@boggo @lritter That being said, a positive version of an "overflow menu" is sometimes referred to as a "junk drawer". It's sometimes handy to have "a bunch of seemingly random tools in one drawer", even though it doesn't look tidy on paper. This kind of 1-deep menu is useful sometimes, but more typically as a user-modified thing (like Favorite Tools). On Dreams we went full extreme and banned junk drawers also :P (I think some may have come back)

@boggo @lritter Of course in a hardcoded app like Dreams we control the sizes of things and they're not user-modifiable, so we won't ever *have* to have overflow, unless it's self imposed. So yeah the "radical" stance is to not allow it :P

@boggo @lritter Yeah "overflow menu" is the common UI term for these, since it's what happens when the item amount "overflows" the allotted container size (say 8 tools, or whatever). It comes from desktop UI where you can resize the window and thus tool panels etc may get even smaller (or larger), unlike Dreams. So when a panel is limited (by user control), there is only one option, which is to "overflow menu".

Since #Barbie is a thing again, here’s a reminder on “shader permutations” (originally I did this almost 10 years ago).

#introductions I am a lead programmer at Sucker Punch Productions. I do bits of everything low level (graphics, physics, scripting VM, build pipeline, etc.).

@demofox have you seen this? https://tellusim.com/improved-blue-noise/

@demofox @lorenschmidt You may find this useful!

@maxliani @michaelbartnett Sounds a lot like the Modeler scenegraph then :) I would just go with an array per property, and just index into that. It's working pretty well for us so far. Maybe do back-of-the-napkin for how much memory it will waste on properties that some nodes use and some don't. Dreams had a similar scheme but compacted some properties and not others. For Big Data (like meshes etc) can use a userdata handle.

@maxliani @michaelbartnett I think it depends a lot on what your "objects" are, ultimately. Classes work just fine unless you have a ton and it gets unruly. Arrays-per-property is fast if your objects are relatively homogeneous, but it's more typing than just classes 🤷‍♂️

Took part in the @blenderartists 10 cube challenge:

Recreate an iconic movie scene by using just 10 cubes!

Here's "The Lion King":

#b3d

@Donzanoid @stephan_gfx @lorenschmidt @jcowles Sames.

@aras @acegikmo But how do you get near the top of an AABB 🤔

@stephan_gfx @lorenschmidt @Donzanoid @jcowles Can I just say that this thread is mega cool, but honestly how did we go from totally readable forums to this madness of "kinda weakly linked and maybe disappearing messages with a bunch of @ incantations" lol
Wouldn't it be great if for each post, whenever someone started replying, it would just make a mini forum post and it was "just form thread" from then on? Feed is great 👍 , replies are a goddamn mess hah

@lorenschmidt what do you mean by complex samples?

@Donzanoid @jcowles @lorenschmidt @stephan_gfx Ah ok so it turns out I already did this :P That's what took me from 10x slower to 3x slower than floats (nowhere near as optimal as your code, but I think I got the "big" gains out already, the next big one I think is to just speed up the hash as you mention).

@jcowles @lorenschmidt @stephan_gfx @Donzanoid I think it's probably doable, but I'm trying to wrap my head around what that looks like for moving multiple noise grids (as in fbm noise) into camera space. They all have different scales (which are at least conceptually doubles) and all have integer offsets. So it's just a lot more "messy" than just doing transform math in doubles, then just converting to float when "stuff is close enough to camera".

@jcowles @lorenschmidt @stephan_gfx Yeah I think it's probably doable, that's what I've been trying prior to giving doubles ago, but as you say you still have precision issues far away. I sort of got it to work, but my textures both up close and far were failing, but I think it's because I didn't move those into camera space as well correctly. Need to try again.

@jcowles @lorenschmidt @stephan_gfx Yeah I think so, but I don't think you can get rid of doubles entirely. I think you still need some initial lookup/offset into the right place, then drop into floats for the rest. But yeah the noise itself needs offsetting, not just the camera (so that the cell hashes are correct for the true place in the world etc). It does seem like a lot of faff though, but gonna read that Don Williamson link!

@Donzanoid @jcowles @lorenschmidt @stephan_gfx oooh thanks for the link @Donzanoid ! Devouring...

@lorenschmidt What's fractal noise?