Update @thebennybox
Update  @thebennybox
Uploaded March 2018 | Updated November 2026, 3 hours ago
I talk about why I've been on hiatus, plans for the future in this channel, and I ramble about a lot of projects and other things.

In short, I do hope to make more videos eventually. I don't know specifically when, but I'm doing multiple projects that will likely lead to more content at some point.

Also, minor note: When I discuss the frustum culling code, I remark that in theory, the technique shown should be cheaper than typical methods. After more thought, I realized it's a little more nuanced than that. In terms of worst case performance, my technique should be better, but the more typical frustum culling technique makes it easier to early out if the mesh is on the outside of one of the first few planes tested. So really, it depends. That said, it's almost entirely a moot point since both techniques are very cheap and almost certainly not part of the critical path. So I'd go with whichever is easier to implement. That said, after even more thought I began experimenting more with frustum culling techniques because I was still not entirely satisfied with the method in the video (namely, the lack of good early out abilities) and I came up with an even better technique. It's 12 dot products and 6 compares for the comparable potion, but I went a step further and tested the AABB planes vs. frustum points because it creates a more accurate cull, the project ran noticably smoother, and even including those calculations the method was still less expensive than what I had in the video. For some reason, youtube does not allow angle brackets in the description, but I've replied the method to one of the comments.

Another note, although I did say the genetic algorithm approach to BSP construction did not work as well as expected, note that I was using pretty close to the most basic possible approach, and only on relatively small to medium sized meshes. Genetic algorithms do have a lower asymptotic time complexity than the heuristics (at least if you're using the heuristics to their full potential) and also should theoretically converge on better solutions given enough iterations; in other words, in theory they should produce better trees in notably less time given a large enough mesh to work with. So if you're working with large, complicated meshes that need very well optimized BSP, I'd say genetic algorithms are still worth exploring further, especially if you find a particularly good technique for crossing two BSP trees. In my case however, the further exploration didn't make sense, as it was becoming obvious at that point that BSP trees would be less useful for my purposes than I had hoped.
Update#5 Intro to Modern OpenGL Tutorial: 3D Motion#4 3D Game Programming Tutorial: Dynamic Components (ECS Part 1)Update: Behind the Scenes#0 3D Software Rendering Tutorial: FAQ#44 3D Game Engine Tutorial: Quaternions & Transformation#13 Wolfenstein 3D Clone Tutorial: Hacked SpritesOpenGL Game Rendering Tutorial: Implementing FXAAUpdate: New stuff, Ludum Dare Post-Mortem, and more.#16 3D Software Rendering Tutorial: Perspective Texture MappingOpenGL Game Rendering Tutorial: Generic Directional Light ShadowsFAQ: How do I debug issues when following your tutorials?
thebennybox |

Update

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER