Binary Greedy Voxel Meshing Algorithm
github.com
github.com
He very arrogantly opens the video by basically saying "your mesher SUCKS compared to mine". What he doesn't tell is that he tested his mesher with 3 block types. His mesher requires separating each block type in the chunk and making copies of the chunk data with only that block type present. Of course, this scales horribly - if you have a reasonable 500 block types in your game, you would run out of memory on any computer which isn't a server with 128GB+ of RAM.
He also strongly implies that greedy meshers are just as fast as regular meshers - which is straight up false. He compares his optimised mesher to his really bad first naive mesher, which is just cheating. He also has zero clue about what he is doing - he is spending the first 22% of the function allocating the vec.... which he could trivially avoid by pre-allocating a vec (or a bunch of them if you want to do meshing in parallel) then just reusing them for every mesh.
He has zero clue about profiling and optimisation or even about how to make an actual game which is not a tech demo, yet he confidently spews bullshit to aspiring gamedevs who believe him because they don't know any better. This is truly sad and con artists like this should be shamed to the ground for misinforming new developers.
the problem isn't with the excitement, the problem is being arrogant and being an asshole but not having the leet skillz to actually justify it....
edit: they are doing harm by creating unrealistic expectations towards new gamedevs and misleading them in what matters for performance. (spoiler alert the bit manipulation looks clever but that isn't the majority of performance optimisation now, this is not 1996)
Video explainer: https://www.youtube.com/watch?v=qnGoGq7DWMc