@demofox@mastodon.gamedev.place
@demofox@mastodon.gamedev.place avatar

demofox

@demofox@mastodon.gamedev.place

Graphics and game dev research. previously nvidia, blizzard, monolith, others. graphics, audio synth, exotic computation. No gods, no masters. http://blog.demofox.org.

This profile is from a federated server and may be incomplete. Browse more on the original instance.

jeffjarvis, to random
@jeffjarvis@mastodon.social avatar

Google blew it. They could have stood back and said: "Ha! Microsoft is so desperate for hype, it is irresponsibly linking LLMs to Bing. Google instead stands for reliable search and won't do that." But instead, Google did. They knew better. They all knew better. LLMs have no sense of meaning. They should be nowhere near the expectation of credibility.

demofox,
@demofox@mastodon.gamedev.place avatar

@nosmaharba @rodhilton @jeffjarvis if they don't want their jobs and don't want to run companies, they could just quit. Has anyone told them?

brainwagon, to random
@brainwagon@mastodon.social avatar

Struggling with the realization that while a lot of people are really kind and smart, a lot of people are neither.

demofox,
@demofox@mastodon.gamedev.place avatar

@brainwagon Sorry to hear that. Agreed. It baffles me when folks reward people like that.
Like "surely you notice right? no? weird..."

demofox, to random
@demofox@mastodon.gamedev.place avatar

For some reason people are talking about homomorphic encryption.
In game dev, for deterministic simulations like you see in RTSs, you can prevent write cheats by ensuring that the hash of the deterministic sim state matches across players.
You can't easily prevent read cheats though - being able to see the entire map when you shouldn't eyc.
Homomorphic encryption is a way to prevent read cheats.
Fyi!

demofox,
@demofox@mastodon.gamedev.place avatar

Last I checked, fully homomorphic encryption was much too slow to use in game dev. There are partially homomorphic encryption schemes though, where you have a subset of operations you can do, vs FHE which can do any operation.

demofox,
@demofox@mastodon.gamedev.place avatar

@dougmerritt I saw it and put it on the todo list and it scrolled away before I got to it! I just had a look, and it seems like "bootstrapping" has been replaced?! If so that's great news. But yeah there are a lot of new things since I last looked. That's pretty cool.

demofox,
@demofox@mastodon.gamedev.place avatar

@lunarood I'm not in charge of what people are into 😀

demofox, to random
@demofox@mastodon.gamedev.place avatar

Does anyone know of any code laying around the net that distributes points on a mesh in a blue noise distribution?
A student intern i work with is looking for this. It's tempting to write it, but im also kinda swamped :X

demofox,
@demofox@mastodon.gamedev.place avatar

Mitchell's best candidate would work well here, and would give a sequence instead of a set (use the first N of M total points, for any N), but what makes it more than a 1 hour task is the mesh connectivity, and multiple paths through triangles to the same points that you need to calculate distance between.

demofox,
@demofox@mastodon.gamedev.place avatar

@lritter well i can choose a triangle uniform randomly based on size, and can uniformly generate a point in a triangle.
That'd give you candidates for MBC.
But now you have to calculate distance between points :/

demofox,
@demofox@mastodon.gamedev.place avatar

@lritter so we can keep whichever white noise candidate is farthest away from whatever blue noise point it's closest to.
Then rinse / repeat :P

demofox,
@demofox@mastodon.gamedev.place avatar

@breakin @lritter you still need distance between points on mesh for lloyd relaxation. im more a fan of MBC, but they both have the requirement of needing to be able to calculate distance between points on a mesh, which makes it a longer task hehe

demofox,
@demofox@mastodon.gamedev.place avatar

@breakin @lritter if dense enough, yeah, i bet you could just deal with points within a triangle and the immediate neighbors. I think this is all solvable but not ~1 hour of work haha. sigh...

demofox,
@demofox@mastodon.gamedev.place avatar

@breakin @lritter basically like.. why didn't one of us already write this?! past selves were slacking

demofox,
@demofox@mastodon.gamedev.place avatar

@breakin @lritter yeah that'd work, and xatlas is good at that. But, unless you deal with triangle connectivity, the seams of triangles will be fubar

demofox,
@demofox@mastodon.gamedev.place avatar

@breakin oh yeah, that's true. not good enough for me. sorry for posing an unsolvable problem. Was just hoping some code existed somewhere hehe. TinyOBJ + mitchell's best candidate.

demofox,
@demofox@mastodon.gamedev.place avatar

@breakin oh interesting!
https://github.com/BrunoLevy/geogram

demofox,
@demofox@mastodon.gamedev.place avatar

@breakin it looks really promising

demofox,
@demofox@mastodon.gamedev.place avatar

@mmby working locally on the surface is what I'd do, yeah, agreed.
Islands would work themselves out

demofox,
@demofox@mastodon.gamedev.place avatar

@vethanis neat

demofox,
@demofox@mastodon.gamedev.place avatar

@aeva it turns out it's really hard. Sorry for not realizing.

demofox,
@demofox@mastodon.gamedev.place avatar

@breakin it specifically mentions being able to make a voronoi diagram and using Lloyd relaxation to optimize the point sets, which is an algorithm for making blue noise point sets.
I think this is the winner.
Thanks a lot!
https://github.com/BrunoLevy/geogram/wiki/Delaunay2D

demofox,
@demofox@mastodon.gamedev.place avatar

@breakin oh darn it. It's not on a mesh.

demofox,
@demofox@mastodon.gamedev.place avatar

@mmby @breakin eesh, i wish this problem was simpler hehe.
finding the distance between 2 points on a mesh doesn't SOUND that hard. but then you think of the details and urgh...

demofox,
@demofox@mastodon.gamedev.place avatar

@breakin @mmby the hardest problem i can see with the "shortest path between points on mesh" is: what is the straight line as you move between triangles that go concave, convex etc. am i overthinking that? it sounds hard.
The "which triangles should i take to go the shortest path" maze solving seems easier at least.

demofox,
@demofox@mastodon.gamedev.place avatar

@litherum @breakin @mmby thank you :)

demofox, to random
@demofox@mastodon.gamedev.place avatar

When AI hype has settled some, I'd like to see neural primitives be considered to be part of standard CS education along with other ADTs.
Hype makes ML look like too good to be true magical algorithms, and then fails because it was a grift all along. But, there is legit value.
When you watch educational videos on auto encoders, U nets, etc etc, they talk about specific things they are good at to fit in a larger solution.

demofox,
@demofox@mastodon.gamedev.place avatar

@gpakosz yeah!
I saw an interesting article the other day related to bloom filters, hyperloglog etc.
https://www.quantamagazine.org/computer-scientists-invent-an-efficient-new-way-to-count-20240516/

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • provamag3
  • kavyap
  • DreamBathrooms
  • cisconetworking
  • magazineikmin
  • InstantRegret
  • Durango
  • thenastyranch
  • Youngstown
  • rosin
  • slotface
  • mdbf
  • khanakhh
  • tacticalgear
  • JUstTest
  • everett
  • modclub
  • Leos
  • cubers
  • ngwrru68w68
  • ethstaker
  • osvaldo12
  • GTA5RPClips
  • anitta
  • megavids
  • normalnudes
  • tester
  • lostlight
  • All magazines
    •