite having changed more than 300 lines over 13 files)

So, if you are trying to implement FPS modes/RPG classes/RTS units, etc: do break their features down to the simplest on/off switches and let them emerge from their interaction, don't try to code each separately!
Highly recommended

[goshki]

Finally implemented Bezier curves to draw the pole in Pole Vaultage! It's HaxeFlixel plus Nape so there was some local to world point conversion to find all the required points on pole fragments (the pole consist of 4 fragments). I've still got to implement the case when the pole breaks and splits into several fragments.

[Dr. Cooldude]

I just compiled babby's first Mono/Android app.

[EdgeOfProphecy]

After futzing around for a couple days and being hopelessly incompetent with math, I got bullets working properly in this pet FPS project I sometimes work on.  Suuuuuper thrilled

Next I need to make some modifications to the character controller, but I don't know when I'll be able to get around to it.

[_Tommo_]

Delaunay didn't stand a chance



still to do: hole detection via contour parity (already somewhat working but disabled here), and extruding.

I remade the thing from scratch because FTGL was too big and required gluTesselator, and OpenGL ES doesn't have that.

[ALourenco]

YAY! I finished one more game. This was a simple asessement work for college and I had to send it today. I worked yesterday untill 3am of today and it's worth it because its working quite good. At least I already have some "GameJam" experience XD.

[Geti]

Babby's first pathtracer as a coding exercise.

Rendering in "check number of steps per pixel" mode because that looks coolest at the moment.

Fractal simplex noise and ray surfing. Need to thread it, then find more ways to optimise without resorting to the GPU because I'm a stubborn fucker/masochist/dont have a new gpu - will probably have to see if its faster on integrated gpu or not at some point though

Will continue later probably, would like to try out arbitrary geometry with distance fields at some point but then the ray surfing wont work and I'll really have to get it running on the GPU, or else cut the res.

[ericmbernier]

I recently had a problem that took me from very grumpy programmer to very happy programmer. I just finished up a Flash game, and this game has an intro, with a mute button in the top right corner of the screen. If the player played the first level, and then quit the game to the level select menu there were now two mute buttons in the top right corner, on top of one another. This caused the mute buttons to both register mouse events, thus canceling each other out, making the mute button seem broken. FGL found this bug in my review, and really slammed the game's quality, but gave me a window to fix it. Turns out I was merely missing a removeAll() call in my code when transitioning between the Intro Screens and Game World. I must've thought my mute logic was broken for at least 45 minutes to an hour 

[powly]

Babby's first pathtracer as a coding exercise.

Don't know if it's strictly a path tracer but it does look pretty cool! In my (admittedly limited) experience distance fields are mostly good for analytic surfaces as calculating the distances into a grid takes a long time, stepping along it isn't the fastest thing ever and you still get some artifacts. For ray traversal and triangle meshes you'd probably want some kind of a tree structure instead.

[Geti]

I'm talking about going the completely procedural route for the geometry, defining the field rather than trying to generate it or store it as anything other than code. I know it's possible to store it in voxels but A) the overhead!!! and B) modifying the voxels is then more or less intractible as far as i can see, at least for the case of removal and C) there's probably better ways to render voxels.
The distance fields are purely there for the purpose of speeding up the stepping though so I'm not sure what you mean by not the fastest thing ever - faster than regular raymarching by anyone's standards I'm sure

Working on fiddling with the distance fields now, defining primitives and fiddling with displacements on them. We'll see what I can get done by the end of the hour when amy gets out of her tutorial B^)

@eric: haha I know that feeling, funny that they picked up on it and slammed it - might hint that you need better sound, or might just be something they make a point of checking

[powly]

Ah, I thought that by arbitrary geometry you meant voxels or triangles where marching a distance field can result in quite a few steps around corners compared to octrees or such. If you intend to keep it analytical, ray marching with a distance field is the best option and is pretty magical - very little code can produce wonderful outcome.

[Geti]

lol current implementation needs a lot of work :D same "time to render" metric as before, one tenth of the rendered distance, half the res, still choppy. Got to look at where I'm optimising badly

Looks kinda neat though.



Playing with domain distortions, this time just rendering the z buffer. Gonna see how slow it gets dropping some noise in there now, then i should think about cooking dinner.



did the noise, performance seems to not really care too much about the cost of the distance function so much as casting all those damned rays (especially the ones near-parallel to the ground for obvious reasons). Might need to add some clever world helper structures soon, it's already got a sort of LOD built into the pillar function (to avoid evaluating all the boxes, the noise etc for rays that aren't even close to intersection) but that could probably be extended somewhat. I could also just make stuff occlude that ground plane, or remove it entirely

[Geti]

Sorry for the double post, made some more progress though:

lighting, including rudimentary soft shadows and cheap as chips geometric ambient occlusion. Gonna build some more interesting geometry and figure out colouring tomorrow probably, dinner and then work time.



Cranked up the settings to show the AO and lighting as much as possible, you can see the start of bump mapping too, in the grainy AO (without this it's smooth as butter, though with this geometry just looks like a bunch of square gradients :D).

The corners show up one of the probably bad optimisations, actually - the squares return an incorrect distance metric at this point. Got to figure out a way to make that more correct without making it slow before I start adding more primitives...

Gotta say it feels weird doing things outside of scripts these days, but it's fun trying to optimise this lower level stuff

edit: I initially lied about dinner, fixed the distance thing after reading some of inigo quilez's stuff again. Got a feeling that guy's maths knowledge is a step or three above mine.



I'll re-optimise later on, glad the AO is all glittery now and the soft shadows are behaving even better.

[Quarry]

This is way too cool, does it work well enough when you move the camera around realtime?

[Geti]

The camera in all of those shots was moved there in real time, though for the most recent one with 8fps it was pretty fucking choppy getting there. It works with however you set up the rays to be traced, at the moment its a simple lerped frustrum (ie normalise(-1->1,-1->1,1); ) with a transformation on it, could be orthographic or whatever easily.

it's also not got a proper camera rotation in the sense of a transformation matrix as I haven't actually written that code yet. it's all just inline rotations and whatnot.

I'd get a new screenshot from where I got mixing functions working properly and made the shadowing more elite, but that would require admitting I'd kept working on it when I shouldn't have.

Needs more optimisation, and a proper way of defining objects in the scene sounds like both are in opposition of each other. I'll grab a screenshot of where I got to once I've pushed this KAG code.

[Geti]

Double post to make sure you see this quarry since ITS FOR YOU MAAAAN

The function here is a sphere deformed with soft, medium frequency simplex noise and the previous cube tower thing, blended together using a smoothstep. Figured it was a nice one to end the night with.




fps is getting real low in these ones but as I said, needs optimising, including tiled threaded rendering and potentially biting the bullet and harnessing the parallel winning of the GPU.

Really happy with how the shadows are working out now, would love to make them cheaper though. This is heading quickly towards non-realtime, which I dont really want it to do. We'll see how it goes long term as a potential pet project

[Quarry]

Cannot contain my boner

[Schrompf]

This is so copulating cool.

[sublinimal]

I really like the visual style in those. They look like cryptic transmissions from outer space or something.