RAYTRACING SCIENTIFIC DATA IN (R) NVIDIA OPTIX WITH TM GVDB - - PowerPoint PPT Presentation

▶
raytracing scientific data in
SMART_READER_LITE
LIVE PREVIEW

RAYTRACING SCIENTIFIC DATA IN (R) NVIDIA OPTIX WITH TM GVDB - - PowerPoint PPT Presentation

Nov 11, 2023 573 likes •1.1k views

April 4-7, 2016 | Silicon Valley RAYTRACING SCIENTIFIC DATA IN (R) NVIDIA OPTIX WITH TM GVDB SPARSE VOLUMES Rama Karl Hoetzlein, April 6 th 2016 CONVERGENCE Video Motion Games Pictures CONVERGENCE Physical Sciences Video

slide-1
SLIDE 1

April 4-7, 2016 | Silicon Valley

Rama Karl Hoetzlein, April 6th 2016

RAYTRACING SCIENTIFIC DATA IN NVIDIA OPTIX WITH GVDB SPARSE VOLUMES

(R) TM

slide-2
SLIDE 2

CONVERGENCE

Video Games Motion Pictures

slide-3
SLIDE 3

CONVERGENCE

Video Games Motion Pictures Data Visualization Mathematics Physical Sciences SpaceX

slide-4
SLIDE 4

CONVERGENCE

Video Games Motion Pictures Data Visualization ProVis / CAD NVIDIA OptiX

GVDB

Mathematics Physical Sciences SpaceX

slide-5
SLIDE 5
  • Science - SpaceX
  • Volume Visualization
  • Motion Pictures – OpenVDB
  • Provis Graphics – NVIDIA OptiX
  • GVDB S

parse Volumes

  • Results

OUTLINE

slide-6
SLIDE 6

SPACEX

Wavelet Simulation Source Data S paceX Dragon Capsule Re-entry

3D S parse Volume data: 3864 x 944 x 464

Adam Lichtl, S tephen Jones, GPU Technology Conference 2015

slide-7
SLIDE 7
slide-8
SLIDE 8

SPACEX

Data Preparation

Wavelet Cloud Data Visualization

slide-9
SLIDE 9

A Numerical S tudy of High-Pressure Oxygen/ Methane Mixing and Combustion of a S hear Coaxial Inj ector, Nan Zong & Vigor Y ang, AIAA 2005

NVIDIA INDEX HPC Visualization Using NVIDIA Index

Tom-Michael Thamm, Christopher Lux, Marc Nienhaus

S ee show floor!

slide-10
SLIDE 10

VOLUME VISUALIZATION

Dense Volumes

16 x 16 = 256 data values

slide-11
SLIDE 11

VOLUME VISUALIZATION

Dense Volumes

16 x 16 = 256 data values

slide-12
SLIDE 12

VOLUME VISUALIZATION

Dense Volumes

slide-13
SLIDE 13

VOLUME VISUALIZATION

Dense Volumes

8 empty steps 5 sample steps

slide-14
SLIDE 14

VOLUME VISUALIZATION

Sparse Volumes

2 DDA skip steps 5 sample steps 52 data values (instead of 256!)

slide-15
SLIDE 15

VOLUME VISUALIZATION

Topology Value Atlas

slide-16
SLIDE 16

VOLUME VISUALIZATION

Octree Topology Tilemap Topology

Many levels Every level divided in half Only two levels Map has uniform divisions

slide-17
SLIDE 17

SPACEX

Data Preparation with OpenVDB

Wavelet Cloud Ideal for GPU Simulation. Difficult to directly visualize. OpenVDB Grid Sparse, Spatial Grid Hierarchy. Efficent storage and visualization.

slide-18
SLIDE 18

MOTION PICTURES: OPENVDB

slide-19
SLIDE 19

MOTION PICTURES: OPENVDB

S parse Hierarchy of Grids

VDB: High-Resolution Sparse Volumes with Dynamic Topology Ken Muset h, Dreamworks Animation. Transactions on Graphics, 2013

Novel topology for very large volumes. Multi-core CPU Design

slide-20
SLIDE 20

OPENVDB

VDB Topology

Many levels Each level is a grid Each level has its

  • wn resolution

e.g. top = 4x4 mid = 3x3 brick = 4x4

Key features:

Can store

very large volumes

with only a few levels. Efficient to traverse, since every level is a grid.

slide-21
SLIDE 21

OPENVDB VS AMR

S parse Hierarchy of Voxel Grids

OpenVDB: 1) Children are strictly contained 2) Spatial extent is a voxel of parent 3) All siblings have same resolution Adaptive Mesh Refinement (AMR): 1) Children are strictly contained 2) Spatial extent is free in parent 3) Siblings can have different res

slide-22
SLIDE 22

SUMMARY OF GOALS

  • 1. Load and Visualize S

parse Grids

  • 2. Using the VDB Data S

tructure

  • 3. ..with.. Efficient Rendering on GPU
  • 4. ..and.. High qualit y Rayt racing
slide-23
SLIDE 23

1 3

RENDERING MATERIALS DISPLA Y LIVE VIDEO VIRTUAL REALITY

Multi-Display GRID SDK Virtual Reality GPUDirect for Video Video Codec S DK VR Works Iray SDK OptiX VXGI GVDB MDL SDK vMaterials

NVIDIA DesignWorks™

  • Advanced technologies for design applications

https:/ / developer.nvidia.com/ designworks

slide-24
SLIDE 24

1 3

RENDERING MATERIALS DISPLA Y LIVE VIDEO VIRTUAL REALITY

Multi-Display GRID SDK Virtual Reality GPUDirect for Video Video Codec S DK VR Works Iray SDK OptiX VXGI GVDB MDL SDK vMaterials

NVIDIA DesignWorks™

  • Advanced technologies for design applications

https:/ / developer.nvidia.com/ designworks

* NEW *

slide-25
SLIDE 25

OptiX: A General Purpose Ray Tracing Engine, S teven Parker, NVIDIA ACM Transactions on Graphics, Vol. 29, No. 4, July 2010

GTC 2016 Talks: S 6643, S 6244 S 6258, S 6844, S 6261, S 6320

slide-26
SLIDE 26

Acceleration Structures JIT Compiler Scheduler

GPU OptiX API User Application

Mesh Hit Program Ray Shade Program

User Application

Launch!

launch describe

slide-27
SLIDE 27
slide-28
SLIDE 28

NVIDIA GVDB SPARSE VOLUMES

GVDB

NVIDIA OptiX OpenVDB

Efficient GPU Structures, Raytracing and Compute CPU Format and Layout

Data Visualization

High Quality Raytracing

slide-29
SLIDE 29

What is GVDB?

VDB Sparse Voxel Structures entirely on GPU Efficient Direct Raytracing Compatible with OpenVDB files and layout Integrated with NVIDIA OptiX Compute Operations on Sparse Volumes

NVIDIA GVDB SPARSE VOLUMES

slide-30
SLIDE 30

VDB Topology S ame t opology as OpenVDB

Additional features:

Run-time tree configuration Dynamic topology Efficient construction

NVIDIA GVDB SPARSE VOLUMES

slide-31
SLIDE 31

CPU GPU

VDB Topology VDB Topology Value Atlas Brick Voxels

NVIDIA GVDB SPARSE VOLUMES

slide-32
SLIDE 32

624 x 304 x 304 (133 MB/ frame)

RESULTS

A Numerical S tudy of High-Pressure Oxygen/ Methane Mixing and Combustion of a S hear Coaxial Inj ector, Nan Zong & Vigor Y ang, AIAA 2005

slide-33
SLIDE 33

A Numerical S tudy of High-Pressure Oxygen/ Methane Mixing and Combustion of a S hear Coaxial Inj ector, Nan Zong & Vigor Y ang, AIAA 2005

slide-34
SLIDE 34

A Numerical S tudy of High-Pressure Oxygen/ Methane Mixing and Combustion of a S hear Coaxial Inj ector, Nan Zong & Vigor Y ang, AIAA 2005

slide-35
SLIDE 35

Scalability Testing

Fluid simulations from S mooth Particle Hydrodynamics 1,000,000 Particle simulation on GPU using Fluids v.3, http:/ / fluids3.com Generated volumes at multiple resolutions by resampling velocity field Low-res: 512 x 180 x 256, 50 MB/ frame, 60 GB total Med-res: 1024 x 360 x 512, 338 MB/ frame, 408 GB total High-res: 2048 x 720 x 1024, 2.4 GB/ frame, 3 TB total

RESULTS

slide-36
SLIDE 36

Source Data: 1 million Particles Simulated at 20 fps

RESULTS

slide-37
SLIDE 37

RESULTS

512 x 180 x 256 2048 x 720 x 1024

# of Leaves: 12,131 File size: 50 MB/ frame GPU Resampling: 16 seconds/ frame Load Atlas: 450 msec/ frame Raytracing: 27 msec/ frame # of Leaves: 616,444 File size: 2400 MB/ frame GPU Resampling: 804 seconds/ frame Load Atlas: 5420 msec/ frame Raytracing: 64 msec/ frame

High Resolution Low Resolution

38 fps 15 fps

slide-38
SLIDE 38

RESULTS

512 x 180 x 256 2048 x 720 x 1024

# of Leaves: 12,131 File size: 50 MB/ frame GPU Resampling: 16 seconds/ frame Load Atlas: 450 msec/ frame Raytracing: 27 msec/ frame # of Leaves: 616,444 File size: 2400 MB/ frame GPU Resampling: 804 seconds/ frame Load Atlas: 5420 msec/ frame Raytracing: 64 msec/ frame

High Resolution Low Resolution

38 fps 15 fps

48x more data 2x time

slide-39
SLIDE 39

Empty S kipping Isoval Render Volume Render Time (ms)

Low Res 512

6.3 9.8 28.2 26.4 7.4 38.0 43.2 9.6 54.3 58.1 67.4 68.2

35 fps 26 fps 19 fps 16 fps Medium Res 1024 High Res 1536 Highest Res 2048

INTERACTIVE RENDERING

1 ray / pixel

slide-40
SLIDE 40
slide-41
SLIDE 41

check bit check bit j ump down sample

Data Management

GVDB .raw .vdb .obj .vdb .vdb .vdb .vdb .vbx .vbx .vbx OpenVDB Sparse data Dense data Polygonal data Time Series

voxelized resample direct Fast cache format

OptiX

Rendered Images

NVIDIA GVDB SPARSE VOLUMES

slide-42
SLIDE 42

check bit

Data Example: Simulate, Process and Raytrace of 1200 frames

NVIDIA GVDB SPARSE VOLUMES

S PH Simulation

1 min 6 hours

Resample Points to 10243 Grid

On GPU / w Fluids v.3 On GPU S parse resample

1 hours

Write VBX

Disk t ot al 530 GB

9 min Read VBX Raytrace 46 min Rendering: 1200 frames in <1 hour wit h shadows and scat t ering Processing S imulat e Render 4 seconds/ frame Total:

192 rays / pixel

slide-43
SLIDE 43

Acceleration Structures JIT Compiler Scheduler

GPU OptiX API User Application

Mesh Hit Program Ray Shade Program

User Application

Launch!

launch describe

slide-44
SLIDE 44

Acceleration Structures JIT Compiler Scheduler

GPU OptiX API User Application

Mesh Hit Program Ray Shade Program

User Application

Launch!

launch describe

GVDB Hit Program GVDB S hade Program GVDB Structure

describe

* NEW *

slide-45
SLIDE 45
slide-46
SLIDE 46

Poly-to-Voxel Scattering

slide-47
SLIDE 47

Poly-to-Poly Scattering

slide-48
SLIDE 48
slide-49
SLIDE 49

NVIDIA GVDB SPARSE VOLUMES

Interested in GVDB? Let us know! http:/ / developer.nvidia.com/ gvdb

Rama Karl Hoetzlein NVIDIA

slide-50
SLIDE 50

S paceX Stephen Jones Victor X. Guerrero Andre Kessler Ken Museth, Dreamworks Animation NVIDIA NVIDIA IndeX

Tom Fogal Christopher Lux Tristan Lorach Marc Nienhaus Craig Fullman Tom-Michael Thamm Steven Parker

THANKS! TO..

This research used resources of the Oak Ridge Leadership Computing Facility, which is a DOE Office of S cience User Facility supported under Contract DE-AC05-00OR22725

slide-51
SLIDE 51

April 4-7, 2016 | Silicon Valley

THANK YOU

JOIN THE NVIDIA DEVELOPER PROGRAM AT developer.nvidia.com/ j oin GVDB INTEREST SURVEY

developer.nvidia.com/ gvdb