3 New Notifications

New Badge Earned
Get 1K upvotes on your post
Life choices of my cat
Earned 210

Drag Images here or Browse from your computer.

Trending Posts
Sorted by Newest First
t
trishnafile5 25.10.20 06:27 pm

Raytracing Scene on GPU?

So the name of the species sums it up, but I've been struggling with this issue for a while, because it seems that most broadcasting documents just take it for granted. How are you going to broadcast a scene filled https://shareit.onl/ individual tops/triangle buffers and matrixes for each model? I know how to effectively trace the rays of the triangular mesh octra, but how do I access any number of models in the scene for any given beam that can affect any of them. You can't have buffer pointers on other buffers, can you? Do I need to store the entire scene in one large continuous buffer? As you effectively add or remove the grid, you will have to remake the entire buffer...

On CPU it's easy because you can just have some sort of data structure for the scene that contains pointers to the data for each mix, you can then do a wide-phase and narrow phase of ray broadcasting to bounce around the stage. Is it possible on the GPU? https://appvn.onl/
1 Comments
Sort by:
M
MunchkiN 616 25.10.20

as I understand the problem and the question - no way because the philosophy of working something on the gpu is ultra parallelism. breaking up the evictions into non-intersecting computational patches and then subsequent iterations to combine the data into the array and on the already deferred calculations etc. until some necessary set of branch trees for reciprocal reflections of the indirect bounce of the indirect bounce of the etd there should not know
on this if I were a cool programmer I would have sawed the breaking scene on the coal miners and the irarch relationship of each charcoal relative to the next en-rays on the triangle unrelated to the other where the rays that do not fall are culled and crushed only with some density in the coal mine, the next logical operation is to search for 4s (let's say) neighboring coal miners and tracing operations with them, further the next iteration already with the set of the previous as a whole itp and complete accuracy will never be achieved and if there is it there will be a completely linear calculation as a spu and will take all the time to gpu as if it is a stump 4 with a frequency of 800mhz about like that of a gpu. so there will always be glitches and noise on the gpu. and every development of thought in this direction.
I'm not a programmer.