SLIDE 1
Traditional BVH Traversal
Two BVHs are traversed
Using either a stack or a queue Using a descend rule descending either tree Descend both trees simultainiously
For each descend, the BVs in the nodes are compared for
- verlap
2
GPU Accelerated Tandem Traversal of Blocked Bounding Volume - - PowerPoint PPT Presentation
GPU Accelerated Tandem Traversal of Blocked Bounding Volume - - PowerPoint PPT Presentation
GPU Accelerated Tandem Traversal of Blocked Bounding Volume Hierarchies Jesper Damkjr and Kenny Erleben { damkjaer,kenny } @diku.dk Department of Computer Science University of Copenhagen October 2009 Traditional BVH Traversal Two BVHs are
SLIDE 2
SLIDE 3
Naive BVH on GPU
One pair of BVHs per Thread Upper space bound for stack k (c − 1) max (height(A), height(B)) ,
- max. cardinality, c, and size of two BV node references, k.
Shared memory too small and global memory too slow
3
SLIDE 4
Use Blocks
1 Block ≡ Each node has 4 children If overlap ⇒ 16 new overlaps Less data to transfer and more work per thread
4
SLIDE 5
Use Double Buffered List
Stack/Queue ⇒ Double buffered list Swap input/output paris for next pass
5
SLIDE 6
Memory Trick Needed
6
SLIDE 7
Need Imaginary Nodes
Less than 4 children ⇒ fill with imaginary nodes Fills up space ⇒ part of calculation time ⇒ use sparesly
7
SLIDE 8
Blocks with Mixed Internal or Leaf Nodes
Not allowed ⇒ Simpler code
8
SLIDE 9
Internal Block versus Leaf Block
if collide (a, k) ⇒ push (e, k) if collide (a, l) collision ⇒ push (e, k) if collide (a, m) collision ⇒ push (e, k) if collide (a, n) collision ⇒ push (e, k) Redundant results ⇒ add extra check to code
9
SLIDE 10
The Test Setup
Three different configuration types Structured stack Unstructured Pile Rock Slide
10
SLIDE 11
The Test Setup (Cont’d)
For each configuration type
Increasing number of triangles in objects Increasing number of objects
Test against Rapid
Rapid uses OBBs we use AABBs
No optimization of imaginary nodes in BVHs (upto 33%)
11
SLIDE 12
Results
Rapid on Intel Quad CPU using one core
216 343 512 729 1000 192 48 12 1 2 3 Number of objects Stack: Rapid Triangles per object Time in seconds 216 343 512 729 1000 24000 6000 1500 1 2 3 4 5 Number of objects Pile: Rapid Triangles per object Time in seconds 500 1000 1500 2000 2500 24000 6000 1500 0.1 0.2 0.3 Number of objects Rockslide: Rapid Triangles per object Time in seconds
Cuda on ge9800 GX2 using one core
216 343 512 729 1000 192 48 12 1 2 3 Number of objects Stack: Cuda only Triangles per object Time in seconds 216 343 512 729 1000 24000 6000 1500 1 2 3 4 5 Number of objects Pile: Cuda only Triangles per object Time in seconds 500 1000 1500 2000 2500 24000 6000 1500 0.1 0.2 0.3 Number of objects Rockslide: Cuda only Triangles per object Time in seconds
Stack (5-8) Pile (3-7) Slide (2)
12
SLIDE 13
Thanks Questions?
13