Syoyo FUJITA Takashi KANAI - SourceForge

Post on 19-May-2022

6 views 0 download

Transcript of Syoyo FUJITA Takashi KANAI - SourceForge

Syoyo FUJITA and Takashi KANAIKeio Univ.

lucille

• Parallel Global Illumination Renderer

• Open source

• Algorithm testbed and reference implementation

Similar works

• Radiance (California Univ.)

• RenderPark (Leuven Univ.)

• lrt (Stanford Univ.)

Features

• Basic Monte Carlo Ray Tracing

• Recent Global Illumination algorithm

Precomputed Radiance Transfer, Photon maping, Image BasedLighting, etc.

• Parallel Rendering

• SIMD Ray Tracing

• Mac OS X, Windows, linux

Design and Architecture

• RIB file format

• Only Triangle geometory

• Uniform Grid

• Screen space Parallel Ray traicing

• HDR image format

Image Based Lightingwith HDRI

IBL Sampling Methods

Naive Cosweight Importance

Stratified Structured

BRDF Models

Diffuse Modified Phong

Ward Anisotropic Ashikhmin

Parallel Rendering

Parallel Photon tracing

Parallel Ray tracing

MPI(inter-node) thread(Multiprocessor)

SIMD Ray tracing

1 ray - 4 tris. simulateously

x86 SSE, PowerPC AltiVec

2.5x faster

Kilauea vs. lucilleKilauea(*) lucille

Parallelism High Low

File Format Shotdata(original format) RenderMan

Thread pthread pthread & Win32 thread

Message Passing Original MPI mpich, etc.

SIMD ? SSE and AltiVec

Shader SPOT Engine(executed in parallel)

RenderMan SL(future plan)

Motion blur & Animation O X

Scene data Distributed Copied

Process Daemon Batch

(*) Reference: “Practical Parallel Rendering”, A.K. Peters, and SIGGRAPH 2002 Course note.

Cornell box

Photon mapping(Diffuse interreflection)

Area light

Sponza Atrium

Famous GI test scene

Photon mapping

texture mapping

Sun light

Teapots2 million tris.

dome light

5 min.(512x512 16 CPUs)

Stanford lucy28 million Tris

Image Based Lighting

1.7 GB memory footprint

2 hours 40 min.(1024x1024, 16 CPUs)

Future work

• Documenation

• Programmable Shader Language

• Modeler Integration, Tools

• quasi-Monte Carlo Rendering

• Implement various G.I. algorithms

http://web.sfc.keio.ac.jp/~syoyo/lucille/