I think I may have a “new” rendering method that beats traditional rasterization and ray tracing in most general cases, yet utilizes aspects of both. TBD how well it works, I’ll post the results regardless.
-
-
Replying to @gavanw
There are definitely some ways to out perform rasterization, but not on mobile GPUs yet.. The advantages only seen to crop up when you're dealing with Geo on the order of a couple pixels, and you can't get near that on a phone. I'm excited to hear what you found!
1 reply 0 retweets 1 like -
Replying to @SculptrVR
I don’t think my method will beat out dedicated hardware of any sort, unless it had its own hardware support.
1 reply 0 retweets 1 like -
Replying to @gavanw
Oooh! Extra interesting. So it's a theoretical model that can more quickly populate visible surfaces with correctly rendered pixels if given as much hardware support as triangles have.
1 reply 0 retweets 1 like -
Replying to @SculptrVR
yep, in essence a very fast and accurate hashing algorithm that determines where things will appear or are likely to appear (not related to my previous method). Unlike (traditional) rasterization, no overdraw. Unlike (traditional) ray tracing, very little iteration over objects.
2 replies 0 retweets 2 likes -
Replying to @gavanw
Hell yeah! I've spent a lot of time thinking about hashing after switching to chunked hash sets for our octree format. On the CPU you have to be very careful around cache sizes. Access time is actually O(sqrt (N)), but it's discretized with the cache hierarchy.
1 reply 0 retweets 1 like -
Replying to @SculptrVR
Yep - I have been avoiding methods that span large areas of memory or attempt to cache results using lots of data. The "near miss" cost of rays against octrees or SDFs was also a big inspiration for finding a better solution.
1 reply 0 retweets 2 likes -
Replying to @gavanw
Well I definitely give you the benefit of the doubt. It sounds awesome and I'm excited to learn more!
1 reply 0 retweets 1 like
Thanks - I’m still slightly doubtful myself, I usually fail to account for something
-
-
Replying to @gavanw
Yeah I'm with you. I've written something like 10 meshing algorithms for SculptrVR certain that each one would be the new best. But only 3 ever were good enough to see the light of day..
0 replies 0 retweets 2 likesThanks. Twitter will use this to make your timeline better. UndoUndo
-
Loading seems to be taking a while.
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.