So it seems likely I will be interning at Microsoft and then Facebook starting in May! Pretty excited about the opportunities, but moving away from home is never too easy, especially for an 8 month stretch.
That's next on the chopping block, since afaik it'll need a different shader which will shake up my current rendering system a bit
Made my first contribution to an open source repo. Don't know if they'll merge my changes though. Or if the author is even still watching that repo (like 3 months since they updated the readme and 1 year since they added any code)
i've made pull requests to repos that haven't been touched for 2 years but the author still merged them
You won't need a different shader. Render the opaque geometry first, then disable depth writing and then draw the transparent geometry with blending enabled with SRC_ALPHA and ONE_MINUS_SRC_ALPHA blending functions.
Don't you still have to depth sort the transparent geometry ahead of time so that if a few of them overlap they blend properly?
I haven't even tried yet
Just tried setting up MonoGame, but it looks like the VS 15/17 templates are broken... I can't create new projects I've even tried installing the old XNA 3.0, the templates get created but it won't build.
Mostly looking to do 2D/2.5D orthographic. Suggestions for managed game platforms along those lines? .NET Core would be cool, but I doubt there's any good frameworks there yet. If there isn't anything managed, I've played with OpenTK before so unmanaged stuff isn't too bad either.
inb4 Unity3D.
That's why you disable depth writes first, if you use a function which produces the same result irrelevant of the draw order, you basically got order independent transparency. Granted, it doesn't look that great but it's really simple to implement.
Oh yeah forgot to mention, culling should be disabled too.
At work I've begun abusing bot frameworks to create two way chat integrations for Salesforce. I want to post about the general idea, but I'm going to wait until I start my new job before I do
https://github.com/luca-piccioni/OpenGL.Net
https://github.com/Chman/Glfw.Net
These two coupled together work like a charm
Oh, cool. Thanks for the tip!
https://files.facepunch.com/forum/upload/111926/cc17e9f6-b725-42db-97f3-522bacd4a331/image.png
I really want to get good pixel lighting working next but from what I've tried it's kind of an insane pita. What's the commonly-used solution for having n vertex lights?
What do you mean by pixel lighting? Per fragment lighting is pretty standard stuff. Vertex lights? Like directional lights? Either way, the good way to do a lot of lights is deferred rendering.
Sorry, should have specified (I've been using Unity for years so my terminology is completely fucked)
By pixel lights I meant per-fragment forward lights including shadows and all that fun stuff. By vertex lights, I just meant simple lights that calculate per-vertex and ignore light occlusion etc.
Deferred rendering did seem interesting, but it also seemed like it would be quite the undertaking. I might implement that once I get the other stuff done for this project.
Both deferred rendering and the alternative (fewer, larger monolithic shaders) have their advantages
If you know the number of lights you will need ahead of time, and have more static geometry and just a relatively static scene in general, you can probably get away with forward rendering and optimize the hell out of it and get great performance.
If you need more dynamic scenes with items coming and going, an indeterminate number of anything, deferred rendering can be great. Plus it gets you to learn framebuffers and how post processing techniques work.
https://youtu.be/3WaDQoEzIIE
Wtf is going on here…
Forward does come with the immense benefit that you can have an unlimited number of shading models without any overhead of material buffers etc, which is one of the reasons modern engines are going back to more forward based approaches, with compute based light clustering etc. for lots of light sources and decals, as you said.
WRT per-pixel vs per-vertex lighting, the latter is only used as an optimization for low LOD levels these days, there's no reason really to use per-vertex lighting anymore.
https://files.facepunch.com/forum/upload/149288/d404e206-f47b-4ce1-b123-20247aec860f/a bit off.png
I'm starting to understand what's going on here, but objects held by the gravity gun spin and shake like crazy Also no idea why it refuses to load more than 1 ledge in each VPHY.
Is that from Jcw87 and party's work from years ago? Neat to see someone work on it again if so.
hey assuming you're an SDE who will have a bachelors in CS...
if you're considering going back to msft as an FTE after your internship at fb, negotiate:
fb should give you in total at sign-on around $100k
bounce that off msft and they should give you fbsignon*2 in stock _grants_
msft level determines your base, +/-5k
try to get on a team in azure, avoid wdg, avoid msit, avoid office
No, while I do use DrChat/Gmod-vphysics code occasionally for reference, it's not a fork, I'm redoing it from scratch. Things like multithreading and soft bodies are out of the scope of my project though, but my aim is to cover as much of the API as possible, including functions only used by VBSP and StudioMDL like CollideGetOrthographicAreas.
I went into a job interview. After a few rounds of different people interviewing me, the lead engineer came in and asked me some questions. One was like "What are the object oriented properties of a book?" I kind walked through that and he said "lets say you've got a book, like a small paperback book, and you want to resize it to the size of a textbook. How would you reflow the pages, since there will be more words per page now?"
Then I thought about it and said "... I actually did this..."
He says "Oh, you did?" and I say "Yeah, in one of my projects".
He laughs and says "I don't suppose you'd be able to show me?"
"Yeah I have it right here", pointing to my laptop.
Then I show him my books project, and he says "Nobody's ever had a project for that question". He calls in another engineer and tells me "show him what you just showed me". They both laugh
2 days later they offered me a job
Is my strategy of reading/watching several tutorials on Django over and over again until I slowly start to get it effective or am I just retarded
Keep in mind I've never done backend web before
Addendum: I'm not going to lie, reading documentation on the same framework over a dozen times and still not getting it is seriously discouraging. It's a project I've been wanting to do for a long time but it's just not going very well.
I mean I wouldn't say that's necessarily the case anymore, since deferred has a number of disadvantages over the latent resurgance of forward techniques. Breaking down things into an offscreen render, some compute work for binning/sorting lights, then finally just rendering to the framebuffer is great. It also better leverages the compute power of most modern GPUs, uses less texture bandwidth/memory, and lets you still use nice and simple MSAA techniques. I had planned on supporting deferred and forward+ originally, but now that I'm working on it I'm definitely thinking of just sticking with Clustered Forward as my go-to.
Changing the number of lights is also fairly easy for me, and the lights themselves are moving. It also supports particle effects and lighting, really it's pretty great and I haven't run into many major disadvantages yet. The sheer quantity of dynamic lights you can have is just kinda silly, too
You're right, I meant to say "a good way" instead of "the good way", but yeah there have been some surgence with forward+ that you can't ignore. I just meant it as a stepping point to look at, but it might be not worthwhile enough and one should just look at forward+ instead.
I think it's fair to still have an understanding of deferred - even if only to understand why there's a movement away from it (GPU computational growth far outpacing memory improvements). I got a fair way through implementing it before realizing "lol yeah this kinda sucks".
Also still a bit difficult to find examples for Clustered Forward - Forward+ is good, but Clustered Forward is even better. If one is interested in just having the most performant setup, then I'd definitely recommend Clustered Forward over deferred.
The best way to learn is by doing, in my opinion. For good frameworks, you will be able to understand the why from the how if you work on a few projects that utilize it.
https://i.imgur.com/rX4oY10.png
OOoh thanks! Wasn't too hard to get set up.
https://youtu.be/sa7Rrukd1VI
Absolutely no idea what to do with these bouncy collisions, CCD doesn't help, increasing the frame rate from 66 to 264 helps a bit, but still props fly through walls
i can be your angel or your devil
https://files.facepunch.com/forum/upload/109806/43400caa-6ebd-4174-86a1-f179d3b82530/image.png
Sorry, you need to Log In to post a reply to this thread.
What are you working on? v67 - March 2017
