myth0genesis

I'd like to see assignment of multi-valued swizzles, as well. I don't think looking at the percentage of lines in shaders using a given functionality is a good metric to...

How so? I've used that feature in nearly every shader I've written and not once have I ever experienced a bug because of it. If you know a particular implementation...

I can confirm the fix above works for me on Windows 11 and Neovim 0.10.0.

Just bumping this to say this still works and wondering if there has been any testing done elsewhere to see if this works for others.

Here's the double XOR-rotate inversion function I forgot to include: ``` uvec4 dblxoro128_inv(uvec4 x) { for (int i = 0; i < 7; i++) x ^= x.yzwx ^ x.zwxy; return...

Okay. I found that only three, instead of seven, iterations of the rotation inversions above will work. I collapsed the inversion function by simplifying the cyclic permutations: (I+C³²+C⁶⁴)(I+C³²+C⁶⁴)(I+C³²+C⁶⁴) (I+C³²+C⁶⁴)(I(I+C³²+C⁶⁴)+C³²(I+C³²+C⁶⁴)+C⁶⁴(I+C³²+C⁶⁴)) (I+C³²+C⁶⁴)((2I+C³²+C⁶⁴)+(C³²+C⁶⁴+C⁹⁶)+(C⁶⁴+C⁹⁶+C⁰))...

Yeah, the constants were arbitrarily chosen from the list in #19 just to get something to start off with. There was no statistical motivation for choosing them, but I'm reasonably...

The shifts being linear were also an arbitrary choice, as I swizzle the constant vector that contains them between steps, hoping that would be enough. I also thought about using...

If you wanted to extend the hash above to higher bit counts, it would probably work fine for any power-of-2-times-32-bit inputs, but in order to ensure bits propagate across the...

Just thought I'd give an update on my GPU hash test suite progress. I'm too clumsy with GitHub to really maintain a codebase here yet, as I'm still pretty new...