Видео

Well, I have a Windows project as well. That one is DX12 though.

@@pm-u3b i have followed you man everywhere when in future when i am done learning basic of computer graphics with opengl as my starting api, i would definitely look at how you do stuff in your marvelous renderers but that is in very distant future.

@@123drakus Best of luck ;)

Hey. This is just a toy renderer, not a real "engine" by any means. It has 0 tooling. If you want to make levels, just use the established industry engines.

@@pavlo2692 they all suck gaming is dead

Well, I'm still here, even though the video is old ;) The scene is called Sponza. The single most popular scene between studying graphics programmers :P

@@pavlo2692 thank you! Also, nice renderer, very inspiring work. I'd like to create my own someday

Thanks :) Started to learn coding in 2011. First job - 2014.

Whatever has more resolution, has less light leaks. Tracing rays per pixel, maybe using a screen space trace for acceleration would give 0 light leakage. Otherwise sometimes, somewhere it'll leak.

@@pavlo2692 ohh I understand now, thats why Lumen is said to use screenspace + other methods, so the areas near camera will have no leak but the areas outside of camera can have leak that we dont see. Thats smart :)) so screenspace is still needed to make a good hybrid GI solution. Thank you so much for ur simple explanation on me cuz I m just interested in this topic even though i m just architect :D

The industry (AAA that is) is moving away from baking and been doing that for years. Sure, baking is still a good tool in certain scenarios, but generally it's all about ray tracing in real time now. And what's typically called GI requires not that many rays.

It's been forever. I don't remember anymore :(. Intel TBB was required, but I can't guarantee that the new versions are compatible.

@@pavlo2692 like basic information. Do we have to run it Visual studio ?

Hey, thanks a lot. I'm glad it was helpful to you.

I don't remember whether those exact papers discuss the sky. Probably they do. But maybe it was some Call of Duty paper which discussed the sky issue, can't remember. Regarding the second question, I just don't remember anymore :(

@@pavlo2692 it’s all good! I’m sure with enough googling I can figure something out! I’m still so impressed! Are you done with the world of graphics these days?

@@thehambone1454 Not at all. It's a full time job now. Making an engine at work, so there's no time and no need for side projects :) Look into spherical harmonic product to apply occlusion data, where sky is obstructed. Something that Activision put out years ago. I think that you project the skybox to the spherical harmonics, but to avoid leakage you bake occlusion data into another spherical harmonics and then you take a product of them, effectively shadowing the light in the directions where it's obstructed. Look for SH code in my repo, maybe there are some comments left for that.

@@pavlo2692 will do! Thank you!!

@@pavlo2692 follow-up to the follow-up question - You say you have a job now working on an engine?! How cool, would you say learning this low-level rendering code is still very valuable (in terms of getting a job and such), it is not why I am learning it but if it means securing other employment ops then that is just icing on the cake. :)

No in this project, no. But generally yes, you can use gbuffer data to quickly generate surfels in real time. Check this out ruclips.net/video/h1ocYFrtsM4/видео.html

@@pavlo2692 Wow, that's pretty impressive and very elaborate. I feel like by the time it'd take me to implement it GPUs would have advanced pass the point of needing it.

Thanks. No personal projects atm. Only a job now. Was lucky to find one where I don't need personal projects to do something interesting

@@pavlo2692 That's great. I'm working my first serious graphics programming job now too, and Iove it, but man do I miss writing my engine.

Sponza

There is no license. It's free.

@@pavlo2692 so is it possible for other developers to benefit from it to implement in open source projects like BLENDER ?

Sure. This really was just a passion project. I learned a lot from other people, might as well give something back to the community@@mcan-piano4718

Yes you are really passionate in ur profession, I admire that :)))))) I wish there were passionate developers like you in open source projects. Ur awesome :)) @@pavlo2692

If you want to make games then I recommend using existing engines. If you want make engines for people that make games then you could start learning dx12 and other related tech. How many hours? As many as you have if you’re serious about it. This particular video was achieved after about ~2 years of work, on and off, I didn’t work full time on it. I also spent a lot of time on architecture, memory management and things like that, you can get a pretty picture faster, that just wasn’t my goal the first year

Yeah, I targeted 1080p 60 frames on my RTX 2070S

@@pavlo2692 man is it hard to implant this? I’m really trying to beef up my rendering in my little engine

@@thehambone1454 It was challenging for sure. I think it took half a year (not full time) since I first seen this tech presented by Nvidia somewhere to a debugged and "done" implementation that you see in the video. And still a lot of optimizations that it should have in a real world I skipped, because I don't have so much time to invest in it. Take your time, don't give up, and you'll progress bit by bit.

oh, looks like others asked, I will refence those comments.

​@@thehambone1454If you want ray traced GI, then check out my other project :) ruclips.net/video/1oZw7PCOC2E/видео.html

Can you tech me how to implement it. I'm a beginner.

Not sure I understand the question

​@@pavlo2692 I meant the probe shadow bias. float3 selfShadowBias = (surfaceNormal * 0.2 + viewDirection * 0.8) * (0.75 * D) * B; Great implementation! I don't see any light leaks or shadow leaks in the video :D , but it didn't give me great results. Did you find this pretty robust?

@@ravikhannawalia3107 Oh, this bias is developed by Nvidia, arxiv.org/pdf/2009.10796.pdf While these little formulas can look easy, it's often not easy to come up with them.. even for Nvidia

@@pavlo2692 Yes totally agreed! Looking forward to your dynamic surfel GI implementation hahaa

Thanks. What you see is exactly the thing you mention. I use their ratio estimator and denoise 2 images separately before dividing.

@@pavlo2692 Yes looked like that seeing the good results. What did you use for ghosting and disocclusion? It is pretty damn robust.

@@ravikhannawalia3107 developer.nvidia.com/gtc/2020/video/s22699-vid + ruclips.net/video/U5LXWt2dCXs/видео.html

Rendering techniques are independent of graphic APIs, it doesn't have to be DX12. Regarding the DX12, I started with Frank Luna's book, but quickly switched to reading docs and talking to people (on the internet and colleagues), discussing DX12 engine architecture approaches with them, writing my own engine and seeing what works and what doesn't

Getting to know that all of this is done by a single person in amazing! You seem very passionated about global illumination. First I tried learning DX12, but didn't get any resources, but then I tried to learn DX11, but I couldn't find them too. Can you suggest me a few resources? Do you really feel the performance difference in a lower level api?

@@alexpal51 Hi. Try Frank Luna's book, he has both DX11 and DX12 versions. Other than that I can only think of random tutorials scattered around the internet. Regarding performance, getting the same performance as DX11 on DX12 is actually more work. I also wouldn't say it's that important on such a small project like this. In modern big games though it's worth it if you have the manpower for all the 'proper' engineering :)

@@pavlo2692 Thank You very much, I'll surely check those books out. Your renderer is amazing, I can't appreciate enough

@@pavlo2692 don't look at my dp, that's not me I am a not a girl

This particular project took 2 years. Although, I wasn’t working full time on it since I had a job at the same time

@@pavlo2692 cool! How much OpenGL and c++ experience did you have before this?

@@JaimieX OpenGL - none. C++ absolute basics from university times

Thank you very much. The hardest was robust denoising, took so long. And keep going as well, the pretty picture at the end if worth it imo :)

Thanks. Pretty much all pipeline is a collection of compute shaders. I only use vertex/pixel when I need to render geometry - in GBuffer pass. Also, shading pass is done in RayGeneration shader, everything else is compute.

Hi. I'd recommend to start with this www.davidkuri.de/downloads/SIC-GI.pdf and then proceed to jcgt.org/published/0005/04/02/. It is what I've implemented, but not to the full extent though. Ubisoft's The Division uses a similar probe-surfel system, which is described here: ruclips.net/video/04YUZ3bWAyg/видео.html

@@pavlo2692 Wow, nice information, thank you and well done :D

This was run on GTX 1080, 1080p@60

Egleson Beltran yes

Hi. Check out a topic I created www.reddit.com/r/opengl/comments/andaa1/want_to_share_a_renderer_i_made_to_educate_myself/ You also may want to look at the comment section, I've answered several questions there

@@pavlo2692 This is funny. I stumbled across this video while searching for some OpenGL PBR rendering, turns out I already saw your reddit post back when you posted that.

Hello. Probably not, because I've never worked with Cinema 4D. Though I'm not completely sure what you meant there, specifically.