Scalable Ray Tracing Using the Distributed FrameBuffer

Usher, Will; Wald, Ingo; Amstutz, Jefferson; Günther, Johannes; Brownlee, Carson; Pascucci, Valerio

doi:10.1111/cgf.13702

Cited by 13 publications

(7 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It was selected due to its uniqueness and wide acceptance by both industry and the research community as a key production grade benchmark. The second one is based on the Natural History Museum scene [Birn 2015], and is extended by sculpture models [Threedscans 2020] to increase the geometric complexity, and by paintings [The Art Institute of Chicago 2020] to increase the size and number of textures. Finally, the remaining two scenes come from the recent open movies produced by the animation studio of the Blender Institute, namely, Agent 327: Operation Barbershop [Studio 2020a] and Spring [Studio 2020b].…”

Section: Benchmark Scenesmentioning

confidence: 99%

GPU Accelerated Path Tracing of Massive Scenes

et al. 2021

View full text Add to dashboard Cite

This article presents a solution to path tracing of massive scenes on multiple GPUs. Our approach analyzes the memory access pattern of a path tracer and defines how the scene data should be distributed across up to 16 GPUs with minimal effect on performance. The key concept is that the parts of the scene that have the highest amount of memory accesses are replicated on all GPUs. We propose two methods for maximizing the performance of path tracing when working with partially distributed scene data. Both methods work on the memory management level and therefore path tracer data structures do not have to be redesigned, making our approach applicable to other path tracers with only minor changes in their code. As a proof of concept, we have enhanced the open-source Blender Cycles path tracer. The approach was validated on scenes of sizes up to 169 GB. We show that only 1–5% of the scene data needs to be replicated to all machines for such large scenes. On smaller scenes we have verified that the performance is very close to rendering a fully replicated scene. In terms of scalability we have achieved a parallel efficiency of over 94% using up to 16 GPUs.

show abstract

Section: Benchmark Scenesmentioning

confidence: 99%

GPU Accelerated Path Tracing of Massive Scenes

et al. 2021

View full text Add to dashboard Cite

show abstract

“…To efficiently distribute rendering work and communicate across multiple nodes on a cluster, OSPRay leverages a Distributed Frame-Buffer [Usher et al 2019] and MPI. Finally, to provide a high-quality image at low sample counts OSPRay uses Intel's Open Image Denoise library [Intel 2019] for post-process denoising.…”

Section: Performancementioning

confidence: 99%

“…The Distributed FrameBuffer. To efficiently distribute rendering work among nodes and combine the partial results produced by each node, OSPRay uses a Distributed FrameBuffer [Usher et al 2019]. The Distributed FrameBuffer (DFB) is a general framework for executing image compositing and processing tasks for distributed renderers through a distributed, asynchronous tile processing pipeline.…”

Section: Performancementioning

confidence: 99%

Digesting the Elephant -- Experiences with Interactive Production Quality Path Tracing of the Moana Island Scene

Wald¹,

Cherniak²,

Usher³

et al. 2020

Preprint

Self Cite

View full text Add to dashboard Cite

Fig. 1. Left: An overview of the Moana Island Scene (39.3 million instances, 261.1 million unique quads, and 82.4 billion instanced quads). Right: A close-up of the beach, with each primitive colored individually to try to convey the detail of this model. In this paper, we describe the steps taken to allow us to render this model-in its entirety, without simplification, and with a high-quality path tracer-at interactive frame rates.

show abstract

“…3D rendering complexity is dependent on scene complexity as well as the output resolution. With displays becoming more pixel-dense, current doublebuffered techniques demand more compute power [2], [3]. Also, double-buffered rendering keeps an off-screen buffer for the image under construction, and by the time it is displayed, the information is at least two frames old.…”

Section: Introductionmentioning

confidence: 99%

Coherence and Adaptivity in Frameless Rendering-A Practical and Information Theoretic Analysis

Dayal¹,

Gupta²,

Ponnada³

et al. 2021

IEEE Access

View full text Add to dashboard Cite

Virtual and Augmented Reality applications demand low-latency rendering. Adaptive Frameless Rendering (AFR) techniques offer a potential alternative to currently dominating non-adaptive doublebuffered real-time 3D rendering techniques. However, due to their adaptive non-contiguous importance sampling, AFR does not benefit fully from the performance improvements offered by caching the local scene information, resulting in a lower sampling rate. This article presents an experimental, followed by an information-theoretic analysis, of this coherence vs. adaptivity trade-off in AFR. It introduces a novel way to utilize entropy to evaluate the efficiency of a rendering technique. It also presents various spatial coherence exploiting techniques in 3D path tracing and their application to AFR, paving the way to future research in the field.

show abstract

Scalable Ray Tracing Using the Distributed FrameBuffer

Cited by 13 publications

References 22 publications

GPU Accelerated Path Tracing of Massive Scenes

GPU Accelerated Path Tracing of Massive Scenes

Digesting the Elephant -- Experiences with Interactive Production Quality Path Tracing of the Moana Island Scene

Coherence and Adaptivity in Frameless Rendering-A Practical and Information Theoretic Analysis

Contact Info

Product

Resources

About