Hacker News new | past | comments | ask | show | jobs | submit login

I've recently re-watched John Carmack's 2004 Quakecon keynote (https://www.youtube.com/watch?time_continue=602&v=OOVUZZr655...), which happened during the Doom 3 launch. In that keynote, he mentions that he didn't see any differences between stencil and shadow maps. He mentions that he experimented a lot with both approaches, before settling for stencil shadows which were much faster.

I got curious about that statement, since shadow maps tend to look much different. I also knew that he left part of the experimental renderer in the GPLed code. So, I've decided to go on that rabbit hole of Doom 3 graphics, specifically Carmack's experimental renderer and ended up implementing his approach, as well as adding some poisson disc sampling and fixed the peter panning: https://github.com/klaussilveira/exp-dhewm3

I also spend a lot of time on archaic game engines. I like to call it software archeology.




Can you post screenshots or gifs/video comparisons of your updates to the renderer vs the original? Either directly in the readme or links? Keep up the good work!


Oh, sorry about that! This is D3 with shadow mapping:

https://i.imgur.com/TvnFuDG.jpeg

You can only notice it is not stencil when you get up close:

https://i.imgur.com/WtWojG2.jpeg

https://i.imgur.com/4dniMOT.jpeg

In Carmack's experimental renderer, peter panning was quite high:

https://i.imgur.com/u2ZZJTR.jpeg

This is my version with tweaks and poisson sampling:

https://i.imgur.com/nPI6H5a.jpeg




Consider applying for YC's Fall 2025 batch! Applications are open till Aug 4

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: